ref: e50f96a4a3c4974f784fa764a29c873a58cca150
parent: cbcfbe1b2987386ff21075d32f12633458556708
author: Johann <johannkoenig@google.com>
date: Mon Mar 5 11:50:33 EST 2012
Move SAD and variance functions to common The MFQE function of the postprocessor depends on these Change-Id: I256a37c6de079fe92ce744b1f11e16526d06b50a
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_sad16x16_armv6.asm
@@ -1,0 +1,96 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_sad16x16_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 const unsigned char *src_ptr
+; r1 int src_stride
+; r2 const unsigned char *ref_ptr
+; r3 int ref_stride
+; stack max_sad (not used)
+|vp8_sad16x16_armv6| PROC
+ stmfd sp!, {r4-r12, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+ pld [r0, r1, lsl #1]
+ pld [r2, r3, lsl #1]
+
+ mov r4, #0 ; sad = 0;
+ mov r5, #8 ; loop count
+
+loop
+ ; 1st row
+ ldr r6, [r0, #0x0] ; load 4 src pixels (1A)
+ ldr r8, [r2, #0x0] ; load 4 ref pixels (1A)
+ ldr r7, [r0, #0x4] ; load 4 src pixels (1A)
+ ldr r9, [r2, #0x4] ; load 4 ref pixels (1A)
+ ldr r10, [r0, #0x8] ; load 4 src pixels (1B)
+ ldr r11, [r0, #0xC] ; load 4 src pixels (1B)
+
+ usada8 r4, r8, r6, r4 ; calculate sad for 4 pixels
+ usad8 r8, r7, r9 ; calculate sad for 4 pixels
+
+ ldr r12, [r2, #0x8] ; load 4 ref pixels (1B)
+ ldr lr, [r2, #0xC] ; load 4 ref pixels (1B)
+
+ add r0, r0, r1 ; set src pointer to next row
+ add r2, r2, r3 ; set dst pointer to next row
+
+ pld [r0, r1, lsl #1]
+ pld [r2, r3, lsl #1]
+
+ usada8 r4, r10, r12, r4 ; calculate sad for 4 pixels
+ usada8 r8, r11, lr, r8 ; calculate sad for 4 pixels
+
+ ldr r6, [r0, #0x0] ; load 4 src pixels (2A)
+ ldr r7, [r0, #0x4] ; load 4 src pixels (2A)
+ add r4, r4, r8 ; add partial sad values
+
+ ; 2nd row
+ ldr r8, [r2, #0x0] ; load 4 ref pixels (2A)
+ ldr r9, [r2, #0x4] ; load 4 ref pixels (2A)
+ ldr r10, [r0, #0x8] ; load 4 src pixels (2B)
+ ldr r11, [r0, #0xC] ; load 4 src pixels (2B)
+
+ usada8 r4, r6, r8, r4 ; calculate sad for 4 pixels
+ usad8 r8, r7, r9 ; calculate sad for 4 pixels
+
+ ldr r12, [r2, #0x8] ; load 4 ref pixels (2B)
+ ldr lr, [r2, #0xC] ; load 4 ref pixels (2B)
+
+ add r0, r0, r1 ; set src pointer to next row
+ add r2, r2, r3 ; set dst pointer to next row
+
+ usada8 r4, r10, r12, r4 ; calculate sad for 4 pixels
+ usada8 r8, r11, lr, r8 ; calculate sad for 4 pixels
+
+ pld [r0, r1, lsl #1]
+ pld [r2, r3, lsl #1]
+
+ subs r5, r5, #1 ; decrement loop counter
+ add r4, r4, r8 ; add partial sad values
+
+ bne loop
+
+ mov r0, r4 ; return sad
+ ldmfd sp!, {r4-r12, pc}+
+ ENDP
+
+ END
+
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_variance16x16_armv6.asm
@@ -1,0 +1,154 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance16x16_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance16x16_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+
+ mov r8, #0 ; initialize sum = 0
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+
+loop
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r5, [r2, #0] ; load 4 ref pixels
+
+ mov lr, #0 ; constant zero
+
+ usub8 r6, r4, r5 ; calculate difference
+ pld [r0, r1, lsl #1]
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r9, r5, r4 ; calculate difference with reversed operands
+ pld [r2, r3, lsl #1]
+ sel r6, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r5, [r2, #4] ; load 4 ref pixels
+ smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r9, r5, r4 ; calculate difference with reversed operands
+ sel r6, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r5, [r2, #8] ; load 4 ref pixels
+ smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r9, r5, r4 ; calculate difference with reversed operands
+ sel r6, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r5, [r2, #12] ; load 4 ref pixels
+ smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r9, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
+
+
+ subs r12, r12, #1
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}+
+ ENDP
+
+ END
+
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_variance8x8_armv6.asm
@@ -1,0 +1,101 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance8x8_armv6|
+
+ ARM
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance8x8_armv6| PROC
+
+ push {r4-r10, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+
+ mov r12, #8 ; set loop counter to 8 (=block height)
+ mov r4, #0 ; initialize sum = 0
+ mov r5, #0 ; initialize sse = 0
+
+loop
+ ; 1st 4 pixels
+ ldr r6, [r0, #0x0] ; load 4 src pixels
+ ldr r7, [r2, #0x0] ; load 4 ref pixels
+
+ mov lr, #0 ; constant zero
+
+ usub8 r8, r6, r7 ; calculate difference
+ pld [r0, r1, lsl #1]
+ sel r10, r8, lr ; select bytes with positive difference
+ usub8 r9, r7, r6 ; calculate difference with reversed operands
+ pld [r2, r3, lsl #1]
+ sel r8, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r6, r10, lr ; calculate sum of positive differences
+ usad8 r7, r8, lr ; calculate sum of negative differences
+ orr r8, r8, r10 ; differences of all 4 pixels
+ ; calculate total sum
+ add r4, r4, r6 ; add positive differences to sum
+ sub r4, r4, r7 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r7, r8 ; byte (two pixels) to halfwords
+ uxtb16 r10, r8, ror #8 ; another two pixels to halfwords
+ smlad r5, r7, r7, r5 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r6, [r0, #0x4] ; load 4 src pixels
+ ldr r7, [r2, #0x4] ; load 4 ref pixels
+ smlad r5, r10, r10, r5 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r8, r6, r7 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r10, r8, lr ; select bytes with positive difference
+ usub8 r9, r7, r6 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r8, r9, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r6, r10, lr ; calculate sum of positive differences
+ usad8 r7, r8, lr ; calculate sum of negative differences
+ orr r8, r8, r10 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r4, r4, r6 ; add positive differences to sum
+ sub r4, r4, r7 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r7, r8 ; byte (two pixels) to halfwords
+ uxtb16 r10, r8, ror #8 ; another two pixels to halfwords
+ smlad r5, r7, r7, r5 ; dual signed multiply, add and accumulate (1)
+ subs r12, r12, #1 ; next row
+ smlad r5, r10, r10, r5 ; dual signed multiply, add and accumulate (2)
+
+ bne loop
+
+ ; return stuff
+ ldr r8, [sp, #32] ; get address of sse
+ mul r1, r4, r4 ; sum * sum
+ str r5, [r8] ; store sse
+ sub r0, r5, r1, ASR #6 ; return (sse - ((sum * sum) >> 6))
+
+ pop {r4-r10, pc}+
+ ENDP
+
+ END
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6.asm
@@ -1,0 +1,182 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_h_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_h_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r6, [r0, #1] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #0] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ pld [r0, r1, lsl #1]
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ pld [r2, r3, lsl #1]
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r6, [r0, #5] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #4] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r6, [r0, #9] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #8] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r6, [r0, #13] ; load 4 src pixels with 1 byte offset
+ ldr r5, [r2, #12] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ subs r12, r12, #1
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
+
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6.asm
@@ -1,0 +1,222 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_hv_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_hv_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ add r9, r0, r1 ; pointer to pixels on the next row
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load source pixels a, row N
+ ldr r6, [r0, #1] ; load source pixels b, row N
+ ldr r5, [r9, #0] ; load source pixels c, row N+1
+ ldr r7, [r9, #1] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #0] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ pld [r0, r1, lsl #1]
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ pld [r2, r3, lsl #1]
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load source pixels a, row N
+ ldr r6, [r0, #5] ; load source pixels b, row N
+ ldr r5, [r9, #4] ; load source pixels c, row N+1
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ ldr r7, [r9, #5] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #4] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load source pixels a, row N
+ ldr r6, [r0, #9] ; load source pixels b, row N
+ ldr r5, [r9, #8] ; load source pixels c, row N+1
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ ldr r7, [r9, #9] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #8] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load source pixels a, row N
+ ldr r6, [r0, #13] ; load source pixels b, row N
+ ldr r5, [r9, #12] ; load source pixels c, row N+1
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+ ldr r7, [r9, #13] ; load source pixels d, row N+1
+
+ ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+ ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
+ mvn r7, r7
+ uhsub8 r5, r5, r7
+ eor r5, r5, r10
+ ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
+ mvn r5, r5
+ uhsub8 r4, r4, r5
+ ldr r5, [r2, #12] ; load 4 ref pixels
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ subs r12, r12, #1
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
--- /dev/null
+++ b/vp8/common/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6.asm
@@ -1,0 +1,184 @@
+;
+; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_v_armv6|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance_halfpixvar16x16_v_armv6| PROC
+
+ stmfd sp!, {r4-r12, lr}+
+ pld [r0, r1, lsl #0]
+ pld [r2, r3, lsl #0]
+
+ mov r8, #0 ; initialize sum = 0
+ ldr r10, c80808080
+ mov r11, #0 ; initialize sse = 0
+ mov r12, #16 ; set loop counter to 16 (=block height)
+ mov lr, #0 ; constant zero
+loop
+ add r9, r0, r1 ; set src pointer to next row
+ ; 1st 4 pixels
+ ldr r4, [r0, #0] ; load 4 src pixels
+ ldr r6, [r9, #0] ; load 4 src pixels from next row
+ ldr r5, [r2, #0] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ usub8 r6, r4, r5 ; calculate difference
+ pld [r0, r1, lsl #1]
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ pld [r2, r3, lsl #1]
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+ ; calculate total sum
+ adds r8, r8, r4 ; add positive differences to sum
+ subs r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 2nd 4 pixels
+ ldr r4, [r0, #4] ; load 4 src pixels
+ ldr r6, [r9, #4] ; load 4 src pixels from next row
+ ldr r5, [r2, #4] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 3rd 4 pixels
+ ldr r4, [r0, #8] ; load 4 src pixels
+ ldr r6, [r9, #8] ; load 4 src pixels from next row
+ ldr r5, [r2, #8] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+
+ ; 4th 4 pixels
+ ldr r4, [r0, #12] ; load 4 src pixels
+ ldr r6, [r9, #12] ; load 4 src pixels from next row
+ ldr r5, [r2, #12] ; load 4 ref pixels
+
+ ; bilinear interpolation
+ mvn r6, r6
+ uhsub8 r4, r4, r6
+ eor r4, r4, r10
+
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+ usub8 r6, r4, r5 ; calculate difference
+ add r0, r0, r1 ; set src_ptr to next row
+ sel r7, r6, lr ; select bytes with positive difference
+ usub8 r6, r5, r4 ; calculate difference with reversed operands
+ add r2, r2, r3 ; set dst_ptr to next row
+ sel r6, r6, lr ; select bytes with negative difference
+
+ ; calculate partial sums
+ usad8 r4, r7, lr ; calculate sum of positive differences
+ usad8 r5, r6, lr ; calculate sum of negative differences
+ orr r6, r6, r7 ; differences of all 4 pixels
+
+ ; calculate total sum
+ add r8, r8, r4 ; add positive differences to sum
+ sub r8, r8, r5 ; substract negative differences from sum
+
+ ; calculate sse
+ uxtb16 r5, r6 ; byte (two pixels) to halfwords
+ uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
+ smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
+ smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
+
+
+ subs r12, r12, #1
+
+ bne loop
+
+ ; return stuff
+ ldr r6, [sp, #40] ; get address of sse
+ mul r0, r8, r8 ; sum * sum
+ str r11, [r6] ; store sse
+ sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
+
+ ldmfd sp!, {r4-r12, pc}+
+ ENDP
+
+c80808080
+ DCD 0x80808080
+
+ END
+
--- /dev/null
+++ b/vp8/common/arm/neon/sad16_neon.asm
@@ -1,0 +1,207 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_sad16x16_neon|
+ EXPORT |vp8_sad16x8_neon|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int src_stride
+; r2 unsigned char *ref_ptr
+; r3 int ref_stride
+|vp8_sad16x16_neon| PROC
+;;
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabdl.u8 q12, d0, d8
+ vabdl.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0], r1+ vld1.8 {q7}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+;;
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabal.u8 q12, d0, d8
+ vabal.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0], r1+ vld1.8 {q7}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+;;
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabal.u8 q12, d0, d8
+ vabal.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0], r1+ vld1.8 {q7}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+;;
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabal.u8 q12, d0, d8
+ vabal.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0]+ vld1.8 {q7}, [r2]+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vadd.u16 q0, q12, q13
+
+ vpaddl.u16 q1, q0
+ vpaddl.u32 q0, q1
+
+ vadd.u32 d0, d0, d1
+
+ vmov.32 r0, d0[0]
+
+ bx lr
+
+ ENDP
+
+;==============================
+;unsigned int vp8_sad16x8_c(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+|vp8_sad16x8_neon| PROC
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabdl.u8 q12, d0, d8
+ vabdl.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0], r1+ vld1.8 {q7}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+ vld1.8 {q0}, [r0], r1+ vld1.8 {q4}, [r2], r3+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vld1.8 {q1}, [r0], r1+ vld1.8 {q5}, [r2], r3+
+ vabal.u8 q12, d0, d8
+ vabal.u8 q13, d1, d9
+
+ vld1.8 {q2}, [r0], r1+ vld1.8 {q6}, [r2], r3+
+ vabal.u8 q12, d2, d10
+ vabal.u8 q13, d3, d11
+
+ vld1.8 {q3}, [r0], r1+ vld1.8 {q7}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q13, d5, d13
+
+ vabal.u8 q12, d6, d14
+ vabal.u8 q13, d7, d15
+
+ vadd.u16 q0, q12, q13
+
+ vpaddl.u16 q1, q0
+ vpaddl.u32 q0, q1
+
+ vadd.u32 d0, d0, d1
+
+ vmov.32 r0, d0[0]
+
+ bx lr
+
+ ENDP
+
+ END
--- /dev/null
+++ b/vp8/common/arm/neon/sad8_neon.asm
@@ -1,0 +1,209 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_sad8x8_neon|
+ EXPORT |vp8_sad8x16_neon|
+ EXPORT |vp8_sad4x4_neon|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+; unsigned int vp8_sad8x8_c(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+
+|vp8_sad8x8_neon| PROC
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabdl.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vabal.u8 q12, d6, d14
+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabal.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q12, d6, d14
+
+ vpaddl.u16 q1, q12
+ vpaddl.u32 q0, q1
+ vadd.u32 d0, d0, d1
+
+ vmov.32 r0, d0[0]
+
+ bx lr
+
+ ENDP
+
+;============================
+;unsigned int vp8_sad8x16_c(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+
+|vp8_sad8x16_neon| PROC
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabdl.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vabal.u8 q12, d6, d14
+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabal.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vabal.u8 q12, d6, d14
+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabal.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vabal.u8 q12, d6, d14
+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabal.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q12, d6, d14
+
+ vpaddl.u16 q1, q12
+ vpaddl.u32 q0, q1
+ vadd.u32 d0, d0, d1
+
+ vmov.32 r0, d0[0]
+
+ bx lr
+
+ ENDP
+
+;===========================
+;unsigned int vp8_sad4x4_c(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+
+|vp8_sad4x4_neon| PROC
+ vld1.8 {d0}, [r0], r1+ vld1.8 {d8}, [r2], r3+
+ vld1.8 {d2}, [r0], r1+ vld1.8 {d10}, [r2], r3+
+ vabdl.u8 q12, d0, d8
+
+ vld1.8 {d4}, [r0], r1+ vld1.8 {d12}, [r2], r3+
+ vabal.u8 q12, d2, d10
+
+ vld1.8 {d6}, [r0], r1+ vld1.8 {d14}, [r2], r3+
+ vabal.u8 q12, d4, d12
+ vabal.u8 q12, d6, d14
+
+ vpaddl.u16 d1, d24
+ vpaddl.u32 d0, d1
+ vmov.32 r0, d0[0]
+
+ bx lr
+
+ ENDP
+
+ END
--- /dev/null
+++ b/vp8/common/arm/neon/variance_neon.asm
@@ -1,0 +1,276 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance16x16_neon|
+ EXPORT |vp8_variance16x8_neon|
+ EXPORT |vp8_variance8x16_neon|
+ EXPORT |vp8_variance8x8_neon|
+
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance16x16_neon| PROC
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ mov r12, #8
+
+variance16x16_neon_loop
+ vld1.8 {q0}, [r0], r1 ;Load up source and reference+ vld1.8 {q2}, [r2], r3+ vld1.8 {q1}, [r0], r1+ vld1.8 {q3}, [r2], r3+
+ vsubl.u8 q11, d0, d4 ;calculate diff
+ vsubl.u8 q12, d1, d5
+ vsubl.u8 q13, d2, d6
+ vsubl.u8 q14, d3, d7
+
+ ;VPADAL adds adjacent pairs of elements of a vector, and accumulates
+ ;the results into the elements of the destination vector. The explanation
+ ;in ARM guide is wrong.
+ vpadal.s16 q8, q11 ;calculate sum
+ vmlal.s16 q9, d22, d22 ;calculate sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ bne variance16x16_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ ldr r12, [sp] ;load *sse from stack
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ ;vmov.32 r0, d0[0] ;this instruction costs a lot
+ ;vmov.32 r1, d1[0]
+ ;mul r0, r0, r0
+ ;str r1, [r12]
+ ;sub r0, r1, r0, lsr #8
+
+ ; while sum is signed, sum * sum is always positive and must be treated as
+ ; unsigned to avoid propagating the sign bit.
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [r12] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ bx lr
+
+ ENDP
+
+;================================
+;unsigned int vp8_variance16x8_c(
+; unsigned char *src_ptr,
+; int source_stride,
+; unsigned char *ref_ptr,
+; int recon_stride,
+; unsigned int *sse)
+|vp8_variance16x8_neon| PROC
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ mov r12, #4
+
+variance16x8_neon_loop
+ vld1.8 {q0}, [r0], r1 ;Load up source and reference+ vld1.8 {q2}, [r2], r3+ vld1.8 {q1}, [r0], r1+ vld1.8 {q3}, [r2], r3+
+ vsubl.u8 q11, d0, d4 ;calculate diff
+ vsubl.u8 q12, d1, d5
+ vsubl.u8 q13, d2, d6
+ vsubl.u8 q14, d3, d7
+
+ vpadal.s16 q8, q11 ;calculate sum
+ vmlal.s16 q9, d22, d22 ;calculate sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ bne variance16x8_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ ldr r12, [sp] ;load *sse from stack
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [r12] ;store sse+ vshr.u32 d10, d10, #7
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ bx lr
+
+ ENDP
+
+;=================================
+;unsigned int vp8_variance8x16_c(
+; unsigned char *src_ptr,
+; int source_stride,
+; unsigned char *ref_ptr,
+; int recon_stride,
+; unsigned int *sse)
+
+|vp8_variance8x16_neon| PROC
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ mov r12, #8
+
+variance8x16_neon_loop
+ vld1.8 {d0}, [r0], r1 ;Load up source and reference+ vld1.8 {d4}, [r2], r3+ vld1.8 {d2}, [r0], r1+ vld1.8 {d6}, [r2], r3+
+ vsubl.u8 q11, d0, d4 ;calculate diff
+ vsubl.u8 q12, d2, d6
+
+ vpadal.s16 q8, q11 ;calculate sum
+ vmlal.s16 q9, d22, d22 ;calculate sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+
+ bne variance8x16_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ ldr r12, [sp] ;load *sse from stack
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [r12] ;store sse+ vshr.u32 d10, d10, #7
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ bx lr
+
+ ENDP
+
+;==================================
+; r0 unsigned char *src_ptr
+; r1 int source_stride
+; r2 unsigned char *ref_ptr
+; r3 int recon_stride
+; stack unsigned int *sse
+|vp8_variance8x8_neon| PROC
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ mov r12, #2
+
+variance8x8_neon_loop
+ vld1.8 {d0}, [r0], r1 ;Load up source and reference+ vld1.8 {d4}, [r2], r3+ vld1.8 {d1}, [r0], r1+ vld1.8 {d5}, [r2], r3+ vld1.8 {d2}, [r0], r1+ vld1.8 {d6}, [r2], r3+ vld1.8 {d3}, [r0], r1+ vld1.8 {d7}, [r2], r3+
+ vsubl.u8 q11, d0, d4 ;calculate diff
+ vsubl.u8 q12, d1, d5
+ vsubl.u8 q13, d2, d6
+ vsubl.u8 q14, d3, d7
+
+ vpadal.s16 q8, q11 ;calculate sum
+ vmlal.s16 q9, d22, d22 ;calculate sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ bne variance8x8_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ ldr r12, [sp] ;load *sse from stack
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [r12] ;store sse+ vshr.u32 d10, d10, #6
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ bx lr
+
+ ENDP
+
+ END
--- /dev/null
+++ b/vp8/common/arm/neon/vp8_subpixelvariance16x16_neon.asm
@@ -1,0 +1,425 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_sub_pixel_variance16x16_neon_func|
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+; r0 unsigned char *src_ptr,
+; r1 int src_pixels_per_line,
+; r2 int xoffset,
+; r3 int yoffset,
+; stack(r4) unsigned char *dst_ptr,
+; stack(r5) int dst_pixels_per_line,
+; stack(r6) unsigned int *sse
+;note: most of the code is copied from bilinear_predict16x16_neon and vp8_variance16x16_neon.
+
+|vp8_sub_pixel_variance16x16_neon_func| PROC
+ push {r4-r6, lr}+
+ ldr r12, _BilinearTaps_coeff_
+ ldr r4, [sp, #16] ;load *dst_ptr from stack
+ ldr r5, [sp, #20] ;load dst_pixels_per_line from stack
+ ldr r6, [sp, #24] ;load *sse from stack
+
+ cmp r2, #0 ;skip first_pass filter if xoffset=0
+ beq secondpass_bfilter16x16_only
+
+ add r2, r12, r2, lsl #3 ;calculate filter location
+
+ cmp r3, #0 ;skip second_pass filter if yoffset=0
+
+ vld1.s32 {d31}, [r2] ;load first_pass filter+
+ beq firstpass_bfilter16x16_only
+
+ sub sp, sp, #272 ;reserve space on stack for temporary storage
+ vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data+ mov lr, sp
+ vld1.u8 {d5, d6, d7}, [r0], r1+
+ mov r2, #3 ;loop counter
+ vld1.u8 {d8, d9, d10}, [r0], r1+
+ vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
+ vld1.u8 {d11, d12, d13}, [r0], r1+
+ vdup.8 d1, d31[4]
+
+;First Pass: output_height lines x output_width columns (17x16)
+vp8e_filt_blk2d_fp16x16_loop_neon
+ pld [r0]
+ pld [r0, r1]
+ pld [r0, r1, lsl #1]
+
+ vmull.u8 q7, d2, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d5, d0
+ vmull.u8 q10, d6, d0
+ vmull.u8 q11, d8, d0
+ vmull.u8 q12, d9, d0
+ vmull.u8 q13, d11, d0
+ vmull.u8 q14, d12, d0
+
+ vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
+ vext.8 d5, d5, d6, #1
+ vext.8 d8, d8, d9, #1
+ vext.8 d11, d11, d12, #1
+
+ vmlal.u8 q7, d2, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q9, d5, d1
+ vmlal.u8 q11, d8, d1
+ vmlal.u8 q13, d11, d1
+
+ vext.8 d3, d3, d4, #1
+ vext.8 d6, d6, d7, #1
+ vext.8 d9, d9, d10, #1
+ vext.8 d12, d12, d13, #1
+
+ vmlal.u8 q8, d3, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q10, d6, d1
+ vmlal.u8 q12, d9, d1
+ vmlal.u8 q14, d12, d1
+
+ subs r2, r2, #1
+
+ vqrshrn.u16 d14, q7, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d15, q8, #7
+ vqrshrn.u16 d16, q9, #7
+ vqrshrn.u16 d17, q10, #7
+ vqrshrn.u16 d18, q11, #7
+ vqrshrn.u16 d19, q12, #7
+ vqrshrn.u16 d20, q13, #7
+
+ vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data+ vqrshrn.u16 d21, q14, #7
+ vld1.u8 {d5, d6, d7}, [r0], r1+
+ vst1.u8 {d14, d15, d16, d17}, [lr]! ;store result+ vld1.u8 {d8, d9, d10}, [r0], r1+ vst1.u8 {d18, d19, d20, d21}, [lr]!+ vld1.u8 {d11, d12, d13}, [r0], r1+
+ bne vp8e_filt_blk2d_fp16x16_loop_neon
+
+;First-pass filtering for rest 5 lines
+ vld1.u8 {d14, d15, d16}, [r0], r1+
+ vmull.u8 q9, d2, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q10, d3, d0
+ vmull.u8 q11, d5, d0
+ vmull.u8 q12, d6, d0
+ vmull.u8 q13, d8, d0
+ vmull.u8 q14, d9, d0
+
+ vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
+ vext.8 d5, d5, d6, #1
+ vext.8 d8, d8, d9, #1
+
+ vmlal.u8 q9, d2, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q11, d5, d1
+ vmlal.u8 q13, d8, d1
+
+ vext.8 d3, d3, d4, #1
+ vext.8 d6, d6, d7, #1
+ vext.8 d9, d9, d10, #1
+
+ vmlal.u8 q10, d3, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q12, d6, d1
+ vmlal.u8 q14, d9, d1
+
+ vmull.u8 q1, d11, d0
+ vmull.u8 q2, d12, d0
+ vmull.u8 q3, d14, d0
+ vmull.u8 q4, d15, d0
+
+ vext.8 d11, d11, d12, #1 ;construct src_ptr[1]
+ vext.8 d14, d14, d15, #1
+
+ vmlal.u8 q1, d11, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q3, d14, d1
+
+ vext.8 d12, d12, d13, #1
+ vext.8 d15, d15, d16, #1
+
+ vmlal.u8 q2, d12, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q4, d15, d1
+
+ vqrshrn.u16 d10, q9, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d11, q10, #7
+ vqrshrn.u16 d12, q11, #7
+ vqrshrn.u16 d13, q12, #7
+ vqrshrn.u16 d14, q13, #7
+ vqrshrn.u16 d15, q14, #7
+ vqrshrn.u16 d16, q1, #7
+ vqrshrn.u16 d17, q2, #7
+ vqrshrn.u16 d18, q3, #7
+ vqrshrn.u16 d19, q4, #7
+
+ vst1.u8 {d10, d11, d12, d13}, [lr]! ;store result+ vst1.u8 {d14, d15, d16, d17}, [lr]!+ vst1.u8 {d18, d19}, [lr]!+
+;Second pass: 16x16
+;secondpass_filter
+ add r3, r12, r3, lsl #3
+ sub lr, lr, #272
+
+ vld1.u32 {d31}, [r3] ;load second_pass filter+
+ sub sp, sp, #256
+ mov r3, sp
+
+ vld1.u8 {d22, d23}, [lr]! ;load src data+
+ vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
+ vdup.8 d1, d31[4]
+ mov r12, #4 ;loop counter
+
+vp8e_filt_blk2d_sp16x16_loop_neon
+ vld1.u8 {d24, d25}, [lr]!+ vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
+ vld1.u8 {d26, d27}, [lr]!+ vmull.u8 q2, d23, d0
+ vld1.u8 {d28, d29}, [lr]!+ vmull.u8 q3, d24, d0
+ vld1.u8 {d30, d31}, [lr]!+
+ vmull.u8 q4, d25, d0
+ vmull.u8 q5, d26, d0
+ vmull.u8 q6, d27, d0
+ vmull.u8 q7, d28, d0
+ vmull.u8 q8, d29, d0
+
+ vmlal.u8 q1, d24, d1 ;(src_ptr[pixel_step] * Filter[1])
+ vmlal.u8 q2, d25, d1
+ vmlal.u8 q3, d26, d1
+ vmlal.u8 q4, d27, d1
+ vmlal.u8 q5, d28, d1
+ vmlal.u8 q6, d29, d1
+ vmlal.u8 q7, d30, d1
+ vmlal.u8 q8, d31, d1
+
+ subs r12, r12, #1
+
+ vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d3, q2, #7
+ vqrshrn.u16 d4, q3, #7
+ vqrshrn.u16 d5, q4, #7
+ vqrshrn.u16 d6, q5, #7
+ vqrshrn.u16 d7, q6, #7
+ vqrshrn.u16 d8, q7, #7
+ vqrshrn.u16 d9, q8, #7
+
+ vst1.u8 {d2, d3}, [r3]! ;store result+ vst1.u8 {d4, d5}, [r3]!+ vst1.u8 {d6, d7}, [r3]!+ vmov q11, q15
+ vst1.u8 {d8, d9}, [r3]!+
+ bne vp8e_filt_blk2d_sp16x16_loop_neon
+
+ b sub_pixel_variance16x16_neon
+
+;--------------------
+firstpass_bfilter16x16_only
+ mov r2, #4 ;loop counter
+ sub sp, sp, #528 ;reserve space on stack for temporary storage
+ vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
+ vdup.8 d1, d31[4]
+ mov r3, sp
+
+;First Pass: output_height lines x output_width columns (16x16)
+vp8e_filt_blk2d_fpo16x16_loop_neon
+ vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data+ vld1.u8 {d5, d6, d7}, [r0], r1+ vld1.u8 {d8, d9, d10}, [r0], r1+ vld1.u8 {d11, d12, d13}, [r0], r1+
+ pld [r0]
+ pld [r0, r1]
+ pld [r0, r1, lsl #1]
+
+ vmull.u8 q7, d2, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q8, d3, d0
+ vmull.u8 q9, d5, d0
+ vmull.u8 q10, d6, d0
+ vmull.u8 q11, d8, d0
+ vmull.u8 q12, d9, d0
+ vmull.u8 q13, d11, d0
+ vmull.u8 q14, d12, d0
+
+ vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
+ vext.8 d5, d5, d6, #1
+ vext.8 d8, d8, d9, #1
+ vext.8 d11, d11, d12, #1
+
+ vmlal.u8 q7, d2, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q9, d5, d1
+ vmlal.u8 q11, d8, d1
+ vmlal.u8 q13, d11, d1
+
+ vext.8 d3, d3, d4, #1
+ vext.8 d6, d6, d7, #1
+ vext.8 d9, d9, d10, #1
+ vext.8 d12, d12, d13, #1
+
+ vmlal.u8 q8, d3, d1 ;(src_ptr[0] * Filter[1])
+ vmlal.u8 q10, d6, d1
+ vmlal.u8 q12, d9, d1
+ vmlal.u8 q14, d12, d1
+
+ subs r2, r2, #1
+
+ vqrshrn.u16 d14, q7, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d15, q8, #7
+ vqrshrn.u16 d16, q9, #7
+ vqrshrn.u16 d17, q10, #7
+ vqrshrn.u16 d18, q11, #7
+ vqrshrn.u16 d19, q12, #7
+ vqrshrn.u16 d20, q13, #7
+ vst1.u8 {d14, d15}, [r3]! ;store result+ vqrshrn.u16 d21, q14, #7
+
+ vst1.u8 {d16, d17}, [r3]!+ vst1.u8 {d18, d19}, [r3]!+ vst1.u8 {d20, d21}, [r3]!+
+ bne vp8e_filt_blk2d_fpo16x16_loop_neon
+
+ b sub_pixel_variance16x16_neon
+
+;---------------------
+secondpass_bfilter16x16_only
+;Second pass: 16x16
+;secondpass_filter
+ sub sp, sp, #528 ;reserve space on stack for temporary storage
+ add r3, r12, r3, lsl #3
+ mov r12, #4 ;loop counter
+ vld1.u32 {d31}, [r3] ;load second_pass filter+ vld1.u8 {d22, d23}, [r0], r1 ;load src data+ mov r3, sp
+
+ vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
+ vdup.8 d1, d31[4]
+
+vp8e_filt_blk2d_spo16x16_loop_neon
+ vld1.u8 {d24, d25}, [r0], r1+ vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
+ vld1.u8 {d26, d27}, [r0], r1+ vmull.u8 q2, d23, d0
+ vld1.u8 {d28, d29}, [r0], r1+ vmull.u8 q3, d24, d0
+ vld1.u8 {d30, d31}, [r0], r1+
+ vmull.u8 q4, d25, d0
+ vmull.u8 q5, d26, d0
+ vmull.u8 q6, d27, d0
+ vmull.u8 q7, d28, d0
+ vmull.u8 q8, d29, d0
+
+ vmlal.u8 q1, d24, d1 ;(src_ptr[pixel_step] * Filter[1])
+ vmlal.u8 q2, d25, d1
+ vmlal.u8 q3, d26, d1
+ vmlal.u8 q4, d27, d1
+ vmlal.u8 q5, d28, d1
+ vmlal.u8 q6, d29, d1
+ vmlal.u8 q7, d30, d1
+ vmlal.u8 q8, d31, d1
+
+ vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d3, q2, #7
+ vqrshrn.u16 d4, q3, #7
+ vqrshrn.u16 d5, q4, #7
+ vqrshrn.u16 d6, q5, #7
+ vqrshrn.u16 d7, q6, #7
+ vqrshrn.u16 d8, q7, #7
+ vqrshrn.u16 d9, q8, #7
+
+ vst1.u8 {d2, d3}, [r3]! ;store result+ subs r12, r12, #1
+ vst1.u8 {d4, d5}, [r3]!+ vmov q11, q15
+ vst1.u8 {d6, d7}, [r3]!+ vst1.u8 {d8, d9}, [r3]!+
+ bne vp8e_filt_blk2d_spo16x16_loop_neon
+
+ b sub_pixel_variance16x16_neon
+
+;----------------------------
+;variance16x16
+sub_pixel_variance16x16_neon
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ sub r3, r3, #256
+ mov r12, #8
+
+sub_pixel_variance16x16_neon_loop
+ vld1.8 {q0}, [r3]! ;Load up source and reference+ vld1.8 {q2}, [r4], r5+ vld1.8 {q1}, [r3]!+ vld1.8 {q3}, [r4], r5+
+ vsubl.u8 q11, d0, d4 ;diff
+ vsubl.u8 q12, d1, d5
+ vsubl.u8 q13, d2, d6
+ vsubl.u8 q14, d3, d7
+
+ vpadal.s16 q8, q11 ;sum
+ vmlal.s16 q9, d22, d22 ;sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ bne sub_pixel_variance16x16_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [r6] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ add sp, sp, #528
+ vmov.32 r0, d0[0] ;return
+
+ pop {r4-r6,pc}+
+ ENDP
+
+;-----------------
+
+_BilinearTaps_coeff_
+ DCD bilinear_taps_coeff
+bilinear_taps_coeff
+ DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
+
+ END
--- /dev/null
+++ b/vp8/common/arm/neon/vp8_subpixelvariance16x16s_neon.asm
@@ -1,0 +1,572 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_variance_halfpixvar16x16_h_neon|
+ EXPORT |vp8_variance_halfpixvar16x16_v_neon|
+ EXPORT |vp8_variance_halfpixvar16x16_hv_neon|
+ EXPORT |vp8_sub_pixel_variance16x16s_neon|
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+
+;================================================
+;unsigned int vp8_variance_halfpixvar16x16_h_neon
+;(
+; unsigned char *src_ptr, r0
+; int src_pixels_per_line, r1
+; unsigned char *dst_ptr, r2
+; int dst_pixels_per_line, r3
+; unsigned int *sse
+;);
+;================================================
+|vp8_variance_halfpixvar16x16_h_neon| PROC
+ push {lr}+
+ mov r12, #4 ;loop counter
+ ldr lr, [sp, #4] ;load *sse from stack
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+;First Pass: output_height lines x output_width columns (16x16)
+vp8_filt_fpo16x16s_4_0_loop_neon
+ vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data+ vld1.8 {q11}, [r2], r3+ vld1.u8 {d4, d5, d6, d7}, [r0], r1+ vld1.8 {q12}, [r2], r3+ vld1.u8 {d8, d9, d10, d11}, [r0], r1+ vld1.8 {q13}, [r2], r3+ vld1.u8 {d12, d13, d14, d15}, [r0], r1+
+ ;pld [r0]
+ ;pld [r0, r1]
+ ;pld [r0, r1, lsl #1]
+
+ vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
+ vext.8 q3, q2, q3, #1
+ vext.8 q5, q4, q5, #1
+ vext.8 q7, q6, q7, #1
+
+ vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+ vld1.8 {q14}, [r2], r3+ vrhadd.u8 q1, q2, q3
+ vrhadd.u8 q2, q4, q5
+ vrhadd.u8 q3, q6, q7
+
+ vsubl.u8 q4, d0, d22 ;diff
+ vsubl.u8 q5, d1, d23
+ vsubl.u8 q6, d2, d24
+ vsubl.u8 q7, d3, d25
+ vsubl.u8 q0, d4, d26
+ vsubl.u8 q1, d5, d27
+ vsubl.u8 q2, d6, d28
+ vsubl.u8 q3, d7, d29
+
+ vpadal.s16 q8, q4 ;sum
+ vmlal.s16 q9, d8, d8 ;sse
+ vmlal.s16 q10, d9, d9
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q5
+ vmlal.s16 q9, d10, d10
+ vmlal.s16 q10, d11, d11
+ vpadal.s16 q8, q6
+ vmlal.s16 q9, d12, d12
+ vmlal.s16 q10, d13, d13
+ vpadal.s16 q8, q7
+ vmlal.s16 q9, d14, d14
+ vmlal.s16 q10, d15, d15
+
+ vpadal.s16 q8, q0 ;sum
+ vmlal.s16 q9, d0, d0 ;sse
+ vmlal.s16 q10, d1, d1
+ vpadal.s16 q8, q1
+ vmlal.s16 q9, d2, d2
+ vmlal.s16 q10, d3, d3
+ vpadal.s16 q8, q2
+ vmlal.s16 q9, d4, d4
+ vmlal.s16 q10, d5, d5
+ vpadal.s16 q8, q3
+ vmlal.s16 q9, d6, d6
+ vmlal.s16 q10, d7, d7
+
+ bne vp8_filt_fpo16x16s_4_0_loop_neon
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [lr] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ pop {pc}+ ENDP
+
+;================================================
+;unsigned int vp8_variance_halfpixvar16x16_v_neon
+;(
+; unsigned char *src_ptr, r0
+; int src_pixels_per_line, r1
+; unsigned char *dst_ptr, r2
+; int dst_pixels_per_line, r3
+; unsigned int *sse
+;);
+;================================================
+|vp8_variance_halfpixvar16x16_v_neon| PROC
+ push {lr}+
+ mov r12, #4 ;loop counter
+
+ vld1.u8 {q0}, [r0], r1 ;load src data+ ldr lr, [sp, #4] ;load *sse from stack
+
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+vp8_filt_spo16x16s_0_4_loop_neon
+ vld1.u8 {q2}, [r0], r1+ vld1.8 {q1}, [r2], r3+ vld1.u8 {q4}, [r0], r1+ vld1.8 {q3}, [r2], r3+ vld1.u8 {q6}, [r0], r1+ vld1.8 {q5}, [r2], r3+ vld1.u8 {q15}, [r0], r1+
+ vrhadd.u8 q0, q0, q2
+ vld1.8 {q7}, [r2], r3+ vrhadd.u8 q2, q2, q4
+ vrhadd.u8 q4, q4, q6
+ vrhadd.u8 q6, q6, q15
+
+ vsubl.u8 q11, d0, d2 ;diff
+ vsubl.u8 q12, d1, d3
+ vsubl.u8 q13, d4, d6
+ vsubl.u8 q14, d5, d7
+ vsubl.u8 q0, d8, d10
+ vsubl.u8 q1, d9, d11
+ vsubl.u8 q2, d12, d14
+ vsubl.u8 q3, d13, d15
+
+ vpadal.s16 q8, q11 ;sum
+ vmlal.s16 q9, d22, d22 ;sse
+ vmlal.s16 q10, d23, d23
+
+ subs r12, r12, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ vpadal.s16 q8, q0 ;sum
+ vmlal.s16 q9, d0, d0 ;sse
+ vmlal.s16 q10, d1, d1
+ vpadal.s16 q8, q1
+ vmlal.s16 q9, d2, d2
+ vmlal.s16 q10, d3, d3
+ vpadal.s16 q8, q2
+ vmlal.s16 q9, d4, d4
+ vmlal.s16 q10, d5, d5
+
+ vmov q0, q15
+
+ vpadal.s16 q8, q3
+ vmlal.s16 q9, d6, d6
+ vmlal.s16 q10, d7, d7
+
+ bne vp8_filt_spo16x16s_0_4_loop_neon
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [lr] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ pop {pc}+ ENDP
+
+;================================================
+;unsigned int vp8_variance_halfpixvar16x16_hv_neon
+;(
+; unsigned char *src_ptr, r0
+; int src_pixels_per_line, r1
+; unsigned char *dst_ptr, r2
+; int dst_pixels_per_line, r3
+; unsigned int *sse
+;);
+;================================================
+|vp8_variance_halfpixvar16x16_hv_neon| PROC
+ push {lr}+
+ vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data+
+ ldr lr, [sp, #4] ;load *sse from stack
+ vmov.i8 q13, #0 ;q8 - sum
+ vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
+
+ vmov.i8 q14, #0 ;q9, q10 - sse
+ vmov.i8 q15, #0
+
+ mov r12, #4 ;loop counter
+ vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+
+;First Pass: output_height lines x output_width columns (17x16)
+vp8_filt16x16s_4_4_loop_neon
+ vld1.u8 {d4, d5, d6, d7}, [r0], r1+ vld1.u8 {d8, d9, d10, d11}, [r0], r1+ vld1.u8 {d12, d13, d14, d15}, [r0], r1+ vld1.u8 {d16, d17, d18, d19}, [r0], r1+
+ ;pld [r0]
+ ;pld [r0, r1]
+ ;pld [r0, r1, lsl #1]
+
+ vext.8 q3, q2, q3, #1 ;construct src_ptr[1]
+ vext.8 q5, q4, q5, #1
+ vext.8 q7, q6, q7, #1
+ vext.8 q9, q8, q9, #1
+
+ vrhadd.u8 q1, q2, q3 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+ vrhadd.u8 q2, q4, q5
+ vrhadd.u8 q3, q6, q7
+ vrhadd.u8 q4, q8, q9
+
+ vld1.8 {q5}, [r2], r3+ vrhadd.u8 q0, q0, q1
+ vld1.8 {q6}, [r2], r3+ vrhadd.u8 q1, q1, q2
+ vld1.8 {q7}, [r2], r3+ vrhadd.u8 q2, q2, q3
+ vld1.8 {q8}, [r2], r3+ vrhadd.u8 q3, q3, q4
+
+ vsubl.u8 q9, d0, d10 ;diff
+ vsubl.u8 q10, d1, d11
+ vsubl.u8 q11, d2, d12
+ vsubl.u8 q12, d3, d13
+
+ vsubl.u8 q0, d4, d14 ;diff
+ vsubl.u8 q1, d5, d15
+ vsubl.u8 q5, d6, d16
+ vsubl.u8 q6, d7, d17
+
+ vpadal.s16 q13, q9 ;sum
+ vmlal.s16 q14, d18, d18 ;sse
+ vmlal.s16 q15, d19, d19
+
+ vpadal.s16 q13, q10 ;sum
+ vmlal.s16 q14, d20, d20 ;sse
+ vmlal.s16 q15, d21, d21
+
+ vpadal.s16 q13, q11 ;sum
+ vmlal.s16 q14, d22, d22 ;sse
+ vmlal.s16 q15, d23, d23
+
+ vpadal.s16 q13, q12 ;sum
+ vmlal.s16 q14, d24, d24 ;sse
+ vmlal.s16 q15, d25, d25
+
+ subs r12, r12, #1
+
+ vpadal.s16 q13, q0 ;sum
+ vmlal.s16 q14, d0, d0 ;sse
+ vmlal.s16 q15, d1, d1
+
+ vpadal.s16 q13, q1 ;sum
+ vmlal.s16 q14, d2, d2 ;sse
+ vmlal.s16 q15, d3, d3
+
+ vpadal.s16 q13, q5 ;sum
+ vmlal.s16 q14, d10, d10 ;sse
+ vmlal.s16 q15, d11, d11
+
+ vmov q0, q4
+
+ vpadal.s16 q13, q6 ;sum
+ vmlal.s16 q14, d12, d12 ;sse
+ vmlal.s16 q15, d13, d13
+
+ bne vp8_filt16x16s_4_4_loop_neon
+
+ vadd.u32 q15, q14, q15 ;accumulate sse
+ vpaddl.s32 q0, q13 ;accumulate sum
+
+ vpaddl.u32 q1, q15
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [lr] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ pop {pc}+ ENDP
+
+;==============================
+; r0 unsigned char *src_ptr,
+; r1 int src_pixels_per_line,
+; r2 int xoffset,
+; r3 int yoffset,
+; stack unsigned char *dst_ptr,
+; stack int dst_pixels_per_line,
+; stack unsigned int *sse
+;note: in vp8_find_best_half_pixel_step()(called when 8<Speed<15), and first call of vp8_find_best_sub_pixel_step()
+;(called when speed<=8). xoffset/yoffset can only be 4 or 0, which means either by pass the filter,
+;or filter coeff is {64, 64}. This simplified program only works in this situation.+;note: It happens that both xoffset and yoffset are zero. This can be handled in c code later.
+
+|vp8_sub_pixel_variance16x16s_neon| PROC
+ push {r4, lr}+
+ ldr r4, [sp, #8] ;load *dst_ptr from stack
+ ldr r12, [sp, #12] ;load dst_pixels_per_line from stack
+ ldr lr, [sp, #16] ;load *sse from stack
+
+ cmp r2, #0 ;skip first_pass filter if xoffset=0
+ beq secondpass_bfilter16x16s_only
+
+ cmp r3, #0 ;skip second_pass filter if yoffset=0
+ beq firstpass_bfilter16x16s_only
+
+ vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data+ sub sp, sp, #256 ;reserve space on stack for temporary storage
+ vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
+ mov r3, sp
+ mov r2, #4 ;loop counter
+ vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+
+;First Pass: output_height lines x output_width columns (17x16)
+vp8e_filt_blk2d_fp16x16s_loop_neon
+ vld1.u8 {d4, d5, d6, d7}, [r0], r1+ vld1.u8 {d8, d9, d10, d11}, [r0], r1+ vld1.u8 {d12, d13, d14, d15}, [r0], r1+ vld1.u8 {d16, d17, d18, d19}, [r0], r1+
+ ;pld [r0]
+ ;pld [r0, r1]
+ ;pld [r0, r1, lsl #1]
+
+ vext.8 q3, q2, q3, #1 ;construct src_ptr[1]
+ vext.8 q5, q4, q5, #1
+ vext.8 q7, q6, q7, #1
+ vext.8 q9, q8, q9, #1
+
+ vrhadd.u8 q1, q2, q3 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+ vrhadd.u8 q2, q4, q5
+ vrhadd.u8 q3, q6, q7
+ vrhadd.u8 q4, q8, q9
+
+ vrhadd.u8 q0, q0, q1
+ vrhadd.u8 q1, q1, q2
+ vrhadd.u8 q2, q2, q3
+ vrhadd.u8 q3, q3, q4
+
+ subs r2, r2, #1
+ vst1.u8 {d0, d1 ,d2, d3}, [r3]! ;store result+ vmov q0, q4
+ vst1.u8 {d4, d5, d6, d7}, [r3]!+
+ bne vp8e_filt_blk2d_fp16x16s_loop_neon
+
+ b sub_pixel_variance16x16s_neon
+
+;--------------------
+firstpass_bfilter16x16s_only
+ mov r2, #2 ;loop counter
+ sub sp, sp, #256 ;reserve space on stack for temporary storage
+ mov r3, sp
+
+;First Pass: output_height lines x output_width columns (16x16)
+vp8e_filt_blk2d_fpo16x16s_loop_neon
+ vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data+ vld1.u8 {d4, d5, d6, d7}, [r0], r1+ vld1.u8 {d8, d9, d10, d11}, [r0], r1+ vld1.u8 {d12, d13, d14, d15}, [r0], r1+
+ ;pld [r0]
+ ;pld [r0, r1]
+ ;pld [r0, r1, lsl #1]
+
+ vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
+ vld1.u8 {d16, d17, d18, d19}, [r0], r1+ vext.8 q3, q2, q3, #1
+ vld1.u8 {d20, d21, d22, d23}, [r0], r1+ vext.8 q5, q4, q5, #1
+ vld1.u8 {d24, d25, d26, d27}, [r0], r1+ vext.8 q7, q6, q7, #1
+ vld1.u8 {d28, d29, d30, d31}, [r0], r1+ vext.8 q9, q8, q9, #1
+ vext.8 q11, q10, q11, #1
+ vext.8 q13, q12, q13, #1
+ vext.8 q15, q14, q15, #1
+
+ vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
+ vrhadd.u8 q1, q2, q3
+ vrhadd.u8 q2, q4, q5
+ vrhadd.u8 q3, q6, q7
+ vrhadd.u8 q4, q8, q9
+ vrhadd.u8 q5, q10, q11
+ vrhadd.u8 q6, q12, q13
+ vrhadd.u8 q7, q14, q15
+
+ subs r2, r2, #1
+
+ vst1.u8 {d0, d1, d2, d3}, [r3]! ;store result+ vst1.u8 {d4, d5, d6, d7}, [r3]!+ vst1.u8 {d8, d9, d10, d11}, [r3]!+ vst1.u8 {d12, d13, d14, d15}, [r3]!+
+ bne vp8e_filt_blk2d_fpo16x16s_loop_neon
+
+ b sub_pixel_variance16x16s_neon
+
+;---------------------
+secondpass_bfilter16x16s_only
+ sub sp, sp, #256 ;reserve space on stack for temporary storage
+
+ mov r2, #2 ;loop counter
+ vld1.u8 {d0, d1}, [r0], r1 ;load src data+ mov r3, sp
+
+vp8e_filt_blk2d_spo16x16s_loop_neon
+ vld1.u8 {d2, d3}, [r0], r1+ vld1.u8 {d4, d5}, [r0], r1+ vld1.u8 {d6, d7}, [r0], r1+ vld1.u8 {d8, d9}, [r0], r1+
+ vrhadd.u8 q0, q0, q1
+ vld1.u8 {d10, d11}, [r0], r1+ vrhadd.u8 q1, q1, q2
+ vld1.u8 {d12, d13}, [r0], r1+ vrhadd.u8 q2, q2, q3
+ vld1.u8 {d14, d15}, [r0], r1+ vrhadd.u8 q3, q3, q4
+ vld1.u8 {d16, d17}, [r0], r1+ vrhadd.u8 q4, q4, q5
+ vrhadd.u8 q5, q5, q6
+ vrhadd.u8 q6, q6, q7
+ vrhadd.u8 q7, q7, q8
+
+ subs r2, r2, #1
+
+ vst1.u8 {d0, d1, d2, d3}, [r3]! ;store result+ vmov q0, q8
+ vst1.u8 {d4, d5, d6, d7}, [r3]!+ vst1.u8 {d8, d9, d10, d11}, [r3]! ;store result+ vst1.u8 {d12, d13, d14, d15}, [r3]!+
+ bne vp8e_filt_blk2d_spo16x16s_loop_neon
+
+ b sub_pixel_variance16x16s_neon
+
+;----------------------------
+;variance16x16
+sub_pixel_variance16x16s_neon
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ sub r3, r3, #256
+ mov r2, #4
+
+sub_pixel_variance16x16s_neon_loop
+ vld1.8 {q0}, [r3]! ;Load up source and reference+ vld1.8 {q1}, [r4], r12+ vld1.8 {q2}, [r3]!+ vld1.8 {q3}, [r4], r12+ vld1.8 {q4}, [r3]!+ vld1.8 {q5}, [r4], r12+ vld1.8 {q6}, [r3]!+ vld1.8 {q7}, [r4], r12+
+ vsubl.u8 q11, d0, d2 ;diff
+ vsubl.u8 q12, d1, d3
+ vsubl.u8 q13, d4, d6
+ vsubl.u8 q14, d5, d7
+ vsubl.u8 q0, d8, d10
+ vsubl.u8 q1, d9, d11
+ vsubl.u8 q2, d12, d14
+ vsubl.u8 q3, d13, d15
+
+ vpadal.s16 q8, q11 ;sum
+ vmlal.s16 q9, d22, d22 ;sse
+ vmlal.s16 q10, d23, d23
+
+ subs r2, r2, #1
+
+ vpadal.s16 q8, q12
+ vmlal.s16 q9, d24, d24
+ vmlal.s16 q10, d25, d25
+ vpadal.s16 q8, q13
+ vmlal.s16 q9, d26, d26
+ vmlal.s16 q10, d27, d27
+ vpadal.s16 q8, q14
+ vmlal.s16 q9, d28, d28
+ vmlal.s16 q10, d29, d29
+
+ vpadal.s16 q8, q0 ;sum
+ vmlal.s16 q9, d0, d0 ;sse
+ vmlal.s16 q10, d1, d1
+ vpadal.s16 q8, q1
+ vmlal.s16 q9, d2, d2
+ vmlal.s16 q10, d3, d3
+ vpadal.s16 q8, q2
+ vmlal.s16 q9, d4, d4
+ vmlal.s16 q10, d5, d5
+ vpadal.s16 q8, q3
+ vmlal.s16 q9, d6, d6
+ vmlal.s16 q10, d7, d7
+
+ bne sub_pixel_variance16x16s_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [lr] ;store sse+ vshr.u32 d10, d10, #8
+ vsub.u32 d0, d1, d10
+
+ add sp, sp, #256
+ vmov.32 r0, d0[0] ;return
+
+ pop {r4, pc}+ ENDP
+
+ END
--- /dev/null
+++ b/vp8/common/arm/neon/vp8_subpixelvariance8x8_neon.asm
@@ -1,0 +1,224 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ EXPORT |vp8_sub_pixel_variance8x8_neon|
+ ARM
+ REQUIRE8
+ PRESERVE8
+
+ AREA ||.text||, CODE, READONLY, ALIGN=2
+; r0 unsigned char *src_ptr,
+; r1 int src_pixels_per_line,
+; r2 int xoffset,
+; r3 int yoffset,
+; stack(r4) unsigned char *dst_ptr,
+; stack(r5) int dst_pixels_per_line,
+; stack(r6) unsigned int *sse
+;note: most of the code is copied from bilinear_predict8x8_neon and vp8_variance8x8_neon.
+
+|vp8_sub_pixel_variance8x8_neon| PROC
+ push {r4-r5, lr}+
+ ldr r12, _BilinearTaps_coeff_
+ ldr r4, [sp, #12] ;load *dst_ptr from stack
+ ldr r5, [sp, #16] ;load dst_pixels_per_line from stack
+ ldr lr, [sp, #20] ;load *sse from stack
+
+ cmp r2, #0 ;skip first_pass filter if xoffset=0
+ beq skip_firstpass_filter
+
+;First pass: output_height lines x output_width columns (9x8)
+ add r2, r12, r2, lsl #3 ;calculate filter location
+
+ vld1.u8 {q1}, [r0], r1 ;load src data+ vld1.u32 {d31}, [r2] ;load first_pass filter+ vld1.u8 {q2}, [r0], r1+ vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
+ vld1.u8 {q3}, [r0], r1+ vdup.8 d1, d31[4]
+ vld1.u8 {q4}, [r0], r1+
+ vmull.u8 q6, d2, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q7, d4, d0
+ vmull.u8 q8, d6, d0
+ vmull.u8 q9, d8, d0
+
+ vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
+ vext.8 d5, d4, d5, #1
+ vext.8 d7, d6, d7, #1
+ vext.8 d9, d8, d9, #1
+
+ vmlal.u8 q6, d3, d1 ;(src_ptr[1] * Filter[1])
+ vmlal.u8 q7, d5, d1
+ vmlal.u8 q8, d7, d1
+ vmlal.u8 q9, d9, d1
+
+ vld1.u8 {q1}, [r0], r1 ;load src data+ vqrshrn.u16 d22, q6, #7 ;shift/round/saturate to u8
+ vld1.u8 {q2}, [r0], r1+ vqrshrn.u16 d23, q7, #7
+ vld1.u8 {q3}, [r0], r1+ vqrshrn.u16 d24, q8, #7
+ vld1.u8 {q4}, [r0], r1+ vqrshrn.u16 d25, q9, #7
+
+ ;first_pass filtering on the rest 5-line data
+ vld1.u8 {q5}, [r0], r1+
+ vmull.u8 q6, d2, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q7, d4, d0
+ vmull.u8 q8, d6, d0
+ vmull.u8 q9, d8, d0
+ vmull.u8 q10, d10, d0
+
+ vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
+ vext.8 d5, d4, d5, #1
+ vext.8 d7, d6, d7, #1
+ vext.8 d9, d8, d9, #1
+ vext.8 d11, d10, d11, #1
+
+ vmlal.u8 q6, d3, d1 ;(src_ptr[1] * Filter[1])
+ vmlal.u8 q7, d5, d1
+ vmlal.u8 q8, d7, d1
+ vmlal.u8 q9, d9, d1
+ vmlal.u8 q10, d11, d1
+
+ vqrshrn.u16 d26, q6, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d27, q7, #7
+ vqrshrn.u16 d28, q8, #7
+ vqrshrn.u16 d29, q9, #7
+ vqrshrn.u16 d30, q10, #7
+
+;Second pass: 8x8
+secondpass_filter
+ cmp r3, #0 ;skip second_pass filter if yoffset=0
+ ;skip_secondpass_filter
+ beq sub_pixel_variance8x8_neon
+
+ add r3, r12, r3, lsl #3
+
+ vld1.u32 {d31}, [r3] ;load second_pass filter+
+ vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
+ vdup.8 d1, d31[4]
+
+ vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
+ vmull.u8 q2, d23, d0
+ vmull.u8 q3, d24, d0
+ vmull.u8 q4, d25, d0
+ vmull.u8 q5, d26, d0
+ vmull.u8 q6, d27, d0
+ vmull.u8 q7, d28, d0
+ vmull.u8 q8, d29, d0
+
+ vmlal.u8 q1, d23, d1 ;(src_ptr[pixel_step] * Filter[1])
+ vmlal.u8 q2, d24, d1
+ vmlal.u8 q3, d25, d1
+ vmlal.u8 q4, d26, d1
+ vmlal.u8 q5, d27, d1
+ vmlal.u8 q6, d28, d1
+ vmlal.u8 q7, d29, d1
+ vmlal.u8 q8, d30, d1
+
+ vqrshrn.u16 d22, q1, #7 ;shift/round/saturate to u8
+ vqrshrn.u16 d23, q2, #7
+ vqrshrn.u16 d24, q3, #7
+ vqrshrn.u16 d25, q4, #7
+ vqrshrn.u16 d26, q5, #7
+ vqrshrn.u16 d27, q6, #7
+ vqrshrn.u16 d28, q7, #7
+ vqrshrn.u16 d29, q8, #7
+
+ b sub_pixel_variance8x8_neon
+
+;--------------------
+skip_firstpass_filter
+ vld1.u8 {d22}, [r0], r1 ;load src data+ vld1.u8 {d23}, [r0], r1+ vld1.u8 {d24}, [r0], r1+ vld1.u8 {d25}, [r0], r1+ vld1.u8 {d26}, [r0], r1+ vld1.u8 {d27}, [r0], r1+ vld1.u8 {d28}, [r0], r1+ vld1.u8 {d29}, [r0], r1+ vld1.u8 {d30}, [r0], r1+
+ b secondpass_filter
+
+;----------------------
+;vp8_variance8x8_neon
+sub_pixel_variance8x8_neon
+ vmov.i8 q8, #0 ;q8 - sum
+ vmov.i8 q9, #0 ;q9, q10 - sse
+ vmov.i8 q10, #0
+
+ mov r12, #2
+
+sub_pixel_variance8x8_neon_loop
+ vld1.8 {d0}, [r4], r5 ;load dst data+ subs r12, r12, #1
+ vld1.8 {d1}, [r4], r5+ vld1.8 {d2}, [r4], r5+ vsubl.u8 q4, d22, d0 ;calculate diff
+ vld1.8 {d3}, [r4], r5+
+ vsubl.u8 q5, d23, d1
+ vsubl.u8 q6, d24, d2
+
+ vpadal.s16 q8, q4 ;sum
+ vmlal.s16 q9, d8, d8 ;sse
+ vmlal.s16 q10, d9, d9
+
+ vsubl.u8 q7, d25, d3
+
+ vpadal.s16 q8, q5
+ vmlal.s16 q9, d10, d10
+ vmlal.s16 q10, d11, d11
+
+ vmov q11, q13
+
+ vpadal.s16 q8, q6
+ vmlal.s16 q9, d12, d12
+ vmlal.s16 q10, d13, d13
+
+ vmov q12, q14
+
+ vpadal.s16 q8, q7
+ vmlal.s16 q9, d14, d14
+ vmlal.s16 q10, d15, d15
+
+ bne sub_pixel_variance8x8_neon_loop
+
+ vadd.u32 q10, q9, q10 ;accumulate sse
+ vpaddl.s32 q0, q8 ;accumulate sum
+
+ vpaddl.u32 q1, q10
+ vadd.s64 d0, d0, d1
+ vadd.u64 d1, d2, d3
+
+ vmull.s32 q5, d0, d0
+ vst1.32 {d1[0]}, [lr] ;store sse+ vshr.u32 d10, d10, #6
+ vsub.u32 d0, d1, d10
+
+ vmov.32 r0, d0[0] ;return
+ pop {r4-r5, pc}+
+ ENDP
+
+;-----------------
+
+_BilinearTaps_coeff_
+ DCD bilinear_taps_coeff
+bilinear_taps_coeff
+ DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
+
+ END
--- /dev/null
+++ b/vp8/common/arm/variance_arm.c
@@ -1,0 +1,121 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_config.h"
+#include "vpx_rtcd.h"
+#include "vp8/common/variance.h"
+#include "vp8/common/filter.h"
+
+#if HAVE_MEDIA
+#include "vp8/common/arm/bilinearfilter_arm.h"
+
+unsigned int vp8_sub_pixel_variance8x8_armv6
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short first_pass[10*8];
+ unsigned char second_pass[8*8];
+ const short *HFilter, *VFilter;
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
+ src_pixels_per_line,
+ 9, 8, HFilter);
+ vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
+ 8, 8, 8, VFilter);
+
+ return vp8_variance8x8_armv6(second_pass, 8, dst_ptr,
+ dst_pixels_per_line, sse);
+}
+
+unsigned int vp8_sub_pixel_variance16x16_armv6
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short first_pass[36*16];
+ unsigned char second_pass[20*16];
+ const short *HFilter, *VFilter;
+ unsigned int var;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ var = vp8_variance_halfpixvar16x16_h_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ var = vp8_variance_halfpixvar16x16_v_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ var = vp8_variance_halfpixvar16x16_hv_armv6(src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, sse);
+ }
+ else
+ {+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
+ src_pixels_per_line,
+ 17, 16, HFilter);
+ vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
+ 16, 16, 16, VFilter);
+
+ var = vp8_variance16x16_armv6(second_pass, 16, dst_ptr,
+ dst_pixels_per_line, sse);
+ }
+ return var;
+}
+
+#endif /* HAVE_MEDIA */
+
+
+#if HAVE_NEON
+
+unsigned int vp8_sub_pixel_variance16x16_neon
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ if (xoffset == 4 && yoffset == 0)
+ return vp8_variance_halfpixvar16x16_h_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
+ else if (xoffset == 0 && yoffset == 4)
+ return vp8_variance_halfpixvar16x16_v_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
+ else if (xoffset == 4 && yoffset == 4)
+ return vp8_variance_halfpixvar16x16_hv_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
+ else
+ return vp8_sub_pixel_variance16x16_neon_func(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
+}
+
+#endif
--- a/vp8/common/postproc.c
+++ b/vp8/common/postproc.c
@@ -17,7 +17,7 @@
#include "vpx_scale/yv12extend.h"
#include "vpx_scale/vpxscale.h"
#include "systemdependent.h"
-#include "../encoder/variance.h"
+#include "variance.h"
#include <limits.h>
#include <math.h>
--- /dev/null
+++ b/vp8/common/ppc/sad_altivec.asm
@@ -1,0 +1,277 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ .globl vp8_sad16x16_ppc
+ .globl vp8_sad16x8_ppc
+ .globl vp8_sad8x16_ppc
+ .globl vp8_sad8x8_ppc
+ .globl vp8_sad4x4_ppc
+
+.macro load_aligned_16 V R O
+ lvsl v3, 0, \R ;# permutate value for alignment
+
+ lvx v1, 0, \R
+ lvx v2, \O, \R
+
+ vperm \V, v1, v2, v3
+.endm
+
+.macro prologue
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xffc0
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1, -32(r1) ;# create space on the stack
+
+ li r10, 16 ;# load offset and loop counter
+
+ vspltisw v8, 0 ;# zero out total to start
+.endm
+
+.macro epilogue
+ addi r1, r1, 32 ;# recover stack
+
+ mtspr 256, r11 ;# reset old VRSAVE
+.endm
+
+.macro SAD_16
+ ;# v6 = abs (v4 - v5)
+ vsububs v6, v4, v5
+ vsububs v7, v5, v4
+ vor v6, v6, v7
+
+ ;# v8 += abs (v4 - v5)
+ vsum4ubs v8, v6, v8
+.endm
+
+.macro sad_16_loop loop_label
+ lvsl v3, 0, r5 ;# only needs to be done once per block
+
+ ;# preload a line of data before getting into the loop
+ lvx v4, 0, r3
+ lvx v1, 0, r5
+ lvx v2, r10, r5
+
+ add r5, r5, r6
+ add r3, r3, r4
+
+ vperm v5, v1, v2, v3
+
+ .align 4
+\loop_label:
+ ;# compute difference on first row
+ vsububs v6, v4, v5
+ vsububs v7, v5, v4
+
+ ;# load up next set of data
+ lvx v9, 0, r3
+ lvx v1, 0, r5
+ lvx v2, r10, r5
+
+ ;# perform abs() of difference
+ vor v6, v6, v7
+ add r3, r3, r4
+
+ ;# add to the running tally
+ vsum4ubs v8, v6, v8
+
+ ;# now onto the next line
+ vperm v5, v1, v2, v3
+ add r5, r5, r6
+ lvx v4, 0, r3
+
+ ;# compute difference on second row
+ vsububs v6, v9, v5
+ lvx v1, 0, r5
+ vsububs v7, v5, v9
+ lvx v2, r10, r5
+ vor v6, v6, v7
+ add r3, r3, r4
+ vsum4ubs v8, v6, v8
+ vperm v5, v1, v2, v3
+ add r5, r5, r6
+
+ bdnz \loop_label
+
+ vspltisw v7, 0
+
+ vsumsws v8, v8, v7
+
+ stvx v8, 0, r1
+ lwz r3, 12(r1)
+.endm
+
+.macro sad_8_loop loop_label
+ .align 4
+\loop_label:
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v4, r3, r10
+ load_aligned_16 v5, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v6, r3, r10
+ load_aligned_16 v7, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ vmrghb v4, v4, v6
+ vmrghb v5, v5, v7
+
+ SAD_16
+
+ bdnz \loop_label
+
+ vspltisw v7, 0
+
+ vsumsws v8, v8, v7
+
+ stvx v8, 0, r1
+ lwz r3, 12(r1)
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int ref_stride
+;#
+;# r3 return value
+vp8_sad16x16_ppc:
+
+ prologue
+
+ li r9, 8
+ mtctr r9
+
+ sad_16_loop sad16x16_loop
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int ref_stride
+;#
+;# r3 return value
+vp8_sad16x8_ppc:
+
+ prologue
+
+ li r9, 4
+ mtctr r9
+
+ sad_16_loop sad16x8_loop
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int ref_stride
+;#
+;# r3 return value
+vp8_sad8x16_ppc:
+
+ prologue
+
+ li r9, 8
+ mtctr r9
+
+ sad_8_loop sad8x16_loop
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int ref_stride
+;#
+;# r3 return value
+vp8_sad8x8_ppc:
+
+ prologue
+
+ li r9, 4
+ mtctr r9
+
+ sad_8_loop sad8x8_loop
+
+ epilogue
+
+ blr
+
+.macro transfer_4x4 I P
+ lwz r0, 0(\I)
+ add \I, \I, \P
+
+ lwz r7, 0(\I)
+ add \I, \I, \P
+
+ lwz r8, 0(\I)
+ add \I, \I, \P
+
+ lwz r9, 0(\I)
+
+ stw r0, 0(r1)
+ stw r7, 4(r1)
+ stw r8, 8(r1)
+ stw r9, 12(r1)
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int ref_stride
+;#
+;# r3 return value
+vp8_sad4x4_ppc:
+
+ prologue
+
+ transfer_4x4 r3, r4
+ lvx v4, 0, r1
+
+ transfer_4x4 r5, r6
+ lvx v5, 0, r1
+
+ vspltisw v8, 0 ;# zero out total to start
+
+ ;# v6 = abs (v4 - v5)
+ vsububs v6, v4, v5
+ vsububs v7, v5, v4
+ vor v6, v6, v7
+
+ ;# v8 += abs (v4 - v5)
+ vsum4ubs v7, v6, v8
+ vsumsws v7, v7, v8
+
+ stvx v7, 0, r1
+ lwz r3, 12(r1)
+
+ epilogue
+
+ blr
--- /dev/null
+++ b/vp8/common/ppc/variance_altivec.asm
@@ -1,0 +1,375 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ .globl vp8_get8x8var_ppc
+ .globl vp8_get16x16var_ppc
+ .globl vp8_mse16x16_ppc
+ .globl vp8_variance16x16_ppc
+ .globl vp8_variance16x8_ppc
+ .globl vp8_variance8x16_ppc
+ .globl vp8_variance8x8_ppc
+ .globl vp8_variance4x4_ppc
+
+.macro load_aligned_16 V R O
+ lvsl v3, 0, \R ;# permutate value for alignment
+
+ lvx v1, 0, \R
+ lvx v2, \O, \R
+
+ vperm \V, v1, v2, v3
+.endm
+
+.macro prologue
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xffc0
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1, -32(r1) ;# create space on the stack
+
+ li r10, 16 ;# load offset and loop counter
+
+ vspltisw v7, 0 ;# zero for merging
+ vspltisw v8, 0 ;# zero out total to start
+ vspltisw v9, 0 ;# zero out total for dif^2
+.endm
+
+.macro epilogue
+ addi r1, r1, 32 ;# recover stack
+
+ mtspr 256, r11 ;# reset old VRSAVE
+.endm
+
+.macro compute_sum_sse
+ ;# Compute sum first. Unpack to so signed subract
+ ;# can be used. Only have a half word signed
+ ;# subract. Do high, then low.
+ vmrghb v2, v7, v4
+ vmrghb v3, v7, v5
+ vsubshs v2, v2, v3
+ vsum4shs v8, v2, v8
+
+ vmrglb v2, v7, v4
+ vmrglb v3, v7, v5
+ vsubshs v2, v2, v3
+ vsum4shs v8, v2, v8
+
+ ;# Now compute sse.
+ vsububs v2, v4, v5
+ vsububs v3, v5, v4
+ vor v2, v2, v3
+
+ vmsumubm v9, v2, v2, v9
+.endm
+
+.macro variance_16 DS loop_label store_sum
+\loop_label:
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v4, r3, r10
+ load_aligned_16 v5, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ compute_sum_sse
+
+ bdnz \loop_label
+
+ vsumsws v8, v8, v7
+ vsumsws v9, v9, v7
+
+ stvx v8, 0, r1
+ lwz r3, 12(r1)
+
+ stvx v9, 0, r1
+ lwz r4, 12(r1)
+
+.if \store_sum
+ stw r3, 0(r8) ;# sum
+.endif
+ stw r4, 0(r7) ;# sse
+
+ mullw r3, r3, r3 ;# sum*sum
+ srlwi r3, r3, \DS ;# (sum*sum) >> DS
+ subf r3, r3, r4 ;# sse - ((sum*sum) >> DS)
+.endm
+
+.macro variance_8 DS loop_label store_sum
+\loop_label:
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v4, r3, r10
+ load_aligned_16 v5, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v6, r3, r10
+ load_aligned_16 v0, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ vmrghb v4, v4, v6
+ vmrghb v5, v5, v0
+
+ compute_sum_sse
+
+ bdnz \loop_label
+
+ vsumsws v8, v8, v7
+ vsumsws v9, v9, v7
+
+ stvx v8, 0, r1
+ lwz r3, 12(r1)
+
+ stvx v9, 0, r1
+ lwz r4, 12(r1)
+
+.if \store_sum
+ stw r3, 0(r8) ;# sum
+.endif
+ stw r4, 0(r7) ;# sse
+
+ mullw r3, r3, r3 ;# sum*sum
+ srlwi r3, r3, \DS ;# (sum*sum) >> 8
+ subf r3, r3, r4 ;# sse - ((sum*sum) >> 8)
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *SSE
+;# r8 int *Sum
+;#
+;# r3 return value
+vp8_get8x8var_ppc:
+
+ prologue
+
+ li r9, 4
+ mtctr r9
+
+ variance_8 6, get8x8var_loop, 1
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *SSE
+;# r8 int *Sum
+;#
+;# r3 return value
+vp8_get16x16var_ppc:
+
+ prologue
+
+ mtctr r10
+
+ variance_16 8, get16x16var_loop, 1
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r 3 return value
+vp8_mse16x16_ppc:
+ prologue
+
+ mtctr r10
+
+mse16x16_loop:
+ ;# only one of the inputs should need to be aligned.
+ load_aligned_16 v4, r3, r10
+ load_aligned_16 v5, r5, r10
+
+ ;# move onto the next line
+ add r3, r3, r4
+ add r5, r5, r6
+
+ ;# Now compute sse.
+ vsububs v2, v4, v5
+ vsububs v3, v5, v4
+ vor v2, v2, v3
+
+ vmsumubm v9, v2, v2, v9
+
+ bdnz mse16x16_loop
+
+ vsumsws v9, v9, v7
+
+ stvx v9, 0, r1
+ lwz r3, 12(r1)
+
+ stvx v9, 0, r1
+ lwz r3, 12(r1)
+
+ stw r3, 0(r7) ;# sse
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r3 return value
+vp8_variance16x16_ppc:
+
+ prologue
+
+ mtctr r10
+
+ variance_16 8, variance16x16_loop, 0
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r3 return value
+vp8_variance16x8_ppc:
+
+ prologue
+
+ li r9, 8
+ mtctr r9
+
+ variance_16 7, variance16x8_loop, 0
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r3 return value
+vp8_variance8x16_ppc:
+
+ prologue
+
+ li r9, 8
+ mtctr r9
+
+ variance_8 7, variance8x16_loop, 0
+
+ epilogue
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r3 return value
+vp8_variance8x8_ppc:
+
+ prologue
+
+ li r9, 4
+ mtctr r9
+
+ variance_8 6, variance8x8_loop, 0
+
+ epilogue
+
+ blr
+
+.macro transfer_4x4 I P
+ lwz r0, 0(\I)
+ add \I, \I, \P
+
+ lwz r10,0(\I)
+ add \I, \I, \P
+
+ lwz r8, 0(\I)
+ add \I, \I, \P
+
+ lwz r9, 0(\I)
+
+ stw r0, 0(r1)
+ stw r10, 4(r1)
+ stw r8, 8(r1)
+ stw r9, 12(r1)
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int source_stride
+;# r5 unsigned char *ref_ptr
+;# r6 int recon_stride
+;# r7 unsigned int *sse
+;#
+;# r3 return value
+vp8_variance4x4_ppc:
+
+ prologue
+
+ transfer_4x4 r3, r4
+ lvx v4, 0, r1
+
+ transfer_4x4 r5, r6
+ lvx v5, 0, r1
+
+ compute_sum_sse
+
+ vsumsws v8, v8, v7
+ vsumsws v9, v9, v7
+
+ stvx v8, 0, r1
+ lwz r3, 12(r1)
+
+ stvx v9, 0, r1
+ lwz r4, 12(r1)
+
+ stw r4, 0(r7) ;# sse
+
+ mullw r3, r3, r3 ;# sum*sum
+ srlwi r3, r3, 4 ;# (sum*sum) >> 4
+ subf r3, r3, r4 ;# sse - ((sum*sum) >> 4)
+
+ epilogue
+
+ blr
--- /dev/null
+++ b/vp8/common/ppc/variance_subpixel_altivec.asm
@@ -1,0 +1,865 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+ .globl vp8_sub_pixel_variance4x4_ppc
+ .globl vp8_sub_pixel_variance8x8_ppc
+ .globl vp8_sub_pixel_variance8x16_ppc
+ .globl vp8_sub_pixel_variance16x8_ppc
+ .globl vp8_sub_pixel_variance16x16_ppc
+
+.macro load_c V, LABEL, OFF, R0, R1
+ lis \R0, \LABEL@ha
+ la \R1, \LABEL@l(\R0)
+ lvx \V, \OFF, \R1
+.endm
+
+.macro load_vfilter V0, V1
+ load_c \V0, vfilter_b, r6, r12, r10
+
+ addi r6, r6, 16
+ lvx \V1, r6, r10
+.endm
+
+.macro HProlog jump_label
+ ;# load up horizontal filter
+ slwi. r5, r5, 4 ;# index into horizontal filter array
+
+ ;# index to the next set of vectors in the row.
+ li r10, 16
+
+ ;# downshift by 7 ( divide by 128 ) at the end
+ vspltish v19, 7
+
+ ;# If there isn't any filtering to be done for the horizontal, then
+ ;# just skip to the second pass.
+ beq \jump_label
+
+ load_c v20, hfilter_b, r5, r12, r0
+
+ ;# setup constants
+ ;# v14 permutation value for alignment
+ load_c v28, b_hperm_b, 0, r12, r0
+
+ ;# index to the next set of vectors in the row.
+ li r12, 32
+
+ ;# rounding added in on the multiply
+ vspltisw v21, 8
+ vspltisw v18, 3
+ vslw v18, v21, v18 ;# 0x00000040000000400000004000000040
+
+ slwi. r6, r6, 5 ;# index into vertical filter array
+.endm
+
+;# Filters a horizontal line
+;# expects:
+;# r3 src_ptr
+;# r4 pitch
+;# r10 16
+;# r12 32
+;# v17 perm intput
+;# v18 rounding
+;# v19 shift
+;# v20 filter taps
+;# v21 tmp
+;# v22 tmp
+;# v23 tmp
+;# v24 tmp
+;# v25 tmp
+;# v26 tmp
+;# v27 tmp
+;# v28 perm output
+;#
+
+.macro hfilter_8 V, hp, lp, increment_counter
+ lvsl v17, 0, r3 ;# permutate value for alignment
+
+ ;# input to filter is 9 bytes wide, output is 8 bytes.
+ lvx v21, 0, r3
+ lvx v22, r10, r3
+
+.if \increment_counter
+ add r3, r3, r4
+.endif
+ vperm v21, v21, v22, v17
+
+ vperm v24, v21, v21, \hp ;# v20 = 0123 1234 2345 3456
+ vperm v25, v21, v21, \lp ;# v21 = 4567 5678 6789 789A
+
+ vmsummbm v24, v20, v24, v18
+ vmsummbm v25, v20, v25, v18
+
+ vpkswus v24, v24, v25 ;# v24 = 0 4 8 C 1 5 9 D (16-bit)
+
+ vsrh v24, v24, v19 ;# divide v0, v1 by 128
+
+ vpkuhus \V, v24, v24 ;# \V = scrambled 8-bit result
+.endm
+
+.macro vfilter_16 P0 P1
+ vmuleub v22, \P0, v20 ;# 64 + 4 positive taps
+ vadduhm v22, v18, v22
+ vmuloub v23, \P0, v20
+ vadduhm v23, v18, v23
+
+ vmuleub v24, \P1, v21
+ vadduhm v22, v22, v24 ;# Re = evens, saturation unnecessary
+ vmuloub v25, \P1, v21
+ vadduhm v23, v23, v25 ;# Ro = odds
+
+ vsrh v22, v22, v19 ;# divide by 128
+ vsrh v23, v23, v19 ;# v16 v17 = evens, odds
+ vmrghh \P0, v22, v23 ;# v18 v19 = 16-bit result in order
+ vmrglh v23, v22, v23
+ vpkuhus \P0, \P0, v23 ;# P0 = 8-bit result
+.endm
+
+.macro compute_sum_sse src, ref, sum, sse, t1, t2, z0
+ ;# Compute sum first. Unpack to so signed subract
+ ;# can be used. Only have a half word signed
+ ;# subract. Do high, then low.
+ vmrghb \t1, \z0, \src
+ vmrghb \t2, \z0, \ref
+ vsubshs \t1, \t1, \t2
+ vsum4shs \sum, \t1, \sum
+
+ vmrglb \t1, \z0, \src
+ vmrglb \t2, \z0, \ref
+ vsubshs \t1, \t1, \t2
+ vsum4shs \sum, \t1, \sum
+
+ ;# Now compute sse.
+ vsububs \t1, \src, \ref
+ vsububs \t2, \ref, \src
+ vor \t1, \t1, \t2
+
+ vmsumubm \sse, \t1, \t1, \sse
+.endm
+
+.macro variance_final sum, sse, z0, DS
+ vsumsws \sum, \sum, \z0
+ vsumsws \sse, \sse, \z0
+
+ stvx \sum, 0, r1
+ lwz r3, 12(r1)
+
+ stvx \sse, 0, r1
+ lwz r4, 12(r1)
+
+ stw r4, 0(r9) ;# sse
+
+ mullw r3, r3, r3 ;# sum*sum
+ srlwi r3, r3, \DS ;# (sum*sum) >> 8
+ subf r3, r3, r4 ;# sse - ((sum*sum) >> 8)
+.endm
+
+.macro compute_sum_sse_16 V, increment_counter
+ load_and_align_16 v16, r7, r8, \increment_counter
+ compute_sum_sse \V, v16, v18, v19, v20, v21, v23
+.endm
+
+.macro load_and_align_16 V, R, P, increment_counter
+ lvsl v17, 0, \R ;# permutate value for alignment
+
+ ;# input to filter is 21 bytes wide, output is 16 bytes.
+ ;# input will can span three vectors if not aligned correctly.
+ lvx v21, 0, \R
+ lvx v22, r10, \R
+
+.if \increment_counter
+ add \R, \R, \P
+.endif
+
+ vperm \V, v21, v22, v17
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_pixels_per_line
+;# r5 int xoffset
+;# r6 int yoffset
+;# r7 unsigned char *dst_ptr
+;# r8 int dst_pixels_per_line
+;# r9 unsigned int *sse
+;#
+;# r3 return value
+vp8_sub_pixel_variance4x4_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xf830
+ ori r12, r12, 0xfff8
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1,-32(r1) ;# create space on the stack
+
+ HProlog second_pass_4x4_pre_copy_b
+
+ ;# Load up permutation constants
+ load_c v10, b_0123_b, 0, r12, r0
+ load_c v11, b_4567_b, 0, r12, r0
+
+ hfilter_8 v0, v10, v11, 1
+ hfilter_8 v1, v10, v11, 1
+ hfilter_8 v2, v10, v11, 1
+ hfilter_8 v3, v10, v11, 1
+
+ ;# Finished filtering main horizontal block. If there is no
+ ;# vertical filtering, jump to storing the data. Otherwise
+ ;# load up and filter the additional line that is needed
+ ;# for the vertical filter.
+ beq compute_sum_sse_4x4_b
+
+ hfilter_8 v4, v10, v11, 0
+
+ b second_pass_4x4_b
+
+second_pass_4x4_pre_copy_b:
+ slwi r6, r6, 5 ;# index into vertical filter array
+
+ load_and_align_16 v0, r3, r4, 1
+ load_and_align_16 v1, r3, r4, 1
+ load_and_align_16 v2, r3, r4, 1
+ load_and_align_16 v3, r3, r4, 1
+ load_and_align_16 v4, r3, r4, 0
+
+second_pass_4x4_b:
+ vspltish v20, 8
+ vspltish v18, 3
+ vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
+
+ load_vfilter v20, v21
+
+ vfilter_16 v0, v1
+ vfilter_16 v1, v2
+ vfilter_16 v2, v3
+ vfilter_16 v3, v4
+
+compute_sum_sse_4x4_b:
+ vspltish v18, 0 ;# sum
+ vspltish v19, 0 ;# sse
+ vspltish v23, 0 ;# unpack
+ li r10, 16
+
+ load_and_align_16 v4, r7, r8, 1
+ load_and_align_16 v5, r7, r8, 1
+ load_and_align_16 v6, r7, r8, 1
+ load_and_align_16 v7, r7, r8, 1
+
+ vmrghb v0, v0, v1
+ vmrghb v1, v2, v3
+
+ vmrghb v2, v4, v5
+ vmrghb v3, v6, v7
+
+ load_c v10, b_hilo_b, 0, r12, r0
+
+ vperm v0, v0, v1, v10
+ vperm v1, v2, v3, v10
+
+ compute_sum_sse v0, v1, v18, v19, v20, v21, v23
+
+ variance_final v18, v19, v23, 4
+
+ addi r1, r1, 32 ;# recover stack
+ mtspr 256, r11 ;# reset old VRSAVE
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_pixels_per_line
+;# r5 int xoffset
+;# r6 int yoffset
+;# r7 unsigned char *dst_ptr
+;# r8 int dst_pixels_per_line
+;# r9 unsigned int *sse
+;#
+;# r3 return value
+vp8_sub_pixel_variance8x8_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xfff0
+ ori r12, r12, 0xffff
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1,-32(r1) ;# create space on the stack
+
+ HProlog second_pass_8x8_pre_copy_b
+
+ ;# Load up permutation constants
+ load_c v10, b_0123_b, 0, r12, r0
+ load_c v11, b_4567_b, 0, r12, r0
+
+ hfilter_8 v0, v10, v11, 1
+ hfilter_8 v1, v10, v11, 1
+ hfilter_8 v2, v10, v11, 1
+ hfilter_8 v3, v10, v11, 1
+ hfilter_8 v4, v10, v11, 1
+ hfilter_8 v5, v10, v11, 1
+ hfilter_8 v6, v10, v11, 1
+ hfilter_8 v7, v10, v11, 1
+
+ ;# Finished filtering main horizontal block. If there is no
+ ;# vertical filtering, jump to storing the data. Otherwise
+ ;# load up and filter the additional line that is needed
+ ;# for the vertical filter.
+ beq compute_sum_sse_8x8_b
+
+ hfilter_8 v8, v10, v11, 0
+
+ b second_pass_8x8_b
+
+second_pass_8x8_pre_copy_b:
+ slwi. r6, r6, 5 ;# index into vertical filter array
+
+ load_and_align_16 v0, r3, r4, 1
+ load_and_align_16 v1, r3, r4, 1
+ load_and_align_16 v2, r3, r4, 1
+ load_and_align_16 v3, r3, r4, 1
+ load_and_align_16 v4, r3, r4, 1
+ load_and_align_16 v5, r3, r4, 1
+ load_and_align_16 v6, r3, r4, 1
+ load_and_align_16 v7, r3, r4, 1
+ load_and_align_16 v8, r3, r4, 0
+
+ beq compute_sum_sse_8x8_b
+
+second_pass_8x8_b:
+ vspltish v20, 8
+ vspltish v18, 3
+ vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
+
+ load_vfilter v20, v21
+
+ vfilter_16 v0, v1
+ vfilter_16 v1, v2
+ vfilter_16 v2, v3
+ vfilter_16 v3, v4
+ vfilter_16 v4, v5
+ vfilter_16 v5, v6
+ vfilter_16 v6, v7
+ vfilter_16 v7, v8
+
+compute_sum_sse_8x8_b:
+ vspltish v18, 0 ;# sum
+ vspltish v19, 0 ;# sse
+ vspltish v23, 0 ;# unpack
+ li r10, 16
+
+ vmrghb v0, v0, v1
+ vmrghb v1, v2, v3
+ vmrghb v2, v4, v5
+ vmrghb v3, v6, v7
+
+ load_and_align_16 v4, r7, r8, 1
+ load_and_align_16 v5, r7, r8, 1
+ load_and_align_16 v6, r7, r8, 1
+ load_and_align_16 v7, r7, r8, 1
+ load_and_align_16 v8, r7, r8, 1
+ load_and_align_16 v9, r7, r8, 1
+ load_and_align_16 v10, r7, r8, 1
+ load_and_align_16 v11, r7, r8, 0
+
+ vmrghb v4, v4, v5
+ vmrghb v5, v6, v7
+ vmrghb v6, v8, v9
+ vmrghb v7, v10, v11
+
+ compute_sum_sse v0, v4, v18, v19, v20, v21, v23
+ compute_sum_sse v1, v5, v18, v19, v20, v21, v23
+ compute_sum_sse v2, v6, v18, v19, v20, v21, v23
+ compute_sum_sse v3, v7, v18, v19, v20, v21, v23
+
+ variance_final v18, v19, v23, 6
+
+ addi r1, r1, 32 ;# recover stack
+ mtspr 256, r11 ;# reset old VRSAVE
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_pixels_per_line
+;# r5 int xoffset
+;# r6 int yoffset
+;# r7 unsigned char *dst_ptr
+;# r8 int dst_pixels_per_line
+;# r9 unsigned int *sse
+;#
+;# r3 return value
+vp8_sub_pixel_variance8x16_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xffff
+ ori r12, r12, 0xfffc
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1,-32(r1) ;# create space on the stack
+
+ HProlog second_pass_8x16_pre_copy_b
+
+ ;# Load up permutation constants
+ load_c v29, b_0123_b, 0, r12, r0
+ load_c v30, b_4567_b, 0, r12, r0
+
+ hfilter_8 v0, v29, v30, 1
+ hfilter_8 v1, v29, v30, 1
+ hfilter_8 v2, v29, v30, 1
+ hfilter_8 v3, v29, v30, 1
+ hfilter_8 v4, v29, v30, 1
+ hfilter_8 v5, v29, v30, 1
+ hfilter_8 v6, v29, v30, 1
+ hfilter_8 v7, v29, v30, 1
+ hfilter_8 v8, v29, v30, 1
+ hfilter_8 v9, v29, v30, 1
+ hfilter_8 v10, v29, v30, 1
+ hfilter_8 v11, v29, v30, 1
+ hfilter_8 v12, v29, v30, 1
+ hfilter_8 v13, v29, v30, 1
+ hfilter_8 v14, v29, v30, 1
+ hfilter_8 v15, v29, v30, 1
+
+ ;# Finished filtering main horizontal block. If there is no
+ ;# vertical filtering, jump to storing the data. Otherwise
+ ;# load up and filter the additional line that is needed
+ ;# for the vertical filter.
+ beq compute_sum_sse_8x16_b
+
+ hfilter_8 v16, v29, v30, 0
+
+ b second_pass_8x16_b
+
+second_pass_8x16_pre_copy_b:
+ slwi. r6, r6, 5 ;# index into vertical filter array
+
+ load_and_align_16 v0, r3, r4, 1
+ load_and_align_16 v1, r3, r4, 1
+ load_and_align_16 v2, r3, r4, 1
+ load_and_align_16 v3, r3, r4, 1
+ load_and_align_16 v4, r3, r4, 1
+ load_and_align_16 v5, r3, r4, 1
+ load_and_align_16 v6, r3, r4, 1
+ load_and_align_16 v7, r3, r4, 1
+ load_and_align_16 v8, r3, r4, 1
+ load_and_align_16 v9, r3, r4, 1
+ load_and_align_16 v10, r3, r4, 1
+ load_and_align_16 v11, r3, r4, 1
+ load_and_align_16 v12, r3, r4, 1
+ load_and_align_16 v13, r3, r4, 1
+ load_and_align_16 v14, r3, r4, 1
+ load_and_align_16 v15, r3, r4, 1
+ load_and_align_16 v16, r3, r4, 0
+
+ beq compute_sum_sse_8x16_b
+
+second_pass_8x16_b:
+ vspltish v20, 8
+ vspltish v18, 3
+ vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
+
+ load_vfilter v20, v21
+
+ vfilter_16 v0, v1
+ vfilter_16 v1, v2
+ vfilter_16 v2, v3
+ vfilter_16 v3, v4
+ vfilter_16 v4, v5
+ vfilter_16 v5, v6
+ vfilter_16 v6, v7
+ vfilter_16 v7, v8
+ vfilter_16 v8, v9
+ vfilter_16 v9, v10
+ vfilter_16 v10, v11
+ vfilter_16 v11, v12
+ vfilter_16 v12, v13
+ vfilter_16 v13, v14
+ vfilter_16 v14, v15
+ vfilter_16 v15, v16
+
+compute_sum_sse_8x16_b:
+ vspltish v18, 0 ;# sum
+ vspltish v19, 0 ;# sse
+ vspltish v23, 0 ;# unpack
+ li r10, 16
+
+ vmrghb v0, v0, v1
+ vmrghb v1, v2, v3
+ vmrghb v2, v4, v5
+ vmrghb v3, v6, v7
+ vmrghb v4, v8, v9
+ vmrghb v5, v10, v11
+ vmrghb v6, v12, v13
+ vmrghb v7, v14, v15
+
+ load_and_align_16 v8, r7, r8, 1
+ load_and_align_16 v9, r7, r8, 1
+ load_and_align_16 v10, r7, r8, 1
+ load_and_align_16 v11, r7, r8, 1
+ load_and_align_16 v12, r7, r8, 1
+ load_and_align_16 v13, r7, r8, 1
+ load_and_align_16 v14, r7, r8, 1
+ load_and_align_16 v15, r7, r8, 1
+
+ vmrghb v8, v8, v9
+ vmrghb v9, v10, v11
+ vmrghb v10, v12, v13
+ vmrghb v11, v14, v15
+
+ compute_sum_sse v0, v8, v18, v19, v20, v21, v23
+ compute_sum_sse v1, v9, v18, v19, v20, v21, v23
+ compute_sum_sse v2, v10, v18, v19, v20, v21, v23
+ compute_sum_sse v3, v11, v18, v19, v20, v21, v23
+
+ load_and_align_16 v8, r7, r8, 1
+ load_and_align_16 v9, r7, r8, 1
+ load_and_align_16 v10, r7, r8, 1
+ load_and_align_16 v11, r7, r8, 1
+ load_and_align_16 v12, r7, r8, 1
+ load_and_align_16 v13, r7, r8, 1
+ load_and_align_16 v14, r7, r8, 1
+ load_and_align_16 v15, r7, r8, 0
+
+ vmrghb v8, v8, v9
+ vmrghb v9, v10, v11
+ vmrghb v10, v12, v13
+ vmrghb v11, v14, v15
+
+ compute_sum_sse v4, v8, v18, v19, v20, v21, v23
+ compute_sum_sse v5, v9, v18, v19, v20, v21, v23
+ compute_sum_sse v6, v10, v18, v19, v20, v21, v23
+ compute_sum_sse v7, v11, v18, v19, v20, v21, v23
+
+ variance_final v18, v19, v23, 7
+
+ addi r1, r1, 32 ;# recover stack
+ mtspr 256, r11 ;# reset old VRSAVE
+ blr
+
+;# Filters a horizontal line
+;# expects:
+;# r3 src_ptr
+;# r4 pitch
+;# r10 16
+;# r12 32
+;# v17 perm intput
+;# v18 rounding
+;# v19 shift
+;# v20 filter taps
+;# v21 tmp
+;# v22 tmp
+;# v23 tmp
+;# v24 tmp
+;# v25 tmp
+;# v26 tmp
+;# v27 tmp
+;# v28 perm output
+;#
+.macro hfilter_16 V, increment_counter
+
+ lvsl v17, 0, r3 ;# permutate value for alignment
+
+ ;# input to filter is 21 bytes wide, output is 16 bytes.
+ ;# input will can span three vectors if not aligned correctly.
+ lvx v21, 0, r3
+ lvx v22, r10, r3
+ lvx v23, r12, r3
+
+.if \increment_counter
+ add r3, r3, r4
+.endif
+ vperm v21, v21, v22, v17
+ vperm v22, v22, v23, v17 ;# v8 v9 = 21 input pixels left-justified
+
+ ;# set 0
+ vmsummbm v24, v20, v21, v18 ;# taps times elements
+
+ ;# set 1
+ vsldoi v23, v21, v22, 1
+ vmsummbm v25, v20, v23, v18
+
+ ;# set 2
+ vsldoi v23, v21, v22, 2
+ vmsummbm v26, v20, v23, v18
+
+ ;# set 3
+ vsldoi v23, v21, v22, 3
+ vmsummbm v27, v20, v23, v18
+
+ vpkswus v24, v24, v25 ;# v24 = 0 4 8 C 1 5 9 D (16-bit)
+ vpkswus v25, v26, v27 ;# v25 = 2 6 A E 3 7 B F
+
+ vsrh v24, v24, v19 ;# divide v0, v1 by 128
+ vsrh v25, v25, v19
+
+ vpkuhus \V, v24, v25 ;# \V = scrambled 8-bit result
+ vperm \V, \V, v0, v28 ;# \V = correctly-ordered result
+.endm
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_pixels_per_line
+;# r5 int xoffset
+;# r6 int yoffset
+;# r7 unsigned char *dst_ptr
+;# r8 int dst_pixels_per_line
+;# r9 unsigned int *sse
+;#
+;# r3 return value
+vp8_sub_pixel_variance16x8_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xffff
+ ori r12, r12, 0xfff8
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1, -32(r1) ;# create space on the stack
+
+ HProlog second_pass_16x8_pre_copy_b
+
+ hfilter_16 v0, 1
+ hfilter_16 v1, 1
+ hfilter_16 v2, 1
+ hfilter_16 v3, 1
+ hfilter_16 v4, 1
+ hfilter_16 v5, 1
+ hfilter_16 v6, 1
+ hfilter_16 v7, 1
+
+ ;# Finished filtering main horizontal block. If there is no
+ ;# vertical filtering, jump to storing the data. Otherwise
+ ;# load up and filter the additional line that is needed
+ ;# for the vertical filter.
+ beq compute_sum_sse_16x8_b
+
+ hfilter_16 v8, 0
+
+ b second_pass_16x8_b
+
+second_pass_16x8_pre_copy_b:
+ slwi. r6, r6, 5 ;# index into vertical filter array
+
+ load_and_align_16 v0, r3, r4, 1
+ load_and_align_16 v1, r3, r4, 1
+ load_and_align_16 v2, r3, r4, 1
+ load_and_align_16 v3, r3, r4, 1
+ load_and_align_16 v4, r3, r4, 1
+ load_and_align_16 v5, r3, r4, 1
+ load_and_align_16 v6, r3, r4, 1
+ load_and_align_16 v7, r3, r4, 1
+ load_and_align_16 v8, r3, r4, 1
+
+ beq compute_sum_sse_16x8_b
+
+second_pass_16x8_b:
+ vspltish v20, 8
+ vspltish v18, 3
+ vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
+
+ load_vfilter v20, v21
+
+ vfilter_16 v0, v1
+ vfilter_16 v1, v2
+ vfilter_16 v2, v3
+ vfilter_16 v3, v4
+ vfilter_16 v4, v5
+ vfilter_16 v5, v6
+ vfilter_16 v6, v7
+ vfilter_16 v7, v8
+
+compute_sum_sse_16x8_b:
+ vspltish v18, 0 ;# sum
+ vspltish v19, 0 ;# sse
+ vspltish v23, 0 ;# unpack
+ li r10, 16
+
+ compute_sum_sse_16 v0, 1
+ compute_sum_sse_16 v1, 1
+ compute_sum_sse_16 v2, 1
+ compute_sum_sse_16 v3, 1
+ compute_sum_sse_16 v4, 1
+ compute_sum_sse_16 v5, 1
+ compute_sum_sse_16 v6, 1
+ compute_sum_sse_16 v7, 0
+
+ variance_final v18, v19, v23, 7
+
+ addi r1, r1, 32 ;# recover stack
+
+ mtspr 256, r11 ;# reset old VRSAVE
+
+ blr
+
+ .align 2
+;# r3 unsigned char *src_ptr
+;# r4 int src_pixels_per_line
+;# r5 int xoffset
+;# r6 int yoffset
+;# r7 unsigned char *dst_ptr
+;# r8 int dst_pixels_per_line
+;# r9 unsigned int *sse
+;#
+;# r3 return value
+vp8_sub_pixel_variance16x16_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xffff
+ ori r12, r12, 0xfff8
+ mtspr 256, r12 ;# set VRSAVE
+
+ stwu r1, -32(r1) ;# create space on the stack
+
+ HProlog second_pass_16x16_pre_copy_b
+
+ hfilter_16 v0, 1
+ hfilter_16 v1, 1
+ hfilter_16 v2, 1
+ hfilter_16 v3, 1
+ hfilter_16 v4, 1
+ hfilter_16 v5, 1
+ hfilter_16 v6, 1
+ hfilter_16 v7, 1
+ hfilter_16 v8, 1
+ hfilter_16 v9, 1
+ hfilter_16 v10, 1
+ hfilter_16 v11, 1
+ hfilter_16 v12, 1
+ hfilter_16 v13, 1
+ hfilter_16 v14, 1
+ hfilter_16 v15, 1
+
+ ;# Finished filtering main horizontal block. If there is no
+ ;# vertical filtering, jump to storing the data. Otherwise
+ ;# load up and filter the additional line that is needed
+ ;# for the vertical filter.
+ beq compute_sum_sse_16x16_b
+
+ hfilter_16 v16, 0
+
+ b second_pass_16x16_b
+
+second_pass_16x16_pre_copy_b:
+ slwi. r6, r6, 5 ;# index into vertical filter array
+
+ load_and_align_16 v0, r3, r4, 1
+ load_and_align_16 v1, r3, r4, 1
+ load_and_align_16 v2, r3, r4, 1
+ load_and_align_16 v3, r3, r4, 1
+ load_and_align_16 v4, r3, r4, 1
+ load_and_align_16 v5, r3, r4, 1
+ load_and_align_16 v6, r3, r4, 1
+ load_and_align_16 v7, r3, r4, 1
+ load_and_align_16 v8, r3, r4, 1
+ load_and_align_16 v9, r3, r4, 1
+ load_and_align_16 v10, r3, r4, 1
+ load_and_align_16 v11, r3, r4, 1
+ load_and_align_16 v12, r3, r4, 1
+ load_and_align_16 v13, r3, r4, 1
+ load_and_align_16 v14, r3, r4, 1
+ load_and_align_16 v15, r3, r4, 1
+ load_and_align_16 v16, r3, r4, 0
+
+ beq compute_sum_sse_16x16_b
+
+second_pass_16x16_b:
+ vspltish v20, 8
+ vspltish v18, 3
+ vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
+
+ load_vfilter v20, v21
+
+ vfilter_16 v0, v1
+ vfilter_16 v1, v2
+ vfilter_16 v2, v3
+ vfilter_16 v3, v4
+ vfilter_16 v4, v5
+ vfilter_16 v5, v6
+ vfilter_16 v6, v7
+ vfilter_16 v7, v8
+ vfilter_16 v8, v9
+ vfilter_16 v9, v10
+ vfilter_16 v10, v11
+ vfilter_16 v11, v12
+ vfilter_16 v12, v13
+ vfilter_16 v13, v14
+ vfilter_16 v14, v15
+ vfilter_16 v15, v16
+
+compute_sum_sse_16x16_b:
+ vspltish v18, 0 ;# sum
+ vspltish v19, 0 ;# sse
+ vspltish v23, 0 ;# unpack
+ li r10, 16
+
+ compute_sum_sse_16 v0, 1
+ compute_sum_sse_16 v1, 1
+ compute_sum_sse_16 v2, 1
+ compute_sum_sse_16 v3, 1
+ compute_sum_sse_16 v4, 1
+ compute_sum_sse_16 v5, 1
+ compute_sum_sse_16 v6, 1
+ compute_sum_sse_16 v7, 1
+ compute_sum_sse_16 v8, 1
+ compute_sum_sse_16 v9, 1
+ compute_sum_sse_16 v10, 1
+ compute_sum_sse_16 v11, 1
+ compute_sum_sse_16 v12, 1
+ compute_sum_sse_16 v13, 1
+ compute_sum_sse_16 v14, 1
+ compute_sum_sse_16 v15, 0
+
+ variance_final v18, v19, v23, 8
+
+ addi r1, r1, 32 ;# recover stack
+
+ mtspr 256, r11 ;# reset old VRSAVE
+
+ blr
+
+ .data
+
+ .align 4
+hfilter_b:
+ .byte 128, 0, 0, 0,128, 0, 0, 0,128, 0, 0, 0,128, 0, 0, 0
+ .byte 112, 16, 0, 0,112, 16, 0, 0,112, 16, 0, 0,112, 16, 0, 0
+ .byte 96, 32, 0, 0, 96, 32, 0, 0, 96, 32, 0, 0, 96, 32, 0, 0
+ .byte 80, 48, 0, 0, 80, 48, 0, 0, 80, 48, 0, 0, 80, 48, 0, 0
+ .byte 64, 64, 0, 0, 64, 64, 0, 0, 64, 64, 0, 0, 64, 64, 0, 0
+ .byte 48, 80, 0, 0, 48, 80, 0, 0, 48, 80, 0, 0, 48, 80, 0, 0
+ .byte 32, 96, 0, 0, 32, 96, 0, 0, 32, 96, 0, 0, 32, 96, 0, 0
+ .byte 16,112, 0, 0, 16,112, 0, 0, 16,112, 0, 0, 16,112, 0, 0
+
+ .align 4
+vfilter_b:
+ .byte 128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128
+ .byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+ .byte 112,112,112,112,112,112,112,112,112,112,112,112,112,112,112,112
+ .byte 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+ .byte 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96
+ .byte 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
+ .byte 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80
+ .byte 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48
+ .byte 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
+ .byte 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
+ .byte 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48
+ .byte 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80
+ .byte 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
+ .byte 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96
+ .byte 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+ .byte 112,112,112,112,112,112,112,112,112,112,112,112,112,112,112,112
+
+ .align 4
+b_hperm_b:
+ .byte 0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15
+
+ .align 4
+b_0123_b:
+ .byte 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
+
+ .align 4
+b_4567_b:
+ .byte 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10
+
+b_hilo_b:
+ .byte 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23
--- /dev/null
+++ b/vp8/common/sad_c.c
@@ -1,0 +1,395 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+
+#include <stdlib.h>
+#include "vpx_config.h"
+#include "vpx/vpx_integer.h"
+
+static __inline
+unsigned int sad_mx_n_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad,
+ int m,
+ int n)
+{+
+ int r, c;
+ unsigned int sad = 0;
+
+ for (r = 0; r < n; r++)
+ {+ for (c = 0; c < m; c++)
+ {+ sad += abs(src_ptr[c] - ref_ptr[c]);
+ }
+
+ if (sad > max_sad)
+ break;
+
+ src_ptr += src_stride;
+ ref_ptr += ref_stride;
+ }
+
+ return sad;
+}
+
+/* max_sad is provided as an optional optimization point. Alternative
+ * implementations of these functions are not required to check it.
+ */
+
+unsigned int vp8_sad16x16_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad)
+{+
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 16);
+}
+
+
+unsigned int vp8_sad8x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad)
+{+
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 8);
+}
+
+
+unsigned int vp8_sad16x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad)
+{+
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 8);
+
+}
+
+
+unsigned int vp8_sad8x16_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad)
+{+
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 16);
+}
+
+
+unsigned int vp8_sad4x4_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad)
+{+
+ return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 4, 4);
+}
+
+void vp8_sad16x16x3_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad16x16x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+)
+{+ sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+ sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
+ sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
+ sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
+ sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
+ sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad16x8x3_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad16x8x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+)
+{+ sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+ sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
+ sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
+ sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
+ sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
+ sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x8x3_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x8x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+)
+{+ sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+ sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
+ sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
+ sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
+ sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
+ sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x16x3_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x16x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+)
+{+ sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+ sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
+ sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
+ sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
+ sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
+ sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad4x4x3_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad4x4x8_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+)
+{+ sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
+ sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
+ sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
+ sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
+ sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
+ sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
+ sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
+ sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
+}
+
+void vp8_sad16x16x4d_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *ref_ptr[],
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
+ sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+}
+
+void vp8_sad16x8x4d_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *ref_ptr[],
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
+ sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x8x4d_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *ref_ptr[],
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
+ sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+}
+
+void vp8_sad8x16x4d_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *ref_ptr[],
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
+ sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+}
+
+void vp8_sad4x4x4d_c(
+ const unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *ref_ptr[],
+ int ref_stride,
+ unsigned int *sad_array
+)
+{+ sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
+ sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
+ sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
+ sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
+}
+
+/* Copy 2 macroblocks to a buffer */
+void vp8_copy32xn_c(
+ unsigned char *src_ptr,
+ int src_stride,
+ unsigned char *dst_ptr,
+ int dst_stride,
+ int height)
+{+ int r;
+
+ for (r = 0; r < height; r++)
+ {+#if !(CONFIG_FAST_UNALIGNED)
+ dst_ptr[0] = src_ptr[0];
+ dst_ptr[1] = src_ptr[1];
+ dst_ptr[2] = src_ptr[2];
+ dst_ptr[3] = src_ptr[3];
+ dst_ptr[4] = src_ptr[4];
+ dst_ptr[5] = src_ptr[5];
+ dst_ptr[6] = src_ptr[6];
+ dst_ptr[7] = src_ptr[7];
+ dst_ptr[8] = src_ptr[8];
+ dst_ptr[9] = src_ptr[9];
+ dst_ptr[10] = src_ptr[10];
+ dst_ptr[11] = src_ptr[11];
+ dst_ptr[12] = src_ptr[12];
+ dst_ptr[13] = src_ptr[13];
+ dst_ptr[14] = src_ptr[14];
+ dst_ptr[15] = src_ptr[15];
+ dst_ptr[16] = src_ptr[16];
+ dst_ptr[17] = src_ptr[17];
+ dst_ptr[18] = src_ptr[18];
+ dst_ptr[19] = src_ptr[19];
+ dst_ptr[20] = src_ptr[20];
+ dst_ptr[21] = src_ptr[21];
+ dst_ptr[22] = src_ptr[22];
+ dst_ptr[23] = src_ptr[23];
+ dst_ptr[24] = src_ptr[24];
+ dst_ptr[25] = src_ptr[25];
+ dst_ptr[26] = src_ptr[26];
+ dst_ptr[27] = src_ptr[27];
+ dst_ptr[28] = src_ptr[28];
+ dst_ptr[29] = src_ptr[29];
+ dst_ptr[30] = src_ptr[30];
+ dst_ptr[31] = src_ptr[31];
+#else
+ ((uint32_t *)dst_ptr)[0] = ((uint32_t *)src_ptr)[0] ;
+ ((uint32_t *)dst_ptr)[1] = ((uint32_t *)src_ptr)[1] ;
+ ((uint32_t *)dst_ptr)[2] = ((uint32_t *)src_ptr)[2] ;
+ ((uint32_t *)dst_ptr)[3] = ((uint32_t *)src_ptr)[3] ;
+ ((uint32_t *)dst_ptr)[4] = ((uint32_t *)src_ptr)[4] ;
+ ((uint32_t *)dst_ptr)[5] = ((uint32_t *)src_ptr)[5] ;
+ ((uint32_t *)dst_ptr)[6] = ((uint32_t *)src_ptr)[6] ;
+ ((uint32_t *)dst_ptr)[7] = ((uint32_t *)src_ptr)[7] ;
+#endif
+ src_ptr += src_stride;
+ dst_ptr += dst_stride;
+
+ }
+}
--- /dev/null
+++ b/vp8/common/variance.h
@@ -1,0 +1,115 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+
+#ifndef VARIANCE_H
+#define VARIANCE_H
+
+typedef unsigned int(*vp8_sad_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int max_sad
+ );
+
+typedef void (*vp8_copy32xn_fn_t)(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ int n);
+
+typedef void (*vp8_sad_multi_fn_t)(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sad_array);
+
+typedef void (*vp8_sad_multi1_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned short *sad_array
+ );
+
+typedef void (*vp8_sad_multi_d_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ unsigned char *ref_ptr[4],
+ int ref_stride,
+ unsigned int *sad_array
+ );
+
+typedef unsigned int (*vp8_variance_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride,
+ unsigned int *sse
+ );
+
+typedef unsigned int (*vp8_subpixvariance_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ int xoffset,
+ int yoffset,
+ const unsigned char *ref_ptr,
+ int Refstride,
+ unsigned int *sse
+ );
+
+typedef void (*vp8_ssimpf_fn_t)
+ (
+ unsigned char *s,
+ int sp,
+ unsigned char *r,
+ int rp,
+ unsigned long *sum_s,
+ unsigned long *sum_r,
+ unsigned long *sum_sq_s,
+ unsigned long *sum_sq_r,
+ unsigned long *sum_sxr
+ );
+
+typedef unsigned int (*vp8_getmbss_fn_t)(const short *);
+
+typedef unsigned int (*vp8_get16x16prederror_fn_t)
+ (
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int ref_stride
+ );
+
+typedef struct variance_vtable
+{+ vp8_sad_fn_t sdf;
+ vp8_variance_fn_t vf;
+ vp8_subpixvariance_fn_t svf;
+ vp8_variance_fn_t svf_halfpix_h;
+ vp8_variance_fn_t svf_halfpix_v;
+ vp8_variance_fn_t svf_halfpix_hv;
+ vp8_sad_multi_fn_t sdx3f;
+ vp8_sad_multi1_fn_t sdx8f;
+ vp8_sad_multi_d_fn_t sdx4df;
+#if ARCH_X86 || ARCH_X86_64
+ vp8_copy32xn_fn_t copymem;
+#endif
+} vp8_variance_fn_ptr_t;
+
+#endif
--- /dev/null
+++ b/vp8/common/variance_c.c
@@ -1,0 +1,458 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+
+#include "variance.h"
+#include "filter.h"
+
+
+unsigned int vp8_get_mb_ss_c
+(
+ const short *src_ptr
+)
+{+ unsigned int i = 0, sum = 0;
+
+ do
+ {+ sum += (src_ptr[i] * src_ptr[i]);
+ i++;
+ }
+ while (i < 256);
+
+ return sum;
+}
+
+
+static void variance(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ int w,
+ int h,
+ unsigned int *sse,
+ int *sum)
+{+ int i, j;
+ int diff;
+
+ *sum = 0;
+ *sse = 0;
+
+ for (i = 0; i < h; i++)
+ {+ for (j = 0; j < w; j++)
+ {+ diff = src_ptr[j] - ref_ptr[j];
+ *sum += diff;
+ *sse += diff * diff;
+ }
+
+ src_ptr += source_stride;
+ ref_ptr += recon_stride;
+ }
+}
+
+
+unsigned int vp8_variance16x16_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 8));
+}
+
+unsigned int vp8_variance8x16_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16, &var, &avg);
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 7));
+}
+
+unsigned int vp8_variance16x8_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8, &var, &avg);
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 7));
+}
+
+
+unsigned int vp8_variance8x8_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, &var, &avg);
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 6));
+}
+
+unsigned int vp8_variance4x4_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 4, &var, &avg);
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 4));
+}
+
+
+unsigned int vp8_mse16x16_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+ variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
+ *sse = var;
+ return var;
+}
+
+
+/****************************************************************************
+ *
+ * ROUTINE : filter_block2d_bil_first_pass
+ *
+ * INPUTS : UINT8 *src_ptr : Pointer to source block.
+ * UINT32 src_pixels_per_line : Stride of input block.
+ * UINT32 pixel_step : Offset between filter input samples (see notes).
+ * UINT32 output_height : Input block height.
+ * UINT32 output_width : Input block width.
+ * INT32 *vp8_filter : Array of 2 bi-linear filter taps.
+ *
+ * OUTPUTS : INT32 *output_ptr : Pointer to filtered block.
+ *
+ * RETURNS : void
+ *
+ * FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
+ * either horizontal or vertical direction to produce the
+ * filtered output block. Used to implement first-pass
+ * of 2-D separable filter.
+ *
+ * SPECIAL NOTES : Produces INT32 output to retain precision for next pass.
+ * Two filter taps should sum to VP8_FILTER_WEIGHT.
+ * pixel_step defines whether the filter is applied
+ * horizontally (pixel_step=1) or vertically (pixel_step=stride).
+ * It defines the offset required to move from one input
+ * to the next.
+ *
+ ****************************************************************************/
+static void var_filter_block2d_bil_first_pass
+(
+ const unsigned char *src_ptr,
+ unsigned short *output_ptr,
+ unsigned int src_pixels_per_line,
+ int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ const short *vp8_filter
+)
+{+ unsigned int i, j;
+
+ for (i = 0; i < output_height; i++)
+ {+ for (j = 0; j < output_width; j++)
+ {+ // Apply bilinear filter
+ output_ptr[j] = (((int)src_ptr[0] * vp8_filter[0]) +
+ ((int)src_ptr[pixel_step] * vp8_filter[1]) +
+ (VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
+ src_ptr++;
+ }
+
+ // Next row...
+ src_ptr += src_pixels_per_line - output_width;
+ output_ptr += output_width;
+ }
+}
+
+/****************************************************************************
+ *
+ * ROUTINE : filter_block2d_bil_second_pass
+ *
+ * INPUTS : INT32 *src_ptr : Pointer to source block.
+ * UINT32 src_pixels_per_line : Stride of input block.
+ * UINT32 pixel_step : Offset between filter input samples (see notes).
+ * UINT32 output_height : Input block height.
+ * UINT32 output_width : Input block width.
+ * INT32 *vp8_filter : Array of 2 bi-linear filter taps.
+ *
+ * OUTPUTS : UINT16 *output_ptr : Pointer to filtered block.
+ *
+ * RETURNS : void
+ *
+ * FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
+ * either horizontal or vertical direction to produce the
+ * filtered output block. Used to implement second-pass
+ * of 2-D separable filter.
+ *
+ * SPECIAL NOTES : Requires 32-bit input as produced by filter_block2d_bil_first_pass.
+ * Two filter taps should sum to VP8_FILTER_WEIGHT.
+ * pixel_step defines whether the filter is applied
+ * horizontally (pixel_step=1) or vertically (pixel_step=stride).
+ * It defines the offset required to move from one input
+ * to the next.
+ *
+ ****************************************************************************/
+static void var_filter_block2d_bil_second_pass
+(
+ const unsigned short *src_ptr,
+ unsigned char *output_ptr,
+ unsigned int src_pixels_per_line,
+ unsigned int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ const short *vp8_filter
+)
+{+ unsigned int i, j;
+ int Temp;
+
+ for (i = 0; i < output_height; i++)
+ {+ for (j = 0; j < output_width; j++)
+ {+ // Apply filter
+ Temp = ((int)src_ptr[0] * vp8_filter[0]) +
+ ((int)src_ptr[pixel_step] * vp8_filter[1]) +
+ (VP8_FILTER_WEIGHT / 2);
+ output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
+ src_ptr++;
+ }
+
+ // Next row...
+ src_ptr += src_pixels_per_line - output_width;
+ output_ptr += output_width;
+ }
+}
+
+
+unsigned int vp8_sub_pixel_variance4x4_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned char temp2[20*16];
+ const short *HFilter, *VFilter;
+ unsigned short FData3[5*4]; // Temp data bufffer used in filtering
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ // First filter 1d Horizontal
+ var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 5, 4, HFilter);
+
+ // Now filter Verticaly
+ var_filter_block2d_bil_second_pass(FData3, temp2, 4, 4, 4, 4, VFilter);
+
+ return vp8_variance4x4_c(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
+}
+
+
+unsigned int vp8_sub_pixel_variance8x8_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short FData3[9*8]; // Temp data bufffer used in filtering
+ unsigned char temp2[20*16];
+ const short *HFilter, *VFilter;
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 8, HFilter);
+ var_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 8, 8, VFilter);
+
+ return vp8_variance8x8_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
+}
+
+unsigned int vp8_sub_pixel_variance16x16_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short FData3[17*16]; // Temp data bufffer used in filtering
+ unsigned char temp2[20*16];
+ const short *HFilter, *VFilter;
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 16, HFilter);
+ var_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 16, 16, VFilter);
+
+ return vp8_variance16x16_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_h_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 0,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_v_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 0, 4,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_hv_c(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 4,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_sub_pixel_mse16x16_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ vp8_sub_pixel_variance16x16_c(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
+ return *sse;
+}
+
+unsigned int vp8_sub_pixel_variance16x8_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short FData3[16*9]; // Temp data bufffer used in filtering
+ unsigned char temp2[20*16];
+ const short *HFilter, *VFilter;
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+ var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 16, HFilter);
+ var_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 8, 16, VFilter);
+
+ return vp8_variance16x8_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
+}
+
+unsigned int vp8_sub_pixel_variance8x16_c
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ unsigned short FData3[9*16]; // Temp data bufffer used in filtering
+ unsigned char temp2[20*16];
+ const short *HFilter, *VFilter;
+
+
+ HFilter = vp8_bilinear_filters[xoffset];
+ VFilter = vp8_bilinear_filters[yoffset];
+
+
+ var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 8, HFilter);
+ var_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 16, 8, VFilter);
+
+ return vp8_variance8x16_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
+}
--- /dev/null
+++ b/vp8/common/x86/sad_mmx.asm
@@ -1,0 +1,427 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+global sym(vp8_sad16x16_mmx)
+global sym(vp8_sad8x16_mmx)
+global sym(vp8_sad8x8_mmx)
+global sym(vp8_sad4x4_mmx)
+global sym(vp8_sad16x8_mmx)
+
+;unsigned int vp8_sad16x16_mmx(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+sym(vp8_sad16x16_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rax*8]
+
+ lea rcx, [rcx+rax*8]
+ pxor mm7, mm7
+
+ pxor mm6, mm6
+
+.x16x16sad_mmx_loop:
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm2, QWORD PTR [rsi+8]
+
+ movq mm1, QWORD PTR [rdi]
+ movq mm3, QWORD PTR [rdi+8]
+
+ movq mm4, mm0
+ movq mm5, mm2
+
+ psubusb mm0, mm1
+ psubusb mm1, mm4
+
+ psubusb mm2, mm3
+ psubusb mm3, mm5
+
+ por mm0, mm1
+ por mm2, mm3
+
+ movq mm1, mm0
+ movq mm3, mm2
+
+ punpcklbw mm0, mm6
+ punpcklbw mm2, mm6
+
+ punpckhbw mm1, mm6
+ punpckhbw mm3, mm6
+
+ paddw mm0, mm2
+ paddw mm1, mm3
+
+
+ lea rsi, [rsi+rax]
+ add rdi, rdx
+
+ paddw mm7, mm0
+ paddw mm7, mm1
+
+ cmp rsi, rcx
+ jne .x16x16sad_mmx_loop
+
+
+ movq mm0, mm7
+
+ punpcklwd mm0, mm6
+ punpckhwd mm7, mm6
+
+ paddw mm0, mm7
+ movq mm7, mm0
+
+
+ psrlq mm0, 32
+ paddw mm7, mm0
+
+ movq rax, mm7
+
+ pop rdi
+ pop rsi
+ mov rsp, rbp
+ ; begin epilog
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad8x16_mmx(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+sym(vp8_sad8x16_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rax*8]
+
+ lea rcx, [rcx+rax*8]
+ pxor mm7, mm7
+
+ pxor mm6, mm6
+
+.x8x16sad_mmx_loop:
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm1, QWORD PTR [rdi]
+
+ movq mm2, mm0
+ psubusb mm0, mm1
+
+ psubusb mm1, mm2
+ por mm0, mm1
+
+ movq mm2, mm0
+ punpcklbw mm0, mm6
+
+ punpckhbw mm2, mm6
+ lea rsi, [rsi+rax]
+
+ add rdi, rdx
+ paddw mm7, mm0
+
+ paddw mm7, mm2
+ cmp rsi, rcx
+
+ jne .x8x16sad_mmx_loop
+
+ movq mm0, mm7
+ punpcklwd mm0, mm6
+
+ punpckhwd mm7, mm6
+ paddw mm0, mm7
+
+ movq mm7, mm0
+ psrlq mm0, 32
+
+ paddw mm7, mm0
+ movq rax, mm7
+
+ pop rdi
+ pop rsi
+ mov rsp, rbp
+ ; begin epilog
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad8x8_mmx(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+sym(vp8_sad8x8_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rax*8]
+ pxor mm7, mm7
+
+ pxor mm6, mm6
+
+.x8x8sad_mmx_loop:
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm1, QWORD PTR [rdi]
+
+ movq mm2, mm0
+ psubusb mm0, mm1
+
+ psubusb mm1, mm2
+ por mm0, mm1
+
+ movq mm2, mm0
+ punpcklbw mm0, mm6
+
+ punpckhbw mm2, mm6
+ paddw mm0, mm2
+
+ lea rsi, [rsi+rax]
+ add rdi, rdx
+
+ paddw mm7, mm0
+ cmp rsi, rcx
+
+ jne .x8x8sad_mmx_loop
+
+ movq mm0, mm7
+ punpcklwd mm0, mm6
+
+ punpckhwd mm7, mm6
+ paddw mm0, mm7
+
+ movq mm7, mm0
+ psrlq mm0, 32
+
+ paddw mm7, mm0
+ movq rax, mm7
+
+ pop rdi
+ pop rsi
+ mov rsp, rbp
+ ; begin epilog
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad4x4_mmx(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+sym(vp8_sad4x4_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ movd mm0, DWORD PTR [rsi]
+ movd mm1, DWORD PTR [rdi]
+
+ movd mm2, DWORD PTR [rsi+rax]
+ movd mm3, DWORD PTR [rdi+rdx]
+
+ punpcklbw mm0, mm2
+ punpcklbw mm1, mm3
+
+ movq mm2, mm0
+ psubusb mm0, mm1
+
+ psubusb mm1, mm2
+ por mm0, mm1
+
+ movq mm2, mm0
+ pxor mm3, mm3
+
+ punpcklbw mm0, mm3
+ punpckhbw mm2, mm3
+
+ paddw mm0, mm2
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ movd mm4, DWORD PTR [rsi]
+ movd mm5, DWORD PTR [rdi]
+
+ movd mm6, DWORD PTR [rsi+rax]
+ movd mm7, DWORD PTR [rdi+rdx]
+
+ punpcklbw mm4, mm6
+ punpcklbw mm5, mm7
+
+ movq mm6, mm4
+ psubusb mm4, mm5
+
+ psubusb mm5, mm6
+ por mm4, mm5
+
+ movq mm5, mm4
+ punpcklbw mm4, mm3
+
+ punpckhbw mm5, mm3
+ paddw mm4, mm5
+
+ paddw mm0, mm4
+ movq mm1, mm0
+
+ punpcklwd mm0, mm3
+ punpckhwd mm1, mm3
+
+ paddw mm0, mm1
+ movq mm1, mm0
+
+ psrlq mm0, 32
+ paddw mm0, mm1
+
+ movq rax, mm0
+
+ pop rdi
+ pop rsi
+ mov rsp, rbp
+ ; begin epilog
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad16x8_mmx(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+sym(vp8_sad16x8_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rax*8]
+ pxor mm7, mm7
+
+ pxor mm6, mm6
+
+.x16x8sad_mmx_loop:
+
+ movq mm0, [rsi]
+ movq mm1, [rdi]
+
+ movq mm2, [rsi+8]
+ movq mm3, [rdi+8]
+
+ movq mm4, mm0
+ movq mm5, mm2
+
+ psubusb mm0, mm1
+ psubusb mm1, mm4
+
+ psubusb mm2, mm3
+ psubusb mm3, mm5
+
+ por mm0, mm1
+ por mm2, mm3
+
+ movq mm1, mm0
+ movq mm3, mm2
+
+ punpcklbw mm0, mm6
+ punpckhbw mm1, mm6
+
+ punpcklbw mm2, mm6
+ punpckhbw mm3, mm6
+
+
+ paddw mm0, mm2
+ paddw mm1, mm3
+
+ paddw mm0, mm1
+ lea rsi, [rsi+rax]
+
+ add rdi, rdx
+ paddw mm7, mm0
+
+ cmp rsi, rcx
+ jne .x16x8sad_mmx_loop
+
+ movq mm0, mm7
+ punpcklwd mm0, mm6
+
+ punpckhwd mm7, mm6
+ paddw mm0, mm7
+
+ movq mm7, mm0
+ psrlq mm0, 32
+
+ paddw mm7, mm0
+ movq rax, mm7
+
+ pop rdi
+ pop rsi
+ mov rsp, rbp
+ ; begin epilog
+ UNSHADOW_ARGS
+ pop rbp
+ ret
--- /dev/null
+++ b/vp8/common/x86/sad_sse2.asm
@@ -1,0 +1,410 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+;unsigned int vp8_sad16x16_wmt(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+global sym(vp8_sad16x16_wmt)
+sym(vp8_sad16x16_wmt):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ SAVE_XMM 6
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rax*8]
+
+ lea rcx, [rcx+rax*8]
+ pxor xmm6, xmm6
+
+.x16x16sad_wmt_loop:
+
+ movq xmm0, QWORD PTR [rsi]
+ movq xmm2, QWORD PTR [rsi+8]
+
+ movq xmm1, QWORD PTR [rdi]
+ movq xmm3, QWORD PTR [rdi+8]
+
+ movq xmm4, QWORD PTR [rsi+rax]
+ movq xmm5, QWORD PTR [rdi+rdx]
+
+
+ punpcklbw xmm0, xmm2
+ punpcklbw xmm1, xmm3
+
+ psadbw xmm0, xmm1
+ movq xmm2, QWORD PTR [rsi+rax+8]
+
+ movq xmm3, QWORD PTR [rdi+rdx+8]
+ lea rsi, [rsi+rax*2]
+
+ lea rdi, [rdi+rdx*2]
+ punpcklbw xmm4, xmm2
+
+ punpcklbw xmm5, xmm3
+ psadbw xmm4, xmm5
+
+ paddw xmm6, xmm0
+ paddw xmm6, xmm4
+
+ cmp rsi, rcx
+ jne .x16x16sad_wmt_loop
+
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movq rax, xmm0
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;unsigned int vp8_sad8x16_wmt(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int max_sad)
+global sym(vp8_sad8x16_wmt)
+sym(vp8_sad8x16_wmt):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rbx, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rbx*8]
+
+ lea rcx, [rcx+rbx*8]
+ pxor mm7, mm7
+
+.x8x16sad_wmt_loop:
+
+ movq rax, mm7
+ cmp eax, arg(4)
+ jg .x8x16sad_wmt_early_exit
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm1, QWORD PTR [rdi]
+
+ movq mm2, QWORD PTR [rsi+rbx]
+ movq mm3, QWORD PTR [rdi+rdx]
+
+ psadbw mm0, mm1
+ psadbw mm2, mm3
+
+ lea rsi, [rsi+rbx*2]
+ lea rdi, [rdi+rdx*2]
+
+ paddw mm7, mm0
+ paddw mm7, mm2
+
+ cmp rsi, rcx
+ jne .x8x16sad_wmt_loop
+
+ movq rax, mm7
+
+.x8x16sad_wmt_early_exit:
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ pop rbx
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad8x8_wmt(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+global sym(vp8_sad8x8_wmt)
+sym(vp8_sad8x8_wmt):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rbx, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rbx*8]
+ pxor mm7, mm7
+
+.x8x8sad_wmt_loop:
+
+ movq rax, mm7
+ cmp eax, arg(4)
+ jg .x8x8sad_wmt_early_exit
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm1, QWORD PTR [rdi]
+
+ psadbw mm0, mm1
+ lea rsi, [rsi+rbx]
+
+ add rdi, rdx
+ paddw mm7, mm0
+
+ cmp rsi, rcx
+ jne .x8x8sad_wmt_loop
+
+ movq rax, mm7
+.x8x8sad_wmt_early_exit:
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ pop rbx
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;unsigned int vp8_sad4x4_wmt(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+global sym(vp8_sad4x4_wmt)
+sym(vp8_sad4x4_wmt):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ movd mm0, DWORD PTR [rsi]
+ movd mm1, DWORD PTR [rdi]
+
+ movd mm2, DWORD PTR [rsi+rax]
+ movd mm3, DWORD PTR [rdi+rdx]
+
+ punpcklbw mm0, mm2
+ punpcklbw mm1, mm3
+
+ psadbw mm0, mm1
+ lea rsi, [rsi+rax*2]
+
+ lea rdi, [rdi+rdx*2]
+ movd mm4, DWORD PTR [rsi]
+
+ movd mm5, DWORD PTR [rdi]
+ movd mm6, DWORD PTR [rsi+rax]
+
+ movd mm7, DWORD PTR [rdi+rdx]
+ punpcklbw mm4, mm6
+
+ punpcklbw mm5, mm7
+ psadbw mm4, mm5
+
+ paddw mm0, mm4
+ movq rax, mm0
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_sad16x8_wmt(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride)
+global sym(vp8_sad16x8_wmt)
+sym(vp8_sad16x8_wmt):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rbx, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ lea rcx, [rsi+rbx*8]
+ pxor mm7, mm7
+
+.x16x8sad_wmt_loop:
+
+ movq rax, mm7
+ cmp eax, arg(4)
+ jg .x16x8sad_wmt_early_exit
+
+ movq mm0, QWORD PTR [rsi]
+ movq mm2, QWORD PTR [rsi+8]
+
+ movq mm1, QWORD PTR [rdi]
+ movq mm3, QWORD PTR [rdi+8]
+
+ movq mm4, QWORD PTR [rsi+rbx]
+ movq mm5, QWORD PTR [rdi+rdx]
+
+ psadbw mm0, mm1
+ psadbw mm2, mm3
+
+ movq mm1, QWORD PTR [rsi+rbx+8]
+ movq mm3, QWORD PTR [rdi+rdx+8]
+
+ psadbw mm4, mm5
+ psadbw mm1, mm3
+
+ lea rsi, [rsi+rbx*2]
+ lea rdi, [rdi+rdx*2]
+
+ paddw mm0, mm2
+ paddw mm4, mm1
+
+ paddw mm7, mm0
+ paddw mm7, mm4
+
+ cmp rsi, rcx
+ jne .x16x8sad_wmt_loop
+
+ movq rax, mm7
+
+.x16x8sad_wmt_early_exit:
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ pop rbx
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp8_copy32xn_sse2(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *dst_ptr,
+; int dst_stride,
+; int height);
+global sym(vp8_copy32xn_sse2)
+sym(vp8_copy32xn_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ SAVE_XMM 7
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;dst_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;dst_stride
+ movsxd rcx, dword ptr arg(4) ;height
+
+.block_copy_sse2_loopx4:
+ movdqu xmm0, XMMWORD PTR [rsi]
+ movdqu xmm1, XMMWORD PTR [rsi + 16]
+ movdqu xmm2, XMMWORD PTR [rsi + rax]
+ movdqu xmm3, XMMWORD PTR [rsi + rax + 16]
+
+ lea rsi, [rsi+rax*2]
+
+ movdqu xmm4, XMMWORD PTR [rsi]
+ movdqu xmm5, XMMWORD PTR [rsi + 16]
+ movdqu xmm6, XMMWORD PTR [rsi + rax]
+ movdqu xmm7, XMMWORD PTR [rsi + rax + 16]
+
+ lea rsi, [rsi+rax*2]
+
+ movdqa XMMWORD PTR [rdi], xmm0
+ movdqa XMMWORD PTR [rdi + 16], xmm1
+ movdqa XMMWORD PTR [rdi + rdx], xmm2
+ movdqa XMMWORD PTR [rdi + rdx + 16], xmm3
+
+ lea rdi, [rdi+rdx*2]
+
+ movdqa XMMWORD PTR [rdi], xmm4
+ movdqa XMMWORD PTR [rdi + 16], xmm5
+ movdqa XMMWORD PTR [rdi + rdx], xmm6
+ movdqa XMMWORD PTR [rdi + rdx + 16], xmm7
+
+ lea rdi, [rdi+rdx*2]
+
+ sub rcx, 4
+ cmp rcx, 4
+ jge .block_copy_sse2_loopx4
+
+ cmp rcx, 0
+ je .copy_is_done
+
+.block_copy_sse2_loop:
+ movdqu xmm0, XMMWORD PTR [rsi]
+ movdqu xmm1, XMMWORD PTR [rsi + 16]
+ lea rsi, [rsi+rax]
+
+ movdqa XMMWORD PTR [rdi], xmm0
+ movdqa XMMWORD PTR [rdi + 16], xmm1
+ lea rdi, [rdi+rdx]
+
+ sub rcx, 1
+ jne .block_copy_sse2_loop
+
+.copy_is_done:
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
--- /dev/null
+++ b/vp8/common/x86/sad_sse3.asm
@@ -1,0 +1,960 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%macro STACK_FRAME_CREATE_X3 0
+%if ABI_IS_32BIT
+ %define src_ptr rsi
+ %define src_stride rax
+ %define ref_ptr rdi
+ %define ref_stride rdx
+ %define end_ptr rcx
+ %define ret_var rbx
+ %define result_ptr arg(4)
+ %define max_sad arg(4)
+ %define height dword ptr arg(4)
+ push rbp
+ mov rbp, rsp
+ push rsi
+ push rdi
+ push rbx
+
+ mov rsi, arg(0) ; src_ptr
+ mov rdi, arg(2) ; ref_ptr
+
+ movsxd rax, dword ptr arg(1) ; src_stride
+ movsxd rdx, dword ptr arg(3) ; ref_stride
+%else
+ %ifidn __OUTPUT_FORMAT__,x64
+ SAVE_XMM 7, u
+ %define src_ptr rcx
+ %define src_stride rdx
+ %define ref_ptr r8
+ %define ref_stride r9
+ %define end_ptr r10
+ %define ret_var r11
+ %define result_ptr [rsp+xmm_stack_space+8+4*8]
+ %define max_sad [rsp+xmm_stack_space+8+4*8]
+ %define height dword ptr [rsp+xmm_stack_space+8+4*8]
+ %else
+ %define src_ptr rdi
+ %define src_stride rsi
+ %define ref_ptr rdx
+ %define ref_stride rcx
+ %define end_ptr r9
+ %define ret_var r10
+ %define result_ptr r8
+ %define max_sad r8
+ %define height r8
+ %endif
+%endif
+
+%endmacro
+
+%macro STACK_FRAME_DESTROY_X3 0
+ %define src_ptr
+ %define src_stride
+ %define ref_ptr
+ %define ref_stride
+ %define end_ptr
+ %define ret_var
+ %define result_ptr
+ %define max_sad
+ %define height
+
+%if ABI_IS_32BIT
+ pop rbx
+ pop rdi
+ pop rsi
+ pop rbp
+%else
+ %ifidn __OUTPUT_FORMAT__,x64
+ RESTORE_XMM
+ %endif
+%endif
+ ret
+%endmacro
+
+%macro STACK_FRAME_CREATE_X4 0
+%if ABI_IS_32BIT
+ %define src_ptr rsi
+ %define src_stride rax
+ %define r0_ptr rcx
+ %define r1_ptr rdx
+ %define r2_ptr rbx
+ %define r3_ptr rdi
+ %define ref_stride rbp
+ %define result_ptr arg(4)
+ push rbp
+ mov rbp, rsp
+ push rsi
+ push rdi
+ push rbx
+
+ push rbp
+ mov rdi, arg(2) ; ref_ptr_base
+
+ LOAD_X4_ADDRESSES rdi, rcx, rdx, rax, rdi
+
+ mov rsi, arg(0) ; src_ptr
+
+ movsxd rbx, dword ptr arg(1) ; src_stride
+ movsxd rbp, dword ptr arg(3) ; ref_stride
+
+ xchg rbx, rax
+%else
+ %ifidn __OUTPUT_FORMAT__,x64
+ SAVE_XMM 7, u
+ %define src_ptr rcx
+ %define src_stride rdx
+ %define r0_ptr rsi
+ %define r1_ptr r10
+ %define r2_ptr r11
+ %define r3_ptr r8
+ %define ref_stride r9
+ %define result_ptr [rsp+xmm_stack_space+16+4*8]
+ push rsi
+
+ LOAD_X4_ADDRESSES r8, r0_ptr, r1_ptr, r2_ptr, r3_ptr
+ %else
+ %define src_ptr rdi
+ %define src_stride rsi
+ %define r0_ptr r9
+ %define r1_ptr r10
+ %define r2_ptr r11
+ %define r3_ptr rdx
+ %define ref_stride rcx
+ %define result_ptr r8
+
+ LOAD_X4_ADDRESSES rdx, r0_ptr, r1_ptr, r2_ptr, r3_ptr
+
+ %endif
+%endif
+%endmacro
+
+%macro STACK_FRAME_DESTROY_X4 0
+ %define src_ptr
+ %define src_stride
+ %define r0_ptr
+ %define r1_ptr
+ %define r2_ptr
+ %define r3_ptr
+ %define ref_stride
+ %define result_ptr
+
+%if ABI_IS_32BIT
+ pop rbx
+ pop rdi
+ pop rsi
+ pop rbp
+%else
+ %ifidn __OUTPUT_FORMAT__,x64
+ pop rsi
+ RESTORE_XMM
+ %endif
+%endif
+ ret
+%endmacro
+
+%macro PROCESS_16X2X3 5
+%if %1==0
+ movdqa xmm0, XMMWORD PTR [%2]
+ lddqu xmm5, XMMWORD PTR [%3]
+ lddqu xmm6, XMMWORD PTR [%3+1]
+ lddqu xmm7, XMMWORD PTR [%3+2]
+
+ psadbw xmm5, xmm0
+ psadbw xmm6, xmm0
+ psadbw xmm7, xmm0
+%else
+ movdqa xmm0, XMMWORD PTR [%2]
+ lddqu xmm1, XMMWORD PTR [%3]
+ lddqu xmm2, XMMWORD PTR [%3+1]
+ lddqu xmm3, XMMWORD PTR [%3+2]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endif
+ movdqa xmm0, XMMWORD PTR [%2+%4]
+ lddqu xmm1, XMMWORD PTR [%3+%5]
+ lddqu xmm2, XMMWORD PTR [%3+%5+1]
+ lddqu xmm3, XMMWORD PTR [%3+%5+2]
+
+%if %1==0 || %1==1
+ lea %2, [%2+%4*2]
+ lea %3, [%3+%5*2]
+%endif
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endmacro
+
+%macro PROCESS_8X2X3 5
+%if %1==0
+ movq mm0, QWORD PTR [%2]
+ movq mm5, QWORD PTR [%3]
+ movq mm6, QWORD PTR [%3+1]
+ movq mm7, QWORD PTR [%3+2]
+
+ psadbw mm5, mm0
+ psadbw mm6, mm0
+ psadbw mm7, mm0
+%else
+ movq mm0, QWORD PTR [%2]
+ movq mm1, QWORD PTR [%3]
+ movq mm2, QWORD PTR [%3+1]
+ movq mm3, QWORD PTR [%3+2]
+
+ psadbw mm1, mm0
+ psadbw mm2, mm0
+ psadbw mm3, mm0
+
+ paddw mm5, mm1
+ paddw mm6, mm2
+ paddw mm7, mm3
+%endif
+ movq mm0, QWORD PTR [%2+%4]
+ movq mm1, QWORD PTR [%3+%5]
+ movq mm2, QWORD PTR [%3+%5+1]
+ movq mm3, QWORD PTR [%3+%5+2]
+
+%if %1==0 || %1==1
+ lea %2, [%2+%4*2]
+ lea %3, [%3+%5*2]
+%endif
+
+ psadbw mm1, mm0
+ psadbw mm2, mm0
+ psadbw mm3, mm0
+
+ paddw mm5, mm1
+ paddw mm6, mm2
+ paddw mm7, mm3
+%endmacro
+
+%macro LOAD_X4_ADDRESSES 5
+ mov %2, [%1+REG_SZ_BYTES*0]
+ mov %3, [%1+REG_SZ_BYTES*1]
+
+ mov %4, [%1+REG_SZ_BYTES*2]
+ mov %5, [%1+REG_SZ_BYTES*3]
+%endmacro
+
+%macro PROCESS_16X2X4 8
+%if %1==0
+ movdqa xmm0, XMMWORD PTR [%2]
+ lddqu xmm4, XMMWORD PTR [%3]
+ lddqu xmm5, XMMWORD PTR [%4]
+ lddqu xmm6, XMMWORD PTR [%5]
+ lddqu xmm7, XMMWORD PTR [%6]
+
+ psadbw xmm4, xmm0
+ psadbw xmm5, xmm0
+ psadbw xmm6, xmm0
+ psadbw xmm7, xmm0
+%else
+ movdqa xmm0, XMMWORD PTR [%2]
+ lddqu xmm1, XMMWORD PTR [%3]
+ lddqu xmm2, XMMWORD PTR [%4]
+ lddqu xmm3, XMMWORD PTR [%5]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm4, xmm1
+ lddqu xmm1, XMMWORD PTR [%6]
+ paddw xmm5, xmm2
+ paddw xmm6, xmm3
+
+ psadbw xmm1, xmm0
+ paddw xmm7, xmm1
+%endif
+ movdqa xmm0, XMMWORD PTR [%2+%7]
+ lddqu xmm1, XMMWORD PTR [%3+%8]
+ lddqu xmm2, XMMWORD PTR [%4+%8]
+ lddqu xmm3, XMMWORD PTR [%5+%8]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm4, xmm1
+ lddqu xmm1, XMMWORD PTR [%6+%8]
+ paddw xmm5, xmm2
+ paddw xmm6, xmm3
+
+%if %1==0 || %1==1
+ lea %2, [%2+%7*2]
+ lea %3, [%3+%8*2]
+
+ lea %4, [%4+%8*2]
+ lea %5, [%5+%8*2]
+
+ lea %6, [%6+%8*2]
+%endif
+ psadbw xmm1, xmm0
+ paddw xmm7, xmm1
+
+%endmacro
+
+%macro PROCESS_8X2X4 8
+%if %1==0
+ movq mm0, QWORD PTR [%2]
+ movq mm4, QWORD PTR [%3]
+ movq mm5, QWORD PTR [%4]
+ movq mm6, QWORD PTR [%5]
+ movq mm7, QWORD PTR [%6]
+
+ psadbw mm4, mm0
+ psadbw mm5, mm0
+ psadbw mm6, mm0
+ psadbw mm7, mm0
+%else
+ movq mm0, QWORD PTR [%2]
+ movq mm1, QWORD PTR [%3]
+ movq mm2, QWORD PTR [%4]
+ movq mm3, QWORD PTR [%5]
+
+ psadbw mm1, mm0
+ psadbw mm2, mm0
+ psadbw mm3, mm0
+
+ paddw mm4, mm1
+ movq mm1, QWORD PTR [%6]
+ paddw mm5, mm2
+ paddw mm6, mm3
+
+ psadbw mm1, mm0
+ paddw mm7, mm1
+%endif
+ movq mm0, QWORD PTR [%2+%7]
+ movq mm1, QWORD PTR [%3+%8]
+ movq mm2, QWORD PTR [%4+%8]
+ movq mm3, QWORD PTR [%5+%8]
+
+ psadbw mm1, mm0
+ psadbw mm2, mm0
+ psadbw mm3, mm0
+
+ paddw mm4, mm1
+ movq mm1, QWORD PTR [%6+%8]
+ paddw mm5, mm2
+ paddw mm6, mm3
+
+%if %1==0 || %1==1
+ lea %2, [%2+%7*2]
+ lea %3, [%3+%8*2]
+
+ lea %4, [%4+%8*2]
+ lea %5, [%5+%8*2]
+
+ lea %6, [%6+%8*2]
+%endif
+ psadbw mm1, mm0
+ paddw mm7, mm1
+
+%endmacro
+
+;void int vp8_sad16x16x3_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x16x3_sse3)
+sym(vp8_sad16x16x3_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
+
+ mov rcx, result_ptr
+
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rcx], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rcx+4], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rcx+8], xmm0
+
+ STACK_FRAME_DESTROY_X3
+
+;void int vp8_sad16x8x3_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x8x3_sse3)
+sym(vp8_sad16x8x3_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
+
+ mov rcx, result_ptr
+
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rcx], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rcx+4], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rcx+8], xmm0
+
+ STACK_FRAME_DESTROY_X3
+
+;void int vp8_sad8x16x3_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad8x16x3_sse3)
+sym(vp8_sad8x16x3_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
+
+ mov rcx, result_ptr
+
+ punpckldq mm5, mm6
+
+ movq [rcx], mm5
+ movd [rcx+8], mm7
+
+ STACK_FRAME_DESTROY_X3
+
+;void int vp8_sad8x8x3_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad8x8x3_sse3)
+sym(vp8_sad8x8x3_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
+ PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
+
+ mov rcx, result_ptr
+
+ punpckldq mm5, mm6
+
+ movq [rcx], mm5
+ movd [rcx+8], mm7
+
+ STACK_FRAME_DESTROY_X3
+
+;void int vp8_sad4x4x3_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad4x4x3_sse3)
+sym(vp8_sad4x4x3_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ movd mm0, DWORD PTR [src_ptr]
+ movd mm1, DWORD PTR [ref_ptr]
+
+ movd mm2, DWORD PTR [src_ptr+src_stride]
+ movd mm3, DWORD PTR [ref_ptr+ref_stride]
+
+ punpcklbw mm0, mm2
+ punpcklbw mm1, mm3
+
+ movd mm4, DWORD PTR [ref_ptr+1]
+ movd mm5, DWORD PTR [ref_ptr+2]
+
+ movd mm2, DWORD PTR [ref_ptr+ref_stride+1]
+ movd mm3, DWORD PTR [ref_ptr+ref_stride+2]
+
+ psadbw mm1, mm0
+
+ punpcklbw mm4, mm2
+ punpcklbw mm5, mm3
+
+ psadbw mm4, mm0
+ psadbw mm5, mm0
+
+ lea src_ptr, [src_ptr+src_stride*2]
+ lea ref_ptr, [ref_ptr+ref_stride*2]
+
+ movd mm0, DWORD PTR [src_ptr]
+ movd mm2, DWORD PTR [ref_ptr]
+
+ movd mm3, DWORD PTR [src_ptr+src_stride]
+ movd mm6, DWORD PTR [ref_ptr+ref_stride]
+
+ punpcklbw mm0, mm3
+ punpcklbw mm2, mm6
+
+ movd mm3, DWORD PTR [ref_ptr+1]
+ movd mm7, DWORD PTR [ref_ptr+2]
+
+ psadbw mm2, mm0
+
+ paddw mm1, mm2
+
+ movd mm2, DWORD PTR [ref_ptr+ref_stride+1]
+ movd mm6, DWORD PTR [ref_ptr+ref_stride+2]
+
+ punpcklbw mm3, mm2
+ punpcklbw mm7, mm6
+
+ psadbw mm3, mm0
+ psadbw mm7, mm0
+
+ paddw mm3, mm4
+ paddw mm7, mm5
+
+ mov rcx, result_ptr
+
+ punpckldq mm1, mm3
+
+ movq [rcx], mm1
+ movd [rcx+8], mm7
+
+ STACK_FRAME_DESTROY_X3
+
+;unsigned int vp8_sad16x16_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int max_sad)
+;%define lddqu movdqu
+global sym(vp8_sad16x16_sse3)
+sym(vp8_sad16x16_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+ mov end_ptr, 4
+ pxor xmm7, xmm7
+
+.vp8_sad16x16_sse3_loop:
+ movdqa xmm0, XMMWORD PTR [src_ptr]
+ movdqu xmm1, XMMWORD PTR [ref_ptr]
+ movdqa xmm2, XMMWORD PTR [src_ptr+src_stride]
+ movdqu xmm3, XMMWORD PTR [ref_ptr+ref_stride]
+
+ lea src_ptr, [src_ptr+src_stride*2]
+ lea ref_ptr, [ref_ptr+ref_stride*2]
+
+ movdqa xmm4, XMMWORD PTR [src_ptr]
+ movdqu xmm5, XMMWORD PTR [ref_ptr]
+ movdqa xmm6, XMMWORD PTR [src_ptr+src_stride]
+
+ psadbw xmm0, xmm1
+
+ movdqu xmm1, XMMWORD PTR [ref_ptr+ref_stride]
+
+ psadbw xmm2, xmm3
+ psadbw xmm4, xmm5
+ psadbw xmm6, xmm1
+
+ lea src_ptr, [src_ptr+src_stride*2]
+ lea ref_ptr, [ref_ptr+ref_stride*2]
+
+ paddw xmm7, xmm0
+ paddw xmm7, xmm2
+ paddw xmm7, xmm4
+ paddw xmm7, xmm6
+
+ sub end_ptr, 1
+ jne .vp8_sad16x16_sse3_loop
+
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+ paddw xmm0, xmm7
+ movq rax, xmm0
+
+ STACK_FRAME_DESTROY_X3
+
+;void vp8_copy32xn_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *dst_ptr,
+; int dst_stride,
+; int height);
+global sym(vp8_copy32xn_sse3)
+sym(vp8_copy32xn_sse3):
+
+ STACK_FRAME_CREATE_X3
+
+.block_copy_sse3_loopx4:
+ lea end_ptr, [src_ptr+src_stride*2]
+
+ movdqu xmm0, XMMWORD PTR [src_ptr]
+ movdqu xmm1, XMMWORD PTR [src_ptr + 16]
+ movdqu xmm2, XMMWORD PTR [src_ptr + src_stride]
+ movdqu xmm3, XMMWORD PTR [src_ptr + src_stride + 16]
+ movdqu xmm4, XMMWORD PTR [end_ptr]
+ movdqu xmm5, XMMWORD PTR [end_ptr + 16]
+ movdqu xmm6, XMMWORD PTR [end_ptr + src_stride]
+ movdqu xmm7, XMMWORD PTR [end_ptr + src_stride + 16]
+
+ lea src_ptr, [src_ptr+src_stride*4]
+
+ lea end_ptr, [ref_ptr+ref_stride*2]
+
+ movdqa XMMWORD PTR [ref_ptr], xmm0
+ movdqa XMMWORD PTR [ref_ptr + 16], xmm1
+ movdqa XMMWORD PTR [ref_ptr + ref_stride], xmm2
+ movdqa XMMWORD PTR [ref_ptr + ref_stride + 16], xmm3
+ movdqa XMMWORD PTR [end_ptr], xmm4
+ movdqa XMMWORD PTR [end_ptr + 16], xmm5
+ movdqa XMMWORD PTR [end_ptr + ref_stride], xmm6
+ movdqa XMMWORD PTR [end_ptr + ref_stride + 16], xmm7
+
+ lea ref_ptr, [ref_ptr+ref_stride*4]
+
+ sub height, 4
+ cmp height, 4
+ jge .block_copy_sse3_loopx4
+
+ ;Check to see if there is more rows need to be copied.
+ cmp height, 0
+ je .copy_is_done
+
+.block_copy_sse3_loop:
+ movdqu xmm0, XMMWORD PTR [src_ptr]
+ movdqu xmm1, XMMWORD PTR [src_ptr + 16]
+ lea src_ptr, [src_ptr+src_stride]
+
+ movdqa XMMWORD PTR [ref_ptr], xmm0
+ movdqa XMMWORD PTR [ref_ptr + 16], xmm1
+ lea ref_ptr, [ref_ptr+ref_stride]
+
+ sub height, 1
+ jne .block_copy_sse3_loop
+
+.copy_is_done:
+ STACK_FRAME_DESTROY_X3
+
+;void vp8_sad16x16x4d_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr_base,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x16x4d_sse3)
+sym(vp8_sad16x16x4d_sse3):
+
+ STACK_FRAME_CREATE_X4
+
+ PROCESS_16X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+
+%if ABI_IS_32BIT
+ pop rbp
+%endif
+ mov rcx, result_ptr
+
+ movq xmm0, xmm4
+ psrldq xmm4, 8
+
+ paddw xmm0, xmm4
+ movd [rcx], xmm0
+;-
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rcx+4], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rcx+8], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rcx+12], xmm0
+
+ STACK_FRAME_DESTROY_X4
+
+;void vp8_sad16x8x4d_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr_base,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x8x4d_sse3)
+sym(vp8_sad16x8x4d_sse3):
+
+ STACK_FRAME_CREATE_X4
+
+ PROCESS_16X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_16X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+
+%if ABI_IS_32BIT
+ pop rbp
+%endif
+ mov rcx, result_ptr
+
+ movq xmm0, xmm4
+ psrldq xmm4, 8
+
+ paddw xmm0, xmm4
+ movd [rcx], xmm0
+;-
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rcx+4], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rcx+8], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rcx+12], xmm0
+
+ STACK_FRAME_DESTROY_X4
+
+;void int vp8_sad8x16x4d_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad8x16x4d_sse3)
+sym(vp8_sad8x16x4d_sse3):
+
+ STACK_FRAME_CREATE_X4
+
+ PROCESS_8X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+
+%if ABI_IS_32BIT
+ pop rbp
+%endif
+ mov rcx, result_ptr
+
+ punpckldq mm4, mm5
+ punpckldq mm6, mm7
+
+ movq [rcx], mm4
+ movq [rcx+8], mm6
+
+ STACK_FRAME_DESTROY_X4
+
+;void int vp8_sad8x8x4d_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad8x8x4d_sse3)
+sym(vp8_sad8x8x4d_sse3):
+
+ STACK_FRAME_CREATE_X4
+
+ PROCESS_8X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+ PROCESS_8X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
+
+%if ABI_IS_32BIT
+ pop rbp
+%endif
+ mov rcx, result_ptr
+
+ punpckldq mm4, mm5
+ punpckldq mm6, mm7
+
+ movq [rcx], mm4
+ movq [rcx+8], mm6
+
+ STACK_FRAME_DESTROY_X4
+
+;void int vp8_sad4x4x4d_sse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad4x4x4d_sse3)
+sym(vp8_sad4x4x4d_sse3):
+
+ STACK_FRAME_CREATE_X4
+
+ movd mm0, DWORD PTR [src_ptr]
+ movd mm1, DWORD PTR [r0_ptr]
+
+ movd mm2, DWORD PTR [src_ptr+src_stride]
+ movd mm3, DWORD PTR [r0_ptr+ref_stride]
+
+ punpcklbw mm0, mm2
+ punpcklbw mm1, mm3
+
+ movd mm4, DWORD PTR [r1_ptr]
+ movd mm5, DWORD PTR [r2_ptr]
+
+ movd mm6, DWORD PTR [r3_ptr]
+ movd mm2, DWORD PTR [r1_ptr+ref_stride]
+
+ movd mm3, DWORD PTR [r2_ptr+ref_stride]
+ movd mm7, DWORD PTR [r3_ptr+ref_stride]
+
+ psadbw mm1, mm0
+
+ punpcklbw mm4, mm2
+ punpcklbw mm5, mm3
+
+ punpcklbw mm6, mm7
+ psadbw mm4, mm0
+
+ psadbw mm5, mm0
+ psadbw mm6, mm0
+
+
+
+ lea src_ptr, [src_ptr+src_stride*2]
+ lea r0_ptr, [r0_ptr+ref_stride*2]
+
+ lea r1_ptr, [r1_ptr+ref_stride*2]
+ lea r2_ptr, [r2_ptr+ref_stride*2]
+
+ lea r3_ptr, [r3_ptr+ref_stride*2]
+
+ movd mm0, DWORD PTR [src_ptr]
+ movd mm2, DWORD PTR [r0_ptr]
+
+ movd mm3, DWORD PTR [src_ptr+src_stride]
+ movd mm7, DWORD PTR [r0_ptr+ref_stride]
+
+ punpcklbw mm0, mm3
+ punpcklbw mm2, mm7
+
+ movd mm3, DWORD PTR [r1_ptr]
+ movd mm7, DWORD PTR [r2_ptr]
+
+ psadbw mm2, mm0
+%if ABI_IS_32BIT
+ mov rax, rbp
+
+ pop rbp
+%define ref_stride rax
+%endif
+ mov rsi, result_ptr
+
+ paddw mm1, mm2
+ movd [rsi], mm1
+
+ movd mm2, DWORD PTR [r1_ptr+ref_stride]
+ movd mm1, DWORD PTR [r2_ptr+ref_stride]
+
+ punpcklbw mm3, mm2
+ punpcklbw mm7, mm1
+
+ psadbw mm3, mm0
+ psadbw mm7, mm0
+
+ movd mm2, DWORD PTR [r3_ptr]
+ movd mm1, DWORD PTR [r3_ptr+ref_stride]
+
+ paddw mm3, mm4
+ paddw mm7, mm5
+
+ movd [rsi+4], mm3
+ punpcklbw mm2, mm1
+
+ movd [rsi+8], mm7
+ psadbw mm2, mm0
+
+ paddw mm2, mm6
+ movd [rsi+12], mm2
+
+
+ STACK_FRAME_DESTROY_X4
+
--- /dev/null
+++ b/vp8/common/x86/sad_sse4.asm
@@ -1,0 +1,353 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%macro PROCESS_16X2X8 1
+%if %1
+ movdqa xmm0, XMMWORD PTR [rsi]
+ movq xmm1, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ movq xmm2, MMWORD PTR [rdi+16]
+ punpcklqdq xmm1, xmm3
+ punpcklqdq xmm3, xmm2
+
+ movdqa xmm2, xmm1
+ mpsadbw xmm1, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+
+ psrldq xmm0, 8
+
+ movdqa xmm4, xmm3
+ mpsadbw xmm3, xmm0, 0x0
+ mpsadbw xmm4, xmm0, 0x5
+
+ paddw xmm1, xmm2
+ paddw xmm1, xmm3
+ paddw xmm1, xmm4
+%else
+ movdqa xmm0, XMMWORD PTR [rsi]
+ movq xmm5, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ movq xmm2, MMWORD PTR [rdi+16]
+ punpcklqdq xmm5, xmm3
+ punpcklqdq xmm3, xmm2
+
+ movdqa xmm2, xmm5
+ mpsadbw xmm5, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+
+ psrldq xmm0, 8
+
+ movdqa xmm4, xmm3
+ mpsadbw xmm3, xmm0, 0x0
+ mpsadbw xmm4, xmm0, 0x5
+
+ paddw xmm5, xmm2
+ paddw xmm5, xmm3
+ paddw xmm5, xmm4
+
+ paddw xmm1, xmm5
+%endif
+ movdqa xmm0, XMMWORD PTR [rsi + rax]
+ movq xmm5, MMWORD PTR [rdi+ rdx]
+ movq xmm3, MMWORD PTR [rdi+ rdx+8]
+ movq xmm2, MMWORD PTR [rdi+ rdx+16]
+ punpcklqdq xmm5, xmm3
+ punpcklqdq xmm3, xmm2
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ movdqa xmm2, xmm5
+ mpsadbw xmm5, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+
+ psrldq xmm0, 8
+ movdqa xmm4, xmm3
+ mpsadbw xmm3, xmm0, 0x0
+ mpsadbw xmm4, xmm0, 0x5
+
+ paddw xmm5, xmm2
+ paddw xmm5, xmm3
+ paddw xmm5, xmm4
+
+ paddw xmm1, xmm5
+%endmacro
+
+%macro PROCESS_8X2X8 1
+%if %1
+ movq xmm0, MMWORD PTR [rsi]
+ movq xmm1, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ punpcklqdq xmm1, xmm3
+
+ movdqa xmm2, xmm1
+ mpsadbw xmm1, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+ paddw xmm1, xmm2
+%else
+ movq xmm0, MMWORD PTR [rsi]
+ movq xmm5, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ punpcklqdq xmm5, xmm3
+
+ movdqa xmm2, xmm5
+ mpsadbw xmm5, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+ paddw xmm5, xmm2
+
+ paddw xmm1, xmm5
+%endif
+ movq xmm0, MMWORD PTR [rsi + rax]
+ movq xmm5, MMWORD PTR [rdi+ rdx]
+ movq xmm3, MMWORD PTR [rdi+ rdx+8]
+ punpcklqdq xmm5, xmm3
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ movdqa xmm2, xmm5
+ mpsadbw xmm5, xmm0, 0x0
+ mpsadbw xmm2, xmm0, 0x5
+ paddw xmm5, xmm2
+
+ paddw xmm1, xmm5
+%endmacro
+
+%macro PROCESS_4X2X8 1
+%if %1
+ movd xmm0, [rsi]
+ movq xmm1, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ punpcklqdq xmm1, xmm3
+
+ mpsadbw xmm1, xmm0, 0x0
+%else
+ movd xmm0, [rsi]
+ movq xmm5, MMWORD PTR [rdi]
+ movq xmm3, MMWORD PTR [rdi+8]
+ punpcklqdq xmm5, xmm3
+
+ mpsadbw xmm5, xmm0, 0x0
+
+ paddw xmm1, xmm5
+%endif
+ movd xmm0, [rsi + rax]
+ movq xmm5, MMWORD PTR [rdi+ rdx]
+ movq xmm3, MMWORD PTR [rdi+ rdx+8]
+ punpcklqdq xmm5, xmm3
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ mpsadbw xmm5, xmm0, 0x0
+
+ paddw xmm1, xmm5
+%endmacro
+
+
+;void vp8_sad16x16x8_sse4(
+; const unsigned char *src_ptr,
+; int src_stride,
+; const unsigned char *ref_ptr,
+; int ref_stride,
+; unsigned short *sad_array);
+global sym(vp8_sad16x16x8_sse4)
+sym(vp8_sad16x16x8_sse4):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ PROCESS_16X2X8 1
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+
+ mov rdi, arg(4) ;Results
+ movdqa XMMWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_sad16x8x8_sse4(
+; const unsigned char *src_ptr,
+; int src_stride,
+; const unsigned char *ref_ptr,
+; int ref_stride,
+; unsigned short *sad_array
+;);
+global sym(vp8_sad16x8x8_sse4)
+sym(vp8_sad16x8x8_sse4):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ PROCESS_16X2X8 1
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+ PROCESS_16X2X8 0
+
+ mov rdi, arg(4) ;Results
+ movdqa XMMWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_sad8x8x8_sse4(
+; const unsigned char *src_ptr,
+; int src_stride,
+; const unsigned char *ref_ptr,
+; int ref_stride,
+; unsigned short *sad_array
+;);
+global sym(vp8_sad8x8x8_sse4)
+sym(vp8_sad8x8x8_sse4):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ PROCESS_8X2X8 1
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+
+ mov rdi, arg(4) ;Results
+ movdqa XMMWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_sad8x16x8_sse4(
+; const unsigned char *src_ptr,
+; int src_stride,
+; const unsigned char *ref_ptr,
+; int ref_stride,
+; unsigned short *sad_array
+;);
+global sym(vp8_sad8x16x8_sse4)
+sym(vp8_sad8x16x8_sse4):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ PROCESS_8X2X8 1
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ PROCESS_8X2X8 0
+ mov rdi, arg(4) ;Results
+ movdqa XMMWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_sad4x4x8_c(
+; const unsigned char *src_ptr,
+; int src_stride,
+; const unsigned char *ref_ptr,
+; int ref_stride,
+; unsigned short *sad_array
+;);
+global sym(vp8_sad4x4x8_sse4)
+sym(vp8_sad4x4x8_sse4):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ PROCESS_4X2X8 1
+ PROCESS_4X2X8 0
+
+ mov rdi, arg(4) ;Results
+ movdqa XMMWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+
--- /dev/null
+++ b/vp8/common/x86/sad_ssse3.asm
@@ -1,0 +1,370 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%macro PROCESS_16X2X3 1
+%if %1
+ movdqa xmm0, XMMWORD PTR [rsi]
+ lddqu xmm5, XMMWORD PTR [rdi]
+ lddqu xmm6, XMMWORD PTR [rdi+1]
+ lddqu xmm7, XMMWORD PTR [rdi+2]
+
+ psadbw xmm5, xmm0
+ psadbw xmm6, xmm0
+ psadbw xmm7, xmm0
+%else
+ movdqa xmm0, XMMWORD PTR [rsi]
+ lddqu xmm1, XMMWORD PTR [rdi]
+ lddqu xmm2, XMMWORD PTR [rdi+1]
+ lddqu xmm3, XMMWORD PTR [rdi+2]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endif
+ movdqa xmm0, XMMWORD PTR [rsi+rax]
+ lddqu xmm1, XMMWORD PTR [rdi+rdx]
+ lddqu xmm2, XMMWORD PTR [rdi+rdx+1]
+ lddqu xmm3, XMMWORD PTR [rdi+rdx+2]
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endmacro
+
+%macro PROCESS_16X2X3_OFFSET 2
+%if %1
+ movdqa xmm0, XMMWORD PTR [rsi]
+ movdqa xmm4, XMMWORD PTR [rdi]
+ movdqa xmm7, XMMWORD PTR [rdi+16]
+
+ movdqa xmm5, xmm7
+ palignr xmm5, xmm4, %2
+
+ movdqa xmm6, xmm7
+ palignr xmm6, xmm4, (%2+1)
+
+ palignr xmm7, xmm4, (%2+2)
+
+ psadbw xmm5, xmm0
+ psadbw xmm6, xmm0
+ psadbw xmm7, xmm0
+%else
+ movdqa xmm0, XMMWORD PTR [rsi]
+ movdqa xmm4, XMMWORD PTR [rdi]
+ movdqa xmm3, XMMWORD PTR [rdi+16]
+
+ movdqa xmm1, xmm3
+ palignr xmm1, xmm4, %2
+
+ movdqa xmm2, xmm3
+ palignr xmm2, xmm4, (%2+1)
+
+ palignr xmm3, xmm4, (%2+2)
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endif
+ movdqa xmm0, XMMWORD PTR [rsi+rax]
+ movdqa xmm4, XMMWORD PTR [rdi+rdx]
+ movdqa xmm3, XMMWORD PTR [rdi+rdx+16]
+
+ movdqa xmm1, xmm3
+ palignr xmm1, xmm4, %2
+
+ movdqa xmm2, xmm3
+ palignr xmm2, xmm4, (%2+1)
+
+ palignr xmm3, xmm4, (%2+2)
+
+ lea rsi, [rsi+rax*2]
+ lea rdi, [rdi+rdx*2]
+
+ psadbw xmm1, xmm0
+ psadbw xmm2, xmm0
+ psadbw xmm3, xmm0
+
+ paddw xmm5, xmm1
+ paddw xmm6, xmm2
+ paddw xmm7, xmm3
+%endmacro
+
+%macro PROCESS_16X16X3_OFFSET 2
+%2_aligned_by_%1:
+
+ sub rdi, %1
+
+ PROCESS_16X2X3_OFFSET 1, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+
+ jmp %2_store_off
+
+%endmacro
+
+%macro PROCESS_16X8X3_OFFSET 2
+%2_aligned_by_%1:
+
+ sub rdi, %1
+
+ PROCESS_16X2X3_OFFSET 1, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+ PROCESS_16X2X3_OFFSET 0, %1
+
+ jmp %2_store_off
+
+%endmacro
+
+;void int vp8_sad16x16x3_ssse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x16x3_ssse3)
+sym(vp8_sad16x16x3_ssse3):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ SAVE_XMM 7
+ push rsi
+ push rdi
+ push rcx
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ mov rdx, 0xf
+ and rdx, rdi
+
+ jmp .vp8_sad16x16x3_ssse3_skiptable
+.vp8_sad16x16x3_ssse3_jumptable:
+ dd .vp8_sad16x16x3_ssse3_aligned_by_0 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_1 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_2 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_3 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_4 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_5 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_6 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_7 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_8 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_9 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_10 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_11 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_12 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_13 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_14 - .vp8_sad16x16x3_ssse3_do_jump
+ dd .vp8_sad16x16x3_ssse3_aligned_by_15 - .vp8_sad16x16x3_ssse3_do_jump
+.vp8_sad16x16x3_ssse3_skiptable:
+
+ call .vp8_sad16x16x3_ssse3_do_jump
+.vp8_sad16x16x3_ssse3_do_jump:
+ pop rcx ; get the address of do_jump
+ mov rax, .vp8_sad16x16x3_ssse3_jumptable - .vp8_sad16x16x3_ssse3_do_jump
+ add rax, rcx ; get the absolute address of vp8_sad16x16x3_ssse3_jumptable
+
+ movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
+ add rcx, rax
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ jmp rcx
+
+ PROCESS_16X16X3_OFFSET 0, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 1, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 2, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 3, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 4, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 5, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 6, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 7, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 8, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 9, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 10, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 11, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 12, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 13, .vp8_sad16x16x3_ssse3
+ PROCESS_16X16X3_OFFSET 14, .vp8_sad16x16x3_ssse3
+
+.vp8_sad16x16x3_ssse3_aligned_by_15:
+ PROCESS_16X2X3 1
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+
+.vp8_sad16x16x3_ssse3_store_off:
+ mov rdi, arg(4) ;Results
+
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rdi], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rdi+4], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rdi+8], xmm0
+
+ ; begin epilog
+ pop rcx
+ pop rdi
+ pop rsi
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void int vp8_sad16x8x3_ssse3(
+; unsigned char *src_ptr,
+; int src_stride,
+; unsigned char *ref_ptr,
+; int ref_stride,
+; int *results)
+global sym(vp8_sad16x8x3_ssse3)
+sym(vp8_sad16x8x3_ssse3):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 5
+ SAVE_XMM 7
+ push rsi
+ push rdi
+ push rcx
+ ; end prolog
+
+ mov rsi, arg(0) ;src_ptr
+ mov rdi, arg(2) ;ref_ptr
+
+ mov rdx, 0xf
+ and rdx, rdi
+
+ jmp .vp8_sad16x8x3_ssse3_skiptable
+.vp8_sad16x8x3_ssse3_jumptable:
+ dd .vp8_sad16x8x3_ssse3_aligned_by_0 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_1 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_2 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_3 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_4 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_5 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_6 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_7 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_8 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_9 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_10 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_11 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_12 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_13 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_14 - .vp8_sad16x8x3_ssse3_do_jump
+ dd .vp8_sad16x8x3_ssse3_aligned_by_15 - .vp8_sad16x8x3_ssse3_do_jump
+.vp8_sad16x8x3_ssse3_skiptable:
+
+ call .vp8_sad16x8x3_ssse3_do_jump
+.vp8_sad16x8x3_ssse3_do_jump:
+ pop rcx ; get the address of do_jump
+ mov rax, .vp8_sad16x8x3_ssse3_jumptable - .vp8_sad16x8x3_ssse3_do_jump
+ add rax, rcx ; get the absolute address of vp8_sad16x8x3_ssse3_jumptable
+
+ movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
+ add rcx, rax
+
+ movsxd rax, dword ptr arg(1) ;src_stride
+ movsxd rdx, dword ptr arg(3) ;ref_stride
+
+ jmp rcx
+
+ PROCESS_16X8X3_OFFSET 0, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 1, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 2, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 3, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 4, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 5, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 6, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 7, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 8, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 9, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 10, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 11, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 12, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 13, .vp8_sad16x8x3_ssse3
+ PROCESS_16X8X3_OFFSET 14, .vp8_sad16x8x3_ssse3
+
+.vp8_sad16x8x3_ssse3_aligned_by_15:
+
+ PROCESS_16X2X3 1
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+ PROCESS_16X2X3 0
+
+.vp8_sad16x8x3_ssse3_store_off:
+ mov rdi, arg(4) ;Results
+
+ movq xmm0, xmm5
+ psrldq xmm5, 8
+
+ paddw xmm0, xmm5
+ movd [rdi], xmm0
+;-
+ movq xmm0, xmm6
+ psrldq xmm6, 8
+
+ paddw xmm0, xmm6
+ movd [rdi+4], xmm0
+;-
+ movq xmm0, xmm7
+ psrldq xmm7, 8
+
+ paddw xmm0, xmm7
+ movd [rdi+8], xmm0
+
+ ; begin epilog
+ pop rcx
+ pop rdi
+ pop rsi
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
--- /dev/null
+++ b/vp8/common/x86/variance_impl_mmx.asm
@@ -1,0 +1,851 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+;unsigned int vp8_get_mb_ss_mmx( short *src_ptr )
+global sym(vp8_get_mb_ss_mmx)
+sym(vp8_get_mb_ss_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 8
+ ; end prolog
+
+ mov rax, arg(0) ;src_ptr
+ mov rcx, 16
+ pxor mm4, mm4
+
+.NEXTROW:
+ movq mm0, [rax]
+ movq mm1, [rax+8]
+ movq mm2, [rax+16]
+ movq mm3, [rax+24]
+ pmaddwd mm0, mm0
+ pmaddwd mm1, mm1
+ pmaddwd mm2, mm2
+ pmaddwd mm3, mm3
+
+ paddd mm4, mm0
+ paddd mm4, mm1
+ paddd mm4, mm2
+ paddd mm4, mm3
+
+ add rax, 32
+ dec rcx
+ ja .NEXTROW
+ movq QWORD PTR [rsp], mm4
+
+ ;return sum[0]+sum[1];
+ movsxd rax, dword ptr [rsp]
+ movsxd rcx, dword ptr [rsp+4]
+ add rax, rcx
+
+
+ ; begin epilog
+ add rsp, 8
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_get8x8var_mmx
+;(
+; unsigned char *src_ptr,
+; int source_stride,
+; unsigned char *ref_ptr,
+; int recon_stride,
+; unsigned int *SSE,
+; int *Sum
+;)
+global sym(vp8_get8x8var_mmx)
+sym(vp8_get8x8var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 6
+ push rsi
+ push rdi
+ push rbx
+ sub rsp, 16
+ ; end prolog
+
+
+ pxor mm5, mm5 ; Blank mmx6
+ pxor mm6, mm6 ; Blank mmx7
+ pxor mm7, mm7 ; Blank mmx7
+
+ mov rax, arg(0) ;[src_ptr] ; Load base addresses
+ mov rbx, arg(2) ;[ref_ptr]
+ movsxd rcx, dword ptr arg(1) ;[source_stride]
+ movsxd rdx, dword ptr arg(3) ;[recon_stride]
+
+ ; Row 1
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+
+ ; Row 2
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 3
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 4
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 5
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ ; movq mm4, [rbx + rdx]
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 6
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 7
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Row 8
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm2, mm0 ; Take copies
+ movq mm3, mm1 ; Take copies
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ punpckhbw mm2, mm6 ; unpack to higher prrcision
+ punpckhbw mm3, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ psubsw mm2, mm3 ; A-B (high order) to MM2
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+ paddw mm5, mm2 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ pmaddwd mm2, mm2 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ paddd mm7, mm0 ; accumulate in mm7
+ paddd mm7, mm2 ; accumulate in mm7
+
+ ; Now accumulate the final results.
+ movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
+ movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
+ movsx rdx, WORD PTR [rsp+8]
+ movsx rcx, WORD PTR [rsp+10]
+ movsx rbx, WORD PTR [rsp+12]
+ movsx rax, WORD PTR [rsp+14]
+ add rdx, rcx
+ add rbx, rax
+ add rdx, rbx ;XSum
+ movsxd rax, DWORD PTR [rsp]
+ movsxd rcx, DWORD PTR [rsp+4]
+ add rax, rcx ;XXSum
+ mov rsi, arg(4) ;SSE
+ mov rdi, arg(5) ;Sum
+ mov dword ptr [rsi], eax
+ mov dword ptr [rdi], edx
+ xor rax, rax ; return 0
+
+
+ ; begin epilog
+ add rsp, 16
+ pop rbx
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+;unsigned int
+;vp8_get4x4var_mmx
+;(
+; unsigned char *src_ptr,
+; int source_stride,
+; unsigned char *ref_ptr,
+; int recon_stride,
+; unsigned int *SSE,
+; int *Sum
+;)
+global sym(vp8_get4x4var_mmx)
+sym(vp8_get4x4var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 6
+ push rsi
+ push rdi
+ push rbx
+ sub rsp, 16
+ ; end prolog
+
+
+ pxor mm5, mm5 ; Blank mmx6
+ pxor mm6, mm6 ; Blank mmx7
+ pxor mm7, mm7 ; Blank mmx7
+
+ mov rax, arg(0) ;[src_ptr] ; Load base addresses
+ mov rbx, arg(2) ;[ref_ptr]
+ movsxd rcx, dword ptr arg(1) ;[source_stride]
+ movsxd rdx, dword ptr arg(3) ;[recon_stride]
+
+ ; Row 1
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ paddw mm5, mm0 ; accumulate differences in mm5
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+
+ ; Row 2
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ paddw mm5, mm0 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+ ; Row 3
+ movq mm0, [rax] ; Copy eight bytes to mm0
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ paddw mm5, mm0 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movq mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+ ; Row 4
+ movq mm0, [rax] ; Copy eight bytes to mm0
+
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+
+ paddw mm5, mm0 ; accumulate differences in mm5
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ paddd mm7, mm0 ; accumulate in mm7
+
+
+ ; Now accumulate the final results.
+ movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
+ movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
+ movsx rdx, WORD PTR [rsp+8]
+ movsx rcx, WORD PTR [rsp+10]
+ movsx rbx, WORD PTR [rsp+12]
+ movsx rax, WORD PTR [rsp+14]
+ add rdx, rcx
+ add rbx, rax
+ add rdx, rbx ;XSum
+ movsxd rax, DWORD PTR [rsp]
+ movsxd rcx, DWORD PTR [rsp+4]
+ add rax, rcx ;XXSum
+ mov rsi, arg(4) ;SSE
+ mov rdi, arg(5) ;Sum
+ mov dword ptr [rsi], eax
+ mov dword ptr [rdi], edx
+ xor rax, rax ; return 0
+
+
+ ; begin epilog
+ add rsp, 16
+ pop rbx
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+;unsigned int
+;vp8_get4x4sse_cs_mmx
+;(
+; unsigned char *src_ptr,
+; int source_stride,
+; unsigned char *ref_ptr,
+; int recon_stride
+;)
+global sym(vp8_get4x4sse_cs_mmx)
+sym(vp8_get4x4sse_cs_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 4
+ push rsi
+ push rdi
+ push rbx
+ ; end prolog
+
+
+ pxor mm6, mm6 ; Blank mmx7
+ pxor mm7, mm7 ; Blank mmx7
+
+ mov rax, arg(0) ;[src_ptr] ; Load base addresses
+ mov rbx, arg(2) ;[ref_ptr]
+ movsxd rcx, dword ptr arg(1) ;[source_stride]
+ movsxd rdx, dword ptr arg(3) ;[recon_stride]
+ ; Row 1
+ movd mm0, [rax] ; Copy eight bytes to mm0
+ movd mm1, [rbx] ; Copy eight bytes to mm1
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movd mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+ ; Row 2
+ movd mm0, [rax] ; Copy eight bytes to mm0
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movd mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+ ; Row 3
+ movd mm0, [rax] ; Copy eight bytes to mm0
+ punpcklbw mm1, mm6
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+
+ pmaddwd mm0, mm0 ; square and accumulate
+ add rbx,rdx ; Inc pointer into ref data
+ add rax,rcx ; Inc pointer into the new data
+ movd mm1, [rbx] ; Copy eight bytes to mm1
+ paddd mm7, mm0 ; accumulate in mm7
+
+ ; Row 4
+ movd mm0, [rax] ; Copy eight bytes to mm0
+ punpcklbw mm0, mm6 ; unpack to higher prrcision
+ punpcklbw mm1, mm6
+ psubsw mm0, mm1 ; A-B (low order) to MM0
+ pmaddwd mm0, mm0 ; square and accumulate
+ paddd mm7, mm0 ; accumulate in mm7
+
+ movq mm0, mm7 ;
+ psrlq mm7, 32
+
+ paddd mm0, mm7
+ movq rax, mm0
+
+
+ ; begin epilog
+ pop rbx
+ pop rdi
+ pop rsi
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+%define mmx_filter_shift 7
+
+;void vp8_filter_block2d_bil4x4_var_mmx
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned short *HFilter,
+; unsigned short *VFilter,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_filter_block2d_bil4x4_var_mmx)
+sym(vp8_filter_block2d_bil4x4_var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 8
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+
+ pxor mm6, mm6 ;
+ pxor mm7, mm7 ;
+
+ mov rax, arg(4) ;HFilter ;
+ mov rdx, arg(5) ;VFilter ;
+
+ mov rsi, arg(0) ;ref_ptr ;
+ mov rdi, arg(2) ;src_ptr ;
+
+ mov rcx, 4 ;
+ pxor mm0, mm0 ;
+
+ movd mm1, [rsi] ;
+ movd mm3, [rsi+1] ;
+
+ punpcklbw mm1, mm0 ;
+ pmullw mm1, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ movq mm5, mm1
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rsi, r8
+%endif
+
+.filter_block2d_bil4x4_var_mmx_loop:
+
+ movd mm1, [rsi] ;
+ movd mm3, [rsi+1] ;
+
+ punpcklbw mm1, mm0 ;
+ pmullw mm1, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ movq mm3, mm5 ;
+
+ movq mm5, mm1 ;
+ pmullw mm3, [rdx] ;
+
+ pmullw mm1, [rdx+8] ;
+ paddw mm1, mm3 ;
+
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm1, mmx_filter_shift ;
+
+ movd mm3, [rdi] ;
+ punpcklbw mm3, mm0 ;
+
+ psubw mm1, mm3 ;
+ paddw mm6, mm1 ;
+
+ pmaddwd mm1, mm1 ;
+ paddd mm7, mm1 ;
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rdi, dword ptr arg(3) ;src_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+ add rsi, r8
+ add rdi, r9
+%endif
+ sub rcx, 1 ;
+ jnz .filter_block2d_bil4x4_var_mmx_loop ;
+
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rdi, arg(6) ;sum
+ mov rsi, arg(7) ;sumsquared
+
+ movd dword ptr [rdi], mm2 ;
+ movd dword ptr [rsi], mm4 ;
+
+
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+
+;void vp8_filter_block2d_bil_var_mmx
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; unsigned short *HFilter,
+; unsigned short *VFilter,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_filter_block2d_bil_var_mmx)
+sym(vp8_filter_block2d_bil_var_mmx):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 9
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+ pxor mm6, mm6 ;
+ pxor mm7, mm7 ;
+ mov rax, arg(5) ;HFilter ;
+
+ mov rdx, arg(6) ;VFilter ;
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+
+ pxor mm0, mm0 ;
+ movq mm1, [rsi] ;
+
+ movq mm3, [rsi+1] ;
+ movq mm2, mm1 ;
+
+ movq mm4, mm3 ;
+ punpcklbw mm1, mm0 ;
+
+ punpckhbw mm2, mm0 ;
+ pmullw mm1, [rax] ;
+
+ pmullw mm2, [rax] ;
+ punpcklbw mm3, mm0 ;
+
+ punpckhbw mm4, mm0 ;
+ pmullw mm3, [rax+8] ;
+
+ pmullw mm4, [rax+8] ;
+ paddw mm1, mm3 ;
+
+ paddw mm2, mm4 ;
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm2, mmx_filter_shift ;
+ movq mm5, mm1
+
+ packuswb mm5, mm2 ;
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ add rsi, r8
+%endif
+
+.filter_block2d_bil_var_mmx_loop:
+
+ movq mm1, [rsi] ;
+ movq mm3, [rsi+1] ;
+
+ movq mm2, mm1 ;
+ movq mm4, mm3 ;
+
+ punpcklbw mm1, mm0 ;
+ punpckhbw mm2, mm0 ;
+
+ pmullw mm1, [rax] ;
+ pmullw mm2, [rax] ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ pmullw mm3, [rax+8] ;
+ pmullw mm4, [rax+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm2, mm4 ;
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm1, mmx_filter_shift ;
+
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+ psraw mm2, mmx_filter_shift ;
+
+ movq mm3, mm5 ;
+ movq mm4, mm5 ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ movq mm5, mm1 ;
+ packuswb mm5, mm2 ;
+
+ pmullw mm3, [rdx] ;
+ pmullw mm4, [rdx] ;
+
+ pmullw mm1, [rdx+8] ;
+ pmullw mm2, [rdx+8] ;
+
+ paddw mm1, mm3 ;
+ paddw mm2, mm4 ;
+
+ paddw mm1, [GLOBAL(mmx_bi_rd)] ;
+ paddw mm2, [GLOBAL(mmx_bi_rd)] ;
+
+ psraw mm1, mmx_filter_shift ;
+ psraw mm2, mmx_filter_shift ;
+
+ movq mm3, [rdi] ;
+ movq mm4, mm3 ;
+
+ punpcklbw mm3, mm0 ;
+ punpckhbw mm4, mm0 ;
+
+ psubw mm1, mm3 ;
+ psubw mm2, mm4 ;
+
+ paddw mm6, mm1 ;
+ pmaddwd mm1, mm1 ;
+
+ paddw mm6, mm2 ;
+ pmaddwd mm2, mm2 ;
+
+ paddd mm7, mm1 ;
+ paddd mm7, mm2 ;
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
+ add rdi, dword ptr arg(3) ;src_pixels_per_line ;
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line ;
+ add rsi, r8
+ add rdi, r9
+%endif
+ sub rcx, 1 ;
+ jnz .filter_block2d_bil_var_mmx_loop ;
+
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rdi, arg(7) ;sum
+ mov rsi, arg(8) ;sumsquared
+
+ movd dword ptr [rdi], mm2 ;
+ movd dword ptr [rsi], mm4 ;
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+SECTION_RODATA
+;short mmx_bi_rd[4] = { 64, 64, 64, 64};+align 16
+mmx_bi_rd:
+ times 4 dw 64
--- /dev/null
+++ b/vp8/common/x86/variance_impl_sse2.asm
@@ -1,0 +1,1359 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%define xmm_filter_shift 7
+
+;unsigned int vp8_get_mb_ss_sse2
+;(
+; short *src_ptr
+;)
+global sym(vp8_get_mb_ss_sse2)
+sym(vp8_get_mb_ss_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 1
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+
+ mov rax, arg(0) ;[src_ptr]
+ mov rcx, 8
+ pxor xmm4, xmm4
+
+.NEXTROW:
+ movdqa xmm0, [rax]
+ movdqa xmm1, [rax+16]
+ movdqa xmm2, [rax+32]
+ movdqa xmm3, [rax+48]
+ pmaddwd xmm0, xmm0
+ pmaddwd xmm1, xmm1
+ pmaddwd xmm2, xmm2
+ pmaddwd xmm3, xmm3
+
+ paddd xmm0, xmm1
+ paddd xmm2, xmm3
+ paddd xmm4, xmm0
+ paddd xmm4, xmm2
+
+ add rax, 0x40
+ dec rcx
+ ja .NEXTROW
+
+ movdqa xmm3,xmm4
+ psrldq xmm4,8
+ paddd xmm4,xmm3
+ movdqa xmm3,xmm4
+ psrldq xmm4,4
+ paddd xmm4,xmm3
+ movq rax,xmm4
+
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;unsigned int vp8_get16x16var_sse2
+;(
+; unsigned char * src_ptr,
+; int source_stride,
+; unsigned char * ref_ptr,
+; int recon_stride,
+; unsigned int * SSE,
+; int * Sum
+;)
+global sym(vp8_get16x16var_sse2)
+sym(vp8_get16x16var_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 6
+ SAVE_XMM 7
+ push rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ mov rsi, arg(0) ;[src_ptr]
+ mov rdi, arg(2) ;[ref_ptr]
+
+ movsxd rax, DWORD PTR arg(1) ;[source_stride]
+ movsxd rdx, DWORD PTR arg(3) ;[recon_stride]
+
+ ; Prefetch data
+ lea rcx, [rax+rax*2]
+ prefetcht0 [rsi]
+ prefetcht0 [rsi+rax]
+ prefetcht0 [rsi+rax*2]
+ prefetcht0 [rsi+rcx]
+ lea rbx, [rsi+rax*4]
+ prefetcht0 [rbx]
+ prefetcht0 [rbx+rax]
+ prefetcht0 [rbx+rax*2]
+ prefetcht0 [rbx+rcx]
+
+ lea rcx, [rdx+rdx*2]
+ prefetcht0 [rdi]
+ prefetcht0 [rdi+rdx]
+ prefetcht0 [rdi+rdx*2]
+ prefetcht0 [rdi+rcx]
+ lea rbx, [rdi+rdx*4]
+ prefetcht0 [rbx]
+ prefetcht0 [rbx+rdx]
+ prefetcht0 [rbx+rdx*2]
+ prefetcht0 [rbx+rcx]
+
+ pxor xmm0, xmm0 ; clear xmm0 for unpack
+ pxor xmm7, xmm7 ; clear xmm7 for accumulating diffs
+
+ pxor xmm6, xmm6 ; clear xmm6 for accumulating sse
+ mov rcx, 16
+
+.var16loop:
+ movdqu xmm1, XMMWORD PTR [rsi]
+ movdqu xmm2, XMMWORD PTR [rdi]
+
+ prefetcht0 [rsi+rax*8]
+ prefetcht0 [rdi+rdx*8]
+
+ movdqa xmm3, xmm1
+ movdqa xmm4, xmm2
+
+
+ punpcklbw xmm1, xmm0
+ punpckhbw xmm3, xmm0
+
+ punpcklbw xmm2, xmm0
+ punpckhbw xmm4, xmm0
+
+
+ psubw xmm1, xmm2
+ psubw xmm3, xmm4
+
+ paddw xmm7, xmm1
+ pmaddwd xmm1, xmm1
+
+ paddw xmm7, xmm3
+ pmaddwd xmm3, xmm3
+
+ paddd xmm6, xmm1
+ paddd xmm6, xmm3
+
+ add rsi, rax
+ add rdi, rdx
+
+ sub rcx, 1
+ jnz .var16loop
+
+
+ movdqa xmm1, xmm6
+ pxor xmm6, xmm6
+
+ pxor xmm5, xmm5
+ punpcklwd xmm6, xmm7
+
+ punpckhwd xmm5, xmm7
+ psrad xmm5, 16
+
+ psrad xmm6, 16
+ paddd xmm6, xmm5
+
+ movdqa xmm2, xmm1
+ punpckldq xmm1, xmm0
+
+ punpckhdq xmm2, xmm0
+ movdqa xmm7, xmm6
+
+ paddd xmm1, xmm2
+ punpckldq xmm6, xmm0
+
+ punpckhdq xmm7, xmm0
+ paddd xmm6, xmm7
+
+ movdqa xmm2, xmm1
+ movdqa xmm7, xmm6
+
+ psrldq xmm1, 8
+ psrldq xmm6, 8
+
+ paddd xmm7, xmm6
+ paddd xmm1, xmm2
+
+ mov rax, arg(5) ;[Sum]
+ mov rdi, arg(4) ;[SSE]
+
+ movd DWORD PTR [rax], xmm7
+ movd DWORD PTR [rdi], xmm1
+
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ pop rbx
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+
+
+;unsigned int vp8_get8x8var_sse2
+;(
+; unsigned char * src_ptr,
+; int source_stride,
+; unsigned char * ref_ptr,
+; int recon_stride,
+; unsigned int * SSE,
+; int * Sum
+;)
+global sym(vp8_get8x8var_sse2)
+sym(vp8_get8x8var_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 6
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ sub rsp, 16
+ ; end prolog
+
+ mov rsi, arg(0) ;[src_ptr]
+ mov rdi, arg(2) ;[ref_ptr]
+
+ movsxd rax, DWORD PTR arg(1) ;[source_stride]
+ movsxd rdx, DWORD PTR arg(3) ;[recon_stride]
+
+ pxor xmm0, xmm0 ; clear xmm0 for unpack
+ pxor xmm7, xmm7 ; clear xmm7 for accumulating diffs
+
+ movq xmm1, QWORD PTR [rsi]
+ movq xmm2, QWORD PTR [rdi]
+
+ punpcklbw xmm1, xmm0
+ punpcklbw xmm2, xmm0
+
+ psubsw xmm1, xmm2
+ paddw xmm7, xmm1
+
+ pmaddwd xmm1, xmm1
+
+ movq xmm2, QWORD PTR[rsi + rax]
+ movq xmm3, QWORD PTR[rdi + rdx]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+
+ movq xmm2, QWORD PTR[rsi + rax * 2]
+ movq xmm3, QWORD PTR[rdi + rdx * 2]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+
+ lea rsi, [rsi + rax * 2]
+ lea rdi, [rdi + rdx * 2]
+ movq xmm2, QWORD PTR[rsi + rax]
+ movq xmm3, QWORD PTR[rdi + rdx]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+ movq xmm2, QWORD PTR[rsi + rax *2]
+ movq xmm3, QWORD PTR[rdi + rdx *2]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+
+ lea rsi, [rsi + rax * 2]
+ lea rdi, [rdi + rdx * 2]
+
+
+ movq xmm2, QWORD PTR[rsi + rax]
+ movq xmm3, QWORD PTR[rdi + rdx]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+ movq xmm2, QWORD PTR[rsi + rax *2]
+ movq xmm3, QWORD PTR[rdi + rdx *2]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+
+ lea rsi, [rsi + rax * 2]
+ lea rdi, [rdi + rdx * 2]
+
+ movq xmm2, QWORD PTR[rsi + rax]
+ movq xmm3, QWORD PTR[rdi + rdx]
+
+ punpcklbw xmm2, xmm0
+ punpcklbw xmm3, xmm0
+
+ psubsw xmm2, xmm3
+ paddw xmm7, xmm2
+
+ pmaddwd xmm2, xmm2
+ paddd xmm1, xmm2
+
+
+ movdqa xmm6, xmm7
+ punpcklwd xmm6, xmm0
+
+ punpckhwd xmm7, xmm0
+ movdqa xmm2, xmm1
+
+ paddw xmm6, xmm7
+ punpckldq xmm1, xmm0
+
+ punpckhdq xmm2, xmm0
+ movdqa xmm7, xmm6
+
+ paddd xmm1, xmm2
+ punpckldq xmm6, xmm0
+
+ punpckhdq xmm7, xmm0
+ paddw xmm6, xmm7
+
+ movdqa xmm2, xmm1
+ movdqa xmm7, xmm6
+
+ psrldq xmm1, 8
+ psrldq xmm6, 8
+
+ paddw xmm7, xmm6
+ paddd xmm1, xmm2
+
+ mov rax, arg(5) ;[Sum]
+ mov rdi, arg(4) ;[SSE]
+
+ movq rdx, xmm7
+ movsx rcx, dx
+
+ mov dword ptr [rax], ecx
+ movd DWORD PTR [rdi], xmm1
+
+ ; begin epilog
+ add rsp, 16
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp8_filter_block2d_bil_var_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int xoffset,
+; int yoffset,
+; int *sum,
+; unsigned int *sumsquared;;
+;
+;)
+global sym(vp8_filter_block2d_bil_var_sse2)
+sym(vp8_filter_block2d_bil_var_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 9
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ push rbx
+ ; end prolog
+
+ pxor xmm6, xmm6 ;
+ pxor xmm7, xmm7 ;
+
+ lea rsi, [GLOBAL(xmm_bi_rd)] ; rounding
+ movdqa xmm4, XMMWORD PTR [rsi]
+
+ lea rcx, [GLOBAL(vp8_bilinear_filters_sse2)]
+ movsxd rax, dword ptr arg(5) ; xoffset
+
+ cmp rax, 0 ; skip first_pass filter if xoffset=0
+ je filter_block2d_bil_var_sse2_sp_only
+
+ shl rax, 5 ; point to filter coeff with xoffset
+ lea rax, [rax + rcx] ; HFilter
+
+ movsxd rdx, dword ptr arg(6) ; yoffset
+
+ cmp rdx, 0 ; skip second_pass filter if yoffset=0
+ je filter_block2d_bil_var_sse2_fp_only
+
+ shl rdx, 5
+ lea rdx, [rdx + rcx] ; VFilter
+
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+
+ pxor xmm0, xmm0 ;
+ movq xmm1, QWORD PTR [rsi] ;
+ movq xmm3, QWORD PTR [rsi+1] ;
+
+ punpcklbw xmm1, xmm0 ;
+ pmullw xmm1, [rax] ;
+ punpcklbw xmm3, xmm0
+ pmullw xmm3, [rax+16] ;
+
+ paddw xmm1, xmm3 ;
+ paddw xmm1, xmm4 ;
+ psraw xmm1, xmm_filter_shift ;
+ movdqa xmm5, xmm1
+
+ movsxd rbx, dword ptr arg(1) ;ref_pixels_per_line
+ lea rsi, [rsi + rbx]
+%if ABI_IS_32BIT=0
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+filter_block2d_bil_var_sse2_loop:
+ movq xmm1, QWORD PTR [rsi] ;
+ movq xmm3, QWORD PTR [rsi+1] ;
+
+ punpcklbw xmm1, xmm0 ;
+ pmullw xmm1, [rax] ;
+ punpcklbw xmm3, xmm0 ;
+ pmullw xmm3, [rax+16] ;
+
+ paddw xmm1, xmm3 ;
+ paddw xmm1, xmm4 ;
+ psraw xmm1, xmm_filter_shift ;
+
+ movdqa xmm3, xmm5 ;
+ movdqa xmm5, xmm1 ;
+
+ pmullw xmm3, [rdx] ;
+ pmullw xmm1, [rdx+16] ;
+ paddw xmm1, xmm3 ;
+ paddw xmm1, xmm4 ;
+ psraw xmm1, xmm_filter_shift ;
+
+ movq xmm3, QWORD PTR [rdi] ;
+ punpcklbw xmm3, xmm0 ;
+
+ psubw xmm1, xmm3 ;
+ paddw xmm6, xmm1 ;
+
+ pmaddwd xmm1, xmm1 ;
+ paddd xmm7, xmm1 ;
+
+ lea rsi, [rsi + rbx] ;ref_pixels_per_line
+%if ABI_IS_32BIT
+ add rdi, dword ptr arg(3) ;src_pixels_per_line
+%else
+ lea rdi, [rdi + r9]
+%endif
+
+ sub rcx, 1 ;
+ jnz filter_block2d_bil_var_sse2_loop ;
+
+ jmp filter_block2d_bil_variance
+
+filter_block2d_bil_var_sse2_sp_only:
+ movsxd rdx, dword ptr arg(6) ; yoffset
+
+ cmp rdx, 0 ; skip all if both xoffset=0 and yoffset=0
+ je filter_block2d_bil_var_sse2_full_pixel
+
+ shl rdx, 5
+ lea rdx, [rdx + rcx] ; VFilter
+
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+ movq xmm1, QWORD PTR [rsi] ;
+ punpcklbw xmm1, xmm0 ;
+
+ movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
+ lea rsi, [rsi + rax]
+
+filter_block2d_bil_sp_only_loop:
+ movq xmm3, QWORD PTR [rsi] ;
+ punpcklbw xmm3, xmm0 ;
+ movdqa xmm5, xmm3
+
+ pmullw xmm1, [rdx] ;
+ pmullw xmm3, [rdx+16] ;
+ paddw xmm1, xmm3 ;
+ paddw xmm1, xmm4 ;
+ psraw xmm1, xmm_filter_shift ;
+
+ movq xmm3, QWORD PTR [rdi] ;
+ punpcklbw xmm3, xmm0 ;
+
+ psubw xmm1, xmm3 ;
+ paddw xmm6, xmm1 ;
+
+ pmaddwd xmm1, xmm1 ;
+ paddd xmm7, xmm1 ;
+
+ movdqa xmm1, xmm5 ;
+ lea rsi, [rsi + rax] ;ref_pixels_per_line
+ lea rdi, [rdi + rbx] ;src_pixels_per_line
+
+ sub rcx, 1 ;
+ jnz filter_block2d_bil_sp_only_loop ;
+
+ jmp filter_block2d_bil_variance
+
+filter_block2d_bil_var_sse2_full_pixel:
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
+ pxor xmm0, xmm0 ;
+
+filter_block2d_bil_full_pixel_loop:
+ movq xmm1, QWORD PTR [rsi] ;
+ punpcklbw xmm1, xmm0 ;
+
+ movq xmm2, QWORD PTR [rdi] ;
+ punpcklbw xmm2, xmm0 ;
+
+ psubw xmm1, xmm2 ;
+ paddw xmm6, xmm1 ;
+
+ pmaddwd xmm1, xmm1 ;
+ paddd xmm7, xmm1 ;
+
+ lea rsi, [rsi + rax] ;ref_pixels_per_line
+ lea rdi, [rdi + rbx] ;src_pixels_per_line
+
+ sub rcx, 1 ;
+ jnz filter_block2d_bil_full_pixel_loop ;
+
+ jmp filter_block2d_bil_variance
+
+filter_block2d_bil_var_sse2_fp_only:
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+ movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
+
+filter_block2d_bil_fp_only_loop:
+ movq xmm1, QWORD PTR [rsi] ;
+ movq xmm3, QWORD PTR [rsi+1] ;
+
+ punpcklbw xmm1, xmm0 ;
+ pmullw xmm1, [rax] ;
+ punpcklbw xmm3, xmm0 ;
+ pmullw xmm3, [rax+16] ;
+
+ paddw xmm1, xmm3 ;
+ paddw xmm1, xmm4 ;
+ psraw xmm1, xmm_filter_shift ;
+
+ movq xmm3, QWORD PTR [rdi] ;
+ punpcklbw xmm3, xmm0 ;
+
+ psubw xmm1, xmm3 ;
+ paddw xmm6, xmm1 ;
+
+ pmaddwd xmm1, xmm1 ;
+ paddd xmm7, xmm1 ;
+ lea rsi, [rsi + rdx]
+ lea rdi, [rdi + rbx] ;src_pixels_per_line
+
+ sub rcx, 1 ;
+ jnz filter_block2d_bil_fp_only_loop ;
+
+ jmp filter_block2d_bil_variance
+
+filter_block2d_bil_variance:
+ movdq2q mm6, xmm6 ;
+ movdq2q mm7, xmm7 ;
+
+ psrldq xmm6, 8
+ psrldq xmm7, 8
+
+ movdq2q mm2, xmm6
+ movdq2q mm3, xmm7
+
+ paddw mm6, mm2
+ paddd mm7, mm3
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rsi, arg(7) ; sum
+ mov rdi, arg(8) ; sumsquared
+
+ movd [rsi], mm2 ; xsum
+ movd [rdi], mm4 ; xxsum
+
+ ; begin epilog
+ pop rbx
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_half_horiz_vert_variance8x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_horiz_vert_variance8x_h_sse2)
+sym(vp8_half_horiz_vert_variance8x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+%if ABI_IS_32BIT=0
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+
+ movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
+ movq xmm3, QWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s9
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3) horizontal line 1
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line ; next source
+%else
+ add rsi, r8
+%endif
+
+vp8_half_horiz_vert_variance8x_h_1:
+
+ movq xmm1, QWORD PTR [rsi] ;
+ movq xmm2, QWORD PTR [rsi+1] ;
+ pavgb xmm1, xmm2 ; xmm1 = avg(xmm1,xmm3) horizontal line i+1
+
+ pavgb xmm5, xmm1 ; xmm = vertical average of the above
+ punpcklbw xmm5, xmm0 ; xmm5 = words of above
+
+ movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
+ punpcklbw xmm3, xmm0 ; xmm3 = words of above
+
+ psubw xmm5, xmm3 ; xmm5 -= xmm3
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+
+ movdqa xmm5, xmm1 ; save xmm1 for use on the next row
+
+%if ABI_IS_32BIT
+ add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
+ add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
+%else
+ add rsi, r8
+ add rdi, r9
+%endif
+
+ sub rcx, 1 ;
+ jnz vp8_half_horiz_vert_variance8x_h_1 ;
+
+ movdq2q mm6, xmm6 ;
+ movdq2q mm7, xmm7 ;
+
+ psrldq xmm6, 8
+ psrldq xmm7, 8
+
+ movdq2q mm2, xmm6
+ movdq2q mm3, xmm7
+
+ paddw mm6, mm2
+ paddd mm7, mm3
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rsi, arg(5) ; sum
+ mov rdi, arg(6) ; sumsquared
+
+ movd [rsi], mm2 ;
+ movd [rdi], mm4 ;
+
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp8_half_horiz_vert_variance16x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_horiz_vert_variance16x_h_sse2)
+sym(vp8_half_horiz_vert_variance16x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+
+ movdqu xmm5, XMMWORD PTR [rsi]
+ movdqu xmm3, XMMWORD PTR [rsi+1]
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3) horizontal line 1
+
+ lea rsi, [rsi + rax]
+
+vp8_half_horiz_vert_variance16x_h_1:
+ movdqu xmm1, XMMWORD PTR [rsi] ;
+ movdqu xmm2, XMMWORD PTR [rsi+1] ;
+ pavgb xmm1, xmm2 ; xmm1 = avg(xmm1,xmm3) horizontal line i+1
+
+ pavgb xmm5, xmm1 ; xmm = vertical average of the above
+
+ movdqa xmm4, xmm5
+ punpcklbw xmm5, xmm0 ; xmm5 = words of above
+ punpckhbw xmm4, xmm0
+
+ movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
+ punpcklbw xmm3, xmm0 ; xmm3 = words of above
+ psubw xmm5, xmm3 ; xmm5 -= xmm3
+
+ movq xmm3, QWORD PTR [rdi+8]
+ punpcklbw xmm3, xmm0
+ psubw xmm4, xmm3
+
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ paddw xmm6, xmm4
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ pmaddwd xmm4, xmm4
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+ paddd xmm7, xmm4
+
+ movdqa xmm5, xmm1 ; save xmm1 for use on the next row
+
+ lea rsi, [rsi + rax]
+ lea rdi, [rdi + rdx]
+
+ sub rcx, 1 ;
+ jnz vp8_half_horiz_vert_variance16x_h_1 ;
+
+ pxor xmm1, xmm1
+ pxor xmm5, xmm5
+
+ punpcklwd xmm0, xmm6
+ punpckhwd xmm1, xmm6
+ psrad xmm0, 16
+ psrad xmm1, 16
+ paddd xmm0, xmm1
+ movdqa xmm1, xmm0
+
+ movdqa xmm6, xmm7
+ punpckldq xmm6, xmm5
+ punpckhdq xmm7, xmm5
+ paddd xmm6, xmm7
+
+ punpckldq xmm0, xmm5
+ punpckhdq xmm1, xmm5
+ paddd xmm0, xmm1
+
+ movdqa xmm7, xmm6
+ movdqa xmm1, xmm0
+
+ psrldq xmm7, 8
+ psrldq xmm1, 8
+
+ paddd xmm6, xmm7
+ paddd xmm0, xmm1
+
+ mov rsi, arg(5) ;[Sum]
+ mov rdi, arg(6) ;[SSE]
+
+ movd [rsi], xmm0
+ movd [rdi], xmm6
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_half_vert_variance8x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_vert_variance8x_h_sse2)
+sym(vp8_half_vert_variance8x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+%if ABI_IS_32BIT=0
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+vp8_half_vert_variance8x_h_1:
+ movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
+ movq xmm3, QWORD PTR [rsi+rax] ; xmm3 = s1,s2,s3..s9
+
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
+ punpcklbw xmm5, xmm0 ; xmm5 = words of above
+
+ movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
+ punpcklbw xmm3, xmm0 ; xmm3 = words of above
+
+ psubw xmm5, xmm3 ; xmm5 -= xmm3
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+
+%if ABI_IS_32BIT
+ add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
+ add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
+%else
+ add rsi, r8
+ add rdi, r9
+%endif
+
+ sub rcx, 1 ;
+ jnz vp8_half_vert_variance8x_h_1 ;
+
+ movdq2q mm6, xmm6 ;
+ movdq2q mm7, xmm7 ;
+
+ psrldq xmm6, 8
+ psrldq xmm7, 8
+
+ movdq2q mm2, xmm6
+ movdq2q mm3, xmm7
+
+ paddw mm6, mm2
+ paddd mm7, mm3
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rsi, arg(5) ; sum
+ mov rdi, arg(6) ; sumsquared
+
+ movd [rsi], mm2 ;
+ movd [rdi], mm4 ;
+
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp8_half_vert_variance16x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_vert_variance16x_h_sse2)
+sym(vp8_half_vert_variance16x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr
+
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
+
+ movdqu xmm5, XMMWORD PTR [rsi]
+ lea rsi, [rsi + rax ]
+ pxor xmm0, xmm0
+
+vp8_half_vert_variance16x_h_1:
+ movdqu xmm3, XMMWORD PTR [rsi]
+
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
+ movdqa xmm4, xmm5
+ punpcklbw xmm5, xmm0
+ punpckhbw xmm4, xmm0
+
+ movq xmm2, QWORD PTR [rdi]
+ punpcklbw xmm2, xmm0
+ psubw xmm5, xmm2
+ movq xmm2, QWORD PTR [rdi+8]
+ punpcklbw xmm2, xmm0
+ psubw xmm4, xmm2
+
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ paddw xmm6, xmm4
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ pmaddwd xmm4, xmm4
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+ paddd xmm7, xmm4
+
+ movdqa xmm5, xmm3
+
+ lea rsi, [rsi + rax]
+ lea rdi, [rdi + rdx]
+
+ sub rcx, 1
+ jnz vp8_half_vert_variance16x_h_1
+
+ pxor xmm1, xmm1
+ pxor xmm5, xmm5
+
+ punpcklwd xmm0, xmm6
+ punpckhwd xmm1, xmm6
+ psrad xmm0, 16
+ psrad xmm1, 16
+ paddd xmm0, xmm1
+ movdqa xmm1, xmm0
+
+ movdqa xmm6, xmm7
+ punpckldq xmm6, xmm5
+ punpckhdq xmm7, xmm5
+ paddd xmm6, xmm7
+
+ punpckldq xmm0, xmm5
+ punpckhdq xmm1, xmm5
+ paddd xmm0, xmm1
+
+ movdqa xmm7, xmm6
+ movdqa xmm1, xmm0
+
+ psrldq xmm7, 8
+ psrldq xmm1, 8
+
+ paddd xmm6, xmm7
+ paddd xmm0, xmm1
+
+ mov rsi, arg(5) ;[Sum]
+ mov rdi, arg(6) ;[SSE]
+
+ movd [rsi], xmm0
+ movd [rdi], xmm6
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+;void vp8_half_horiz_variance8x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_horiz_variance8x_h_sse2)
+sym(vp8_half_horiz_variance8x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+%if ABI_IS_32BIT=0
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+
+ pxor xmm0, xmm0 ;
+vp8_half_horiz_variance8x_h_1:
+ movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
+ movq xmm3, QWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s9
+
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
+ punpcklbw xmm5, xmm0 ; xmm5 = words of above
+
+ movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
+ punpcklbw xmm3, xmm0 ; xmm3 = words of above
+
+ psubw xmm5, xmm3 ; xmm5 -= xmm3
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+
+%if ABI_IS_32BIT
+ add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
+ add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
+%else
+ add rsi, r8
+ add rdi, r9
+%endif
+ sub rcx, 1 ;
+ jnz vp8_half_horiz_variance8x_h_1 ;
+
+ movdq2q mm6, xmm6 ;
+ movdq2q mm7, xmm7 ;
+
+ psrldq xmm6, 8
+ psrldq xmm7, 8
+
+ movdq2q mm2, xmm6
+ movdq2q mm3, xmm7
+
+ paddw mm6, mm2
+ paddd mm7, mm3
+
+ pxor mm3, mm3 ;
+ pxor mm2, mm2 ;
+
+ punpcklwd mm2, mm6 ;
+ punpckhwd mm3, mm6 ;
+
+ paddd mm2, mm3 ;
+ movq mm6, mm2 ;
+
+ psrlq mm6, 32 ;
+ paddd mm2, mm6 ;
+
+ psrad mm2, 16 ;
+ movq mm4, mm7 ;
+
+ psrlq mm4, 32 ;
+ paddd mm4, mm7 ;
+
+ mov rsi, arg(5) ; sum
+ mov rdi, arg(6) ; sumsquared
+
+ movd [rsi], mm2 ;
+ movd [rdi], mm4 ;
+
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp8_half_horiz_variance16x_h_sse2
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int *sum,
+; unsigned int *sumsquared
+;)
+global sym(vp8_half_horiz_variance16x_h_sse2)
+sym(vp8_half_horiz_variance16x_h_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ pxor xmm6, xmm6 ; error accumulator
+ pxor xmm7, xmm7 ; sse eaccumulator
+ mov rsi, arg(0) ;ref_ptr ;
+
+ mov rdi, arg(2) ;src_ptr ;
+ movsxd rcx, dword ptr arg(4) ;Height ;
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
+
+ pxor xmm0, xmm0 ;
+
+vp8_half_horiz_variance16x_h_1:
+ movdqu xmm5, XMMWORD PTR [rsi] ; xmm5 = s0,s1,s2..s15
+ movdqu xmm3, XMMWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s16
+
+ pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
+ movdqa xmm1, xmm5
+ punpcklbw xmm5, xmm0 ; xmm5 = words of above
+ punpckhbw xmm1, xmm0
+
+ movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
+ punpcklbw xmm3, xmm0 ; xmm3 = words of above
+ movq xmm2, QWORD PTR [rdi+8]
+ punpcklbw xmm2, xmm0
+
+ psubw xmm5, xmm3 ; xmm5 -= xmm3
+ psubw xmm1, xmm2
+ paddw xmm6, xmm5 ; xmm6 += accumulated column differences
+ paddw xmm6, xmm1
+ pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
+ pmaddwd xmm1, xmm1
+ paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
+ paddd xmm7, xmm1
+
+ lea rsi, [rsi + rax]
+ lea rdi, [rdi + rdx]
+
+ sub rcx, 1 ;
+ jnz vp8_half_horiz_variance16x_h_1 ;
+
+ pxor xmm1, xmm1
+ pxor xmm5, xmm5
+
+ punpcklwd xmm0, xmm6
+ punpckhwd xmm1, xmm6
+ psrad xmm0, 16
+ psrad xmm1, 16
+ paddd xmm0, xmm1
+ movdqa xmm1, xmm0
+
+ movdqa xmm6, xmm7
+ punpckldq xmm6, xmm5
+ punpckhdq xmm7, xmm5
+ paddd xmm6, xmm7
+
+ punpckldq xmm0, xmm5
+ punpckhdq xmm1, xmm5
+ paddd xmm0, xmm1
+
+ movdqa xmm7, xmm6
+ movdqa xmm1, xmm0
+
+ psrldq xmm7, 8
+ psrldq xmm1, 8
+
+ paddd xmm6, xmm7
+ paddd xmm0, xmm1
+
+ mov rsi, arg(5) ;[Sum]
+ mov rdi, arg(6) ;[SSE]
+
+ movd [rsi], xmm0
+ movd [rdi], xmm6
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+SECTION_RODATA
+; short xmm_bi_rd[8] = { 64, 64, 64, 64,64, 64, 64, 64};+align 16
+xmm_bi_rd:
+ times 8 dw 64
+align 16
+vp8_bilinear_filters_sse2:
+ dw 128, 128, 128, 128, 128, 128, 128, 128, 0, 0, 0, 0, 0, 0, 0, 0
+ dw 112, 112, 112, 112, 112, 112, 112, 112, 16, 16, 16, 16, 16, 16, 16, 16
+ dw 96, 96, 96, 96, 96, 96, 96, 96, 32, 32, 32, 32, 32, 32, 32, 32
+ dw 80, 80, 80, 80, 80, 80, 80, 80, 48, 48, 48, 48, 48, 48, 48, 48
+ dw 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
+ dw 48, 48, 48, 48, 48, 48, 48, 48, 80, 80, 80, 80, 80, 80, 80, 80
+ dw 32, 32, 32, 32, 32, 32, 32, 32, 96, 96, 96, 96, 96, 96, 96, 96
+ dw 16, 16, 16, 16, 16, 16, 16, 16, 112, 112, 112, 112, 112, 112, 112, 112
--- /dev/null
+++ b/vp8/common/x86/variance_impl_ssse3.asm
@@ -1,0 +1,364 @@
+;
+; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+; Use of this source code is governed by a BSD-style license
+; that can be found in the LICENSE file in the root of the source
+; tree. An additional intellectual property rights grant can be found
+; in the file PATENTS. All contributing project authors may
+; be found in the AUTHORS file in the root of the source tree.
+;
+
+
+%include "vpx_ports/x86_abi_support.asm"
+
+%define xmm_filter_shift 7
+
+
+;void vp8_filter_block2d_bil_var_ssse3
+;(
+; unsigned char *ref_ptr,
+; int ref_pixels_per_line,
+; unsigned char *src_ptr,
+; int src_pixels_per_line,
+; unsigned int Height,
+; int xoffset,
+; int yoffset,
+; int *sum,
+; unsigned int *sumsquared;;
+;
+;)
+;Note: The filter coefficient at offset=0 is 128. Since the second register
+;for Pmaddubsw is signed bytes, we must calculate zero offset seperately.
+global sym(vp8_filter_block2d_bil_var_ssse3)
+sym(vp8_filter_block2d_bil_var_ssse3):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 9
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ pxor xmm6, xmm6
+ pxor xmm7, xmm7
+
+ lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
+ movsxd rax, dword ptr arg(5) ; xoffset
+
+ cmp rax, 0 ; skip first_pass filter if xoffset=0
+ je .filter_block2d_bil_var_ssse3_sp_only
+
+ shl rax, 4 ; point to filter coeff with xoffset
+ lea rax, [rax + rcx] ; HFilter
+
+ movsxd rdx, dword ptr arg(6) ; yoffset
+
+ cmp rdx, 0 ; skip second_pass filter if yoffset=0
+ je .filter_block2d_bil_var_ssse3_fp_only
+
+ shl rdx, 4
+ lea rdx, [rdx + rcx] ; VFilter
+
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+
+ movdqu xmm0, XMMWORD PTR [rsi]
+ movdqu xmm1, XMMWORD PTR [rsi+1]
+ movdqa xmm2, xmm0
+
+ punpcklbw xmm0, xmm1
+ punpckhbw xmm2, xmm1
+ pmaddubsw xmm0, [rax]
+ pmaddubsw xmm2, [rax]
+
+ paddw xmm0, [GLOBAL(xmm_bi_rd)]
+ paddw xmm2, [GLOBAL(xmm_bi_rd)]
+ psraw xmm0, xmm_filter_shift
+ psraw xmm2, xmm_filter_shift
+
+ packuswb xmm0, xmm2
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line
+%else
+ movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+ lea rsi, [rsi + r8]
+%endif
+
+.filter_block2d_bil_var_ssse3_loop:
+ movdqu xmm1, XMMWORD PTR [rsi]
+ movdqu xmm2, XMMWORD PTR [rsi+1]
+ movdqa xmm3, xmm1
+
+ punpcklbw xmm1, xmm2
+ punpckhbw xmm3, xmm2
+ pmaddubsw xmm1, [rax]
+ pmaddubsw xmm3, [rax]
+
+ paddw xmm1, [GLOBAL(xmm_bi_rd)]
+ paddw xmm3, [GLOBAL(xmm_bi_rd)]
+ psraw xmm1, xmm_filter_shift
+ psraw xmm3, xmm_filter_shift
+ packuswb xmm1, xmm3
+
+ movdqa xmm2, xmm0
+ movdqa xmm0, xmm1
+ movdqa xmm3, xmm2
+
+ punpcklbw xmm2, xmm1
+ punpckhbw xmm3, xmm1
+ pmaddubsw xmm2, [rdx]
+ pmaddubsw xmm3, [rdx]
+
+ paddw xmm2, [GLOBAL(xmm_bi_rd)]
+ paddw xmm3, [GLOBAL(xmm_bi_rd)]
+ psraw xmm2, xmm_filter_shift
+ psraw xmm3, xmm_filter_shift
+
+ movq xmm1, QWORD PTR [rdi]
+ pxor xmm4, xmm4
+ punpcklbw xmm1, xmm4
+ movq xmm5, QWORD PTR [rdi+8]
+ punpcklbw xmm5, xmm4
+
+ psubw xmm2, xmm1
+ psubw xmm3, xmm5
+ paddw xmm6, xmm2
+ paddw xmm6, xmm3
+ pmaddwd xmm2, xmm2
+ pmaddwd xmm3, xmm3
+ paddd xmm7, xmm2
+ paddd xmm7, xmm3
+
+%if ABI_IS_32BIT
+ add rsi, dword ptr arg(1) ;ref_pixels_per_line
+ add rdi, dword ptr arg(3) ;src_pixels_per_line
+%else
+ lea rsi, [rsi + r8]
+ lea rdi, [rdi + r9]
+%endif
+
+ sub rcx, 1
+ jnz .filter_block2d_bil_var_ssse3_loop
+
+ jmp .filter_block2d_bil_variance
+
+.filter_block2d_bil_var_ssse3_sp_only:
+ movsxd rdx, dword ptr arg(6) ; yoffset
+
+ cmp rdx, 0 ; Both xoffset =0 and yoffset=0
+ je .filter_block2d_bil_var_ssse3_full_pixel
+
+ shl rdx, 4
+ lea rdx, [rdx + rcx] ; VFilter
+
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+
+ movdqu xmm1, XMMWORD PTR [rsi]
+ movdqa xmm0, xmm1
+
+%if ABI_IS_32BIT=0
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+ lea rsi, [rsi + rax]
+
+.filter_block2d_bil_sp_only_loop:
+ movdqu xmm3, XMMWORD PTR [rsi]
+ movdqa xmm2, xmm1
+ movdqa xmm0, xmm3
+
+ punpcklbw xmm1, xmm3
+ punpckhbw xmm2, xmm3
+ pmaddubsw xmm1, [rdx]
+ pmaddubsw xmm2, [rdx]
+
+ paddw xmm1, [GLOBAL(xmm_bi_rd)]
+ paddw xmm2, [GLOBAL(xmm_bi_rd)]
+ psraw xmm1, xmm_filter_shift
+ psraw xmm2, xmm_filter_shift
+
+ movq xmm3, QWORD PTR [rdi]
+ pxor xmm4, xmm4
+ punpcklbw xmm3, xmm4
+ movq xmm5, QWORD PTR [rdi+8]
+ punpcklbw xmm5, xmm4
+
+ psubw xmm1, xmm3
+ psubw xmm2, xmm5
+ paddw xmm6, xmm1
+ paddw xmm6, xmm2
+ pmaddwd xmm1, xmm1
+ pmaddwd xmm2, xmm2
+ paddd xmm7, xmm1
+ paddd xmm7, xmm2
+
+ movdqa xmm1, xmm0
+ lea rsi, [rsi + rax] ;ref_pixels_per_line
+
+%if ABI_IS_32BIT
+ add rdi, dword ptr arg(3) ;src_pixels_per_line
+%else
+ lea rdi, [rdi + r9]
+%endif
+
+ sub rcx, 1
+ jnz .filter_block2d_bil_sp_only_loop
+
+ jmp .filter_block2d_bil_variance
+
+.filter_block2d_bil_var_ssse3_full_pixel:
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
+ movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
+ pxor xmm0, xmm0
+
+.filter_block2d_bil_full_pixel_loop:
+ movq xmm1, QWORD PTR [rsi]
+ punpcklbw xmm1, xmm0
+ movq xmm2, QWORD PTR [rsi+8]
+ punpcklbw xmm2, xmm0
+
+ movq xmm3, QWORD PTR [rdi]
+ punpcklbw xmm3, xmm0
+ movq xmm4, QWORD PTR [rdi+8]
+ punpcklbw xmm4, xmm0
+
+ psubw xmm1, xmm3
+ psubw xmm2, xmm4
+ paddw xmm6, xmm1
+ paddw xmm6, xmm2
+ pmaddwd xmm1, xmm1
+ pmaddwd xmm2, xmm2
+ paddd xmm7, xmm1
+ paddd xmm7, xmm2
+
+ lea rsi, [rsi + rax] ;ref_pixels_per_line
+ lea rdi, [rdi + rdx] ;src_pixels_per_line
+ sub rcx, 1
+ jnz .filter_block2d_bil_full_pixel_loop
+
+ jmp .filter_block2d_bil_variance
+
+.filter_block2d_bil_var_ssse3_fp_only:
+ mov rsi, arg(0) ;ref_ptr
+ mov rdi, arg(2) ;src_ptr
+ movsxd rcx, dword ptr arg(4) ;Height
+ movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
+
+ pxor xmm0, xmm0
+
+%if ABI_IS_32BIT=0
+ movsxd r9, dword ptr arg(3) ;src_pixels_per_line
+%endif
+
+.filter_block2d_bil_fp_only_loop:
+ movdqu xmm1, XMMWORD PTR [rsi]
+ movdqu xmm2, XMMWORD PTR [rsi+1]
+ movdqa xmm3, xmm1
+
+ punpcklbw xmm1, xmm2
+ punpckhbw xmm3, xmm2
+ pmaddubsw xmm1, [rax]
+ pmaddubsw xmm3, [rax]
+
+ paddw xmm1, [GLOBAL(xmm_bi_rd)]
+ paddw xmm3, [GLOBAL(xmm_bi_rd)]
+ psraw xmm1, xmm_filter_shift
+ psraw xmm3, xmm_filter_shift
+
+ movq xmm2, XMMWORD PTR [rdi]
+ pxor xmm4, xmm4
+ punpcklbw xmm2, xmm4
+ movq xmm5, QWORD PTR [rdi+8]
+ punpcklbw xmm5, xmm4
+
+ psubw xmm1, xmm2
+ psubw xmm3, xmm5
+ paddw xmm6, xmm1
+ paddw xmm6, xmm3
+ pmaddwd xmm1, xmm1
+ pmaddwd xmm3, xmm3
+ paddd xmm7, xmm1
+ paddd xmm7, xmm3
+
+ lea rsi, [rsi + rdx]
+%if ABI_IS_32BIT
+ add rdi, dword ptr arg(3) ;src_pixels_per_line
+%else
+ lea rdi, [rdi + r9]
+%endif
+
+ sub rcx, 1
+ jnz .filter_block2d_bil_fp_only_loop
+
+ jmp .filter_block2d_bil_variance
+
+.filter_block2d_bil_variance:
+ pxor xmm0, xmm0
+ pxor xmm1, xmm1
+ pxor xmm5, xmm5
+
+ punpcklwd xmm0, xmm6
+ punpckhwd xmm1, xmm6
+ psrad xmm0, 16
+ psrad xmm1, 16
+ paddd xmm0, xmm1
+ movdqa xmm1, xmm0
+
+ movdqa xmm6, xmm7
+ punpckldq xmm6, xmm5
+ punpckhdq xmm7, xmm5
+ paddd xmm6, xmm7
+
+ punpckldq xmm0, xmm5
+ punpckhdq xmm1, xmm5
+ paddd xmm0, xmm1
+
+ movdqa xmm7, xmm6
+ movdqa xmm1, xmm0
+
+ psrldq xmm7, 8
+ psrldq xmm1, 8
+
+ paddd xmm6, xmm7
+ paddd xmm0, xmm1
+
+ mov rsi, arg(7) ;[Sum]
+ mov rdi, arg(8) ;[SSE]
+
+ movd [rsi], xmm0
+ movd [rdi], xmm6
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+
+SECTION_RODATA
+align 16
+xmm_bi_rd:
+ times 8 dw 64
+align 16
+vp8_bilinear_filters_ssse3:
+ times 8 db 128, 0
+ times 8 db 112, 16
+ times 8 db 96, 32
+ times 8 db 80, 48
+ times 8 db 64, 64
+ times 8 db 48, 80
+ times 8 db 32, 96
+ times 8 db 16, 112
--- /dev/null
+++ b/vp8/common/x86/variance_mmx.c
@@ -1,0 +1,398 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_config.h"
+#include "vp8/common/variance.h"
+#include "vp8/common/pragmas.h"
+#include "vpx_ports/mem.h"
+#include "vp8/common/x86/filter_x86.h"
+
+extern void filter_block1d_h6_mmx
+(
+ const unsigned char *src_ptr,
+ unsigned short *output_ptr,
+ unsigned int src_pixels_per_line,
+ unsigned int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ short *filter
+);
+extern void filter_block1d_v6_mmx
+(
+ const short *src_ptr,
+ unsigned char *output_ptr,
+ unsigned int pixels_per_line,
+ unsigned int pixel_step,
+ unsigned int output_height,
+ unsigned int output_width,
+ short *filter
+);
+
+extern unsigned int vp8_get_mb_ss_mmx(const short *src_ptr);
+extern unsigned int vp8_get8x8var_mmx
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+extern unsigned int vp8_get4x4var_mmx
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+extern void vp8_filter_block2d_bil4x4_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+extern void vp8_filter_block2d_bil_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+
+unsigned int vp8_variance4x4_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+ vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 4));
+
+}
+
+unsigned int vp8_variance8x8_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+ vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
+ *sse = var;
+
+ return (var - ((unsigned int)(avg * avg) >> 6));
+
+}
+
+unsigned int vp8_mse16x16_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, sse2, sse3, var;
+ int sum0, sum1, sum2, sum3;
+
+
+ vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
+ vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
+ vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
+
+ var = sse0 + sse1 + sse2 + sse3;
+ *sse = var;
+ return var;
+}
+
+
+unsigned int vp8_variance16x16_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, sse2, sse3, var;
+ int sum0, sum1, sum2, sum3, avg;
+
+
+ vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
+ vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
+ vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
+
+ var = sse0 + sse1 + sse2 + sse3;
+ avg = sum0 + sum1 + sum2 + sum3;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 8));
+}
+
+unsigned int vp8_variance16x8_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, var;
+ int sum0, sum1, avg;
+
+ vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
+
+ var = sse0 + sse1;
+ avg = sum0 + sum1;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 7));
+
+}
+
+
+unsigned int vp8_variance8x16_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, var;
+ int sum0, sum1, avg;
+
+ vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
+
+ var = sse0 + sse1;
+ avg = sum0 + sum1;
+ *sse = var;
+
+ return (var - ((unsigned int)(avg * avg) >> 7));
+
+}
+
+
+unsigned int vp8_sub_pixel_variance4x4_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil4x4_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 4));
+}
+
+
+unsigned int vp8_sub_pixel_variance8x8_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+
+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 6));
+}
+
+unsigned int vp8_sub_pixel_variance16x16_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+
+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum0, &xxsum0
+ );
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum1, &xxsum1
+ );
+
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+
+
+}
+
+unsigned int vp8_sub_pixel_mse16x16_mmx(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ vp8_sub_pixel_variance16x16_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
+ return *sse;
+}
+
+unsigned int vp8_sub_pixel_variance16x8_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum0, &xxsum0
+ );
+
+
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 8,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum1, &xxsum1
+ );
+
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
+}
+
+unsigned int vp8_sub_pixel_variance8x16_mmx
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 7));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_h_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 0,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_v_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 0, 4,
+ ref_ptr, recon_stride, sse);
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_hv_mmx(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 4,
+ ref_ptr, recon_stride, sse);
+}
--- /dev/null
+++ b/vp8/common/x86/variance_sse2.c
@@ -1,0 +1,557 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_config.h"
+#include "vp8/common/variance.h"
+#include "vp8/common/pragmas.h"
+#include "vpx_ports/mem.h"
+#include "vp8/common/x86/filter_x86.h"
+
+extern void filter_block1d_h6_mmx(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d_v6_mmx(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d8_h6_sse2(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+extern void filter_block1d8_v6_sse2(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
+
+extern void vp8_filter_block2d_bil4x4_var_mmx
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const short *HFilter,
+ const short *VFilter,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+extern unsigned int vp8_get4x4var_mmx
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+
+unsigned int vp8_get_mb_ss_sse2
+(
+ const short *src_ptr
+);
+unsigned int vp8_get16x16var_sse2
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+unsigned int vp8_get8x8var_sse2
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+void vp8_filter_block2d_bil_var_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int xoffset,
+ int yoffset,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_vert_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_horiz_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_vert_variance8x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+void vp8_half_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+unsigned int vp8_variance4x4_wmt(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+ vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 4));
+
+}
+
+unsigned int vp8_variance8x8_wmt
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int var;
+ int avg;
+
+ vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 6));
+
+}
+
+
+unsigned int vp8_variance16x16_wmt
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0;
+ int sum0;
+
+
+ vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ *sse = sse0;
+ return (sse0 - ((unsigned int)(sum0 * sum0) >> 8));
+}
+unsigned int vp8_mse16x16_wmt(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+
+ unsigned int sse0;
+ int sum0;
+ vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ *sse = sse0;
+ return sse0;
+
+}
+
+
+unsigned int vp8_variance16x8_wmt
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, var;
+ int sum0, sum1, avg;
+
+ vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_sse2(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
+
+ var = sse0 + sse1;
+ avg = sum0 + sum1;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 7));
+
+}
+
+unsigned int vp8_variance8x16_wmt
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *sse)
+{+ unsigned int sse0, sse1, var;
+ int sum0, sum1, avg;
+
+ vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
+ vp8_get8x8var_sse2(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
+
+ var = sse0 + sse1;
+ avg = sum0 + sum1;
+ *sse = var;
+ return (var - ((unsigned int)(avg * avg) >> 7));
+
+}
+
+unsigned int vp8_sub_pixel_variance4x4_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+ vp8_filter_block2d_bil4x4_var_mmx(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line,
+ vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
+ &xsum, &xxsum
+ );
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 4));
+}
+
+
+unsigned int vp8_sub_pixel_variance8x8_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum, &xxsum);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum, &xxsum);
+ }
+
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 6));
+}
+
+unsigned int vp8_sub_pixel_variance16x16_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+
+ // note we could avoid these if statements if the calling function
+ // just called the appropriate functions inside.
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum0, &xxsum0
+ );
+
+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum1, &xxsum1
+ );
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+}
+
+unsigned int vp8_sub_pixel_mse16x16_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ vp8_sub_pixel_variance16x16_wmt(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
+ return *sse;
+}
+
+unsigned int vp8_sub_pixel_variance16x8_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+
+)
+{+ int xsum0, xsum1;
+ unsigned int xxsum0, xxsum1;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum0, &xxsum0);
+
+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr + 8, src_pixels_per_line,
+ dst_ptr + 8, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum1, &xxsum1);
+ xsum0 += xsum1;
+ xxsum0 += xxsum1;
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
+}
+
+unsigned int vp8_sub_pixel_variance8x16_wmt
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum;
+ unsigned int xxsum;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance8x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum, &xxsum);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum, &xxsum);
+ }
+
+ *sse = xxsum;
+ return (xxsum - ((unsigned int)(xsum * xsum) >> 7));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_h_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_v_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+}
+
+
+unsigned int vp8_variance_halfpixvar16x16_hv_wmt(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+}
--- /dev/null
+++ b/vp8/common/x86/variance_ssse3.c
@@ -1,0 +1,165 @@
+/*
+ * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ * Use of this source code is governed by a BSD-style license
+ * that can be found in the LICENSE file in the root of the source
+ * tree. An additional intellectual property rights grant can be found
+ * in the file PATENTS. All contributing project authors may
+ * be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include "vpx_config.h"
+#include "vp8/common/variance.h"
+#include "vp8/common/pragmas.h"
+#include "vpx_ports/mem.h"
+
+extern unsigned int vp8_get16x16var_sse2
+(
+ const unsigned char *src_ptr,
+ int source_stride,
+ const unsigned char *ref_ptr,
+ int recon_stride,
+ unsigned int *SSE,
+ int *Sum
+);
+extern void vp8_half_horiz_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+extern void vp8_half_horiz_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+extern void vp8_half_vert_variance16x_h_sse2
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int *sum,
+ unsigned int *sumsquared
+);
+extern void vp8_filter_block2d_bil_var_ssse3
+(
+ const unsigned char *ref_ptr,
+ int ref_pixels_per_line,
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ unsigned int Height,
+ int xoffset,
+ int yoffset,
+ int *sum,
+ unsigned int *sumsquared
+);
+
+unsigned int vp8_sub_pixel_variance16x16_ssse3
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ // note we could avoid these if statements if the calling function
+ // just called the appropriate functions inside.
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_ssse3(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 16,
+ xoffset, yoffset,
+ &xsum0, &xxsum0);
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
+}
+
+unsigned int vp8_sub_pixel_variance16x8_ssse3
+(
+ const unsigned char *src_ptr,
+ int src_pixels_per_line,
+ int xoffset,
+ int yoffset,
+ const unsigned char *dst_ptr,
+ int dst_pixels_per_line,
+ unsigned int *sse
+
+)
+{+ int xsum0;
+ unsigned int xxsum0;
+
+ if (xoffset == 4 && yoffset == 0)
+ {+ vp8_half_horiz_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 0 && yoffset == 4)
+ {+ vp8_half_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else if (xoffset == 4 && yoffset == 4)
+ {+ vp8_half_horiz_vert_variance16x_h_sse2(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ &xsum0, &xxsum0);
+ }
+ else
+ {+ vp8_filter_block2d_bil_var_ssse3(
+ src_ptr, src_pixels_per_line,
+ dst_ptr, dst_pixels_per_line, 8,
+ xoffset, yoffset,
+ &xsum0, &xxsum0);
+ }
+
+ *sse = xxsum0;
+ return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
+}
--- a/vp8/encoder/arm/armv6/vp8_sad16x16_armv6.asm
+++ /dev/null
@@ -1,96 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_sad16x16_armv6|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 const unsigned char *src_ptr
-; r1 int src_stride
-; r2 const unsigned char *ref_ptr
-; r3 int ref_stride
-; stack max_sad (not used)
-|vp8_sad16x16_armv6| PROC
- stmfd sp!, {r4-r12, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
- pld [r0, r1, lsl #1]
- pld [r2, r3, lsl #1]
-
- mov r4, #0 ; sad = 0;
- mov r5, #8 ; loop count
-
-loop
- ; 1st row
- ldr r6, [r0, #0x0] ; load 4 src pixels (1A)
- ldr r8, [r2, #0x0] ; load 4 ref pixels (1A)
- ldr r7, [r0, #0x4] ; load 4 src pixels (1A)
- ldr r9, [r2, #0x4] ; load 4 ref pixels (1A)
- ldr r10, [r0, #0x8] ; load 4 src pixels (1B)
- ldr r11, [r0, #0xC] ; load 4 src pixels (1B)
-
- usada8 r4, r8, r6, r4 ; calculate sad for 4 pixels
- usad8 r8, r7, r9 ; calculate sad for 4 pixels
-
- ldr r12, [r2, #0x8] ; load 4 ref pixels (1B)
- ldr lr, [r2, #0xC] ; load 4 ref pixels (1B)
-
- add r0, r0, r1 ; set src pointer to next row
- add r2, r2, r3 ; set dst pointer to next row
-
- pld [r0, r1, lsl #1]
- pld [r2, r3, lsl #1]
-
- usada8 r4, r10, r12, r4 ; calculate sad for 4 pixels
- usada8 r8, r11, lr, r8 ; calculate sad for 4 pixels
-
- ldr r6, [r0, #0x0] ; load 4 src pixels (2A)
- ldr r7, [r0, #0x4] ; load 4 src pixels (2A)
- add r4, r4, r8 ; add partial sad values
-
- ; 2nd row
- ldr r8, [r2, #0x0] ; load 4 ref pixels (2A)
- ldr r9, [r2, #0x4] ; load 4 ref pixels (2A)
- ldr r10, [r0, #0x8] ; load 4 src pixels (2B)
- ldr r11, [r0, #0xC] ; load 4 src pixels (2B)
-
- usada8 r4, r6, r8, r4 ; calculate sad for 4 pixels
- usad8 r8, r7, r9 ; calculate sad for 4 pixels
-
- ldr r12, [r2, #0x8] ; load 4 ref pixels (2B)
- ldr lr, [r2, #0xC] ; load 4 ref pixels (2B)
-
- add r0, r0, r1 ; set src pointer to next row
- add r2, r2, r3 ; set dst pointer to next row
-
- usada8 r4, r10, r12, r4 ; calculate sad for 4 pixels
- usada8 r8, r11, lr, r8 ; calculate sad for 4 pixels
-
- pld [r0, r1, lsl #1]
- pld [r2, r3, lsl #1]
-
- subs r5, r5, #1 ; decrement loop counter
- add r4, r4, r8 ; add partial sad values
-
- bne loop
-
- mov r0, r4 ; return sad
- ldmfd sp!, {r4-r12, pc}-
- ENDP
-
- END
-
--- a/vp8/encoder/arm/armv6/vp8_variance16x16_armv6.asm
+++ /dev/null
@@ -1,154 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance16x16_armv6|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance16x16_armv6| PROC
-
- stmfd sp!, {r4-r12, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
-
- mov r8, #0 ; initialize sum = 0
- mov r11, #0 ; initialize sse = 0
- mov r12, #16 ; set loop counter to 16 (=block height)
-
-loop
- ; 1st 4 pixels
- ldr r4, [r0, #0] ; load 4 src pixels
- ldr r5, [r2, #0] ; load 4 ref pixels
-
- mov lr, #0 ; constant zero
-
- usub8 r6, r4, r5 ; calculate difference
- pld [r0, r1, lsl #1]
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r9, r5, r4 ; calculate difference with reversed operands
- pld [r2, r3, lsl #1]
- sel r6, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
- ; calculate total sum
- adds r8, r8, r4 ; add positive differences to sum
- subs r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 2nd 4 pixels
- ldr r4, [r0, #4] ; load 4 src pixels
- ldr r5, [r2, #4] ; load 4 ref pixels
- smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r9, r5, r4 ; calculate difference with reversed operands
- sel r6, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 3rd 4 pixels
- ldr r4, [r0, #8] ; load 4 src pixels
- ldr r5, [r2, #8] ; load 4 ref pixels
- smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r9, r5, r4 ; calculate difference with reversed operands
- sel r6, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 4th 4 pixels
- ldr r4, [r0, #12] ; load 4 src pixels
- ldr r5, [r2, #12] ; load 4 ref pixels
- smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- add r0, r0, r1 ; set src_ptr to next row
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r9, r5, r4 ; calculate difference with reversed operands
- add r2, r2, r3 ; set dst_ptr to next row
- sel r6, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r10, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
- smlad r11, r10, r10, r11 ; dual signed multiply, add and accumulate (2)
-
-
- subs r12, r12, #1
-
- bne loop
-
- ; return stuff
- ldr r6, [sp, #40] ; get address of sse
- mul r0, r8, r8 ; sum * sum
- str r11, [r6] ; store sse
- sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
-
- ldmfd sp!, {r4-r12, pc}-
- ENDP
-
- END
-
--- a/vp8/encoder/arm/armv6/vp8_variance8x8_armv6.asm
+++ /dev/null
@@ -1,101 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance8x8_armv6|
-
- ARM
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance8x8_armv6| PROC
-
- push {r4-r10, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
-
- mov r12, #8 ; set loop counter to 8 (=block height)
- mov r4, #0 ; initialize sum = 0
- mov r5, #0 ; initialize sse = 0
-
-loop
- ; 1st 4 pixels
- ldr r6, [r0, #0x0] ; load 4 src pixels
- ldr r7, [r2, #0x0] ; load 4 ref pixels
-
- mov lr, #0 ; constant zero
-
- usub8 r8, r6, r7 ; calculate difference
- pld [r0, r1, lsl #1]
- sel r10, r8, lr ; select bytes with positive difference
- usub8 r9, r7, r6 ; calculate difference with reversed operands
- pld [r2, r3, lsl #1]
- sel r8, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r6, r10, lr ; calculate sum of positive differences
- usad8 r7, r8, lr ; calculate sum of negative differences
- orr r8, r8, r10 ; differences of all 4 pixels
- ; calculate total sum
- add r4, r4, r6 ; add positive differences to sum
- sub r4, r4, r7 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r7, r8 ; byte (two pixels) to halfwords
- uxtb16 r10, r8, ror #8 ; another two pixels to halfwords
- smlad r5, r7, r7, r5 ; dual signed multiply, add and accumulate (1)
-
- ; 2nd 4 pixels
- ldr r6, [r0, #0x4] ; load 4 src pixels
- ldr r7, [r2, #0x4] ; load 4 ref pixels
- smlad r5, r10, r10, r5 ; dual signed multiply, add and accumulate (2)
-
- usub8 r8, r6, r7 ; calculate difference
- add r0, r0, r1 ; set src_ptr to next row
- sel r10, r8, lr ; select bytes with positive difference
- usub8 r9, r7, r6 ; calculate difference with reversed operands
- add r2, r2, r3 ; set dst_ptr to next row
- sel r8, r9, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r6, r10, lr ; calculate sum of positive differences
- usad8 r7, r8, lr ; calculate sum of negative differences
- orr r8, r8, r10 ; differences of all 4 pixels
-
- ; calculate total sum
- add r4, r4, r6 ; add positive differences to sum
- sub r4, r4, r7 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r7, r8 ; byte (two pixels) to halfwords
- uxtb16 r10, r8, ror #8 ; another two pixels to halfwords
- smlad r5, r7, r7, r5 ; dual signed multiply, add and accumulate (1)
- subs r12, r12, #1 ; next row
- smlad r5, r10, r10, r5 ; dual signed multiply, add and accumulate (2)
-
- bne loop
-
- ; return stuff
- ldr r8, [sp, #32] ; get address of sse
- mul r1, r4, r4 ; sum * sum
- str r5, [r8] ; store sse
- sub r0, r5, r1, ASR #6 ; return (sse - ((sum * sum) >> 6))
-
- pop {r4-r10, pc}-
- ENDP
-
- END
--- a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6.asm
+++ /dev/null
@@ -1,182 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance_halfpixvar16x16_h_armv6|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance_halfpixvar16x16_h_armv6| PROC
-
- stmfd sp!, {r4-r12, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
-
- mov r8, #0 ; initialize sum = 0
- ldr r10, c80808080
- mov r11, #0 ; initialize sse = 0
- mov r12, #16 ; set loop counter to 16 (=block height)
- mov lr, #0 ; constant zero
-loop
- ; 1st 4 pixels
- ldr r4, [r0, #0] ; load 4 src pixels
- ldr r6, [r0, #1] ; load 4 src pixels with 1 byte offset
- ldr r5, [r2, #0] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- pld [r0, r1, lsl #1]
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- pld [r2, r3, lsl #1]
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
- ; calculate total sum
- adds r8, r8, r4 ; add positive differences to sum
- subs r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 2nd 4 pixels
- ldr r4, [r0, #4] ; load 4 src pixels
- ldr r6, [r0, #5] ; load 4 src pixels with 1 byte offset
- ldr r5, [r2, #4] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 3rd 4 pixels
- ldr r4, [r0, #8] ; load 4 src pixels
- ldr r6, [r0, #9] ; load 4 src pixels with 1 byte offset
- ldr r5, [r2, #8] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 4th 4 pixels
- ldr r4, [r0, #12] ; load 4 src pixels
- ldr r6, [r0, #13] ; load 4 src pixels with 1 byte offset
- ldr r5, [r2, #12] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- add r0, r0, r1 ; set src_ptr to next row
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- add r2, r2, r3 ; set dst_ptr to next row
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- subs r12, r12, #1
-
- bne loop
-
- ; return stuff
- ldr r6, [sp, #40] ; get address of sse
- mul r0, r8, r8 ; sum * sum
- str r11, [r6] ; store sse
- sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
-
- ldmfd sp!, {r4-r12, pc}-
- ENDP
-
-c80808080
- DCD 0x80808080
-
- END
-
--- a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6.asm
+++ /dev/null
@@ -1,222 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance_halfpixvar16x16_hv_armv6|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance_halfpixvar16x16_hv_armv6| PROC
-
- stmfd sp!, {r4-r12, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
-
- mov r8, #0 ; initialize sum = 0
- ldr r10, c80808080
- mov r11, #0 ; initialize sse = 0
- mov r12, #16 ; set loop counter to 16 (=block height)
- mov lr, #0 ; constant zero
-loop
- add r9, r0, r1 ; pointer to pixels on the next row
- ; 1st 4 pixels
- ldr r4, [r0, #0] ; load source pixels a, row N
- ldr r6, [r0, #1] ; load source pixels b, row N
- ldr r5, [r9, #0] ; load source pixels c, row N+1
- ldr r7, [r9, #1] ; load source pixels d, row N+1
-
- ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
- ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
- mvn r7, r7
- uhsub8 r5, r5, r7
- eor r5, r5, r10
- ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
- mvn r5, r5
- uhsub8 r4, r4, r5
- ldr r5, [r2, #0] ; load 4 ref pixels
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- pld [r0, r1, lsl #1]
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- pld [r2, r3, lsl #1]
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
- ; calculate total sum
- adds r8, r8, r4 ; add positive differences to sum
- subs r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 2nd 4 pixels
- ldr r4, [r0, #4] ; load source pixels a, row N
- ldr r6, [r0, #5] ; load source pixels b, row N
- ldr r5, [r9, #4] ; load source pixels c, row N+1
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- ldr r7, [r9, #5] ; load source pixels d, row N+1
-
- ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
- ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
- mvn r7, r7
- uhsub8 r5, r5, r7
- eor r5, r5, r10
- ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
- mvn r5, r5
- uhsub8 r4, r4, r5
- ldr r5, [r2, #4] ; load 4 ref pixels
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 3rd 4 pixels
- ldr r4, [r0, #8] ; load source pixels a, row N
- ldr r6, [r0, #9] ; load source pixels b, row N
- ldr r5, [r9, #8] ; load source pixels c, row N+1
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- ldr r7, [r9, #9] ; load source pixels d, row N+1
-
- ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
- ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
- mvn r7, r7
- uhsub8 r5, r5, r7
- eor r5, r5, r10
- ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
- mvn r5, r5
- uhsub8 r4, r4, r5
- ldr r5, [r2, #8] ; load 4 ref pixels
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 4th 4 pixels
- ldr r4, [r0, #12] ; load source pixels a, row N
- ldr r6, [r0, #13] ; load source pixels b, row N
- ldr r5, [r9, #12] ; load source pixels c, row N+1
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
- ldr r7, [r9, #13] ; load source pixels d, row N+1
-
- ; x = (a + b + 1) >> 1, interpolate pixels horizontally on row N
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
- ; y = (c + d + 1) >> 1, interpolate pixels horizontally on row N+1
- mvn r7, r7
- uhsub8 r5, r5, r7
- eor r5, r5, r10
- ; z = (x + y + 1) >> 1, interpolate half pixel values vertically
- mvn r5, r5
- uhsub8 r4, r4, r5
- ldr r5, [r2, #12] ; load 4 ref pixels
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- add r0, r0, r1 ; set src_ptr to next row
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- add r2, r2, r3 ; set dst_ptr to next row
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
- subs r12, r12, #1
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- bne loop
-
- ; return stuff
- ldr r6, [sp, #40] ; get address of sse
- mul r0, r8, r8 ; sum * sum
- str r11, [r6] ; store sse
- sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
-
- ldmfd sp!, {r4-r12, pc}-
- ENDP
-
-c80808080
- DCD 0x80808080
-
- END
--- a/vp8/encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6.asm
+++ /dev/null
@@ -1,184 +1,0 @@
-;
-; Copyright (c) 2011 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance_halfpixvar16x16_v_armv6|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance_halfpixvar16x16_v_armv6| PROC
-
- stmfd sp!, {r4-r12, lr}-
- pld [r0, r1, lsl #0]
- pld [r2, r3, lsl #0]
-
- mov r8, #0 ; initialize sum = 0
- ldr r10, c80808080
- mov r11, #0 ; initialize sse = 0
- mov r12, #16 ; set loop counter to 16 (=block height)
- mov lr, #0 ; constant zero
-loop
- add r9, r0, r1 ; set src pointer to next row
- ; 1st 4 pixels
- ldr r4, [r0, #0] ; load 4 src pixels
- ldr r6, [r9, #0] ; load 4 src pixels from next row
- ldr r5, [r2, #0] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- usub8 r6, r4, r5 ; calculate difference
- pld [r0, r1, lsl #1]
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- pld [r2, r3, lsl #1]
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
- ; calculate total sum
- adds r8, r8, r4 ; add positive differences to sum
- subs r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 2nd 4 pixels
- ldr r4, [r0, #4] ; load 4 src pixels
- ldr r6, [r9, #4] ; load 4 src pixels from next row
- ldr r5, [r2, #4] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 3rd 4 pixels
- ldr r4, [r0, #8] ; load 4 src pixels
- ldr r6, [r9, #8] ; load 4 src pixels from next row
- ldr r5, [r2, #8] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
-
- ; 4th 4 pixels
- ldr r4, [r0, #12] ; load 4 src pixels
- ldr r6, [r9, #12] ; load 4 src pixels from next row
- ldr r5, [r2, #12] ; load 4 ref pixels
-
- ; bilinear interpolation
- mvn r6, r6
- uhsub8 r4, r4, r6
- eor r4, r4, r10
-
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
- usub8 r6, r4, r5 ; calculate difference
- add r0, r0, r1 ; set src_ptr to next row
- sel r7, r6, lr ; select bytes with positive difference
- usub8 r6, r5, r4 ; calculate difference with reversed operands
- add r2, r2, r3 ; set dst_ptr to next row
- sel r6, r6, lr ; select bytes with negative difference
-
- ; calculate partial sums
- usad8 r4, r7, lr ; calculate sum of positive differences
- usad8 r5, r6, lr ; calculate sum of negative differences
- orr r6, r6, r7 ; differences of all 4 pixels
-
- ; calculate total sum
- add r8, r8, r4 ; add positive differences to sum
- sub r8, r8, r5 ; substract negative differences from sum
-
- ; calculate sse
- uxtb16 r5, r6 ; byte (two pixels) to halfwords
- uxtb16 r7, r6, ror #8 ; another two pixels to halfwords
- smlad r11, r5, r5, r11 ; dual signed multiply, add and accumulate (1)
- smlad r11, r7, r7, r11 ; dual signed multiply, add and accumulate (2)
-
-
- subs r12, r12, #1
-
- bne loop
-
- ; return stuff
- ldr r6, [sp, #40] ; get address of sse
- mul r0, r8, r8 ; sum * sum
- str r11, [r6] ; store sse
- sub r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
-
- ldmfd sp!, {r4-r12, pc}-
- ENDP
-
-c80808080
- DCD 0x80808080
-
- END
-
--- a/vp8/encoder/arm/neon/sad16_neon.asm
+++ /dev/null
@@ -1,207 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_sad16x16_neon|
- EXPORT |vp8_sad16x8_neon|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int src_stride
-; r2 unsigned char *ref_ptr
-; r3 int ref_stride
-|vp8_sad16x16_neon| PROC
-;;
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabdl.u8 q12, d0, d8
- vabdl.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0], r1- vld1.8 {q7}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
-;;
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabal.u8 q12, d0, d8
- vabal.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0], r1- vld1.8 {q7}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
-;;
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabal.u8 q12, d0, d8
- vabal.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0], r1- vld1.8 {q7}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
-;;
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabal.u8 q12, d0, d8
- vabal.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0]- vld1.8 {q7}, [r2]-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vadd.u16 q0, q12, q13
-
- vpaddl.u16 q1, q0
- vpaddl.u32 q0, q1
-
- vadd.u32 d0, d0, d1
-
- vmov.32 r0, d0[0]
-
- bx lr
-
- ENDP
-
-;==============================
-;unsigned int vp8_sad16x8_c(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-|vp8_sad16x8_neon| PROC
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabdl.u8 q12, d0, d8
- vabdl.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0], r1- vld1.8 {q7}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
- vld1.8 {q0}, [r0], r1- vld1.8 {q4}, [r2], r3-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vld1.8 {q1}, [r0], r1- vld1.8 {q5}, [r2], r3-
- vabal.u8 q12, d0, d8
- vabal.u8 q13, d1, d9
-
- vld1.8 {q2}, [r0], r1- vld1.8 {q6}, [r2], r3-
- vabal.u8 q12, d2, d10
- vabal.u8 q13, d3, d11
-
- vld1.8 {q3}, [r0], r1- vld1.8 {q7}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q13, d5, d13
-
- vabal.u8 q12, d6, d14
- vabal.u8 q13, d7, d15
-
- vadd.u16 q0, q12, q13
-
- vpaddl.u16 q1, q0
- vpaddl.u32 q0, q1
-
- vadd.u32 d0, d0, d1
-
- vmov.32 r0, d0[0]
-
- bx lr
-
- ENDP
-
- END
--- a/vp8/encoder/arm/neon/sad8_neon.asm
+++ /dev/null
@@ -1,209 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_sad8x8_neon|
- EXPORT |vp8_sad8x16_neon|
- EXPORT |vp8_sad4x4_neon|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-; unsigned int vp8_sad8x8_c(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-
-|vp8_sad8x8_neon| PROC
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabdl.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
-
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vabal.u8 q12, d6, d14
-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabal.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q12, d6, d14
-
- vpaddl.u16 q1, q12
- vpaddl.u32 q0, q1
- vadd.u32 d0, d0, d1
-
- vmov.32 r0, d0[0]
-
- bx lr
-
- ENDP
-
-;============================
-;unsigned int vp8_sad8x16_c(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-
-|vp8_sad8x16_neon| PROC
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabdl.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
-
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vabal.u8 q12, d6, d14
-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabal.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
-
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vabal.u8 q12, d6, d14
-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabal.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
-
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vabal.u8 q12, d6, d14
-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabal.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q12, d6, d14
-
- vpaddl.u16 q1, q12
- vpaddl.u32 q0, q1
- vadd.u32 d0, d0, d1
-
- vmov.32 r0, d0[0]
-
- bx lr
-
- ENDP
-
-;===========================
-;unsigned int vp8_sad4x4_c(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-
-|vp8_sad4x4_neon| PROC
- vld1.8 {d0}, [r0], r1- vld1.8 {d8}, [r2], r3-
- vld1.8 {d2}, [r0], r1- vld1.8 {d10}, [r2], r3-
- vabdl.u8 q12, d0, d8
-
- vld1.8 {d4}, [r0], r1- vld1.8 {d12}, [r2], r3-
- vabal.u8 q12, d2, d10
-
- vld1.8 {d6}, [r0], r1- vld1.8 {d14}, [r2], r3-
- vabal.u8 q12, d4, d12
- vabal.u8 q12, d6, d14
-
- vpaddl.u16 d1, d24
- vpaddl.u32 d0, d1
- vmov.32 r0, d0[0]
-
- bx lr
-
- ENDP
-
- END
--- a/vp8/encoder/arm/neon/variance_neon.asm
+++ /dev/null
@@ -1,276 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance16x16_neon|
- EXPORT |vp8_variance16x8_neon|
- EXPORT |vp8_variance8x16_neon|
- EXPORT |vp8_variance8x8_neon|
-
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance16x16_neon| PROC
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- mov r12, #8
-
-variance16x16_neon_loop
- vld1.8 {q0}, [r0], r1 ;Load up source and reference- vld1.8 {q2}, [r2], r3- vld1.8 {q1}, [r0], r1- vld1.8 {q3}, [r2], r3-
- vsubl.u8 q11, d0, d4 ;calculate diff
- vsubl.u8 q12, d1, d5
- vsubl.u8 q13, d2, d6
- vsubl.u8 q14, d3, d7
-
- ;VPADAL adds adjacent pairs of elements of a vector, and accumulates
- ;the results into the elements of the destination vector. The explanation
- ;in ARM guide is wrong.
- vpadal.s16 q8, q11 ;calculate sum
- vmlal.s16 q9, d22, d22 ;calculate sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- bne variance16x16_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- ldr r12, [sp] ;load *sse from stack
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- ;vmov.32 r0, d0[0] ;this instruction costs a lot
- ;vmov.32 r1, d1[0]
- ;mul r0, r0, r0
- ;str r1, [r12]
- ;sub r0, r1, r0, lsr #8
-
- ; while sum is signed, sum * sum is always positive and must be treated as
- ; unsigned to avoid propagating the sign bit.
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [r12] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- bx lr
-
- ENDP
-
-;================================
-;unsigned int vp8_variance16x8_c(
-; unsigned char *src_ptr,
-; int source_stride,
-; unsigned char *ref_ptr,
-; int recon_stride,
-; unsigned int *sse)
-|vp8_variance16x8_neon| PROC
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- mov r12, #4
-
-variance16x8_neon_loop
- vld1.8 {q0}, [r0], r1 ;Load up source and reference- vld1.8 {q2}, [r2], r3- vld1.8 {q1}, [r0], r1- vld1.8 {q3}, [r2], r3-
- vsubl.u8 q11, d0, d4 ;calculate diff
- vsubl.u8 q12, d1, d5
- vsubl.u8 q13, d2, d6
- vsubl.u8 q14, d3, d7
-
- vpadal.s16 q8, q11 ;calculate sum
- vmlal.s16 q9, d22, d22 ;calculate sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- bne variance16x8_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- ldr r12, [sp] ;load *sse from stack
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [r12] ;store sse- vshr.u32 d10, d10, #7
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- bx lr
-
- ENDP
-
-;=================================
-;unsigned int vp8_variance8x16_c(
-; unsigned char *src_ptr,
-; int source_stride,
-; unsigned char *ref_ptr,
-; int recon_stride,
-; unsigned int *sse)
-
-|vp8_variance8x16_neon| PROC
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- mov r12, #8
-
-variance8x16_neon_loop
- vld1.8 {d0}, [r0], r1 ;Load up source and reference- vld1.8 {d4}, [r2], r3- vld1.8 {d2}, [r0], r1- vld1.8 {d6}, [r2], r3-
- vsubl.u8 q11, d0, d4 ;calculate diff
- vsubl.u8 q12, d2, d6
-
- vpadal.s16 q8, q11 ;calculate sum
- vmlal.s16 q9, d22, d22 ;calculate sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
-
- bne variance8x16_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- ldr r12, [sp] ;load *sse from stack
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [r12] ;store sse- vshr.u32 d10, d10, #7
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- bx lr
-
- ENDP
-
-;==================================
-; r0 unsigned char *src_ptr
-; r1 int source_stride
-; r2 unsigned char *ref_ptr
-; r3 int recon_stride
-; stack unsigned int *sse
-|vp8_variance8x8_neon| PROC
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- mov r12, #2
-
-variance8x8_neon_loop
- vld1.8 {d0}, [r0], r1 ;Load up source and reference- vld1.8 {d4}, [r2], r3- vld1.8 {d1}, [r0], r1- vld1.8 {d5}, [r2], r3- vld1.8 {d2}, [r0], r1- vld1.8 {d6}, [r2], r3- vld1.8 {d3}, [r0], r1- vld1.8 {d7}, [r2], r3-
- vsubl.u8 q11, d0, d4 ;calculate diff
- vsubl.u8 q12, d1, d5
- vsubl.u8 q13, d2, d6
- vsubl.u8 q14, d3, d7
-
- vpadal.s16 q8, q11 ;calculate sum
- vmlal.s16 q9, d22, d22 ;calculate sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- bne variance8x8_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- ldr r12, [sp] ;load *sse from stack
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [r12] ;store sse- vshr.u32 d10, d10, #6
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- bx lr
-
- ENDP
-
- END
--- a/vp8/encoder/arm/neon/vp8_subpixelvariance16x16_neon.asm
+++ /dev/null
@@ -1,425 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_sub_pixel_variance16x16_neon_func|
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-; r0 unsigned char *src_ptr,
-; r1 int src_pixels_per_line,
-; r2 int xoffset,
-; r3 int yoffset,
-; stack(r4) unsigned char *dst_ptr,
-; stack(r5) int dst_pixels_per_line,
-; stack(r6) unsigned int *sse
-;note: most of the code is copied from bilinear_predict16x16_neon and vp8_variance16x16_neon.
-
-|vp8_sub_pixel_variance16x16_neon_func| PROC
- push {r4-r6, lr}-
- ldr r12, _BilinearTaps_coeff_
- ldr r4, [sp, #16] ;load *dst_ptr from stack
- ldr r5, [sp, #20] ;load dst_pixels_per_line from stack
- ldr r6, [sp, #24] ;load *sse from stack
-
- cmp r2, #0 ;skip first_pass filter if xoffset=0
- beq secondpass_bfilter16x16_only
-
- add r2, r12, r2, lsl #3 ;calculate filter location
-
- cmp r3, #0 ;skip second_pass filter if yoffset=0
-
- vld1.s32 {d31}, [r2] ;load first_pass filter-
- beq firstpass_bfilter16x16_only
-
- sub sp, sp, #272 ;reserve space on stack for temporary storage
- vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data- mov lr, sp
- vld1.u8 {d5, d6, d7}, [r0], r1-
- mov r2, #3 ;loop counter
- vld1.u8 {d8, d9, d10}, [r0], r1-
- vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
- vld1.u8 {d11, d12, d13}, [r0], r1-
- vdup.8 d1, d31[4]
-
-;First Pass: output_height lines x output_width columns (17x16)
-vp8e_filt_blk2d_fp16x16_loop_neon
- pld [r0]
- pld [r0, r1]
- pld [r0, r1, lsl #1]
-
- vmull.u8 q7, d2, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q8, d3, d0
- vmull.u8 q9, d5, d0
- vmull.u8 q10, d6, d0
- vmull.u8 q11, d8, d0
- vmull.u8 q12, d9, d0
- vmull.u8 q13, d11, d0
- vmull.u8 q14, d12, d0
-
- vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
- vext.8 d5, d5, d6, #1
- vext.8 d8, d8, d9, #1
- vext.8 d11, d11, d12, #1
-
- vmlal.u8 q7, d2, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q9, d5, d1
- vmlal.u8 q11, d8, d1
- vmlal.u8 q13, d11, d1
-
- vext.8 d3, d3, d4, #1
- vext.8 d6, d6, d7, #1
- vext.8 d9, d9, d10, #1
- vext.8 d12, d12, d13, #1
-
- vmlal.u8 q8, d3, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q10, d6, d1
- vmlal.u8 q12, d9, d1
- vmlal.u8 q14, d12, d1
-
- subs r2, r2, #1
-
- vqrshrn.u16 d14, q7, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d15, q8, #7
- vqrshrn.u16 d16, q9, #7
- vqrshrn.u16 d17, q10, #7
- vqrshrn.u16 d18, q11, #7
- vqrshrn.u16 d19, q12, #7
- vqrshrn.u16 d20, q13, #7
-
- vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data- vqrshrn.u16 d21, q14, #7
- vld1.u8 {d5, d6, d7}, [r0], r1-
- vst1.u8 {d14, d15, d16, d17}, [lr]! ;store result- vld1.u8 {d8, d9, d10}, [r0], r1- vst1.u8 {d18, d19, d20, d21}, [lr]!- vld1.u8 {d11, d12, d13}, [r0], r1-
- bne vp8e_filt_blk2d_fp16x16_loop_neon
-
-;First-pass filtering for rest 5 lines
- vld1.u8 {d14, d15, d16}, [r0], r1-
- vmull.u8 q9, d2, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q10, d3, d0
- vmull.u8 q11, d5, d0
- vmull.u8 q12, d6, d0
- vmull.u8 q13, d8, d0
- vmull.u8 q14, d9, d0
-
- vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
- vext.8 d5, d5, d6, #1
- vext.8 d8, d8, d9, #1
-
- vmlal.u8 q9, d2, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q11, d5, d1
- vmlal.u8 q13, d8, d1
-
- vext.8 d3, d3, d4, #1
- vext.8 d6, d6, d7, #1
- vext.8 d9, d9, d10, #1
-
- vmlal.u8 q10, d3, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q12, d6, d1
- vmlal.u8 q14, d9, d1
-
- vmull.u8 q1, d11, d0
- vmull.u8 q2, d12, d0
- vmull.u8 q3, d14, d0
- vmull.u8 q4, d15, d0
-
- vext.8 d11, d11, d12, #1 ;construct src_ptr[1]
- vext.8 d14, d14, d15, #1
-
- vmlal.u8 q1, d11, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q3, d14, d1
-
- vext.8 d12, d12, d13, #1
- vext.8 d15, d15, d16, #1
-
- vmlal.u8 q2, d12, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q4, d15, d1
-
- vqrshrn.u16 d10, q9, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d11, q10, #7
- vqrshrn.u16 d12, q11, #7
- vqrshrn.u16 d13, q12, #7
- vqrshrn.u16 d14, q13, #7
- vqrshrn.u16 d15, q14, #7
- vqrshrn.u16 d16, q1, #7
- vqrshrn.u16 d17, q2, #7
- vqrshrn.u16 d18, q3, #7
- vqrshrn.u16 d19, q4, #7
-
- vst1.u8 {d10, d11, d12, d13}, [lr]! ;store result- vst1.u8 {d14, d15, d16, d17}, [lr]!- vst1.u8 {d18, d19}, [lr]!-
-;Second pass: 16x16
-;secondpass_filter
- add r3, r12, r3, lsl #3
- sub lr, lr, #272
-
- vld1.u32 {d31}, [r3] ;load second_pass filter-
- sub sp, sp, #256
- mov r3, sp
-
- vld1.u8 {d22, d23}, [lr]! ;load src data-
- vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
- vdup.8 d1, d31[4]
- mov r12, #4 ;loop counter
-
-vp8e_filt_blk2d_sp16x16_loop_neon
- vld1.u8 {d24, d25}, [lr]!- vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
- vld1.u8 {d26, d27}, [lr]!- vmull.u8 q2, d23, d0
- vld1.u8 {d28, d29}, [lr]!- vmull.u8 q3, d24, d0
- vld1.u8 {d30, d31}, [lr]!-
- vmull.u8 q4, d25, d0
- vmull.u8 q5, d26, d0
- vmull.u8 q6, d27, d0
- vmull.u8 q7, d28, d0
- vmull.u8 q8, d29, d0
-
- vmlal.u8 q1, d24, d1 ;(src_ptr[pixel_step] * Filter[1])
- vmlal.u8 q2, d25, d1
- vmlal.u8 q3, d26, d1
- vmlal.u8 q4, d27, d1
- vmlal.u8 q5, d28, d1
- vmlal.u8 q6, d29, d1
- vmlal.u8 q7, d30, d1
- vmlal.u8 q8, d31, d1
-
- subs r12, r12, #1
-
- vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d3, q2, #7
- vqrshrn.u16 d4, q3, #7
- vqrshrn.u16 d5, q4, #7
- vqrshrn.u16 d6, q5, #7
- vqrshrn.u16 d7, q6, #7
- vqrshrn.u16 d8, q7, #7
- vqrshrn.u16 d9, q8, #7
-
- vst1.u8 {d2, d3}, [r3]! ;store result- vst1.u8 {d4, d5}, [r3]!- vst1.u8 {d6, d7}, [r3]!- vmov q11, q15
- vst1.u8 {d8, d9}, [r3]!-
- bne vp8e_filt_blk2d_sp16x16_loop_neon
-
- b sub_pixel_variance16x16_neon
-
-;--------------------
-firstpass_bfilter16x16_only
- mov r2, #4 ;loop counter
- sub sp, sp, #528 ;reserve space on stack for temporary storage
- vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
- vdup.8 d1, d31[4]
- mov r3, sp
-
-;First Pass: output_height lines x output_width columns (16x16)
-vp8e_filt_blk2d_fpo16x16_loop_neon
- vld1.u8 {d2, d3, d4}, [r0], r1 ;load src data- vld1.u8 {d5, d6, d7}, [r0], r1- vld1.u8 {d8, d9, d10}, [r0], r1- vld1.u8 {d11, d12, d13}, [r0], r1-
- pld [r0]
- pld [r0, r1]
- pld [r0, r1, lsl #1]
-
- vmull.u8 q7, d2, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q8, d3, d0
- vmull.u8 q9, d5, d0
- vmull.u8 q10, d6, d0
- vmull.u8 q11, d8, d0
- vmull.u8 q12, d9, d0
- vmull.u8 q13, d11, d0
- vmull.u8 q14, d12, d0
-
- vext.8 d2, d2, d3, #1 ;construct src_ptr[1]
- vext.8 d5, d5, d6, #1
- vext.8 d8, d8, d9, #1
- vext.8 d11, d11, d12, #1
-
- vmlal.u8 q7, d2, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q9, d5, d1
- vmlal.u8 q11, d8, d1
- vmlal.u8 q13, d11, d1
-
- vext.8 d3, d3, d4, #1
- vext.8 d6, d6, d7, #1
- vext.8 d9, d9, d10, #1
- vext.8 d12, d12, d13, #1
-
- vmlal.u8 q8, d3, d1 ;(src_ptr[0] * Filter[1])
- vmlal.u8 q10, d6, d1
- vmlal.u8 q12, d9, d1
- vmlal.u8 q14, d12, d1
-
- subs r2, r2, #1
-
- vqrshrn.u16 d14, q7, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d15, q8, #7
- vqrshrn.u16 d16, q9, #7
- vqrshrn.u16 d17, q10, #7
- vqrshrn.u16 d18, q11, #7
- vqrshrn.u16 d19, q12, #7
- vqrshrn.u16 d20, q13, #7
- vst1.u8 {d14, d15}, [r3]! ;store result- vqrshrn.u16 d21, q14, #7
-
- vst1.u8 {d16, d17}, [r3]!- vst1.u8 {d18, d19}, [r3]!- vst1.u8 {d20, d21}, [r3]!-
- bne vp8e_filt_blk2d_fpo16x16_loop_neon
-
- b sub_pixel_variance16x16_neon
-
-;---------------------
-secondpass_bfilter16x16_only
-;Second pass: 16x16
-;secondpass_filter
- sub sp, sp, #528 ;reserve space on stack for temporary storage
- add r3, r12, r3, lsl #3
- mov r12, #4 ;loop counter
- vld1.u32 {d31}, [r3] ;load second_pass filter- vld1.u8 {d22, d23}, [r0], r1 ;load src data- mov r3, sp
-
- vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
- vdup.8 d1, d31[4]
-
-vp8e_filt_blk2d_spo16x16_loop_neon
- vld1.u8 {d24, d25}, [r0], r1- vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
- vld1.u8 {d26, d27}, [r0], r1- vmull.u8 q2, d23, d0
- vld1.u8 {d28, d29}, [r0], r1- vmull.u8 q3, d24, d0
- vld1.u8 {d30, d31}, [r0], r1-
- vmull.u8 q4, d25, d0
- vmull.u8 q5, d26, d0
- vmull.u8 q6, d27, d0
- vmull.u8 q7, d28, d0
- vmull.u8 q8, d29, d0
-
- vmlal.u8 q1, d24, d1 ;(src_ptr[pixel_step] * Filter[1])
- vmlal.u8 q2, d25, d1
- vmlal.u8 q3, d26, d1
- vmlal.u8 q4, d27, d1
- vmlal.u8 q5, d28, d1
- vmlal.u8 q6, d29, d1
- vmlal.u8 q7, d30, d1
- vmlal.u8 q8, d31, d1
-
- vqrshrn.u16 d2, q1, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d3, q2, #7
- vqrshrn.u16 d4, q3, #7
- vqrshrn.u16 d5, q4, #7
- vqrshrn.u16 d6, q5, #7
- vqrshrn.u16 d7, q6, #7
- vqrshrn.u16 d8, q7, #7
- vqrshrn.u16 d9, q8, #7
-
- vst1.u8 {d2, d3}, [r3]! ;store result- subs r12, r12, #1
- vst1.u8 {d4, d5}, [r3]!- vmov q11, q15
- vst1.u8 {d6, d7}, [r3]!- vst1.u8 {d8, d9}, [r3]!-
- bne vp8e_filt_blk2d_spo16x16_loop_neon
-
- b sub_pixel_variance16x16_neon
-
-;----------------------------
-;variance16x16
-sub_pixel_variance16x16_neon
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- sub r3, r3, #256
- mov r12, #8
-
-sub_pixel_variance16x16_neon_loop
- vld1.8 {q0}, [r3]! ;Load up source and reference- vld1.8 {q2}, [r4], r5- vld1.8 {q1}, [r3]!- vld1.8 {q3}, [r4], r5-
- vsubl.u8 q11, d0, d4 ;diff
- vsubl.u8 q12, d1, d5
- vsubl.u8 q13, d2, d6
- vsubl.u8 q14, d3, d7
-
- vpadal.s16 q8, q11 ;sum
- vmlal.s16 q9, d22, d22 ;sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- bne sub_pixel_variance16x16_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [r6] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- add sp, sp, #528
- vmov.32 r0, d0[0] ;return
-
- pop {r4-r6,pc}-
- ENDP
-
-;-----------------
-
-_BilinearTaps_coeff_
- DCD bilinear_taps_coeff
-bilinear_taps_coeff
- DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
-
- END
--- a/vp8/encoder/arm/neon/vp8_subpixelvariance16x16s_neon.asm
+++ /dev/null
@@ -1,572 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_variance_halfpixvar16x16_h_neon|
- EXPORT |vp8_variance_halfpixvar16x16_v_neon|
- EXPORT |vp8_variance_halfpixvar16x16_hv_neon|
- EXPORT |vp8_sub_pixel_variance16x16s_neon|
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-
-;================================================
-;unsigned int vp8_variance_halfpixvar16x16_h_neon
-;(
-; unsigned char *src_ptr, r0
-; int src_pixels_per_line, r1
-; unsigned char *dst_ptr, r2
-; int dst_pixels_per_line, r3
-; unsigned int *sse
-;);
-;================================================
-|vp8_variance_halfpixvar16x16_h_neon| PROC
- push {lr}-
- mov r12, #4 ;loop counter
- ldr lr, [sp, #4] ;load *sse from stack
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
-;First Pass: output_height lines x output_width columns (16x16)
-vp8_filt_fpo16x16s_4_0_loop_neon
- vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data- vld1.8 {q11}, [r2], r3- vld1.u8 {d4, d5, d6, d7}, [r0], r1- vld1.8 {q12}, [r2], r3- vld1.u8 {d8, d9, d10, d11}, [r0], r1- vld1.8 {q13}, [r2], r3- vld1.u8 {d12, d13, d14, d15}, [r0], r1-
- ;pld [r0]
- ;pld [r0, r1]
- ;pld [r0, r1, lsl #1]
-
- vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
- vext.8 q3, q2, q3, #1
- vext.8 q5, q4, q5, #1
- vext.8 q7, q6, q7, #1
-
- vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
- vld1.8 {q14}, [r2], r3- vrhadd.u8 q1, q2, q3
- vrhadd.u8 q2, q4, q5
- vrhadd.u8 q3, q6, q7
-
- vsubl.u8 q4, d0, d22 ;diff
- vsubl.u8 q5, d1, d23
- vsubl.u8 q6, d2, d24
- vsubl.u8 q7, d3, d25
- vsubl.u8 q0, d4, d26
- vsubl.u8 q1, d5, d27
- vsubl.u8 q2, d6, d28
- vsubl.u8 q3, d7, d29
-
- vpadal.s16 q8, q4 ;sum
- vmlal.s16 q9, d8, d8 ;sse
- vmlal.s16 q10, d9, d9
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q5
- vmlal.s16 q9, d10, d10
- vmlal.s16 q10, d11, d11
- vpadal.s16 q8, q6
- vmlal.s16 q9, d12, d12
- vmlal.s16 q10, d13, d13
- vpadal.s16 q8, q7
- vmlal.s16 q9, d14, d14
- vmlal.s16 q10, d15, d15
-
- vpadal.s16 q8, q0 ;sum
- vmlal.s16 q9, d0, d0 ;sse
- vmlal.s16 q10, d1, d1
- vpadal.s16 q8, q1
- vmlal.s16 q9, d2, d2
- vmlal.s16 q10, d3, d3
- vpadal.s16 q8, q2
- vmlal.s16 q9, d4, d4
- vmlal.s16 q10, d5, d5
- vpadal.s16 q8, q3
- vmlal.s16 q9, d6, d6
- vmlal.s16 q10, d7, d7
-
- bne vp8_filt_fpo16x16s_4_0_loop_neon
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [lr] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- pop {pc}- ENDP
-
-;================================================
-;unsigned int vp8_variance_halfpixvar16x16_v_neon
-;(
-; unsigned char *src_ptr, r0
-; int src_pixels_per_line, r1
-; unsigned char *dst_ptr, r2
-; int dst_pixels_per_line, r3
-; unsigned int *sse
-;);
-;================================================
-|vp8_variance_halfpixvar16x16_v_neon| PROC
- push {lr}-
- mov r12, #4 ;loop counter
-
- vld1.u8 {q0}, [r0], r1 ;load src data- ldr lr, [sp, #4] ;load *sse from stack
-
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
-vp8_filt_spo16x16s_0_4_loop_neon
- vld1.u8 {q2}, [r0], r1- vld1.8 {q1}, [r2], r3- vld1.u8 {q4}, [r0], r1- vld1.8 {q3}, [r2], r3- vld1.u8 {q6}, [r0], r1- vld1.8 {q5}, [r2], r3- vld1.u8 {q15}, [r0], r1-
- vrhadd.u8 q0, q0, q2
- vld1.8 {q7}, [r2], r3- vrhadd.u8 q2, q2, q4
- vrhadd.u8 q4, q4, q6
- vrhadd.u8 q6, q6, q15
-
- vsubl.u8 q11, d0, d2 ;diff
- vsubl.u8 q12, d1, d3
- vsubl.u8 q13, d4, d6
- vsubl.u8 q14, d5, d7
- vsubl.u8 q0, d8, d10
- vsubl.u8 q1, d9, d11
- vsubl.u8 q2, d12, d14
- vsubl.u8 q3, d13, d15
-
- vpadal.s16 q8, q11 ;sum
- vmlal.s16 q9, d22, d22 ;sse
- vmlal.s16 q10, d23, d23
-
- subs r12, r12, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- vpadal.s16 q8, q0 ;sum
- vmlal.s16 q9, d0, d0 ;sse
- vmlal.s16 q10, d1, d1
- vpadal.s16 q8, q1
- vmlal.s16 q9, d2, d2
- vmlal.s16 q10, d3, d3
- vpadal.s16 q8, q2
- vmlal.s16 q9, d4, d4
- vmlal.s16 q10, d5, d5
-
- vmov q0, q15
-
- vpadal.s16 q8, q3
- vmlal.s16 q9, d6, d6
- vmlal.s16 q10, d7, d7
-
- bne vp8_filt_spo16x16s_0_4_loop_neon
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [lr] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- pop {pc}- ENDP
-
-;================================================
-;unsigned int vp8_variance_halfpixvar16x16_hv_neon
-;(
-; unsigned char *src_ptr, r0
-; int src_pixels_per_line, r1
-; unsigned char *dst_ptr, r2
-; int dst_pixels_per_line, r3
-; unsigned int *sse
-;);
-;================================================
-|vp8_variance_halfpixvar16x16_hv_neon| PROC
- push {lr}-
- vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data-
- ldr lr, [sp, #4] ;load *sse from stack
- vmov.i8 q13, #0 ;q8 - sum
- vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
-
- vmov.i8 q14, #0 ;q9, q10 - sse
- vmov.i8 q15, #0
-
- mov r12, #4 ;loop counter
- vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
-
-;First Pass: output_height lines x output_width columns (17x16)
-vp8_filt16x16s_4_4_loop_neon
- vld1.u8 {d4, d5, d6, d7}, [r0], r1- vld1.u8 {d8, d9, d10, d11}, [r0], r1- vld1.u8 {d12, d13, d14, d15}, [r0], r1- vld1.u8 {d16, d17, d18, d19}, [r0], r1-
- ;pld [r0]
- ;pld [r0, r1]
- ;pld [r0, r1, lsl #1]
-
- vext.8 q3, q2, q3, #1 ;construct src_ptr[1]
- vext.8 q5, q4, q5, #1
- vext.8 q7, q6, q7, #1
- vext.8 q9, q8, q9, #1
-
- vrhadd.u8 q1, q2, q3 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
- vrhadd.u8 q2, q4, q5
- vrhadd.u8 q3, q6, q7
- vrhadd.u8 q4, q8, q9
-
- vld1.8 {q5}, [r2], r3- vrhadd.u8 q0, q0, q1
- vld1.8 {q6}, [r2], r3- vrhadd.u8 q1, q1, q2
- vld1.8 {q7}, [r2], r3- vrhadd.u8 q2, q2, q3
- vld1.8 {q8}, [r2], r3- vrhadd.u8 q3, q3, q4
-
- vsubl.u8 q9, d0, d10 ;diff
- vsubl.u8 q10, d1, d11
- vsubl.u8 q11, d2, d12
- vsubl.u8 q12, d3, d13
-
- vsubl.u8 q0, d4, d14 ;diff
- vsubl.u8 q1, d5, d15
- vsubl.u8 q5, d6, d16
- vsubl.u8 q6, d7, d17
-
- vpadal.s16 q13, q9 ;sum
- vmlal.s16 q14, d18, d18 ;sse
- vmlal.s16 q15, d19, d19
-
- vpadal.s16 q13, q10 ;sum
- vmlal.s16 q14, d20, d20 ;sse
- vmlal.s16 q15, d21, d21
-
- vpadal.s16 q13, q11 ;sum
- vmlal.s16 q14, d22, d22 ;sse
- vmlal.s16 q15, d23, d23
-
- vpadal.s16 q13, q12 ;sum
- vmlal.s16 q14, d24, d24 ;sse
- vmlal.s16 q15, d25, d25
-
- subs r12, r12, #1
-
- vpadal.s16 q13, q0 ;sum
- vmlal.s16 q14, d0, d0 ;sse
- vmlal.s16 q15, d1, d1
-
- vpadal.s16 q13, q1 ;sum
- vmlal.s16 q14, d2, d2 ;sse
- vmlal.s16 q15, d3, d3
-
- vpadal.s16 q13, q5 ;sum
- vmlal.s16 q14, d10, d10 ;sse
- vmlal.s16 q15, d11, d11
-
- vmov q0, q4
-
- vpadal.s16 q13, q6 ;sum
- vmlal.s16 q14, d12, d12 ;sse
- vmlal.s16 q15, d13, d13
-
- bne vp8_filt16x16s_4_4_loop_neon
-
- vadd.u32 q15, q14, q15 ;accumulate sse
- vpaddl.s32 q0, q13 ;accumulate sum
-
- vpaddl.u32 q1, q15
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [lr] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- pop {pc}- ENDP
-
-;==============================
-; r0 unsigned char *src_ptr,
-; r1 int src_pixels_per_line,
-; r2 int xoffset,
-; r3 int yoffset,
-; stack unsigned char *dst_ptr,
-; stack int dst_pixels_per_line,
-; stack unsigned int *sse
-;note: in vp8_find_best_half_pixel_step()(called when 8<Speed<15), and first call of vp8_find_best_sub_pixel_step()
-;(called when speed<=8). xoffset/yoffset can only be 4 or 0, which means either by pass the filter,
-;or filter coeff is {64, 64}. This simplified program only works in this situation.-;note: It happens that both xoffset and yoffset are zero. This can be handled in c code later.
-
-|vp8_sub_pixel_variance16x16s_neon| PROC
- push {r4, lr}-
- ldr r4, [sp, #8] ;load *dst_ptr from stack
- ldr r12, [sp, #12] ;load dst_pixels_per_line from stack
- ldr lr, [sp, #16] ;load *sse from stack
-
- cmp r2, #0 ;skip first_pass filter if xoffset=0
- beq secondpass_bfilter16x16s_only
-
- cmp r3, #0 ;skip second_pass filter if yoffset=0
- beq firstpass_bfilter16x16s_only
-
- vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data- sub sp, sp, #256 ;reserve space on stack for temporary storage
- vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
- mov r3, sp
- mov r2, #4 ;loop counter
- vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
-
-;First Pass: output_height lines x output_width columns (17x16)
-vp8e_filt_blk2d_fp16x16s_loop_neon
- vld1.u8 {d4, d5, d6, d7}, [r0], r1- vld1.u8 {d8, d9, d10, d11}, [r0], r1- vld1.u8 {d12, d13, d14, d15}, [r0], r1- vld1.u8 {d16, d17, d18, d19}, [r0], r1-
- ;pld [r0]
- ;pld [r0, r1]
- ;pld [r0, r1, lsl #1]
-
- vext.8 q3, q2, q3, #1 ;construct src_ptr[1]
- vext.8 q5, q4, q5, #1
- vext.8 q7, q6, q7, #1
- vext.8 q9, q8, q9, #1
-
- vrhadd.u8 q1, q2, q3 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
- vrhadd.u8 q2, q4, q5
- vrhadd.u8 q3, q6, q7
- vrhadd.u8 q4, q8, q9
-
- vrhadd.u8 q0, q0, q1
- vrhadd.u8 q1, q1, q2
- vrhadd.u8 q2, q2, q3
- vrhadd.u8 q3, q3, q4
-
- subs r2, r2, #1
- vst1.u8 {d0, d1 ,d2, d3}, [r3]! ;store result- vmov q0, q4
- vst1.u8 {d4, d5, d6, d7}, [r3]!-
- bne vp8e_filt_blk2d_fp16x16s_loop_neon
-
- b sub_pixel_variance16x16s_neon
-
-;--------------------
-firstpass_bfilter16x16s_only
- mov r2, #2 ;loop counter
- sub sp, sp, #256 ;reserve space on stack for temporary storage
- mov r3, sp
-
-;First Pass: output_height lines x output_width columns (16x16)
-vp8e_filt_blk2d_fpo16x16s_loop_neon
- vld1.u8 {d0, d1, d2, d3}, [r0], r1 ;load src data- vld1.u8 {d4, d5, d6, d7}, [r0], r1- vld1.u8 {d8, d9, d10, d11}, [r0], r1- vld1.u8 {d12, d13, d14, d15}, [r0], r1-
- ;pld [r0]
- ;pld [r0, r1]
- ;pld [r0, r1, lsl #1]
-
- vext.8 q1, q0, q1, #1 ;construct src_ptr[1]
- vld1.u8 {d16, d17, d18, d19}, [r0], r1- vext.8 q3, q2, q3, #1
- vld1.u8 {d20, d21, d22, d23}, [r0], r1- vext.8 q5, q4, q5, #1
- vld1.u8 {d24, d25, d26, d27}, [r0], r1- vext.8 q7, q6, q7, #1
- vld1.u8 {d28, d29, d30, d31}, [r0], r1- vext.8 q9, q8, q9, #1
- vext.8 q11, q10, q11, #1
- vext.8 q13, q12, q13, #1
- vext.8 q15, q14, q15, #1
-
- vrhadd.u8 q0, q0, q1 ;(src_ptr[0]+src_ptr[1])/round/shift right 1
- vrhadd.u8 q1, q2, q3
- vrhadd.u8 q2, q4, q5
- vrhadd.u8 q3, q6, q7
- vrhadd.u8 q4, q8, q9
- vrhadd.u8 q5, q10, q11
- vrhadd.u8 q6, q12, q13
- vrhadd.u8 q7, q14, q15
-
- subs r2, r2, #1
-
- vst1.u8 {d0, d1, d2, d3}, [r3]! ;store result- vst1.u8 {d4, d5, d6, d7}, [r3]!- vst1.u8 {d8, d9, d10, d11}, [r3]!- vst1.u8 {d12, d13, d14, d15}, [r3]!-
- bne vp8e_filt_blk2d_fpo16x16s_loop_neon
-
- b sub_pixel_variance16x16s_neon
-
-;---------------------
-secondpass_bfilter16x16s_only
- sub sp, sp, #256 ;reserve space on stack for temporary storage
-
- mov r2, #2 ;loop counter
- vld1.u8 {d0, d1}, [r0], r1 ;load src data- mov r3, sp
-
-vp8e_filt_blk2d_spo16x16s_loop_neon
- vld1.u8 {d2, d3}, [r0], r1- vld1.u8 {d4, d5}, [r0], r1- vld1.u8 {d6, d7}, [r0], r1- vld1.u8 {d8, d9}, [r0], r1-
- vrhadd.u8 q0, q0, q1
- vld1.u8 {d10, d11}, [r0], r1- vrhadd.u8 q1, q1, q2
- vld1.u8 {d12, d13}, [r0], r1- vrhadd.u8 q2, q2, q3
- vld1.u8 {d14, d15}, [r0], r1- vrhadd.u8 q3, q3, q4
- vld1.u8 {d16, d17}, [r0], r1- vrhadd.u8 q4, q4, q5
- vrhadd.u8 q5, q5, q6
- vrhadd.u8 q6, q6, q7
- vrhadd.u8 q7, q7, q8
-
- subs r2, r2, #1
-
- vst1.u8 {d0, d1, d2, d3}, [r3]! ;store result- vmov q0, q8
- vst1.u8 {d4, d5, d6, d7}, [r3]!- vst1.u8 {d8, d9, d10, d11}, [r3]! ;store result- vst1.u8 {d12, d13, d14, d15}, [r3]!-
- bne vp8e_filt_blk2d_spo16x16s_loop_neon
-
- b sub_pixel_variance16x16s_neon
-
-;----------------------------
-;variance16x16
-sub_pixel_variance16x16s_neon
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- sub r3, r3, #256
- mov r2, #4
-
-sub_pixel_variance16x16s_neon_loop
- vld1.8 {q0}, [r3]! ;Load up source and reference- vld1.8 {q1}, [r4], r12- vld1.8 {q2}, [r3]!- vld1.8 {q3}, [r4], r12- vld1.8 {q4}, [r3]!- vld1.8 {q5}, [r4], r12- vld1.8 {q6}, [r3]!- vld1.8 {q7}, [r4], r12-
- vsubl.u8 q11, d0, d2 ;diff
- vsubl.u8 q12, d1, d3
- vsubl.u8 q13, d4, d6
- vsubl.u8 q14, d5, d7
- vsubl.u8 q0, d8, d10
- vsubl.u8 q1, d9, d11
- vsubl.u8 q2, d12, d14
- vsubl.u8 q3, d13, d15
-
- vpadal.s16 q8, q11 ;sum
- vmlal.s16 q9, d22, d22 ;sse
- vmlal.s16 q10, d23, d23
-
- subs r2, r2, #1
-
- vpadal.s16 q8, q12
- vmlal.s16 q9, d24, d24
- vmlal.s16 q10, d25, d25
- vpadal.s16 q8, q13
- vmlal.s16 q9, d26, d26
- vmlal.s16 q10, d27, d27
- vpadal.s16 q8, q14
- vmlal.s16 q9, d28, d28
- vmlal.s16 q10, d29, d29
-
- vpadal.s16 q8, q0 ;sum
- vmlal.s16 q9, d0, d0 ;sse
- vmlal.s16 q10, d1, d1
- vpadal.s16 q8, q1
- vmlal.s16 q9, d2, d2
- vmlal.s16 q10, d3, d3
- vpadal.s16 q8, q2
- vmlal.s16 q9, d4, d4
- vmlal.s16 q10, d5, d5
- vpadal.s16 q8, q3
- vmlal.s16 q9, d6, d6
- vmlal.s16 q10, d7, d7
-
- bne sub_pixel_variance16x16s_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [lr] ;store sse- vshr.u32 d10, d10, #8
- vsub.u32 d0, d1, d10
-
- add sp, sp, #256
- vmov.32 r0, d0[0] ;return
-
- pop {r4, pc}- ENDP
-
- END
--- a/vp8/encoder/arm/neon/vp8_subpixelvariance8x8_neon.asm
+++ /dev/null
@@ -1,224 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- EXPORT |vp8_sub_pixel_variance8x8_neon|
- ARM
- REQUIRE8
- PRESERVE8
-
- AREA ||.text||, CODE, READONLY, ALIGN=2
-; r0 unsigned char *src_ptr,
-; r1 int src_pixels_per_line,
-; r2 int xoffset,
-; r3 int yoffset,
-; stack(r4) unsigned char *dst_ptr,
-; stack(r5) int dst_pixels_per_line,
-; stack(r6) unsigned int *sse
-;note: most of the code is copied from bilinear_predict8x8_neon and vp8_variance8x8_neon.
-
-|vp8_sub_pixel_variance8x8_neon| PROC
- push {r4-r5, lr}-
- ldr r12, _BilinearTaps_coeff_
- ldr r4, [sp, #12] ;load *dst_ptr from stack
- ldr r5, [sp, #16] ;load dst_pixels_per_line from stack
- ldr lr, [sp, #20] ;load *sse from stack
-
- cmp r2, #0 ;skip first_pass filter if xoffset=0
- beq skip_firstpass_filter
-
-;First pass: output_height lines x output_width columns (9x8)
- add r2, r12, r2, lsl #3 ;calculate filter location
-
- vld1.u8 {q1}, [r0], r1 ;load src data- vld1.u32 {d31}, [r2] ;load first_pass filter- vld1.u8 {q2}, [r0], r1- vdup.8 d0, d31[0] ;first_pass filter (d0 d1)
- vld1.u8 {q3}, [r0], r1- vdup.8 d1, d31[4]
- vld1.u8 {q4}, [r0], r1-
- vmull.u8 q6, d2, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q7, d4, d0
- vmull.u8 q8, d6, d0
- vmull.u8 q9, d8, d0
-
- vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
- vext.8 d5, d4, d5, #1
- vext.8 d7, d6, d7, #1
- vext.8 d9, d8, d9, #1
-
- vmlal.u8 q6, d3, d1 ;(src_ptr[1] * Filter[1])
- vmlal.u8 q7, d5, d1
- vmlal.u8 q8, d7, d1
- vmlal.u8 q9, d9, d1
-
- vld1.u8 {q1}, [r0], r1 ;load src data- vqrshrn.u16 d22, q6, #7 ;shift/round/saturate to u8
- vld1.u8 {q2}, [r0], r1- vqrshrn.u16 d23, q7, #7
- vld1.u8 {q3}, [r0], r1- vqrshrn.u16 d24, q8, #7
- vld1.u8 {q4}, [r0], r1- vqrshrn.u16 d25, q9, #7
-
- ;first_pass filtering on the rest 5-line data
- vld1.u8 {q5}, [r0], r1-
- vmull.u8 q6, d2, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q7, d4, d0
- vmull.u8 q8, d6, d0
- vmull.u8 q9, d8, d0
- vmull.u8 q10, d10, d0
-
- vext.8 d3, d2, d3, #1 ;construct src_ptr[-1]
- vext.8 d5, d4, d5, #1
- vext.8 d7, d6, d7, #1
- vext.8 d9, d8, d9, #1
- vext.8 d11, d10, d11, #1
-
- vmlal.u8 q6, d3, d1 ;(src_ptr[1] * Filter[1])
- vmlal.u8 q7, d5, d1
- vmlal.u8 q8, d7, d1
- vmlal.u8 q9, d9, d1
- vmlal.u8 q10, d11, d1
-
- vqrshrn.u16 d26, q6, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d27, q7, #7
- vqrshrn.u16 d28, q8, #7
- vqrshrn.u16 d29, q9, #7
- vqrshrn.u16 d30, q10, #7
-
-;Second pass: 8x8
-secondpass_filter
- cmp r3, #0 ;skip second_pass filter if yoffset=0
- ;skip_secondpass_filter
- beq sub_pixel_variance8x8_neon
-
- add r3, r12, r3, lsl #3
-
- vld1.u32 {d31}, [r3] ;load second_pass filter-
- vdup.8 d0, d31[0] ;second_pass filter parameters (d0 d1)
- vdup.8 d1, d31[4]
-
- vmull.u8 q1, d22, d0 ;(src_ptr[0] * Filter[0])
- vmull.u8 q2, d23, d0
- vmull.u8 q3, d24, d0
- vmull.u8 q4, d25, d0
- vmull.u8 q5, d26, d0
- vmull.u8 q6, d27, d0
- vmull.u8 q7, d28, d0
- vmull.u8 q8, d29, d0
-
- vmlal.u8 q1, d23, d1 ;(src_ptr[pixel_step] * Filter[1])
- vmlal.u8 q2, d24, d1
- vmlal.u8 q3, d25, d1
- vmlal.u8 q4, d26, d1
- vmlal.u8 q5, d27, d1
- vmlal.u8 q6, d28, d1
- vmlal.u8 q7, d29, d1
- vmlal.u8 q8, d30, d1
-
- vqrshrn.u16 d22, q1, #7 ;shift/round/saturate to u8
- vqrshrn.u16 d23, q2, #7
- vqrshrn.u16 d24, q3, #7
- vqrshrn.u16 d25, q4, #7
- vqrshrn.u16 d26, q5, #7
- vqrshrn.u16 d27, q6, #7
- vqrshrn.u16 d28, q7, #7
- vqrshrn.u16 d29, q8, #7
-
- b sub_pixel_variance8x8_neon
-
-;--------------------
-skip_firstpass_filter
- vld1.u8 {d22}, [r0], r1 ;load src data- vld1.u8 {d23}, [r0], r1- vld1.u8 {d24}, [r0], r1- vld1.u8 {d25}, [r0], r1- vld1.u8 {d26}, [r0], r1- vld1.u8 {d27}, [r0], r1- vld1.u8 {d28}, [r0], r1- vld1.u8 {d29}, [r0], r1- vld1.u8 {d30}, [r0], r1-
- b secondpass_filter
-
-;----------------------
-;vp8_variance8x8_neon
-sub_pixel_variance8x8_neon
- vmov.i8 q8, #0 ;q8 - sum
- vmov.i8 q9, #0 ;q9, q10 - sse
- vmov.i8 q10, #0
-
- mov r12, #2
-
-sub_pixel_variance8x8_neon_loop
- vld1.8 {d0}, [r4], r5 ;load dst data- subs r12, r12, #1
- vld1.8 {d1}, [r4], r5- vld1.8 {d2}, [r4], r5- vsubl.u8 q4, d22, d0 ;calculate diff
- vld1.8 {d3}, [r4], r5-
- vsubl.u8 q5, d23, d1
- vsubl.u8 q6, d24, d2
-
- vpadal.s16 q8, q4 ;sum
- vmlal.s16 q9, d8, d8 ;sse
- vmlal.s16 q10, d9, d9
-
- vsubl.u8 q7, d25, d3
-
- vpadal.s16 q8, q5
- vmlal.s16 q9, d10, d10
- vmlal.s16 q10, d11, d11
-
- vmov q11, q13
-
- vpadal.s16 q8, q6
- vmlal.s16 q9, d12, d12
- vmlal.s16 q10, d13, d13
-
- vmov q12, q14
-
- vpadal.s16 q8, q7
- vmlal.s16 q9, d14, d14
- vmlal.s16 q10, d15, d15
-
- bne sub_pixel_variance8x8_neon_loop
-
- vadd.u32 q10, q9, q10 ;accumulate sse
- vpaddl.s32 q0, q8 ;accumulate sum
-
- vpaddl.u32 q1, q10
- vadd.s64 d0, d0, d1
- vadd.u64 d1, d2, d3
-
- vmull.s32 q5, d0, d0
- vst1.32 {d1[0]}, [lr] ;store sse- vshr.u32 d10, d10, #6
- vsub.u32 d0, d1, d10
-
- vmov.32 r0, d0[0] ;return
- pop {r4-r5, pc}-
- ENDP
-
-;-----------------
-
-_BilinearTaps_coeff_
- DCD bilinear_taps_coeff
-bilinear_taps_coeff
- DCD 128, 0, 112, 16, 96, 32, 80, 48, 64, 64, 48, 80, 32, 96, 16, 112
-
- END
--- a/vp8/encoder/arm/variance_arm.c
+++ /dev/null
@@ -1,121 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-#include "vpx_config.h"
-#include "vpx_rtcd.h"
-#include "vp8/encoder/variance.h"
-#include "vp8/common/filter.h"
-
-#if HAVE_MEDIA
-#include "vp8/common/arm/bilinearfilter_arm.h"
-
-unsigned int vp8_sub_pixel_variance8x8_armv6
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short first_pass[10*8];
- unsigned char second_pass[8*8];
- const short *HFilter, *VFilter;
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
- src_pixels_per_line,
- 9, 8, HFilter);
- vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
- 8, 8, 8, VFilter);
-
- return vp8_variance8x8_armv6(second_pass, 8, dst_ptr,
- dst_pixels_per_line, sse);
-}
-
-unsigned int vp8_sub_pixel_variance16x16_armv6
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short first_pass[36*16];
- unsigned char second_pass[20*16];
- const short *HFilter, *VFilter;
- unsigned int var;
-
- if (xoffset == 4 && yoffset == 0)
- {- var = vp8_variance_halfpixvar16x16_h_armv6(src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, sse);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- var = vp8_variance_halfpixvar16x16_v_armv6(src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, sse);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- var = vp8_variance_halfpixvar16x16_hv_armv6(src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, sse);
- }
- else
- {- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- vp8_filter_block2d_bil_first_pass_armv6(src_ptr, first_pass,
- src_pixels_per_line,
- 17, 16, HFilter);
- vp8_filter_block2d_bil_second_pass_armv6(first_pass, second_pass,
- 16, 16, 16, VFilter);
-
- var = vp8_variance16x16_armv6(second_pass, 16, dst_ptr,
- dst_pixels_per_line, sse);
- }
- return var;
-}
-
-#endif /* HAVE_MEDIA */
-
-
-#if HAVE_NEON
-
-unsigned int vp8_sub_pixel_variance16x16_neon
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- if (xoffset == 4 && yoffset == 0)
- return vp8_variance_halfpixvar16x16_h_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
- else if (xoffset == 0 && yoffset == 4)
- return vp8_variance_halfpixvar16x16_v_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
- else if (xoffset == 4 && yoffset == 4)
- return vp8_variance_halfpixvar16x16_hv_neon(src_ptr, src_pixels_per_line, dst_ptr, dst_pixels_per_line, sse);
- else
- return vp8_sub_pixel_variance16x16_neon_func(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
-}
-
-#endif
--- a/vp8/encoder/firstpass.c
+++ b/vp8/encoder/firstpass.c
@@ -12,7 +12,7 @@
#include "limits.h"
#include "block.h"
#include "onyx_int.h"
-#include "variance.h"
+#include "vp8/common/variance.h"
#include "encodeintra.h"
#include "vp8/common/setupintrarecon.h"
#include "mcomp.h"
--- a/vp8/encoder/mcomp.h
+++ b/vp8/encoder/mcomp.h
@@ -13,7 +13,7 @@
#define __INC_MCOMP_H
#include "block.h"
-#include "variance.h"
+#include "vp8/common/variance.h"
#ifdef ENTROPY_STATS
extern void init_mv_ref_counts();
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -18,7 +18,7 @@
#include "treewriter.h"
#include "tokenize.h"
#include "vp8/common/onyxc_int.h"
-#include "variance.h"
+#include "vp8/common/variance.h"
#include "encodemb.h"
#include "quantize.h"
#include "vp8/common/entropy.h"
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -20,7 +20,7 @@
#include "encodemb.h"
#include "vp8/common/reconinter.h"
#include "vp8/common/reconintra4x4.h"
-#include "variance.h"
+#include "vp8/common/variance.h"
#include "mcomp.h"
#include "rdopt.h"
#include "vpx_mem/vpx_mem.h"
--- a/vp8/encoder/ppc/sad_altivec.asm
+++ /dev/null
@@ -1,277 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- .globl vp8_sad16x16_ppc
- .globl vp8_sad16x8_ppc
- .globl vp8_sad8x16_ppc
- .globl vp8_sad8x8_ppc
- .globl vp8_sad4x4_ppc
-
-.macro load_aligned_16 V R O
- lvsl v3, 0, \R ;# permutate value for alignment
-
- lvx v1, 0, \R
- lvx v2, \O, \R
-
- vperm \V, v1, v2, v3
-.endm
-
-.macro prologue
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xffc0
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1, -32(r1) ;# create space on the stack
-
- li r10, 16 ;# load offset and loop counter
-
- vspltisw v8, 0 ;# zero out total to start
-.endm
-
-.macro epilogue
- addi r1, r1, 32 ;# recover stack
-
- mtspr 256, r11 ;# reset old VRSAVE
-.endm
-
-.macro SAD_16
- ;# v6 = abs (v4 - v5)
- vsububs v6, v4, v5
- vsububs v7, v5, v4
- vor v6, v6, v7
-
- ;# v8 += abs (v4 - v5)
- vsum4ubs v8, v6, v8
-.endm
-
-.macro sad_16_loop loop_label
- lvsl v3, 0, r5 ;# only needs to be done once per block
-
- ;# preload a line of data before getting into the loop
- lvx v4, 0, r3
- lvx v1, 0, r5
- lvx v2, r10, r5
-
- add r5, r5, r6
- add r3, r3, r4
-
- vperm v5, v1, v2, v3
-
- .align 4
-\loop_label:
- ;# compute difference on first row
- vsububs v6, v4, v5
- vsububs v7, v5, v4
-
- ;# load up next set of data
- lvx v9, 0, r3
- lvx v1, 0, r5
- lvx v2, r10, r5
-
- ;# perform abs() of difference
- vor v6, v6, v7
- add r3, r3, r4
-
- ;# add to the running tally
- vsum4ubs v8, v6, v8
-
- ;# now onto the next line
- vperm v5, v1, v2, v3
- add r5, r5, r6
- lvx v4, 0, r3
-
- ;# compute difference on second row
- vsububs v6, v9, v5
- lvx v1, 0, r5
- vsububs v7, v5, v9
- lvx v2, r10, r5
- vor v6, v6, v7
- add r3, r3, r4
- vsum4ubs v8, v6, v8
- vperm v5, v1, v2, v3
- add r5, r5, r6
-
- bdnz \loop_label
-
- vspltisw v7, 0
-
- vsumsws v8, v8, v7
-
- stvx v8, 0, r1
- lwz r3, 12(r1)
-.endm
-
-.macro sad_8_loop loop_label
- .align 4
-\loop_label:
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v4, r3, r10
- load_aligned_16 v5, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v6, r3, r10
- load_aligned_16 v7, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- vmrghb v4, v4, v6
- vmrghb v5, v5, v7
-
- SAD_16
-
- bdnz \loop_label
-
- vspltisw v7, 0
-
- vsumsws v8, v8, v7
-
- stvx v8, 0, r1
- lwz r3, 12(r1)
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int ref_stride
-;#
-;# r3 return value
-vp8_sad16x16_ppc:
-
- prologue
-
- li r9, 8
- mtctr r9
-
- sad_16_loop sad16x16_loop
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int ref_stride
-;#
-;# r3 return value
-vp8_sad16x8_ppc:
-
- prologue
-
- li r9, 4
- mtctr r9
-
- sad_16_loop sad16x8_loop
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int ref_stride
-;#
-;# r3 return value
-vp8_sad8x16_ppc:
-
- prologue
-
- li r9, 8
- mtctr r9
-
- sad_8_loop sad8x16_loop
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int ref_stride
-;#
-;# r3 return value
-vp8_sad8x8_ppc:
-
- prologue
-
- li r9, 4
- mtctr r9
-
- sad_8_loop sad8x8_loop
-
- epilogue
-
- blr
-
-.macro transfer_4x4 I P
- lwz r0, 0(\I)
- add \I, \I, \P
-
- lwz r7, 0(\I)
- add \I, \I, \P
-
- lwz r8, 0(\I)
- add \I, \I, \P
-
- lwz r9, 0(\I)
-
- stw r0, 0(r1)
- stw r7, 4(r1)
- stw r8, 8(r1)
- stw r9, 12(r1)
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int ref_stride
-;#
-;# r3 return value
-vp8_sad4x4_ppc:
-
- prologue
-
- transfer_4x4 r3, r4
- lvx v4, 0, r1
-
- transfer_4x4 r5, r6
- lvx v5, 0, r1
-
- vspltisw v8, 0 ;# zero out total to start
-
- ;# v6 = abs (v4 - v5)
- vsububs v6, v4, v5
- vsububs v7, v5, v4
- vor v6, v6, v7
-
- ;# v8 += abs (v4 - v5)
- vsum4ubs v7, v6, v8
- vsumsws v7, v7, v8
-
- stvx v7, 0, r1
- lwz r3, 12(r1)
-
- epilogue
-
- blr
--- a/vp8/encoder/ppc/variance_altivec.asm
+++ /dev/null
@@ -1,375 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- .globl vp8_get8x8var_ppc
- .globl vp8_get16x16var_ppc
- .globl vp8_mse16x16_ppc
- .globl vp8_variance16x16_ppc
- .globl vp8_variance16x8_ppc
- .globl vp8_variance8x16_ppc
- .globl vp8_variance8x8_ppc
- .globl vp8_variance4x4_ppc
-
-.macro load_aligned_16 V R O
- lvsl v3, 0, \R ;# permutate value for alignment
-
- lvx v1, 0, \R
- lvx v2, \O, \R
-
- vperm \V, v1, v2, v3
-.endm
-
-.macro prologue
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xffc0
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1, -32(r1) ;# create space on the stack
-
- li r10, 16 ;# load offset and loop counter
-
- vspltisw v7, 0 ;# zero for merging
- vspltisw v8, 0 ;# zero out total to start
- vspltisw v9, 0 ;# zero out total for dif^2
-.endm
-
-.macro epilogue
- addi r1, r1, 32 ;# recover stack
-
- mtspr 256, r11 ;# reset old VRSAVE
-.endm
-
-.macro compute_sum_sse
- ;# Compute sum first. Unpack to so signed subract
- ;# can be used. Only have a half word signed
- ;# subract. Do high, then low.
- vmrghb v2, v7, v4
- vmrghb v3, v7, v5
- vsubshs v2, v2, v3
- vsum4shs v8, v2, v8
-
- vmrglb v2, v7, v4
- vmrglb v3, v7, v5
- vsubshs v2, v2, v3
- vsum4shs v8, v2, v8
-
- ;# Now compute sse.
- vsububs v2, v4, v5
- vsububs v3, v5, v4
- vor v2, v2, v3
-
- vmsumubm v9, v2, v2, v9
-.endm
-
-.macro variance_16 DS loop_label store_sum
-\loop_label:
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v4, r3, r10
- load_aligned_16 v5, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- compute_sum_sse
-
- bdnz \loop_label
-
- vsumsws v8, v8, v7
- vsumsws v9, v9, v7
-
- stvx v8, 0, r1
- lwz r3, 12(r1)
-
- stvx v9, 0, r1
- lwz r4, 12(r1)
-
-.if \store_sum
- stw r3, 0(r8) ;# sum
-.endif
- stw r4, 0(r7) ;# sse
-
- mullw r3, r3, r3 ;# sum*sum
- srlwi r3, r3, \DS ;# (sum*sum) >> DS
- subf r3, r3, r4 ;# sse - ((sum*sum) >> DS)
-.endm
-
-.macro variance_8 DS loop_label store_sum
-\loop_label:
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v4, r3, r10
- load_aligned_16 v5, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v6, r3, r10
- load_aligned_16 v0, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- vmrghb v4, v4, v6
- vmrghb v5, v5, v0
-
- compute_sum_sse
-
- bdnz \loop_label
-
- vsumsws v8, v8, v7
- vsumsws v9, v9, v7
-
- stvx v8, 0, r1
- lwz r3, 12(r1)
-
- stvx v9, 0, r1
- lwz r4, 12(r1)
-
-.if \store_sum
- stw r3, 0(r8) ;# sum
-.endif
- stw r4, 0(r7) ;# sse
-
- mullw r3, r3, r3 ;# sum*sum
- srlwi r3, r3, \DS ;# (sum*sum) >> 8
- subf r3, r3, r4 ;# sse - ((sum*sum) >> 8)
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *SSE
-;# r8 int *Sum
-;#
-;# r3 return value
-vp8_get8x8var_ppc:
-
- prologue
-
- li r9, 4
- mtctr r9
-
- variance_8 6, get8x8var_loop, 1
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *SSE
-;# r8 int *Sum
-;#
-;# r3 return value
-vp8_get16x16var_ppc:
-
- prologue
-
- mtctr r10
-
- variance_16 8, get16x16var_loop, 1
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r 3 return value
-vp8_mse16x16_ppc:
- prologue
-
- mtctr r10
-
-mse16x16_loop:
- ;# only one of the inputs should need to be aligned.
- load_aligned_16 v4, r3, r10
- load_aligned_16 v5, r5, r10
-
- ;# move onto the next line
- add r3, r3, r4
- add r5, r5, r6
-
- ;# Now compute sse.
- vsububs v2, v4, v5
- vsububs v3, v5, v4
- vor v2, v2, v3
-
- vmsumubm v9, v2, v2, v9
-
- bdnz mse16x16_loop
-
- vsumsws v9, v9, v7
-
- stvx v9, 0, r1
- lwz r3, 12(r1)
-
- stvx v9, 0, r1
- lwz r3, 12(r1)
-
- stw r3, 0(r7) ;# sse
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r3 return value
-vp8_variance16x16_ppc:
-
- prologue
-
- mtctr r10
-
- variance_16 8, variance16x16_loop, 0
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r3 return value
-vp8_variance16x8_ppc:
-
- prologue
-
- li r9, 8
- mtctr r9
-
- variance_16 7, variance16x8_loop, 0
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r3 return value
-vp8_variance8x16_ppc:
-
- prologue
-
- li r9, 8
- mtctr r9
-
- variance_8 7, variance8x16_loop, 0
-
- epilogue
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r3 return value
-vp8_variance8x8_ppc:
-
- prologue
-
- li r9, 4
- mtctr r9
-
- variance_8 6, variance8x8_loop, 0
-
- epilogue
-
- blr
-
-.macro transfer_4x4 I P
- lwz r0, 0(\I)
- add \I, \I, \P
-
- lwz r10,0(\I)
- add \I, \I, \P
-
- lwz r8, 0(\I)
- add \I, \I, \P
-
- lwz r9, 0(\I)
-
- stw r0, 0(r1)
- stw r10, 4(r1)
- stw r8, 8(r1)
- stw r9, 12(r1)
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int source_stride
-;# r5 unsigned char *ref_ptr
-;# r6 int recon_stride
-;# r7 unsigned int *sse
-;#
-;# r3 return value
-vp8_variance4x4_ppc:
-
- prologue
-
- transfer_4x4 r3, r4
- lvx v4, 0, r1
-
- transfer_4x4 r5, r6
- lvx v5, 0, r1
-
- compute_sum_sse
-
- vsumsws v8, v8, v7
- vsumsws v9, v9, v7
-
- stvx v8, 0, r1
- lwz r3, 12(r1)
-
- stvx v9, 0, r1
- lwz r4, 12(r1)
-
- stw r4, 0(r7) ;# sse
-
- mullw r3, r3, r3 ;# sum*sum
- srlwi r3, r3, 4 ;# (sum*sum) >> 4
- subf r3, r3, r4 ;# sse - ((sum*sum) >> 4)
-
- epilogue
-
- blr
--- a/vp8/encoder/ppc/variance_subpixel_altivec.asm
+++ /dev/null
@@ -1,865 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
- .globl vp8_sub_pixel_variance4x4_ppc
- .globl vp8_sub_pixel_variance8x8_ppc
- .globl vp8_sub_pixel_variance8x16_ppc
- .globl vp8_sub_pixel_variance16x8_ppc
- .globl vp8_sub_pixel_variance16x16_ppc
-
-.macro load_c V, LABEL, OFF, R0, R1
- lis \R0, \LABEL@ha
- la \R1, \LABEL@l(\R0)
- lvx \V, \OFF, \R1
-.endm
-
-.macro load_vfilter V0, V1
- load_c \V0, vfilter_b, r6, r12, r10
-
- addi r6, r6, 16
- lvx \V1, r6, r10
-.endm
-
-.macro HProlog jump_label
- ;# load up horizontal filter
- slwi. r5, r5, 4 ;# index into horizontal filter array
-
- ;# index to the next set of vectors in the row.
- li r10, 16
-
- ;# downshift by 7 ( divide by 128 ) at the end
- vspltish v19, 7
-
- ;# If there isn't any filtering to be done for the horizontal, then
- ;# just skip to the second pass.
- beq \jump_label
-
- load_c v20, hfilter_b, r5, r12, r0
-
- ;# setup constants
- ;# v14 permutation value for alignment
- load_c v28, b_hperm_b, 0, r12, r0
-
- ;# index to the next set of vectors in the row.
- li r12, 32
-
- ;# rounding added in on the multiply
- vspltisw v21, 8
- vspltisw v18, 3
- vslw v18, v21, v18 ;# 0x00000040000000400000004000000040
-
- slwi. r6, r6, 5 ;# index into vertical filter array
-.endm
-
-;# Filters a horizontal line
-;# expects:
-;# r3 src_ptr
-;# r4 pitch
-;# r10 16
-;# r12 32
-;# v17 perm intput
-;# v18 rounding
-;# v19 shift
-;# v20 filter taps
-;# v21 tmp
-;# v22 tmp
-;# v23 tmp
-;# v24 tmp
-;# v25 tmp
-;# v26 tmp
-;# v27 tmp
-;# v28 perm output
-;#
-
-.macro hfilter_8 V, hp, lp, increment_counter
- lvsl v17, 0, r3 ;# permutate value for alignment
-
- ;# input to filter is 9 bytes wide, output is 8 bytes.
- lvx v21, 0, r3
- lvx v22, r10, r3
-
-.if \increment_counter
- add r3, r3, r4
-.endif
- vperm v21, v21, v22, v17
-
- vperm v24, v21, v21, \hp ;# v20 = 0123 1234 2345 3456
- vperm v25, v21, v21, \lp ;# v21 = 4567 5678 6789 789A
-
- vmsummbm v24, v20, v24, v18
- vmsummbm v25, v20, v25, v18
-
- vpkswus v24, v24, v25 ;# v24 = 0 4 8 C 1 5 9 D (16-bit)
-
- vsrh v24, v24, v19 ;# divide v0, v1 by 128
-
- vpkuhus \V, v24, v24 ;# \V = scrambled 8-bit result
-.endm
-
-.macro vfilter_16 P0 P1
- vmuleub v22, \P0, v20 ;# 64 + 4 positive taps
- vadduhm v22, v18, v22
- vmuloub v23, \P0, v20
- vadduhm v23, v18, v23
-
- vmuleub v24, \P1, v21
- vadduhm v22, v22, v24 ;# Re = evens, saturation unnecessary
- vmuloub v25, \P1, v21
- vadduhm v23, v23, v25 ;# Ro = odds
-
- vsrh v22, v22, v19 ;# divide by 128
- vsrh v23, v23, v19 ;# v16 v17 = evens, odds
- vmrghh \P0, v22, v23 ;# v18 v19 = 16-bit result in order
- vmrglh v23, v22, v23
- vpkuhus \P0, \P0, v23 ;# P0 = 8-bit result
-.endm
-
-.macro compute_sum_sse src, ref, sum, sse, t1, t2, z0
- ;# Compute sum first. Unpack to so signed subract
- ;# can be used. Only have a half word signed
- ;# subract. Do high, then low.
- vmrghb \t1, \z0, \src
- vmrghb \t2, \z0, \ref
- vsubshs \t1, \t1, \t2
- vsum4shs \sum, \t1, \sum
-
- vmrglb \t1, \z0, \src
- vmrglb \t2, \z0, \ref
- vsubshs \t1, \t1, \t2
- vsum4shs \sum, \t1, \sum
-
- ;# Now compute sse.
- vsububs \t1, \src, \ref
- vsububs \t2, \ref, \src
- vor \t1, \t1, \t2
-
- vmsumubm \sse, \t1, \t1, \sse
-.endm
-
-.macro variance_final sum, sse, z0, DS
- vsumsws \sum, \sum, \z0
- vsumsws \sse, \sse, \z0
-
- stvx \sum, 0, r1
- lwz r3, 12(r1)
-
- stvx \sse, 0, r1
- lwz r4, 12(r1)
-
- stw r4, 0(r9) ;# sse
-
- mullw r3, r3, r3 ;# sum*sum
- srlwi r3, r3, \DS ;# (sum*sum) >> 8
- subf r3, r3, r4 ;# sse - ((sum*sum) >> 8)
-.endm
-
-.macro compute_sum_sse_16 V, increment_counter
- load_and_align_16 v16, r7, r8, \increment_counter
- compute_sum_sse \V, v16, v18, v19, v20, v21, v23
-.endm
-
-.macro load_and_align_16 V, R, P, increment_counter
- lvsl v17, 0, \R ;# permutate value for alignment
-
- ;# input to filter is 21 bytes wide, output is 16 bytes.
- ;# input will can span three vectors if not aligned correctly.
- lvx v21, 0, \R
- lvx v22, r10, \R
-
-.if \increment_counter
- add \R, \R, \P
-.endif
-
- vperm \V, v21, v22, v17
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_pixels_per_line
-;# r5 int xoffset
-;# r6 int yoffset
-;# r7 unsigned char *dst_ptr
-;# r8 int dst_pixels_per_line
-;# r9 unsigned int *sse
-;#
-;# r3 return value
-vp8_sub_pixel_variance4x4_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xf830
- ori r12, r12, 0xfff8
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1,-32(r1) ;# create space on the stack
-
- HProlog second_pass_4x4_pre_copy_b
-
- ;# Load up permutation constants
- load_c v10, b_0123_b, 0, r12, r0
- load_c v11, b_4567_b, 0, r12, r0
-
- hfilter_8 v0, v10, v11, 1
- hfilter_8 v1, v10, v11, 1
- hfilter_8 v2, v10, v11, 1
- hfilter_8 v3, v10, v11, 1
-
- ;# Finished filtering main horizontal block. If there is no
- ;# vertical filtering, jump to storing the data. Otherwise
- ;# load up and filter the additional line that is needed
- ;# for the vertical filter.
- beq compute_sum_sse_4x4_b
-
- hfilter_8 v4, v10, v11, 0
-
- b second_pass_4x4_b
-
-second_pass_4x4_pre_copy_b:
- slwi r6, r6, 5 ;# index into vertical filter array
-
- load_and_align_16 v0, r3, r4, 1
- load_and_align_16 v1, r3, r4, 1
- load_and_align_16 v2, r3, r4, 1
- load_and_align_16 v3, r3, r4, 1
- load_and_align_16 v4, r3, r4, 0
-
-second_pass_4x4_b:
- vspltish v20, 8
- vspltish v18, 3
- vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
-
- load_vfilter v20, v21
-
- vfilter_16 v0, v1
- vfilter_16 v1, v2
- vfilter_16 v2, v3
- vfilter_16 v3, v4
-
-compute_sum_sse_4x4_b:
- vspltish v18, 0 ;# sum
- vspltish v19, 0 ;# sse
- vspltish v23, 0 ;# unpack
- li r10, 16
-
- load_and_align_16 v4, r7, r8, 1
- load_and_align_16 v5, r7, r8, 1
- load_and_align_16 v6, r7, r8, 1
- load_and_align_16 v7, r7, r8, 1
-
- vmrghb v0, v0, v1
- vmrghb v1, v2, v3
-
- vmrghb v2, v4, v5
- vmrghb v3, v6, v7
-
- load_c v10, b_hilo_b, 0, r12, r0
-
- vperm v0, v0, v1, v10
- vperm v1, v2, v3, v10
-
- compute_sum_sse v0, v1, v18, v19, v20, v21, v23
-
- variance_final v18, v19, v23, 4
-
- addi r1, r1, 32 ;# recover stack
- mtspr 256, r11 ;# reset old VRSAVE
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_pixels_per_line
-;# r5 int xoffset
-;# r6 int yoffset
-;# r7 unsigned char *dst_ptr
-;# r8 int dst_pixels_per_line
-;# r9 unsigned int *sse
-;#
-;# r3 return value
-vp8_sub_pixel_variance8x8_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xfff0
- ori r12, r12, 0xffff
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1,-32(r1) ;# create space on the stack
-
- HProlog second_pass_8x8_pre_copy_b
-
- ;# Load up permutation constants
- load_c v10, b_0123_b, 0, r12, r0
- load_c v11, b_4567_b, 0, r12, r0
-
- hfilter_8 v0, v10, v11, 1
- hfilter_8 v1, v10, v11, 1
- hfilter_8 v2, v10, v11, 1
- hfilter_8 v3, v10, v11, 1
- hfilter_8 v4, v10, v11, 1
- hfilter_8 v5, v10, v11, 1
- hfilter_8 v6, v10, v11, 1
- hfilter_8 v7, v10, v11, 1
-
- ;# Finished filtering main horizontal block. If there is no
- ;# vertical filtering, jump to storing the data. Otherwise
- ;# load up and filter the additional line that is needed
- ;# for the vertical filter.
- beq compute_sum_sse_8x8_b
-
- hfilter_8 v8, v10, v11, 0
-
- b second_pass_8x8_b
-
-second_pass_8x8_pre_copy_b:
- slwi. r6, r6, 5 ;# index into vertical filter array
-
- load_and_align_16 v0, r3, r4, 1
- load_and_align_16 v1, r3, r4, 1
- load_and_align_16 v2, r3, r4, 1
- load_and_align_16 v3, r3, r4, 1
- load_and_align_16 v4, r3, r4, 1
- load_and_align_16 v5, r3, r4, 1
- load_and_align_16 v6, r3, r4, 1
- load_and_align_16 v7, r3, r4, 1
- load_and_align_16 v8, r3, r4, 0
-
- beq compute_sum_sse_8x8_b
-
-second_pass_8x8_b:
- vspltish v20, 8
- vspltish v18, 3
- vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
-
- load_vfilter v20, v21
-
- vfilter_16 v0, v1
- vfilter_16 v1, v2
- vfilter_16 v2, v3
- vfilter_16 v3, v4
- vfilter_16 v4, v5
- vfilter_16 v5, v6
- vfilter_16 v6, v7
- vfilter_16 v7, v8
-
-compute_sum_sse_8x8_b:
- vspltish v18, 0 ;# sum
- vspltish v19, 0 ;# sse
- vspltish v23, 0 ;# unpack
- li r10, 16
-
- vmrghb v0, v0, v1
- vmrghb v1, v2, v3
- vmrghb v2, v4, v5
- vmrghb v3, v6, v7
-
- load_and_align_16 v4, r7, r8, 1
- load_and_align_16 v5, r7, r8, 1
- load_and_align_16 v6, r7, r8, 1
- load_and_align_16 v7, r7, r8, 1
- load_and_align_16 v8, r7, r8, 1
- load_and_align_16 v9, r7, r8, 1
- load_and_align_16 v10, r7, r8, 1
- load_and_align_16 v11, r7, r8, 0
-
- vmrghb v4, v4, v5
- vmrghb v5, v6, v7
- vmrghb v6, v8, v9
- vmrghb v7, v10, v11
-
- compute_sum_sse v0, v4, v18, v19, v20, v21, v23
- compute_sum_sse v1, v5, v18, v19, v20, v21, v23
- compute_sum_sse v2, v6, v18, v19, v20, v21, v23
- compute_sum_sse v3, v7, v18, v19, v20, v21, v23
-
- variance_final v18, v19, v23, 6
-
- addi r1, r1, 32 ;# recover stack
- mtspr 256, r11 ;# reset old VRSAVE
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_pixels_per_line
-;# r5 int xoffset
-;# r6 int yoffset
-;# r7 unsigned char *dst_ptr
-;# r8 int dst_pixels_per_line
-;# r9 unsigned int *sse
-;#
-;# r3 return value
-vp8_sub_pixel_variance8x16_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xffff
- ori r12, r12, 0xfffc
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1,-32(r1) ;# create space on the stack
-
- HProlog second_pass_8x16_pre_copy_b
-
- ;# Load up permutation constants
- load_c v29, b_0123_b, 0, r12, r0
- load_c v30, b_4567_b, 0, r12, r0
-
- hfilter_8 v0, v29, v30, 1
- hfilter_8 v1, v29, v30, 1
- hfilter_8 v2, v29, v30, 1
- hfilter_8 v3, v29, v30, 1
- hfilter_8 v4, v29, v30, 1
- hfilter_8 v5, v29, v30, 1
- hfilter_8 v6, v29, v30, 1
- hfilter_8 v7, v29, v30, 1
- hfilter_8 v8, v29, v30, 1
- hfilter_8 v9, v29, v30, 1
- hfilter_8 v10, v29, v30, 1
- hfilter_8 v11, v29, v30, 1
- hfilter_8 v12, v29, v30, 1
- hfilter_8 v13, v29, v30, 1
- hfilter_8 v14, v29, v30, 1
- hfilter_8 v15, v29, v30, 1
-
- ;# Finished filtering main horizontal block. If there is no
- ;# vertical filtering, jump to storing the data. Otherwise
- ;# load up and filter the additional line that is needed
- ;# for the vertical filter.
- beq compute_sum_sse_8x16_b
-
- hfilter_8 v16, v29, v30, 0
-
- b second_pass_8x16_b
-
-second_pass_8x16_pre_copy_b:
- slwi. r6, r6, 5 ;# index into vertical filter array
-
- load_and_align_16 v0, r3, r4, 1
- load_and_align_16 v1, r3, r4, 1
- load_and_align_16 v2, r3, r4, 1
- load_and_align_16 v3, r3, r4, 1
- load_and_align_16 v4, r3, r4, 1
- load_and_align_16 v5, r3, r4, 1
- load_and_align_16 v6, r3, r4, 1
- load_and_align_16 v7, r3, r4, 1
- load_and_align_16 v8, r3, r4, 1
- load_and_align_16 v9, r3, r4, 1
- load_and_align_16 v10, r3, r4, 1
- load_and_align_16 v11, r3, r4, 1
- load_and_align_16 v12, r3, r4, 1
- load_and_align_16 v13, r3, r4, 1
- load_and_align_16 v14, r3, r4, 1
- load_and_align_16 v15, r3, r4, 1
- load_and_align_16 v16, r3, r4, 0
-
- beq compute_sum_sse_8x16_b
-
-second_pass_8x16_b:
- vspltish v20, 8
- vspltish v18, 3
- vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
-
- load_vfilter v20, v21
-
- vfilter_16 v0, v1
- vfilter_16 v1, v2
- vfilter_16 v2, v3
- vfilter_16 v3, v4
- vfilter_16 v4, v5
- vfilter_16 v5, v6
- vfilter_16 v6, v7
- vfilter_16 v7, v8
- vfilter_16 v8, v9
- vfilter_16 v9, v10
- vfilter_16 v10, v11
- vfilter_16 v11, v12
- vfilter_16 v12, v13
- vfilter_16 v13, v14
- vfilter_16 v14, v15
- vfilter_16 v15, v16
-
-compute_sum_sse_8x16_b:
- vspltish v18, 0 ;# sum
- vspltish v19, 0 ;# sse
- vspltish v23, 0 ;# unpack
- li r10, 16
-
- vmrghb v0, v0, v1
- vmrghb v1, v2, v3
- vmrghb v2, v4, v5
- vmrghb v3, v6, v7
- vmrghb v4, v8, v9
- vmrghb v5, v10, v11
- vmrghb v6, v12, v13
- vmrghb v7, v14, v15
-
- load_and_align_16 v8, r7, r8, 1
- load_and_align_16 v9, r7, r8, 1
- load_and_align_16 v10, r7, r8, 1
- load_and_align_16 v11, r7, r8, 1
- load_and_align_16 v12, r7, r8, 1
- load_and_align_16 v13, r7, r8, 1
- load_and_align_16 v14, r7, r8, 1
- load_and_align_16 v15, r7, r8, 1
-
- vmrghb v8, v8, v9
- vmrghb v9, v10, v11
- vmrghb v10, v12, v13
- vmrghb v11, v14, v15
-
- compute_sum_sse v0, v8, v18, v19, v20, v21, v23
- compute_sum_sse v1, v9, v18, v19, v20, v21, v23
- compute_sum_sse v2, v10, v18, v19, v20, v21, v23
- compute_sum_sse v3, v11, v18, v19, v20, v21, v23
-
- load_and_align_16 v8, r7, r8, 1
- load_and_align_16 v9, r7, r8, 1
- load_and_align_16 v10, r7, r8, 1
- load_and_align_16 v11, r7, r8, 1
- load_and_align_16 v12, r7, r8, 1
- load_and_align_16 v13, r7, r8, 1
- load_and_align_16 v14, r7, r8, 1
- load_and_align_16 v15, r7, r8, 0
-
- vmrghb v8, v8, v9
- vmrghb v9, v10, v11
- vmrghb v10, v12, v13
- vmrghb v11, v14, v15
-
- compute_sum_sse v4, v8, v18, v19, v20, v21, v23
- compute_sum_sse v5, v9, v18, v19, v20, v21, v23
- compute_sum_sse v6, v10, v18, v19, v20, v21, v23
- compute_sum_sse v7, v11, v18, v19, v20, v21, v23
-
- variance_final v18, v19, v23, 7
-
- addi r1, r1, 32 ;# recover stack
- mtspr 256, r11 ;# reset old VRSAVE
- blr
-
-;# Filters a horizontal line
-;# expects:
-;# r3 src_ptr
-;# r4 pitch
-;# r10 16
-;# r12 32
-;# v17 perm intput
-;# v18 rounding
-;# v19 shift
-;# v20 filter taps
-;# v21 tmp
-;# v22 tmp
-;# v23 tmp
-;# v24 tmp
-;# v25 tmp
-;# v26 tmp
-;# v27 tmp
-;# v28 perm output
-;#
-.macro hfilter_16 V, increment_counter
-
- lvsl v17, 0, r3 ;# permutate value for alignment
-
- ;# input to filter is 21 bytes wide, output is 16 bytes.
- ;# input will can span three vectors if not aligned correctly.
- lvx v21, 0, r3
- lvx v22, r10, r3
- lvx v23, r12, r3
-
-.if \increment_counter
- add r3, r3, r4
-.endif
- vperm v21, v21, v22, v17
- vperm v22, v22, v23, v17 ;# v8 v9 = 21 input pixels left-justified
-
- ;# set 0
- vmsummbm v24, v20, v21, v18 ;# taps times elements
-
- ;# set 1
- vsldoi v23, v21, v22, 1
- vmsummbm v25, v20, v23, v18
-
- ;# set 2
- vsldoi v23, v21, v22, 2
- vmsummbm v26, v20, v23, v18
-
- ;# set 3
- vsldoi v23, v21, v22, 3
- vmsummbm v27, v20, v23, v18
-
- vpkswus v24, v24, v25 ;# v24 = 0 4 8 C 1 5 9 D (16-bit)
- vpkswus v25, v26, v27 ;# v25 = 2 6 A E 3 7 B F
-
- vsrh v24, v24, v19 ;# divide v0, v1 by 128
- vsrh v25, v25, v19
-
- vpkuhus \V, v24, v25 ;# \V = scrambled 8-bit result
- vperm \V, \V, v0, v28 ;# \V = correctly-ordered result
-.endm
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_pixels_per_line
-;# r5 int xoffset
-;# r6 int yoffset
-;# r7 unsigned char *dst_ptr
-;# r8 int dst_pixels_per_line
-;# r9 unsigned int *sse
-;#
-;# r3 return value
-vp8_sub_pixel_variance16x8_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xffff
- ori r12, r12, 0xfff8
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1, -32(r1) ;# create space on the stack
-
- HProlog second_pass_16x8_pre_copy_b
-
- hfilter_16 v0, 1
- hfilter_16 v1, 1
- hfilter_16 v2, 1
- hfilter_16 v3, 1
- hfilter_16 v4, 1
- hfilter_16 v5, 1
- hfilter_16 v6, 1
- hfilter_16 v7, 1
-
- ;# Finished filtering main horizontal block. If there is no
- ;# vertical filtering, jump to storing the data. Otherwise
- ;# load up and filter the additional line that is needed
- ;# for the vertical filter.
- beq compute_sum_sse_16x8_b
-
- hfilter_16 v8, 0
-
- b second_pass_16x8_b
-
-second_pass_16x8_pre_copy_b:
- slwi. r6, r6, 5 ;# index into vertical filter array
-
- load_and_align_16 v0, r3, r4, 1
- load_and_align_16 v1, r3, r4, 1
- load_and_align_16 v2, r3, r4, 1
- load_and_align_16 v3, r3, r4, 1
- load_and_align_16 v4, r3, r4, 1
- load_and_align_16 v5, r3, r4, 1
- load_and_align_16 v6, r3, r4, 1
- load_and_align_16 v7, r3, r4, 1
- load_and_align_16 v8, r3, r4, 1
-
- beq compute_sum_sse_16x8_b
-
-second_pass_16x8_b:
- vspltish v20, 8
- vspltish v18, 3
- vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
-
- load_vfilter v20, v21
-
- vfilter_16 v0, v1
- vfilter_16 v1, v2
- vfilter_16 v2, v3
- vfilter_16 v3, v4
- vfilter_16 v4, v5
- vfilter_16 v5, v6
- vfilter_16 v6, v7
- vfilter_16 v7, v8
-
-compute_sum_sse_16x8_b:
- vspltish v18, 0 ;# sum
- vspltish v19, 0 ;# sse
- vspltish v23, 0 ;# unpack
- li r10, 16
-
- compute_sum_sse_16 v0, 1
- compute_sum_sse_16 v1, 1
- compute_sum_sse_16 v2, 1
- compute_sum_sse_16 v3, 1
- compute_sum_sse_16 v4, 1
- compute_sum_sse_16 v5, 1
- compute_sum_sse_16 v6, 1
- compute_sum_sse_16 v7, 0
-
- variance_final v18, v19, v23, 7
-
- addi r1, r1, 32 ;# recover stack
-
- mtspr 256, r11 ;# reset old VRSAVE
-
- blr
-
- .align 2
-;# r3 unsigned char *src_ptr
-;# r4 int src_pixels_per_line
-;# r5 int xoffset
-;# r6 int yoffset
-;# r7 unsigned char *dst_ptr
-;# r8 int dst_pixels_per_line
-;# r9 unsigned int *sse
-;#
-;# r3 return value
-vp8_sub_pixel_variance16x16_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xffff
- ori r12, r12, 0xfff8
- mtspr 256, r12 ;# set VRSAVE
-
- stwu r1, -32(r1) ;# create space on the stack
-
- HProlog second_pass_16x16_pre_copy_b
-
- hfilter_16 v0, 1
- hfilter_16 v1, 1
- hfilter_16 v2, 1
- hfilter_16 v3, 1
- hfilter_16 v4, 1
- hfilter_16 v5, 1
- hfilter_16 v6, 1
- hfilter_16 v7, 1
- hfilter_16 v8, 1
- hfilter_16 v9, 1
- hfilter_16 v10, 1
- hfilter_16 v11, 1
- hfilter_16 v12, 1
- hfilter_16 v13, 1
- hfilter_16 v14, 1
- hfilter_16 v15, 1
-
- ;# Finished filtering main horizontal block. If there is no
- ;# vertical filtering, jump to storing the data. Otherwise
- ;# load up and filter the additional line that is needed
- ;# for the vertical filter.
- beq compute_sum_sse_16x16_b
-
- hfilter_16 v16, 0
-
- b second_pass_16x16_b
-
-second_pass_16x16_pre_copy_b:
- slwi. r6, r6, 5 ;# index into vertical filter array
-
- load_and_align_16 v0, r3, r4, 1
- load_and_align_16 v1, r3, r4, 1
- load_and_align_16 v2, r3, r4, 1
- load_and_align_16 v3, r3, r4, 1
- load_and_align_16 v4, r3, r4, 1
- load_and_align_16 v5, r3, r4, 1
- load_and_align_16 v6, r3, r4, 1
- load_and_align_16 v7, r3, r4, 1
- load_and_align_16 v8, r3, r4, 1
- load_and_align_16 v9, r3, r4, 1
- load_and_align_16 v10, r3, r4, 1
- load_and_align_16 v11, r3, r4, 1
- load_and_align_16 v12, r3, r4, 1
- load_and_align_16 v13, r3, r4, 1
- load_and_align_16 v14, r3, r4, 1
- load_and_align_16 v15, r3, r4, 1
- load_and_align_16 v16, r3, r4, 0
-
- beq compute_sum_sse_16x16_b
-
-second_pass_16x16_b:
- vspltish v20, 8
- vspltish v18, 3
- vslh v18, v20, v18 ;# 0x0040 0040 0040 0040 0040 0040 0040 0040
-
- load_vfilter v20, v21
-
- vfilter_16 v0, v1
- vfilter_16 v1, v2
- vfilter_16 v2, v3
- vfilter_16 v3, v4
- vfilter_16 v4, v5
- vfilter_16 v5, v6
- vfilter_16 v6, v7
- vfilter_16 v7, v8
- vfilter_16 v8, v9
- vfilter_16 v9, v10
- vfilter_16 v10, v11
- vfilter_16 v11, v12
- vfilter_16 v12, v13
- vfilter_16 v13, v14
- vfilter_16 v14, v15
- vfilter_16 v15, v16
-
-compute_sum_sse_16x16_b:
- vspltish v18, 0 ;# sum
- vspltish v19, 0 ;# sse
- vspltish v23, 0 ;# unpack
- li r10, 16
-
- compute_sum_sse_16 v0, 1
- compute_sum_sse_16 v1, 1
- compute_sum_sse_16 v2, 1
- compute_sum_sse_16 v3, 1
- compute_sum_sse_16 v4, 1
- compute_sum_sse_16 v5, 1
- compute_sum_sse_16 v6, 1
- compute_sum_sse_16 v7, 1
- compute_sum_sse_16 v8, 1
- compute_sum_sse_16 v9, 1
- compute_sum_sse_16 v10, 1
- compute_sum_sse_16 v11, 1
- compute_sum_sse_16 v12, 1
- compute_sum_sse_16 v13, 1
- compute_sum_sse_16 v14, 1
- compute_sum_sse_16 v15, 0
-
- variance_final v18, v19, v23, 8
-
- addi r1, r1, 32 ;# recover stack
-
- mtspr 256, r11 ;# reset old VRSAVE
-
- blr
-
- .data
-
- .align 4
-hfilter_b:
- .byte 128, 0, 0, 0,128, 0, 0, 0,128, 0, 0, 0,128, 0, 0, 0
- .byte 112, 16, 0, 0,112, 16, 0, 0,112, 16, 0, 0,112, 16, 0, 0
- .byte 96, 32, 0, 0, 96, 32, 0, 0, 96, 32, 0, 0, 96, 32, 0, 0
- .byte 80, 48, 0, 0, 80, 48, 0, 0, 80, 48, 0, 0, 80, 48, 0, 0
- .byte 64, 64, 0, 0, 64, 64, 0, 0, 64, 64, 0, 0, 64, 64, 0, 0
- .byte 48, 80, 0, 0, 48, 80, 0, 0, 48, 80, 0, 0, 48, 80, 0, 0
- .byte 32, 96, 0, 0, 32, 96, 0, 0, 32, 96, 0, 0, 32, 96, 0, 0
- .byte 16,112, 0, 0, 16,112, 0, 0, 16,112, 0, 0, 16,112, 0, 0
-
- .align 4
-vfilter_b:
- .byte 128,128,128,128,128,128,128,128,128,128,128,128,128,128,128,128
- .byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
- .byte 112,112,112,112,112,112,112,112,112,112,112,112,112,112,112,112
- .byte 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
- .byte 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96
- .byte 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
- .byte 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80
- .byte 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48
- .byte 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
- .byte 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
- .byte 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48
- .byte 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80, 80
- .byte 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
- .byte 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96, 96
- .byte 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
- .byte 112,112,112,112,112,112,112,112,112,112,112,112,112,112,112,112
-
- .align 4
-b_hperm_b:
- .byte 0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15
-
- .align 4
-b_0123_b:
- .byte 0, 1, 2, 3, 1, 2, 3, 4, 2, 3, 4, 5, 3, 4, 5, 6
-
- .align 4
-b_4567_b:
- .byte 4, 5, 6, 7, 5, 6, 7, 8, 6, 7, 8, 9, 7, 8, 9, 10
-
-b_hilo_b:
- .byte 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -28,7 +28,7 @@
#include "vp8/common/quant_common.h"
#include "encodemb.h"
#include "quantize.h"
-#include "variance.h"
+#include "vp8/common/variance.h"
#include "mcomp.h"
#include "rdopt.h"
#include "vpx_mem/vpx_mem.h"
--- a/vp8/encoder/sad_c.c
+++ /dev/null
@@ -1,395 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include <stdlib.h>
-#include "vpx_config.h"
-#include "vpx/vpx_integer.h"
-
-static __inline
-unsigned int sad_mx_n_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad,
- int m,
- int n)
-{-
- int r, c;
- unsigned int sad = 0;
-
- for (r = 0; r < n; r++)
- {- for (c = 0; c < m; c++)
- {- sad += abs(src_ptr[c] - ref_ptr[c]);
- }
-
- if (sad > max_sad)
- break;
-
- src_ptr += src_stride;
- ref_ptr += ref_stride;
- }
-
- return sad;
-}
-
-/* max_sad is provided as an optional optimization point. Alternative
- * implementations of these functions are not required to check it.
- */
-
-unsigned int vp8_sad16x16_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad)
-{-
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 16);
-}
-
-
-unsigned int vp8_sad8x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad)
-{-
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 8);
-}
-
-
-unsigned int vp8_sad16x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad)
-{-
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 16, 8);
-
-}
-
-
-unsigned int vp8_sad8x16_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad)
-{-
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 8, 16);
-}
-
-
-unsigned int vp8_sad4x4_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad)
-{-
- return sad_mx_n_c(src_ptr, src_stride, ref_ptr, ref_stride, max_sad, 4, 4);
-}
-
-void vp8_sad16x16x3_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad16x16x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
-)
-{- sad_array[0] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
- sad_array[3] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
- sad_array[4] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
- sad_array[5] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
- sad_array[6] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
- sad_array[7] = (unsigned short)vp8_sad16x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad16x8x3_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad16x8x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
-)
-{- sad_array[0] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
- sad_array[3] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
- sad_array[4] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
- sad_array[5] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
- sad_array[6] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
- sad_array[7] = (unsigned short)vp8_sad16x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x8x3_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x8x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
-)
-{- sad_array[0] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
- sad_array[3] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
- sad_array[4] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
- sad_array[5] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
- sad_array[6] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
- sad_array[7] = (unsigned short)vp8_sad8x8_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x16x3_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x16x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
-)
-{- sad_array[0] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
- sad_array[3] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
- sad_array[4] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
- sad_array[5] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
- sad_array[6] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
- sad_array[7] = (unsigned short)vp8_sad8x16_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad4x4x3_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad4x4x8_c(
- const unsigned char *src_ptr,
- int src_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
-)
-{- sad_array[0] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr , ref_stride, 0x7fffffff);
- sad_array[1] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 1, ref_stride, 0x7fffffff);
- sad_array[2] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 2, ref_stride, 0x7fffffff);
- sad_array[3] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 3 , ref_stride, 0x7fffffff);
- sad_array[4] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 4, ref_stride, 0x7fffffff);
- sad_array[5] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 5, ref_stride, 0x7fffffff);
- sad_array[6] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 6 , ref_stride, 0x7fffffff);
- sad_array[7] = (unsigned short)vp8_sad4x4_c(src_ptr, src_stride, ref_ptr + 7, ref_stride, 0x7fffffff);
-}
-
-void vp8_sad16x16x4d_c(
- const unsigned char *src_ptr,
- int src_stride,
- unsigned char *ref_ptr[],
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
- sad_array[3] = vp8_sad16x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
-}
-
-void vp8_sad16x8x4d_c(
- const unsigned char *src_ptr,
- int src_stride,
- unsigned char *ref_ptr[],
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
- sad_array[3] = vp8_sad16x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x8x4d_c(
- const unsigned char *src_ptr,
- int src_stride,
- unsigned char *ref_ptr[],
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
- sad_array[3] = vp8_sad8x8_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
-}
-
-void vp8_sad8x16x4d_c(
- const unsigned char *src_ptr,
- int src_stride,
- unsigned char *ref_ptr[],
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
- sad_array[3] = vp8_sad8x16_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
-}
-
-void vp8_sad4x4x4d_c(
- const unsigned char *src_ptr,
- int src_stride,
- unsigned char *ref_ptr[],
- int ref_stride,
- unsigned int *sad_array
-)
-{- sad_array[0] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[0], ref_stride, 0x7fffffff);
- sad_array[1] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[1], ref_stride, 0x7fffffff);
- sad_array[2] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[2], ref_stride, 0x7fffffff);
- sad_array[3] = vp8_sad4x4_c(src_ptr, src_stride, ref_ptr[3], ref_stride, 0x7fffffff);
-}
-
-/* Copy 2 macroblocks to a buffer */
-void vp8_copy32xn_c(
- unsigned char *src_ptr,
- int src_stride,
- unsigned char *dst_ptr,
- int dst_stride,
- int height)
-{- int r;
-
- for (r = 0; r < height; r++)
- {-#if !(CONFIG_FAST_UNALIGNED)
- dst_ptr[0] = src_ptr[0];
- dst_ptr[1] = src_ptr[1];
- dst_ptr[2] = src_ptr[2];
- dst_ptr[3] = src_ptr[3];
- dst_ptr[4] = src_ptr[4];
- dst_ptr[5] = src_ptr[5];
- dst_ptr[6] = src_ptr[6];
- dst_ptr[7] = src_ptr[7];
- dst_ptr[8] = src_ptr[8];
- dst_ptr[9] = src_ptr[9];
- dst_ptr[10] = src_ptr[10];
- dst_ptr[11] = src_ptr[11];
- dst_ptr[12] = src_ptr[12];
- dst_ptr[13] = src_ptr[13];
- dst_ptr[14] = src_ptr[14];
- dst_ptr[15] = src_ptr[15];
- dst_ptr[16] = src_ptr[16];
- dst_ptr[17] = src_ptr[17];
- dst_ptr[18] = src_ptr[18];
- dst_ptr[19] = src_ptr[19];
- dst_ptr[20] = src_ptr[20];
- dst_ptr[21] = src_ptr[21];
- dst_ptr[22] = src_ptr[22];
- dst_ptr[23] = src_ptr[23];
- dst_ptr[24] = src_ptr[24];
- dst_ptr[25] = src_ptr[25];
- dst_ptr[26] = src_ptr[26];
- dst_ptr[27] = src_ptr[27];
- dst_ptr[28] = src_ptr[28];
- dst_ptr[29] = src_ptr[29];
- dst_ptr[30] = src_ptr[30];
- dst_ptr[31] = src_ptr[31];
-#else
- ((uint32_t *)dst_ptr)[0] = ((uint32_t *)src_ptr)[0] ;
- ((uint32_t *)dst_ptr)[1] = ((uint32_t *)src_ptr)[1] ;
- ((uint32_t *)dst_ptr)[2] = ((uint32_t *)src_ptr)[2] ;
- ((uint32_t *)dst_ptr)[3] = ((uint32_t *)src_ptr)[3] ;
- ((uint32_t *)dst_ptr)[4] = ((uint32_t *)src_ptr)[4] ;
- ((uint32_t *)dst_ptr)[5] = ((uint32_t *)src_ptr)[5] ;
- ((uint32_t *)dst_ptr)[6] = ((uint32_t *)src_ptr)[6] ;
- ((uint32_t *)dst_ptr)[7] = ((uint32_t *)src_ptr)[7] ;
-#endif
- src_ptr += src_stride;
- dst_ptr += dst_stride;
-
- }
-}
--- a/vp8/encoder/variance.h
+++ /dev/null
@@ -1,115 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#ifndef VARIANCE_H
-#define VARIANCE_H
-
-typedef unsigned int(*vp8_sad_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int max_sad
- );
-
-typedef void (*vp8_copy32xn_fn_t)(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- int n);
-
-typedef void (*vp8_sad_multi_fn_t)(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sad_array);
-
-typedef void (*vp8_sad_multi1_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned short *sad_array
- );
-
-typedef void (*vp8_sad_multi_d_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- unsigned char *ref_ptr[4],
- int ref_stride,
- unsigned int *sad_array
- );
-
-typedef unsigned int (*vp8_variance_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride,
- unsigned int *sse
- );
-
-typedef unsigned int (*vp8_subpixvariance_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- int xoffset,
- int yoffset,
- const unsigned char *ref_ptr,
- int Refstride,
- unsigned int *sse
- );
-
-typedef void (*vp8_ssimpf_fn_t)
- (
- unsigned char *s,
- int sp,
- unsigned char *r,
- int rp,
- unsigned long *sum_s,
- unsigned long *sum_r,
- unsigned long *sum_sq_s,
- unsigned long *sum_sq_r,
- unsigned long *sum_sxr
- );
-
-typedef unsigned int (*vp8_getmbss_fn_t)(const short *);
-
-typedef unsigned int (*vp8_get16x16prederror_fn_t)
- (
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int ref_stride
- );
-
-typedef struct variance_vtable
-{- vp8_sad_fn_t sdf;
- vp8_variance_fn_t vf;
- vp8_subpixvariance_fn_t svf;
- vp8_variance_fn_t svf_halfpix_h;
- vp8_variance_fn_t svf_halfpix_v;
- vp8_variance_fn_t svf_halfpix_hv;
- vp8_sad_multi_fn_t sdx3f;
- vp8_sad_multi1_fn_t sdx8f;
- vp8_sad_multi_d_fn_t sdx4df;
-#if ARCH_X86 || ARCH_X86_64
- vp8_copy32xn_fn_t copymem;
-#endif
-} vp8_variance_fn_ptr_t;
-
-#endif
--- a/vp8/encoder/variance_c.c
+++ /dev/null
@@ -1,458 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "variance.h"
-#include "vp8/common/filter.h"
-
-
-unsigned int vp8_get_mb_ss_c
-(
- const short *src_ptr
-)
-{- unsigned int i = 0, sum = 0;
-
- do
- {- sum += (src_ptr[i] * src_ptr[i]);
- i++;
- }
- while (i < 256);
-
- return sum;
-}
-
-
-static void variance(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- int w,
- int h,
- unsigned int *sse,
- int *sum)
-{- int i, j;
- int diff;
-
- *sum = 0;
- *sse = 0;
-
- for (i = 0; i < h; i++)
- {- for (j = 0; j < w; j++)
- {- diff = src_ptr[j] - ref_ptr[j];
- *sum += diff;
- *sse += diff * diff;
- }
-
- src_ptr += source_stride;
- ref_ptr += recon_stride;
- }
-}
-
-
-unsigned int vp8_variance16x16_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 8));
-}
-
-unsigned int vp8_variance8x16_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 16, &var, &avg);
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 7));
-}
-
-unsigned int vp8_variance16x8_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 8, &var, &avg);
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 7));
-}
-
-
-unsigned int vp8_variance8x8_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 8, 8, &var, &avg);
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 6));
-}
-
-unsigned int vp8_variance4x4_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 4, 4, &var, &avg);
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 4));
-}
-
-
-unsigned int vp8_mse16x16_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
- variance(src_ptr, source_stride, ref_ptr, recon_stride, 16, 16, &var, &avg);
- *sse = var;
- return var;
-}
-
-
-/****************************************************************************
- *
- * ROUTINE : filter_block2d_bil_first_pass
- *
- * INPUTS : UINT8 *src_ptr : Pointer to source block.
- * UINT32 src_pixels_per_line : Stride of input block.
- * UINT32 pixel_step : Offset between filter input samples (see notes).
- * UINT32 output_height : Input block height.
- * UINT32 output_width : Input block width.
- * INT32 *vp8_filter : Array of 2 bi-linear filter taps.
- *
- * OUTPUTS : INT32 *output_ptr : Pointer to filtered block.
- *
- * RETURNS : void
- *
- * FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
- * either horizontal or vertical direction to produce the
- * filtered output block. Used to implement first-pass
- * of 2-D separable filter.
- *
- * SPECIAL NOTES : Produces INT32 output to retain precision for next pass.
- * Two filter taps should sum to VP8_FILTER_WEIGHT.
- * pixel_step defines whether the filter is applied
- * horizontally (pixel_step=1) or vertically (pixel_step=stride).
- * It defines the offset required to move from one input
- * to the next.
- *
- ****************************************************************************/
-static void var_filter_block2d_bil_first_pass
-(
- const unsigned char *src_ptr,
- unsigned short *output_ptr,
- unsigned int src_pixels_per_line,
- int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- const short *vp8_filter
-)
-{- unsigned int i, j;
-
- for (i = 0; i < output_height; i++)
- {- for (j = 0; j < output_width; j++)
- {- // Apply bilinear filter
- output_ptr[j] = (((int)src_ptr[0] * vp8_filter[0]) +
- ((int)src_ptr[pixel_step] * vp8_filter[1]) +
- (VP8_FILTER_WEIGHT / 2)) >> VP8_FILTER_SHIFT;
- src_ptr++;
- }
-
- // Next row...
- src_ptr += src_pixels_per_line - output_width;
- output_ptr += output_width;
- }
-}
-
-/****************************************************************************
- *
- * ROUTINE : filter_block2d_bil_second_pass
- *
- * INPUTS : INT32 *src_ptr : Pointer to source block.
- * UINT32 src_pixels_per_line : Stride of input block.
- * UINT32 pixel_step : Offset between filter input samples (see notes).
- * UINT32 output_height : Input block height.
- * UINT32 output_width : Input block width.
- * INT32 *vp8_filter : Array of 2 bi-linear filter taps.
- *
- * OUTPUTS : UINT16 *output_ptr : Pointer to filtered block.
- *
- * RETURNS : void
- *
- * FUNCTION : Applies a 1-D 2-tap bi-linear filter to the source block in
- * either horizontal or vertical direction to produce the
- * filtered output block. Used to implement second-pass
- * of 2-D separable filter.
- *
- * SPECIAL NOTES : Requires 32-bit input as produced by filter_block2d_bil_first_pass.
- * Two filter taps should sum to VP8_FILTER_WEIGHT.
- * pixel_step defines whether the filter is applied
- * horizontally (pixel_step=1) or vertically (pixel_step=stride).
- * It defines the offset required to move from one input
- * to the next.
- *
- ****************************************************************************/
-static void var_filter_block2d_bil_second_pass
-(
- const unsigned short *src_ptr,
- unsigned char *output_ptr,
- unsigned int src_pixels_per_line,
- unsigned int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- const short *vp8_filter
-)
-{- unsigned int i, j;
- int Temp;
-
- for (i = 0; i < output_height; i++)
- {- for (j = 0; j < output_width; j++)
- {- // Apply filter
- Temp = ((int)src_ptr[0] * vp8_filter[0]) +
- ((int)src_ptr[pixel_step] * vp8_filter[1]) +
- (VP8_FILTER_WEIGHT / 2);
- output_ptr[j] = (unsigned int)(Temp >> VP8_FILTER_SHIFT);
- src_ptr++;
- }
-
- // Next row...
- src_ptr += src_pixels_per_line - output_width;
- output_ptr += output_width;
- }
-}
-
-
-unsigned int vp8_sub_pixel_variance4x4_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned char temp2[20*16];
- const short *HFilter, *VFilter;
- unsigned short FData3[5*4]; // Temp data bufffer used in filtering
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- // First filter 1d Horizontal
- var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 5, 4, HFilter);
-
- // Now filter Verticaly
- var_filter_block2d_bil_second_pass(FData3, temp2, 4, 4, 4, 4, VFilter);
-
- return vp8_variance4x4_c(temp2, 4, dst_ptr, dst_pixels_per_line, sse);
-}
-
-
-unsigned int vp8_sub_pixel_variance8x8_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short FData3[9*8]; // Temp data bufffer used in filtering
- unsigned char temp2[20*16];
- const short *HFilter, *VFilter;
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 8, HFilter);
- var_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 8, 8, VFilter);
-
- return vp8_variance8x8_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
-}
-
-unsigned int vp8_sub_pixel_variance16x16_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short FData3[17*16]; // Temp data bufffer used in filtering
- unsigned char temp2[20*16];
- const short *HFilter, *VFilter;
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 16, HFilter);
- var_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 16, 16, VFilter);
-
- return vp8_variance16x16_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_h_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 0,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_v_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 0, 4,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_hv_c(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_c(src_ptr, source_stride, 4, 4,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_sub_pixel_mse16x16_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- vp8_sub_pixel_variance16x16_c(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
- return *sse;
-}
-
-unsigned int vp8_sub_pixel_variance16x8_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short FData3[16*9]; // Temp data bufffer used in filtering
- unsigned char temp2[20*16];
- const short *HFilter, *VFilter;
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
- var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 9, 16, HFilter);
- var_filter_block2d_bil_second_pass(FData3, temp2, 16, 16, 8, 16, VFilter);
-
- return vp8_variance16x8_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
-}
-
-unsigned int vp8_sub_pixel_variance8x16_c
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- unsigned short FData3[9*16]; // Temp data bufffer used in filtering
- unsigned char temp2[20*16];
- const short *HFilter, *VFilter;
-
-
- HFilter = vp8_bilinear_filters[xoffset];
- VFilter = vp8_bilinear_filters[yoffset];
-
-
- var_filter_block2d_bil_first_pass(src_ptr, FData3, src_pixels_per_line, 1, 17, 8, HFilter);
- var_filter_block2d_bil_second_pass(FData3, temp2, 8, 8, 16, 8, VFilter);
-
- return vp8_variance8x16_c(temp2, 8, dst_ptr, dst_pixels_per_line, sse);
-}
--- a/vp8/encoder/x86/sad_mmx.asm
+++ /dev/null
@@ -1,427 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-global sym(vp8_sad16x16_mmx)
-global sym(vp8_sad8x16_mmx)
-global sym(vp8_sad8x8_mmx)
-global sym(vp8_sad4x4_mmx)
-global sym(vp8_sad16x8_mmx)
-
-;unsigned int vp8_sad16x16_mmx(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-sym(vp8_sad16x16_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rax*8]
-
- lea rcx, [rcx+rax*8]
- pxor mm7, mm7
-
- pxor mm6, mm6
-
-.x16x16sad_mmx_loop:
-
- movq mm0, QWORD PTR [rsi]
- movq mm2, QWORD PTR [rsi+8]
-
- movq mm1, QWORD PTR [rdi]
- movq mm3, QWORD PTR [rdi+8]
-
- movq mm4, mm0
- movq mm5, mm2
-
- psubusb mm0, mm1
- psubusb mm1, mm4
-
- psubusb mm2, mm3
- psubusb mm3, mm5
-
- por mm0, mm1
- por mm2, mm3
-
- movq mm1, mm0
- movq mm3, mm2
-
- punpcklbw mm0, mm6
- punpcklbw mm2, mm6
-
- punpckhbw mm1, mm6
- punpckhbw mm3, mm6
-
- paddw mm0, mm2
- paddw mm1, mm3
-
-
- lea rsi, [rsi+rax]
- add rdi, rdx
-
- paddw mm7, mm0
- paddw mm7, mm1
-
- cmp rsi, rcx
- jne .x16x16sad_mmx_loop
-
-
- movq mm0, mm7
-
- punpcklwd mm0, mm6
- punpckhwd mm7, mm6
-
- paddw mm0, mm7
- movq mm7, mm0
-
-
- psrlq mm0, 32
- paddw mm7, mm0
-
- movq rax, mm7
-
- pop rdi
- pop rsi
- mov rsp, rbp
- ; begin epilog
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad8x16_mmx(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-sym(vp8_sad8x16_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rax*8]
-
- lea rcx, [rcx+rax*8]
- pxor mm7, mm7
-
- pxor mm6, mm6
-
-.x8x16sad_mmx_loop:
-
- movq mm0, QWORD PTR [rsi]
- movq mm1, QWORD PTR [rdi]
-
- movq mm2, mm0
- psubusb mm0, mm1
-
- psubusb mm1, mm2
- por mm0, mm1
-
- movq mm2, mm0
- punpcklbw mm0, mm6
-
- punpckhbw mm2, mm6
- lea rsi, [rsi+rax]
-
- add rdi, rdx
- paddw mm7, mm0
-
- paddw mm7, mm2
- cmp rsi, rcx
-
- jne .x8x16sad_mmx_loop
-
- movq mm0, mm7
- punpcklwd mm0, mm6
-
- punpckhwd mm7, mm6
- paddw mm0, mm7
-
- movq mm7, mm0
- psrlq mm0, 32
-
- paddw mm7, mm0
- movq rax, mm7
-
- pop rdi
- pop rsi
- mov rsp, rbp
- ; begin epilog
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad8x8_mmx(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-sym(vp8_sad8x8_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rax*8]
- pxor mm7, mm7
-
- pxor mm6, mm6
-
-.x8x8sad_mmx_loop:
-
- movq mm0, QWORD PTR [rsi]
- movq mm1, QWORD PTR [rdi]
-
- movq mm2, mm0
- psubusb mm0, mm1
-
- psubusb mm1, mm2
- por mm0, mm1
-
- movq mm2, mm0
- punpcklbw mm0, mm6
-
- punpckhbw mm2, mm6
- paddw mm0, mm2
-
- lea rsi, [rsi+rax]
- add rdi, rdx
-
- paddw mm7, mm0
- cmp rsi, rcx
-
- jne .x8x8sad_mmx_loop
-
- movq mm0, mm7
- punpcklwd mm0, mm6
-
- punpckhwd mm7, mm6
- paddw mm0, mm7
-
- movq mm7, mm0
- psrlq mm0, 32
-
- paddw mm7, mm0
- movq rax, mm7
-
- pop rdi
- pop rsi
- mov rsp, rbp
- ; begin epilog
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad4x4_mmx(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-sym(vp8_sad4x4_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- movd mm0, DWORD PTR [rsi]
- movd mm1, DWORD PTR [rdi]
-
- movd mm2, DWORD PTR [rsi+rax]
- movd mm3, DWORD PTR [rdi+rdx]
-
- punpcklbw mm0, mm2
- punpcklbw mm1, mm3
-
- movq mm2, mm0
- psubusb mm0, mm1
-
- psubusb mm1, mm2
- por mm0, mm1
-
- movq mm2, mm0
- pxor mm3, mm3
-
- punpcklbw mm0, mm3
- punpckhbw mm2, mm3
-
- paddw mm0, mm2
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- movd mm4, DWORD PTR [rsi]
- movd mm5, DWORD PTR [rdi]
-
- movd mm6, DWORD PTR [rsi+rax]
- movd mm7, DWORD PTR [rdi+rdx]
-
- punpcklbw mm4, mm6
- punpcklbw mm5, mm7
-
- movq mm6, mm4
- psubusb mm4, mm5
-
- psubusb mm5, mm6
- por mm4, mm5
-
- movq mm5, mm4
- punpcklbw mm4, mm3
-
- punpckhbw mm5, mm3
- paddw mm4, mm5
-
- paddw mm0, mm4
- movq mm1, mm0
-
- punpcklwd mm0, mm3
- punpckhwd mm1, mm3
-
- paddw mm0, mm1
- movq mm1, mm0
-
- psrlq mm0, 32
- paddw mm0, mm1
-
- movq rax, mm0
-
- pop rdi
- pop rsi
- mov rsp, rbp
- ; begin epilog
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad16x8_mmx(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-sym(vp8_sad16x8_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rax*8]
- pxor mm7, mm7
-
- pxor mm6, mm6
-
-.x16x8sad_mmx_loop:
-
- movq mm0, [rsi]
- movq mm1, [rdi]
-
- movq mm2, [rsi+8]
- movq mm3, [rdi+8]
-
- movq mm4, mm0
- movq mm5, mm2
-
- psubusb mm0, mm1
- psubusb mm1, mm4
-
- psubusb mm2, mm3
- psubusb mm3, mm5
-
- por mm0, mm1
- por mm2, mm3
-
- movq mm1, mm0
- movq mm3, mm2
-
- punpcklbw mm0, mm6
- punpckhbw mm1, mm6
-
- punpcklbw mm2, mm6
- punpckhbw mm3, mm6
-
-
- paddw mm0, mm2
- paddw mm1, mm3
-
- paddw mm0, mm1
- lea rsi, [rsi+rax]
-
- add rdi, rdx
- paddw mm7, mm0
-
- cmp rsi, rcx
- jne .x16x8sad_mmx_loop
-
- movq mm0, mm7
- punpcklwd mm0, mm6
-
- punpckhwd mm7, mm6
- paddw mm0, mm7
-
- movq mm7, mm0
- psrlq mm0, 32
-
- paddw mm7, mm0
- movq rax, mm7
-
- pop rdi
- pop rsi
- mov rsp, rbp
- ; begin epilog
- UNSHADOW_ARGS
- pop rbp
- ret
--- a/vp8/encoder/x86/sad_sse2.asm
+++ /dev/null
@@ -1,410 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-;unsigned int vp8_sad16x16_wmt(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-global sym(vp8_sad16x16_wmt)
-sym(vp8_sad16x16_wmt):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- SAVE_XMM 6
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rax*8]
-
- lea rcx, [rcx+rax*8]
- pxor xmm6, xmm6
-
-.x16x16sad_wmt_loop:
-
- movq xmm0, QWORD PTR [rsi]
- movq xmm2, QWORD PTR [rsi+8]
-
- movq xmm1, QWORD PTR [rdi]
- movq xmm3, QWORD PTR [rdi+8]
-
- movq xmm4, QWORD PTR [rsi+rax]
- movq xmm5, QWORD PTR [rdi+rdx]
-
-
- punpcklbw xmm0, xmm2
- punpcklbw xmm1, xmm3
-
- psadbw xmm0, xmm1
- movq xmm2, QWORD PTR [rsi+rax+8]
-
- movq xmm3, QWORD PTR [rdi+rdx+8]
- lea rsi, [rsi+rax*2]
-
- lea rdi, [rdi+rdx*2]
- punpcklbw xmm4, xmm2
-
- punpcklbw xmm5, xmm3
- psadbw xmm4, xmm5
-
- paddw xmm6, xmm0
- paddw xmm6, xmm4
-
- cmp rsi, rcx
- jne .x16x16sad_wmt_loop
-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movq rax, xmm0
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;unsigned int vp8_sad8x16_wmt(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int max_sad)
-global sym(vp8_sad8x16_wmt)
-sym(vp8_sad8x16_wmt):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rbx
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rbx, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rbx*8]
-
- lea rcx, [rcx+rbx*8]
- pxor mm7, mm7
-
-.x8x16sad_wmt_loop:
-
- movq rax, mm7
- cmp eax, arg(4)
- jg .x8x16sad_wmt_early_exit
-
- movq mm0, QWORD PTR [rsi]
- movq mm1, QWORD PTR [rdi]
-
- movq mm2, QWORD PTR [rsi+rbx]
- movq mm3, QWORD PTR [rdi+rdx]
-
- psadbw mm0, mm1
- psadbw mm2, mm3
-
- lea rsi, [rsi+rbx*2]
- lea rdi, [rdi+rdx*2]
-
- paddw mm7, mm0
- paddw mm7, mm2
-
- cmp rsi, rcx
- jne .x8x16sad_wmt_loop
-
- movq rax, mm7
-
-.x8x16sad_wmt_early_exit:
-
- ; begin epilog
- pop rdi
- pop rsi
- pop rbx
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad8x8_wmt(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-global sym(vp8_sad8x8_wmt)
-sym(vp8_sad8x8_wmt):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rbx
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rbx, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rbx*8]
- pxor mm7, mm7
-
-.x8x8sad_wmt_loop:
-
- movq rax, mm7
- cmp eax, arg(4)
- jg .x8x8sad_wmt_early_exit
-
- movq mm0, QWORD PTR [rsi]
- movq mm1, QWORD PTR [rdi]
-
- psadbw mm0, mm1
- lea rsi, [rsi+rbx]
-
- add rdi, rdx
- paddw mm7, mm0
-
- cmp rsi, rcx
- jne .x8x8sad_wmt_loop
-
- movq rax, mm7
-.x8x8sad_wmt_early_exit:
-
- ; begin epilog
- pop rdi
- pop rsi
- pop rbx
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;unsigned int vp8_sad4x4_wmt(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-global sym(vp8_sad4x4_wmt)
-sym(vp8_sad4x4_wmt):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- movd mm0, DWORD PTR [rsi]
- movd mm1, DWORD PTR [rdi]
-
- movd mm2, DWORD PTR [rsi+rax]
- movd mm3, DWORD PTR [rdi+rdx]
-
- punpcklbw mm0, mm2
- punpcklbw mm1, mm3
-
- psadbw mm0, mm1
- lea rsi, [rsi+rax*2]
-
- lea rdi, [rdi+rdx*2]
- movd mm4, DWORD PTR [rsi]
-
- movd mm5, DWORD PTR [rdi]
- movd mm6, DWORD PTR [rsi+rax]
-
- movd mm7, DWORD PTR [rdi+rdx]
- punpcklbw mm4, mm6
-
- punpcklbw mm5, mm7
- psadbw mm4, mm5
-
- paddw mm0, mm4
- movq rax, mm0
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_sad16x8_wmt(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride)
-global sym(vp8_sad16x8_wmt)
-sym(vp8_sad16x8_wmt):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rbx
- push rsi
- push rdi
- ; end prolog
-
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rbx, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- lea rcx, [rsi+rbx*8]
- pxor mm7, mm7
-
-.x16x8sad_wmt_loop:
-
- movq rax, mm7
- cmp eax, arg(4)
- jg .x16x8sad_wmt_early_exit
-
- movq mm0, QWORD PTR [rsi]
- movq mm2, QWORD PTR [rsi+8]
-
- movq mm1, QWORD PTR [rdi]
- movq mm3, QWORD PTR [rdi+8]
-
- movq mm4, QWORD PTR [rsi+rbx]
- movq mm5, QWORD PTR [rdi+rdx]
-
- psadbw mm0, mm1
- psadbw mm2, mm3
-
- movq mm1, QWORD PTR [rsi+rbx+8]
- movq mm3, QWORD PTR [rdi+rdx+8]
-
- psadbw mm4, mm5
- psadbw mm1, mm3
-
- lea rsi, [rsi+rbx*2]
- lea rdi, [rdi+rdx*2]
-
- paddw mm0, mm2
- paddw mm4, mm1
-
- paddw mm7, mm0
- paddw mm7, mm4
-
- cmp rsi, rcx
- jne .x16x8sad_wmt_loop
-
- movq rax, mm7
-
-.x16x8sad_wmt_early_exit:
-
- ; begin epilog
- pop rdi
- pop rsi
- pop rbx
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_copy32xn_sse2(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *dst_ptr,
-; int dst_stride,
-; int height);
-global sym(vp8_copy32xn_sse2)
-sym(vp8_copy32xn_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- SAVE_XMM 7
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;dst_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;dst_stride
- movsxd rcx, dword ptr arg(4) ;height
-
-.block_copy_sse2_loopx4:
- movdqu xmm0, XMMWORD PTR [rsi]
- movdqu xmm1, XMMWORD PTR [rsi + 16]
- movdqu xmm2, XMMWORD PTR [rsi + rax]
- movdqu xmm3, XMMWORD PTR [rsi + rax + 16]
-
- lea rsi, [rsi+rax*2]
-
- movdqu xmm4, XMMWORD PTR [rsi]
- movdqu xmm5, XMMWORD PTR [rsi + 16]
- movdqu xmm6, XMMWORD PTR [rsi + rax]
- movdqu xmm7, XMMWORD PTR [rsi + rax + 16]
-
- lea rsi, [rsi+rax*2]
-
- movdqa XMMWORD PTR [rdi], xmm0
- movdqa XMMWORD PTR [rdi + 16], xmm1
- movdqa XMMWORD PTR [rdi + rdx], xmm2
- movdqa XMMWORD PTR [rdi + rdx + 16], xmm3
-
- lea rdi, [rdi+rdx*2]
-
- movdqa XMMWORD PTR [rdi], xmm4
- movdqa XMMWORD PTR [rdi + 16], xmm5
- movdqa XMMWORD PTR [rdi + rdx], xmm6
- movdqa XMMWORD PTR [rdi + rdx + 16], xmm7
-
- lea rdi, [rdi+rdx*2]
-
- sub rcx, 4
- cmp rcx, 4
- jge .block_copy_sse2_loopx4
-
- cmp rcx, 0
- je .copy_is_done
-
-.block_copy_sse2_loop:
- movdqu xmm0, XMMWORD PTR [rsi]
- movdqu xmm1, XMMWORD PTR [rsi + 16]
- lea rsi, [rsi+rax]
-
- movdqa XMMWORD PTR [rdi], xmm0
- movdqa XMMWORD PTR [rdi + 16], xmm1
- lea rdi, [rdi+rdx]
-
- sub rcx, 1
- jne .block_copy_sse2_loop
-
-.copy_is_done:
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
--- a/vp8/encoder/x86/sad_sse3.asm
+++ /dev/null
@@ -1,960 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro STACK_FRAME_CREATE_X3 0
-%if ABI_IS_32BIT
- %define src_ptr rsi
- %define src_stride rax
- %define ref_ptr rdi
- %define ref_stride rdx
- %define end_ptr rcx
- %define ret_var rbx
- %define result_ptr arg(4)
- %define max_sad arg(4)
- %define height dword ptr arg(4)
- push rbp
- mov rbp, rsp
- push rsi
- push rdi
- push rbx
-
- mov rsi, arg(0) ; src_ptr
- mov rdi, arg(2) ; ref_ptr
-
- movsxd rax, dword ptr arg(1) ; src_stride
- movsxd rdx, dword ptr arg(3) ; ref_stride
-%else
- %ifidn __OUTPUT_FORMAT__,x64
- SAVE_XMM 7, u
- %define src_ptr rcx
- %define src_stride rdx
- %define ref_ptr r8
- %define ref_stride r9
- %define end_ptr r10
- %define ret_var r11
- %define result_ptr [rsp+xmm_stack_space+8+4*8]
- %define max_sad [rsp+xmm_stack_space+8+4*8]
- %define height dword ptr [rsp+xmm_stack_space+8+4*8]
- %else
- %define src_ptr rdi
- %define src_stride rsi
- %define ref_ptr rdx
- %define ref_stride rcx
- %define end_ptr r9
- %define ret_var r10
- %define result_ptr r8
- %define max_sad r8
- %define height r8
- %endif
-%endif
-
-%endmacro
-
-%macro STACK_FRAME_DESTROY_X3 0
- %define src_ptr
- %define src_stride
- %define ref_ptr
- %define ref_stride
- %define end_ptr
- %define ret_var
- %define result_ptr
- %define max_sad
- %define height
-
-%if ABI_IS_32BIT
- pop rbx
- pop rdi
- pop rsi
- pop rbp
-%else
- %ifidn __OUTPUT_FORMAT__,x64
- RESTORE_XMM
- %endif
-%endif
- ret
-%endmacro
-
-%macro STACK_FRAME_CREATE_X4 0
-%if ABI_IS_32BIT
- %define src_ptr rsi
- %define src_stride rax
- %define r0_ptr rcx
- %define r1_ptr rdx
- %define r2_ptr rbx
- %define r3_ptr rdi
- %define ref_stride rbp
- %define result_ptr arg(4)
- push rbp
- mov rbp, rsp
- push rsi
- push rdi
- push rbx
-
- push rbp
- mov rdi, arg(2) ; ref_ptr_base
-
- LOAD_X4_ADDRESSES rdi, rcx, rdx, rax, rdi
-
- mov rsi, arg(0) ; src_ptr
-
- movsxd rbx, dword ptr arg(1) ; src_stride
- movsxd rbp, dword ptr arg(3) ; ref_stride
-
- xchg rbx, rax
-%else
- %ifidn __OUTPUT_FORMAT__,x64
- SAVE_XMM 7, u
- %define src_ptr rcx
- %define src_stride rdx
- %define r0_ptr rsi
- %define r1_ptr r10
- %define r2_ptr r11
- %define r3_ptr r8
- %define ref_stride r9
- %define result_ptr [rsp+xmm_stack_space+16+4*8]
- push rsi
-
- LOAD_X4_ADDRESSES r8, r0_ptr, r1_ptr, r2_ptr, r3_ptr
- %else
- %define src_ptr rdi
- %define src_stride rsi
- %define r0_ptr r9
- %define r1_ptr r10
- %define r2_ptr r11
- %define r3_ptr rdx
- %define ref_stride rcx
- %define result_ptr r8
-
- LOAD_X4_ADDRESSES rdx, r0_ptr, r1_ptr, r2_ptr, r3_ptr
-
- %endif
-%endif
-%endmacro
-
-%macro STACK_FRAME_DESTROY_X4 0
- %define src_ptr
- %define src_stride
- %define r0_ptr
- %define r1_ptr
- %define r2_ptr
- %define r3_ptr
- %define ref_stride
- %define result_ptr
-
-%if ABI_IS_32BIT
- pop rbx
- pop rdi
- pop rsi
- pop rbp
-%else
- %ifidn __OUTPUT_FORMAT__,x64
- pop rsi
- RESTORE_XMM
- %endif
-%endif
- ret
-%endmacro
-
-%macro PROCESS_16X2X3 5
-%if %1==0
- movdqa xmm0, XMMWORD PTR [%2]
- lddqu xmm5, XMMWORD PTR [%3]
- lddqu xmm6, XMMWORD PTR [%3+1]
- lddqu xmm7, XMMWORD PTR [%3+2]
-
- psadbw xmm5, xmm0
- psadbw xmm6, xmm0
- psadbw xmm7, xmm0
-%else
- movdqa xmm0, XMMWORD PTR [%2]
- lddqu xmm1, XMMWORD PTR [%3]
- lddqu xmm2, XMMWORD PTR [%3+1]
- lddqu xmm3, XMMWORD PTR [%3+2]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endif
- movdqa xmm0, XMMWORD PTR [%2+%4]
- lddqu xmm1, XMMWORD PTR [%3+%5]
- lddqu xmm2, XMMWORD PTR [%3+%5+1]
- lddqu xmm3, XMMWORD PTR [%3+%5+2]
-
-%if %1==0 || %1==1
- lea %2, [%2+%4*2]
- lea %3, [%3+%5*2]
-%endif
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endmacro
-
-%macro PROCESS_8X2X3 5
-%if %1==0
- movq mm0, QWORD PTR [%2]
- movq mm5, QWORD PTR [%3]
- movq mm6, QWORD PTR [%3+1]
- movq mm7, QWORD PTR [%3+2]
-
- psadbw mm5, mm0
- psadbw mm6, mm0
- psadbw mm7, mm0
-%else
- movq mm0, QWORD PTR [%2]
- movq mm1, QWORD PTR [%3]
- movq mm2, QWORD PTR [%3+1]
- movq mm3, QWORD PTR [%3+2]
-
- psadbw mm1, mm0
- psadbw mm2, mm0
- psadbw mm3, mm0
-
- paddw mm5, mm1
- paddw mm6, mm2
- paddw mm7, mm3
-%endif
- movq mm0, QWORD PTR [%2+%4]
- movq mm1, QWORD PTR [%3+%5]
- movq mm2, QWORD PTR [%3+%5+1]
- movq mm3, QWORD PTR [%3+%5+2]
-
-%if %1==0 || %1==1
- lea %2, [%2+%4*2]
- lea %3, [%3+%5*2]
-%endif
-
- psadbw mm1, mm0
- psadbw mm2, mm0
- psadbw mm3, mm0
-
- paddw mm5, mm1
- paddw mm6, mm2
- paddw mm7, mm3
-%endmacro
-
-%macro LOAD_X4_ADDRESSES 5
- mov %2, [%1+REG_SZ_BYTES*0]
- mov %3, [%1+REG_SZ_BYTES*1]
-
- mov %4, [%1+REG_SZ_BYTES*2]
- mov %5, [%1+REG_SZ_BYTES*3]
-%endmacro
-
-%macro PROCESS_16X2X4 8
-%if %1==0
- movdqa xmm0, XMMWORD PTR [%2]
- lddqu xmm4, XMMWORD PTR [%3]
- lddqu xmm5, XMMWORD PTR [%4]
- lddqu xmm6, XMMWORD PTR [%5]
- lddqu xmm7, XMMWORD PTR [%6]
-
- psadbw xmm4, xmm0
- psadbw xmm5, xmm0
- psadbw xmm6, xmm0
- psadbw xmm7, xmm0
-%else
- movdqa xmm0, XMMWORD PTR [%2]
- lddqu xmm1, XMMWORD PTR [%3]
- lddqu xmm2, XMMWORD PTR [%4]
- lddqu xmm3, XMMWORD PTR [%5]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm4, xmm1
- lddqu xmm1, XMMWORD PTR [%6]
- paddw xmm5, xmm2
- paddw xmm6, xmm3
-
- psadbw xmm1, xmm0
- paddw xmm7, xmm1
-%endif
- movdqa xmm0, XMMWORD PTR [%2+%7]
- lddqu xmm1, XMMWORD PTR [%3+%8]
- lddqu xmm2, XMMWORD PTR [%4+%8]
- lddqu xmm3, XMMWORD PTR [%5+%8]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm4, xmm1
- lddqu xmm1, XMMWORD PTR [%6+%8]
- paddw xmm5, xmm2
- paddw xmm6, xmm3
-
-%if %1==0 || %1==1
- lea %2, [%2+%7*2]
- lea %3, [%3+%8*2]
-
- lea %4, [%4+%8*2]
- lea %5, [%5+%8*2]
-
- lea %6, [%6+%8*2]
-%endif
- psadbw xmm1, xmm0
- paddw xmm7, xmm1
-
-%endmacro
-
-%macro PROCESS_8X2X4 8
-%if %1==0
- movq mm0, QWORD PTR [%2]
- movq mm4, QWORD PTR [%3]
- movq mm5, QWORD PTR [%4]
- movq mm6, QWORD PTR [%5]
- movq mm7, QWORD PTR [%6]
-
- psadbw mm4, mm0
- psadbw mm5, mm0
- psadbw mm6, mm0
- psadbw mm7, mm0
-%else
- movq mm0, QWORD PTR [%2]
- movq mm1, QWORD PTR [%3]
- movq mm2, QWORD PTR [%4]
- movq mm3, QWORD PTR [%5]
-
- psadbw mm1, mm0
- psadbw mm2, mm0
- psadbw mm3, mm0
-
- paddw mm4, mm1
- movq mm1, QWORD PTR [%6]
- paddw mm5, mm2
- paddw mm6, mm3
-
- psadbw mm1, mm0
- paddw mm7, mm1
-%endif
- movq mm0, QWORD PTR [%2+%7]
- movq mm1, QWORD PTR [%3+%8]
- movq mm2, QWORD PTR [%4+%8]
- movq mm3, QWORD PTR [%5+%8]
-
- psadbw mm1, mm0
- psadbw mm2, mm0
- psadbw mm3, mm0
-
- paddw mm4, mm1
- movq mm1, QWORD PTR [%6+%8]
- paddw mm5, mm2
- paddw mm6, mm3
-
-%if %1==0 || %1==1
- lea %2, [%2+%7*2]
- lea %3, [%3+%8*2]
-
- lea %4, [%4+%8*2]
- lea %5, [%5+%8*2]
-
- lea %6, [%6+%8*2]
-%endif
- psadbw mm1, mm0
- paddw mm7, mm1
-
-%endmacro
-
-;void int vp8_sad16x16x3_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x16x3_sse3)
-sym(vp8_sad16x16x3_sse3):
-
- STACK_FRAME_CREATE_X3
-
- PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
- mov rcx, result_ptr
-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rcx], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rcx+4], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rcx+8], xmm0
-
- STACK_FRAME_DESTROY_X3
-
-;void int vp8_sad16x8x3_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x8x3_sse3)
-sym(vp8_sad16x8x3_sse3):
-
- STACK_FRAME_CREATE_X3
-
- PROCESS_16X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_16X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
- mov rcx, result_ptr
-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rcx], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rcx+4], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rcx+8], xmm0
-
- STACK_FRAME_DESTROY_X3
-
-;void int vp8_sad8x16x3_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad8x16x3_sse3)
-sym(vp8_sad8x16x3_sse3):
-
- STACK_FRAME_CREATE_X3
-
- PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
- mov rcx, result_ptr
-
- punpckldq mm5, mm6
-
- movq [rcx], mm5
- movd [rcx+8], mm7
-
- STACK_FRAME_DESTROY_X3
-
-;void int vp8_sad8x8x3_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad8x8x3_sse3)
-sym(vp8_sad8x8x3_sse3):
-
- STACK_FRAME_CREATE_X3
-
- PROCESS_8X2X3 0, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 1, src_ptr, ref_ptr, src_stride, ref_stride
- PROCESS_8X2X3 2, src_ptr, ref_ptr, src_stride, ref_stride
-
- mov rcx, result_ptr
-
- punpckldq mm5, mm6
-
- movq [rcx], mm5
- movd [rcx+8], mm7
-
- STACK_FRAME_DESTROY_X3
-
-;void int vp8_sad4x4x3_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad4x4x3_sse3)
-sym(vp8_sad4x4x3_sse3):
-
- STACK_FRAME_CREATE_X3
-
- movd mm0, DWORD PTR [src_ptr]
- movd mm1, DWORD PTR [ref_ptr]
-
- movd mm2, DWORD PTR [src_ptr+src_stride]
- movd mm3, DWORD PTR [ref_ptr+ref_stride]
-
- punpcklbw mm0, mm2
- punpcklbw mm1, mm3
-
- movd mm4, DWORD PTR [ref_ptr+1]
- movd mm5, DWORD PTR [ref_ptr+2]
-
- movd mm2, DWORD PTR [ref_ptr+ref_stride+1]
- movd mm3, DWORD PTR [ref_ptr+ref_stride+2]
-
- psadbw mm1, mm0
-
- punpcklbw mm4, mm2
- punpcklbw mm5, mm3
-
- psadbw mm4, mm0
- psadbw mm5, mm0
-
- lea src_ptr, [src_ptr+src_stride*2]
- lea ref_ptr, [ref_ptr+ref_stride*2]
-
- movd mm0, DWORD PTR [src_ptr]
- movd mm2, DWORD PTR [ref_ptr]
-
- movd mm3, DWORD PTR [src_ptr+src_stride]
- movd mm6, DWORD PTR [ref_ptr+ref_stride]
-
- punpcklbw mm0, mm3
- punpcklbw mm2, mm6
-
- movd mm3, DWORD PTR [ref_ptr+1]
- movd mm7, DWORD PTR [ref_ptr+2]
-
- psadbw mm2, mm0
-
- paddw mm1, mm2
-
- movd mm2, DWORD PTR [ref_ptr+ref_stride+1]
- movd mm6, DWORD PTR [ref_ptr+ref_stride+2]
-
- punpcklbw mm3, mm2
- punpcklbw mm7, mm6
-
- psadbw mm3, mm0
- psadbw mm7, mm0
-
- paddw mm3, mm4
- paddw mm7, mm5
-
- mov rcx, result_ptr
-
- punpckldq mm1, mm3
-
- movq [rcx], mm1
- movd [rcx+8], mm7
-
- STACK_FRAME_DESTROY_X3
-
-;unsigned int vp8_sad16x16_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int max_sad)
-;%define lddqu movdqu
-global sym(vp8_sad16x16_sse3)
-sym(vp8_sad16x16_sse3):
-
- STACK_FRAME_CREATE_X3
-
- mov end_ptr, 4
- pxor xmm7, xmm7
-
-.vp8_sad16x16_sse3_loop:
- movdqa xmm0, XMMWORD PTR [src_ptr]
- movdqu xmm1, XMMWORD PTR [ref_ptr]
- movdqa xmm2, XMMWORD PTR [src_ptr+src_stride]
- movdqu xmm3, XMMWORD PTR [ref_ptr+ref_stride]
-
- lea src_ptr, [src_ptr+src_stride*2]
- lea ref_ptr, [ref_ptr+ref_stride*2]
-
- movdqa xmm4, XMMWORD PTR [src_ptr]
- movdqu xmm5, XMMWORD PTR [ref_ptr]
- movdqa xmm6, XMMWORD PTR [src_ptr+src_stride]
-
- psadbw xmm0, xmm1
-
- movdqu xmm1, XMMWORD PTR [ref_ptr+ref_stride]
-
- psadbw xmm2, xmm3
- psadbw xmm4, xmm5
- psadbw xmm6, xmm1
-
- lea src_ptr, [src_ptr+src_stride*2]
- lea ref_ptr, [ref_ptr+ref_stride*2]
-
- paddw xmm7, xmm0
- paddw xmm7, xmm2
- paddw xmm7, xmm4
- paddw xmm7, xmm6
-
- sub end_ptr, 1
- jne .vp8_sad16x16_sse3_loop
-
- movq xmm0, xmm7
- psrldq xmm7, 8
- paddw xmm0, xmm7
- movq rax, xmm0
-
- STACK_FRAME_DESTROY_X3
-
-;void vp8_copy32xn_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *dst_ptr,
-; int dst_stride,
-; int height);
-global sym(vp8_copy32xn_sse3)
-sym(vp8_copy32xn_sse3):
-
- STACK_FRAME_CREATE_X3
-
-.block_copy_sse3_loopx4:
- lea end_ptr, [src_ptr+src_stride*2]
-
- movdqu xmm0, XMMWORD PTR [src_ptr]
- movdqu xmm1, XMMWORD PTR [src_ptr + 16]
- movdqu xmm2, XMMWORD PTR [src_ptr + src_stride]
- movdqu xmm3, XMMWORD PTR [src_ptr + src_stride + 16]
- movdqu xmm4, XMMWORD PTR [end_ptr]
- movdqu xmm5, XMMWORD PTR [end_ptr + 16]
- movdqu xmm6, XMMWORD PTR [end_ptr + src_stride]
- movdqu xmm7, XMMWORD PTR [end_ptr + src_stride + 16]
-
- lea src_ptr, [src_ptr+src_stride*4]
-
- lea end_ptr, [ref_ptr+ref_stride*2]
-
- movdqa XMMWORD PTR [ref_ptr], xmm0
- movdqa XMMWORD PTR [ref_ptr + 16], xmm1
- movdqa XMMWORD PTR [ref_ptr + ref_stride], xmm2
- movdqa XMMWORD PTR [ref_ptr + ref_stride + 16], xmm3
- movdqa XMMWORD PTR [end_ptr], xmm4
- movdqa XMMWORD PTR [end_ptr + 16], xmm5
- movdqa XMMWORD PTR [end_ptr + ref_stride], xmm6
- movdqa XMMWORD PTR [end_ptr + ref_stride + 16], xmm7
-
- lea ref_ptr, [ref_ptr+ref_stride*4]
-
- sub height, 4
- cmp height, 4
- jge .block_copy_sse3_loopx4
-
- ;Check to see if there is more rows need to be copied.
- cmp height, 0
- je .copy_is_done
-
-.block_copy_sse3_loop:
- movdqu xmm0, XMMWORD PTR [src_ptr]
- movdqu xmm1, XMMWORD PTR [src_ptr + 16]
- lea src_ptr, [src_ptr+src_stride]
-
- movdqa XMMWORD PTR [ref_ptr], xmm0
- movdqa XMMWORD PTR [ref_ptr + 16], xmm1
- lea ref_ptr, [ref_ptr+ref_stride]
-
- sub height, 1
- jne .block_copy_sse3_loop
-
-.copy_is_done:
- STACK_FRAME_DESTROY_X3
-
-;void vp8_sad16x16x4d_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr_base,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x16x4d_sse3)
-sym(vp8_sad16x16x4d_sse3):
-
- STACK_FRAME_CREATE_X4
-
- PROCESS_16X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
-
-%if ABI_IS_32BIT
- pop rbp
-%endif
- mov rcx, result_ptr
-
- movq xmm0, xmm4
- psrldq xmm4, 8
-
- paddw xmm0, xmm4
- movd [rcx], xmm0
-;-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rcx+4], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rcx+8], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rcx+12], xmm0
-
- STACK_FRAME_DESTROY_X4
-
-;void vp8_sad16x8x4d_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr_base,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x8x4d_sse3)
-sym(vp8_sad16x8x4d_sse3):
-
- STACK_FRAME_CREATE_X4
-
- PROCESS_16X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_16X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
-
-%if ABI_IS_32BIT
- pop rbp
-%endif
- mov rcx, result_ptr
-
- movq xmm0, xmm4
- psrldq xmm4, 8
-
- paddw xmm0, xmm4
- movd [rcx], xmm0
-;-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rcx+4], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rcx+8], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rcx+12], xmm0
-
- STACK_FRAME_DESTROY_X4
-
-;void int vp8_sad8x16x4d_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad8x16x4d_sse3)
-sym(vp8_sad8x16x4d_sse3):
-
- STACK_FRAME_CREATE_X4
-
- PROCESS_8X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
-
-%if ABI_IS_32BIT
- pop rbp
-%endif
- mov rcx, result_ptr
-
- punpckldq mm4, mm5
- punpckldq mm6, mm7
-
- movq [rcx], mm4
- movq [rcx+8], mm6
-
- STACK_FRAME_DESTROY_X4
-
-;void int vp8_sad8x8x4d_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad8x8x4d_sse3)
-sym(vp8_sad8x8x4d_sse3):
-
- STACK_FRAME_CREATE_X4
-
- PROCESS_8X2X4 0, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 1, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
- PROCESS_8X2X4 2, src_ptr, r0_ptr, r1_ptr, r2_ptr, r3_ptr, src_stride, ref_stride
-
-%if ABI_IS_32BIT
- pop rbp
-%endif
- mov rcx, result_ptr
-
- punpckldq mm4, mm5
- punpckldq mm6, mm7
-
- movq [rcx], mm4
- movq [rcx+8], mm6
-
- STACK_FRAME_DESTROY_X4
-
-;void int vp8_sad4x4x4d_sse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad4x4x4d_sse3)
-sym(vp8_sad4x4x4d_sse3):
-
- STACK_FRAME_CREATE_X4
-
- movd mm0, DWORD PTR [src_ptr]
- movd mm1, DWORD PTR [r0_ptr]
-
- movd mm2, DWORD PTR [src_ptr+src_stride]
- movd mm3, DWORD PTR [r0_ptr+ref_stride]
-
- punpcklbw mm0, mm2
- punpcklbw mm1, mm3
-
- movd mm4, DWORD PTR [r1_ptr]
- movd mm5, DWORD PTR [r2_ptr]
-
- movd mm6, DWORD PTR [r3_ptr]
- movd mm2, DWORD PTR [r1_ptr+ref_stride]
-
- movd mm3, DWORD PTR [r2_ptr+ref_stride]
- movd mm7, DWORD PTR [r3_ptr+ref_stride]
-
- psadbw mm1, mm0
-
- punpcklbw mm4, mm2
- punpcklbw mm5, mm3
-
- punpcklbw mm6, mm7
- psadbw mm4, mm0
-
- psadbw mm5, mm0
- psadbw mm6, mm0
-
-
-
- lea src_ptr, [src_ptr+src_stride*2]
- lea r0_ptr, [r0_ptr+ref_stride*2]
-
- lea r1_ptr, [r1_ptr+ref_stride*2]
- lea r2_ptr, [r2_ptr+ref_stride*2]
-
- lea r3_ptr, [r3_ptr+ref_stride*2]
-
- movd mm0, DWORD PTR [src_ptr]
- movd mm2, DWORD PTR [r0_ptr]
-
- movd mm3, DWORD PTR [src_ptr+src_stride]
- movd mm7, DWORD PTR [r0_ptr+ref_stride]
-
- punpcklbw mm0, mm3
- punpcklbw mm2, mm7
-
- movd mm3, DWORD PTR [r1_ptr]
- movd mm7, DWORD PTR [r2_ptr]
-
- psadbw mm2, mm0
-%if ABI_IS_32BIT
- mov rax, rbp
-
- pop rbp
-%define ref_stride rax
-%endif
- mov rsi, result_ptr
-
- paddw mm1, mm2
- movd [rsi], mm1
-
- movd mm2, DWORD PTR [r1_ptr+ref_stride]
- movd mm1, DWORD PTR [r2_ptr+ref_stride]
-
- punpcklbw mm3, mm2
- punpcklbw mm7, mm1
-
- psadbw mm3, mm0
- psadbw mm7, mm0
-
- movd mm2, DWORD PTR [r3_ptr]
- movd mm1, DWORD PTR [r3_ptr+ref_stride]
-
- paddw mm3, mm4
- paddw mm7, mm5
-
- movd [rsi+4], mm3
- punpcklbw mm2, mm1
-
- movd [rsi+8], mm7
- psadbw mm2, mm0
-
- paddw mm2, mm6
- movd [rsi+12], mm2
-
-
- STACK_FRAME_DESTROY_X4
-
--- a/vp8/encoder/x86/sad_sse4.asm
+++ /dev/null
@@ -1,353 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro PROCESS_16X2X8 1
-%if %1
- movdqa xmm0, XMMWORD PTR [rsi]
- movq xmm1, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- movq xmm2, MMWORD PTR [rdi+16]
- punpcklqdq xmm1, xmm3
- punpcklqdq xmm3, xmm2
-
- movdqa xmm2, xmm1
- mpsadbw xmm1, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
-
- psrldq xmm0, 8
-
- movdqa xmm4, xmm3
- mpsadbw xmm3, xmm0, 0x0
- mpsadbw xmm4, xmm0, 0x5
-
- paddw xmm1, xmm2
- paddw xmm1, xmm3
- paddw xmm1, xmm4
-%else
- movdqa xmm0, XMMWORD PTR [rsi]
- movq xmm5, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- movq xmm2, MMWORD PTR [rdi+16]
- punpcklqdq xmm5, xmm3
- punpcklqdq xmm3, xmm2
-
- movdqa xmm2, xmm5
- mpsadbw xmm5, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
-
- psrldq xmm0, 8
-
- movdqa xmm4, xmm3
- mpsadbw xmm3, xmm0, 0x0
- mpsadbw xmm4, xmm0, 0x5
-
- paddw xmm5, xmm2
- paddw xmm5, xmm3
- paddw xmm5, xmm4
-
- paddw xmm1, xmm5
-%endif
- movdqa xmm0, XMMWORD PTR [rsi + rax]
- movq xmm5, MMWORD PTR [rdi+ rdx]
- movq xmm3, MMWORD PTR [rdi+ rdx+8]
- movq xmm2, MMWORD PTR [rdi+ rdx+16]
- punpcklqdq xmm5, xmm3
- punpcklqdq xmm3, xmm2
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- movdqa xmm2, xmm5
- mpsadbw xmm5, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
-
- psrldq xmm0, 8
- movdqa xmm4, xmm3
- mpsadbw xmm3, xmm0, 0x0
- mpsadbw xmm4, xmm0, 0x5
-
- paddw xmm5, xmm2
- paddw xmm5, xmm3
- paddw xmm5, xmm4
-
- paddw xmm1, xmm5
-%endmacro
-
-%macro PROCESS_8X2X8 1
-%if %1
- movq xmm0, MMWORD PTR [rsi]
- movq xmm1, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- punpcklqdq xmm1, xmm3
-
- movdqa xmm2, xmm1
- mpsadbw xmm1, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
- paddw xmm1, xmm2
-%else
- movq xmm0, MMWORD PTR [rsi]
- movq xmm5, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- punpcklqdq xmm5, xmm3
-
- movdqa xmm2, xmm5
- mpsadbw xmm5, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
- paddw xmm5, xmm2
-
- paddw xmm1, xmm5
-%endif
- movq xmm0, MMWORD PTR [rsi + rax]
- movq xmm5, MMWORD PTR [rdi+ rdx]
- movq xmm3, MMWORD PTR [rdi+ rdx+8]
- punpcklqdq xmm5, xmm3
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- movdqa xmm2, xmm5
- mpsadbw xmm5, xmm0, 0x0
- mpsadbw xmm2, xmm0, 0x5
- paddw xmm5, xmm2
-
- paddw xmm1, xmm5
-%endmacro
-
-%macro PROCESS_4X2X8 1
-%if %1
- movd xmm0, [rsi]
- movq xmm1, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- punpcklqdq xmm1, xmm3
-
- mpsadbw xmm1, xmm0, 0x0
-%else
- movd xmm0, [rsi]
- movq xmm5, MMWORD PTR [rdi]
- movq xmm3, MMWORD PTR [rdi+8]
- punpcklqdq xmm5, xmm3
-
- mpsadbw xmm5, xmm0, 0x0
-
- paddw xmm1, xmm5
-%endif
- movd xmm0, [rsi + rax]
- movq xmm5, MMWORD PTR [rdi+ rdx]
- movq xmm3, MMWORD PTR [rdi+ rdx+8]
- punpcklqdq xmm5, xmm3
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- mpsadbw xmm5, xmm0, 0x0
-
- paddw xmm1, xmm5
-%endmacro
-
-
-;void vp8_sad16x16x8_sse4(
-; const unsigned char *src_ptr,
-; int src_stride,
-; const unsigned char *ref_ptr,
-; int ref_stride,
-; unsigned short *sad_array);
-global sym(vp8_sad16x16x8_sse4)
-sym(vp8_sad16x16x8_sse4):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- PROCESS_16X2X8 1
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
-
- mov rdi, arg(4) ;Results
- movdqa XMMWORD PTR [rdi], xmm1
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_sad16x8x8_sse4(
-; const unsigned char *src_ptr,
-; int src_stride,
-; const unsigned char *ref_ptr,
-; int ref_stride,
-; unsigned short *sad_array
-;);
-global sym(vp8_sad16x8x8_sse4)
-sym(vp8_sad16x8x8_sse4):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- PROCESS_16X2X8 1
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
- PROCESS_16X2X8 0
-
- mov rdi, arg(4) ;Results
- movdqa XMMWORD PTR [rdi], xmm1
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_sad8x8x8_sse4(
-; const unsigned char *src_ptr,
-; int src_stride,
-; const unsigned char *ref_ptr,
-; int ref_stride,
-; unsigned short *sad_array
-;);
-global sym(vp8_sad8x8x8_sse4)
-sym(vp8_sad8x8x8_sse4):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- PROCESS_8X2X8 1
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
-
- mov rdi, arg(4) ;Results
- movdqa XMMWORD PTR [rdi], xmm1
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_sad8x16x8_sse4(
-; const unsigned char *src_ptr,
-; int src_stride,
-; const unsigned char *ref_ptr,
-; int ref_stride,
-; unsigned short *sad_array
-;);
-global sym(vp8_sad8x16x8_sse4)
-sym(vp8_sad8x16x8_sse4):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- PROCESS_8X2X8 1
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- PROCESS_8X2X8 0
- mov rdi, arg(4) ;Results
- movdqa XMMWORD PTR [rdi], xmm1
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_sad4x4x8_c(
-; const unsigned char *src_ptr,
-; int src_stride,
-; const unsigned char *ref_ptr,
-; int ref_stride,
-; unsigned short *sad_array
-;);
-global sym(vp8_sad4x4x8_sse4)
-sym(vp8_sad4x4x8_sse4):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- PROCESS_4X2X8 1
- PROCESS_4X2X8 0
-
- mov rdi, arg(4) ;Results
- movdqa XMMWORD PTR [rdi], xmm1
-
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-
--- a/vp8/encoder/x86/sad_ssse3.asm
+++ /dev/null
@@ -1,370 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%macro PROCESS_16X2X3 1
-%if %1
- movdqa xmm0, XMMWORD PTR [rsi]
- lddqu xmm5, XMMWORD PTR [rdi]
- lddqu xmm6, XMMWORD PTR [rdi+1]
- lddqu xmm7, XMMWORD PTR [rdi+2]
-
- psadbw xmm5, xmm0
- psadbw xmm6, xmm0
- psadbw xmm7, xmm0
-%else
- movdqa xmm0, XMMWORD PTR [rsi]
- lddqu xmm1, XMMWORD PTR [rdi]
- lddqu xmm2, XMMWORD PTR [rdi+1]
- lddqu xmm3, XMMWORD PTR [rdi+2]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endif
- movdqa xmm0, XMMWORD PTR [rsi+rax]
- lddqu xmm1, XMMWORD PTR [rdi+rdx]
- lddqu xmm2, XMMWORD PTR [rdi+rdx+1]
- lddqu xmm3, XMMWORD PTR [rdi+rdx+2]
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endmacro
-
-%macro PROCESS_16X2X3_OFFSET 2
-%if %1
- movdqa xmm0, XMMWORD PTR [rsi]
- movdqa xmm4, XMMWORD PTR [rdi]
- movdqa xmm7, XMMWORD PTR [rdi+16]
-
- movdqa xmm5, xmm7
- palignr xmm5, xmm4, %2
-
- movdqa xmm6, xmm7
- palignr xmm6, xmm4, (%2+1)
-
- palignr xmm7, xmm4, (%2+2)
-
- psadbw xmm5, xmm0
- psadbw xmm6, xmm0
- psadbw xmm7, xmm0
-%else
- movdqa xmm0, XMMWORD PTR [rsi]
- movdqa xmm4, XMMWORD PTR [rdi]
- movdqa xmm3, XMMWORD PTR [rdi+16]
-
- movdqa xmm1, xmm3
- palignr xmm1, xmm4, %2
-
- movdqa xmm2, xmm3
- palignr xmm2, xmm4, (%2+1)
-
- palignr xmm3, xmm4, (%2+2)
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endif
- movdqa xmm0, XMMWORD PTR [rsi+rax]
- movdqa xmm4, XMMWORD PTR [rdi+rdx]
- movdqa xmm3, XMMWORD PTR [rdi+rdx+16]
-
- movdqa xmm1, xmm3
- palignr xmm1, xmm4, %2
-
- movdqa xmm2, xmm3
- palignr xmm2, xmm4, (%2+1)
-
- palignr xmm3, xmm4, (%2+2)
-
- lea rsi, [rsi+rax*2]
- lea rdi, [rdi+rdx*2]
-
- psadbw xmm1, xmm0
- psadbw xmm2, xmm0
- psadbw xmm3, xmm0
-
- paddw xmm5, xmm1
- paddw xmm6, xmm2
- paddw xmm7, xmm3
-%endmacro
-
-%macro PROCESS_16X16X3_OFFSET 2
-%2_aligned_by_%1:
-
- sub rdi, %1
-
- PROCESS_16X2X3_OFFSET 1, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
-
- jmp %2_store_off
-
-%endmacro
-
-%macro PROCESS_16X8X3_OFFSET 2
-%2_aligned_by_%1:
-
- sub rdi, %1
-
- PROCESS_16X2X3_OFFSET 1, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
- PROCESS_16X2X3_OFFSET 0, %1
-
- jmp %2_store_off
-
-%endmacro
-
-;void int vp8_sad16x16x3_ssse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x16x3_ssse3)
-sym(vp8_sad16x16x3_ssse3):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- SAVE_XMM 7
- push rsi
- push rdi
- push rcx
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- mov rdx, 0xf
- and rdx, rdi
-
- jmp .vp8_sad16x16x3_ssse3_skiptable
-.vp8_sad16x16x3_ssse3_jumptable:
- dd .vp8_sad16x16x3_ssse3_aligned_by_0 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_1 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_2 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_3 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_4 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_5 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_6 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_7 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_8 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_9 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_10 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_11 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_12 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_13 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_14 - .vp8_sad16x16x3_ssse3_do_jump
- dd .vp8_sad16x16x3_ssse3_aligned_by_15 - .vp8_sad16x16x3_ssse3_do_jump
-.vp8_sad16x16x3_ssse3_skiptable:
-
- call .vp8_sad16x16x3_ssse3_do_jump
-.vp8_sad16x16x3_ssse3_do_jump:
- pop rcx ; get the address of do_jump
- mov rax, .vp8_sad16x16x3_ssse3_jumptable - .vp8_sad16x16x3_ssse3_do_jump
- add rax, rcx ; get the absolute address of vp8_sad16x16x3_ssse3_jumptable
-
- movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
- add rcx, rax
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- jmp rcx
-
- PROCESS_16X16X3_OFFSET 0, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 1, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 2, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 3, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 4, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 5, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 6, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 7, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 8, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 9, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 10, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 11, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 12, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 13, .vp8_sad16x16x3_ssse3
- PROCESS_16X16X3_OFFSET 14, .vp8_sad16x16x3_ssse3
-
-.vp8_sad16x16x3_ssse3_aligned_by_15:
- PROCESS_16X2X3 1
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
-
-.vp8_sad16x16x3_ssse3_store_off:
- mov rdi, arg(4) ;Results
-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rdi], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rdi+4], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rdi+8], xmm0
-
- ; begin epilog
- pop rcx
- pop rdi
- pop rsi
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void int vp8_sad16x8x3_ssse3(
-; unsigned char *src_ptr,
-; int src_stride,
-; unsigned char *ref_ptr,
-; int ref_stride,
-; int *results)
-global sym(vp8_sad16x8x3_ssse3)
-sym(vp8_sad16x8x3_ssse3):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- SAVE_XMM 7
- push rsi
- push rdi
- push rcx
- ; end prolog
-
- mov rsi, arg(0) ;src_ptr
- mov rdi, arg(2) ;ref_ptr
-
- mov rdx, 0xf
- and rdx, rdi
-
- jmp .vp8_sad16x8x3_ssse3_skiptable
-.vp8_sad16x8x3_ssse3_jumptable:
- dd .vp8_sad16x8x3_ssse3_aligned_by_0 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_1 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_2 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_3 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_4 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_5 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_6 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_7 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_8 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_9 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_10 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_11 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_12 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_13 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_14 - .vp8_sad16x8x3_ssse3_do_jump
- dd .vp8_sad16x8x3_ssse3_aligned_by_15 - .vp8_sad16x8x3_ssse3_do_jump
-.vp8_sad16x8x3_ssse3_skiptable:
-
- call .vp8_sad16x8x3_ssse3_do_jump
-.vp8_sad16x8x3_ssse3_do_jump:
- pop rcx ; get the address of do_jump
- mov rax, .vp8_sad16x8x3_ssse3_jumptable - .vp8_sad16x8x3_ssse3_do_jump
- add rax, rcx ; get the absolute address of vp8_sad16x8x3_ssse3_jumptable
-
- movsxd rax, dword [rax + 4*rdx] ; get the 32 bit offset from the jumptable
- add rcx, rax
-
- movsxd rax, dword ptr arg(1) ;src_stride
- movsxd rdx, dword ptr arg(3) ;ref_stride
-
- jmp rcx
-
- PROCESS_16X8X3_OFFSET 0, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 1, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 2, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 3, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 4, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 5, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 6, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 7, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 8, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 9, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 10, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 11, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 12, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 13, .vp8_sad16x8x3_ssse3
- PROCESS_16X8X3_OFFSET 14, .vp8_sad16x8x3_ssse3
-
-.vp8_sad16x8x3_ssse3_aligned_by_15:
-
- PROCESS_16X2X3 1
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
- PROCESS_16X2X3 0
-
-.vp8_sad16x8x3_ssse3_store_off:
- mov rdi, arg(4) ;Results
-
- movq xmm0, xmm5
- psrldq xmm5, 8
-
- paddw xmm0, xmm5
- movd [rdi], xmm0
-;-
- movq xmm0, xmm6
- psrldq xmm6, 8
-
- paddw xmm0, xmm6
- movd [rdi+4], xmm0
-;-
- movq xmm0, xmm7
- psrldq xmm7, 8
-
- paddw xmm0, xmm7
- movd [rdi+8], xmm0
-
- ; begin epilog
- pop rcx
- pop rdi
- pop rsi
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
--- a/vp8/encoder/x86/variance_impl_mmx.asm
+++ /dev/null
@@ -1,851 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-;unsigned int vp8_get_mb_ss_mmx( short *src_ptr )
-global sym(vp8_get_mb_ss_mmx)
-sym(vp8_get_mb_ss_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 8
- ; end prolog
-
- mov rax, arg(0) ;src_ptr
- mov rcx, 16
- pxor mm4, mm4
-
-.NEXTROW:
- movq mm0, [rax]
- movq mm1, [rax+8]
- movq mm2, [rax+16]
- movq mm3, [rax+24]
- pmaddwd mm0, mm0
- pmaddwd mm1, mm1
- pmaddwd mm2, mm2
- pmaddwd mm3, mm3
-
- paddd mm4, mm0
- paddd mm4, mm1
- paddd mm4, mm2
- paddd mm4, mm3
-
- add rax, 32
- dec rcx
- ja .NEXTROW
- movq QWORD PTR [rsp], mm4
-
- ;return sum[0]+sum[1];
- movsxd rax, dword ptr [rsp]
- movsxd rcx, dword ptr [rsp+4]
- add rax, rcx
-
-
- ; begin epilog
- add rsp, 8
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_get8x8var_mmx
-;(
-; unsigned char *src_ptr,
-; int source_stride,
-; unsigned char *ref_ptr,
-; int recon_stride,
-; unsigned int *SSE,
-; int *Sum
-;)
-global sym(vp8_get8x8var_mmx)
-sym(vp8_get8x8var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- push rsi
- push rdi
- push rbx
- sub rsp, 16
- ; end prolog
-
-
- pxor mm5, mm5 ; Blank mmx6
- pxor mm6, mm6 ; Blank mmx7
- pxor mm7, mm7 ; Blank mmx7
-
- mov rax, arg(0) ;[src_ptr] ; Load base addresses
- mov rbx, arg(2) ;[ref_ptr]
- movsxd rcx, dword ptr arg(1) ;[source_stride]
- movsxd rdx, dword ptr arg(3) ;[recon_stride]
-
- ; Row 1
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm1, [rbx] ; Copy eight bytes to mm1
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
-
- ; Row 2
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 3
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 4
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 5
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- ; movq mm4, [rbx + rdx]
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 6
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 7
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Row 8
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm2, mm0 ; Take copies
- movq mm3, mm1 ; Take copies
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- punpckhbw mm2, mm6 ; unpack to higher prrcision
- punpckhbw mm3, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- psubsw mm2, mm3 ; A-B (high order) to MM2
-
- paddw mm5, mm0 ; accumulate differences in mm5
- paddw mm5, mm2 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- pmaddwd mm2, mm2 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- paddd mm7, mm0 ; accumulate in mm7
- paddd mm7, mm2 ; accumulate in mm7
-
- ; Now accumulate the final results.
- movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
- movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
- movsx rdx, WORD PTR [rsp+8]
- movsx rcx, WORD PTR [rsp+10]
- movsx rbx, WORD PTR [rsp+12]
- movsx rax, WORD PTR [rsp+14]
- add rdx, rcx
- add rbx, rax
- add rdx, rbx ;XSum
- movsxd rax, DWORD PTR [rsp]
- movsxd rcx, DWORD PTR [rsp+4]
- add rax, rcx ;XXSum
- mov rsi, arg(4) ;SSE
- mov rdi, arg(5) ;Sum
- mov dword ptr [rsi], eax
- mov dword ptr [rdi], edx
- xor rax, rax ; return 0
-
-
- ; begin epilog
- add rsp, 16
- pop rbx
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-;unsigned int
-;vp8_get4x4var_mmx
-;(
-; unsigned char *src_ptr,
-; int source_stride,
-; unsigned char *ref_ptr,
-; int recon_stride,
-; unsigned int *SSE,
-; int *Sum
-;)
-global sym(vp8_get4x4var_mmx)
-sym(vp8_get4x4var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- push rsi
- push rdi
- push rbx
- sub rsp, 16
- ; end prolog
-
-
- pxor mm5, mm5 ; Blank mmx6
- pxor mm6, mm6 ; Blank mmx7
- pxor mm7, mm7 ; Blank mmx7
-
- mov rax, arg(0) ;[src_ptr] ; Load base addresses
- mov rbx, arg(2) ;[ref_ptr]
- movsxd rcx, dword ptr arg(1) ;[source_stride]
- movsxd rdx, dword ptr arg(3) ;[recon_stride]
-
- ; Row 1
- movq mm0, [rax] ; Copy eight bytes to mm0
- movq mm1, [rbx] ; Copy eight bytes to mm1
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- paddw mm5, mm0 ; accumulate differences in mm5
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
-
- ; Row 2
- movq mm0, [rax] ; Copy eight bytes to mm0
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- paddw mm5, mm0 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
- ; Row 3
- movq mm0, [rax] ; Copy eight bytes to mm0
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- paddw mm5, mm0 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movq mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
- ; Row 4
- movq mm0, [rax] ; Copy eight bytes to mm0
-
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
-
- paddw mm5, mm0 ; accumulate differences in mm5
-
- pmaddwd mm0, mm0 ; square and accumulate
- paddd mm7, mm0 ; accumulate in mm7
-
-
- ; Now accumulate the final results.
- movq QWORD PTR [rsp+8], mm5 ; copy back accumulated results into normal memory
- movq QWORD PTR [rsp], mm7 ; copy back accumulated results into normal memory
- movsx rdx, WORD PTR [rsp+8]
- movsx rcx, WORD PTR [rsp+10]
- movsx rbx, WORD PTR [rsp+12]
- movsx rax, WORD PTR [rsp+14]
- add rdx, rcx
- add rbx, rax
- add rdx, rbx ;XSum
- movsxd rax, DWORD PTR [rsp]
- movsxd rcx, DWORD PTR [rsp+4]
- add rax, rcx ;XXSum
- mov rsi, arg(4) ;SSE
- mov rdi, arg(5) ;Sum
- mov dword ptr [rsi], eax
- mov dword ptr [rdi], edx
- xor rax, rax ; return 0
-
-
- ; begin epilog
- add rsp, 16
- pop rbx
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-;unsigned int
-;vp8_get4x4sse_cs_mmx
-;(
-; unsigned char *src_ptr,
-; int source_stride,
-; unsigned char *ref_ptr,
-; int recon_stride
-;)
-global sym(vp8_get4x4sse_cs_mmx)
-sym(vp8_get4x4sse_cs_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 4
- push rsi
- push rdi
- push rbx
- ; end prolog
-
-
- pxor mm6, mm6 ; Blank mmx7
- pxor mm7, mm7 ; Blank mmx7
-
- mov rax, arg(0) ;[src_ptr] ; Load base addresses
- mov rbx, arg(2) ;[ref_ptr]
- movsxd rcx, dword ptr arg(1) ;[source_stride]
- movsxd rdx, dword ptr arg(3) ;[recon_stride]
- ; Row 1
- movd mm0, [rax] ; Copy eight bytes to mm0
- movd mm1, [rbx] ; Copy eight bytes to mm1
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movd mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
- ; Row 2
- movd mm0, [rax] ; Copy eight bytes to mm0
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movd mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
- ; Row 3
- movd mm0, [rax] ; Copy eight bytes to mm0
- punpcklbw mm1, mm6
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- psubsw mm0, mm1 ; A-B (low order) to MM0
-
- pmaddwd mm0, mm0 ; square and accumulate
- add rbx,rdx ; Inc pointer into ref data
- add rax,rcx ; Inc pointer into the new data
- movd mm1, [rbx] ; Copy eight bytes to mm1
- paddd mm7, mm0 ; accumulate in mm7
-
- ; Row 4
- movd mm0, [rax] ; Copy eight bytes to mm0
- punpcklbw mm0, mm6 ; unpack to higher prrcision
- punpcklbw mm1, mm6
- psubsw mm0, mm1 ; A-B (low order) to MM0
- pmaddwd mm0, mm0 ; square and accumulate
- paddd mm7, mm0 ; accumulate in mm7
-
- movq mm0, mm7 ;
- psrlq mm7, 32
-
- paddd mm0, mm7
- movq rax, mm0
-
-
- ; begin epilog
- pop rbx
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
-
-%define mmx_filter_shift 7
-
-;void vp8_filter_block2d_bil4x4_var_mmx
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned short *HFilter,
-; unsigned short *VFilter,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_filter_block2d_bil4x4_var_mmx)
-sym(vp8_filter_block2d_bil4x4_var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 8
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
-
- pxor mm6, mm6 ;
- pxor mm7, mm7 ;
-
- mov rax, arg(4) ;HFilter ;
- mov rdx, arg(5) ;VFilter ;
-
- mov rsi, arg(0) ;ref_ptr ;
- mov rdi, arg(2) ;src_ptr ;
-
- mov rcx, 4 ;
- pxor mm0, mm0 ;
-
- movd mm1, [rsi] ;
- movd mm3, [rsi+1] ;
-
- punpcklbw mm1, mm0 ;
- pmullw mm1, [rax] ;
-
- punpcklbw mm3, mm0 ;
- pmullw mm3, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- movq mm5, mm1
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
- add rsi, r8
-%endif
-
-.filter_block2d_bil4x4_var_mmx_loop:
-
- movd mm1, [rsi] ;
- movd mm3, [rsi+1] ;
-
- punpcklbw mm1, mm0 ;
- pmullw mm1, [rax] ;
-
- punpcklbw mm3, mm0 ;
- pmullw mm3, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- movq mm3, mm5 ;
-
- movq mm5, mm1 ;
- pmullw mm3, [rdx] ;
-
- pmullw mm1, [rdx+8] ;
- paddw mm1, mm3 ;
-
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- psraw mm1, mmx_filter_shift ;
-
- movd mm3, [rdi] ;
- punpcklbw mm3, mm0 ;
-
- psubw mm1, mm3 ;
- paddw mm6, mm1 ;
-
- pmaddwd mm1, mm1 ;
- paddd mm7, mm1 ;
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
- add rdi, dword ptr arg(3) ;src_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
- add rsi, r8
- add rdi, r9
-%endif
- sub rcx, 1 ;
- jnz .filter_block2d_bil4x4_var_mmx_loop ;
-
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rdi, arg(6) ;sum
- mov rsi, arg(7) ;sumsquared
-
- movd dword ptr [rdi], mm2 ;
- movd dword ptr [rsi], mm4 ;
-
-
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-
-;void vp8_filter_block2d_bil_var_mmx
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; unsigned short *HFilter,
-; unsigned short *VFilter,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_filter_block2d_bil_var_mmx)
-sym(vp8_filter_block2d_bil_var_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 9
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
- pxor mm6, mm6 ;
- pxor mm7, mm7 ;
- mov rax, arg(5) ;HFilter ;
-
- mov rdx, arg(6) ;VFilter ;
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
-
- pxor mm0, mm0 ;
- movq mm1, [rsi] ;
-
- movq mm3, [rsi+1] ;
- movq mm2, mm1 ;
-
- movq mm4, mm3 ;
- punpcklbw mm1, mm0 ;
-
- punpckhbw mm2, mm0 ;
- pmullw mm1, [rax] ;
-
- pmullw mm2, [rax] ;
- punpcklbw mm3, mm0 ;
-
- punpckhbw mm4, mm0 ;
- pmullw mm3, [rax+8] ;
-
- pmullw mm4, [rax+8] ;
- paddw mm1, mm3 ;
-
- paddw mm2, mm4 ;
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm2, mmx_filter_shift ;
- movq mm5, mm1
-
- packuswb mm5, mm2 ;
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- add rsi, r8
-%endif
-
-.filter_block2d_bil_var_mmx_loop:
-
- movq mm1, [rsi] ;
- movq mm3, [rsi+1] ;
-
- movq mm2, mm1 ;
- movq mm4, mm3 ;
-
- punpcklbw mm1, mm0 ;
- punpckhbw mm2, mm0 ;
-
- pmullw mm1, [rax] ;
- pmullw mm2, [rax] ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- pmullw mm3, [rax+8] ;
- pmullw mm4, [rax+8] ;
-
- paddw mm1, mm3 ;
- paddw mm2, mm4 ;
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- psraw mm1, mmx_filter_shift ;
-
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
- psraw mm2, mmx_filter_shift ;
-
- movq mm3, mm5 ;
- movq mm4, mm5 ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- movq mm5, mm1 ;
- packuswb mm5, mm2 ;
-
- pmullw mm3, [rdx] ;
- pmullw mm4, [rdx] ;
-
- pmullw mm1, [rdx+8] ;
- pmullw mm2, [rdx+8] ;
-
- paddw mm1, mm3 ;
- paddw mm2, mm4 ;
-
- paddw mm1, [GLOBAL(mmx_bi_rd)] ;
- paddw mm2, [GLOBAL(mmx_bi_rd)] ;
-
- psraw mm1, mmx_filter_shift ;
- psraw mm2, mmx_filter_shift ;
-
- movq mm3, [rdi] ;
- movq mm4, mm3 ;
-
- punpcklbw mm3, mm0 ;
- punpckhbw mm4, mm0 ;
-
- psubw mm1, mm3 ;
- psubw mm2, mm4 ;
-
- paddw mm6, mm1 ;
- pmaddwd mm1, mm1 ;
-
- paddw mm6, mm2 ;
- pmaddwd mm2, mm2 ;
-
- paddd mm7, mm1 ;
- paddd mm7, mm2 ;
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ;
- add rdi, dword ptr arg(3) ;src_pixels_per_line ;
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line ;
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line ;
- add rsi, r8
- add rdi, r9
-%endif
- sub rcx, 1 ;
- jnz .filter_block2d_bil_var_mmx_loop ;
-
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rdi, arg(7) ;sum
- mov rsi, arg(8) ;sumsquared
-
- movd dword ptr [rdi], mm2 ;
- movd dword ptr [rsi], mm4 ;
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-SECTION_RODATA
-;short mmx_bi_rd[4] = { 64, 64, 64, 64};-align 16
-mmx_bi_rd:
- times 4 dw 64
--- a/vp8/encoder/x86/variance_impl_sse2.asm
+++ /dev/null
@@ -1,1359 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%define xmm_filter_shift 7
-
-;unsigned int vp8_get_mb_ss_sse2
-;(
-; short *src_ptr
-;)
-global sym(vp8_get_mb_ss_sse2)
-sym(vp8_get_mb_ss_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 1
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
-
- mov rax, arg(0) ;[src_ptr]
- mov rcx, 8
- pxor xmm4, xmm4
-
-.NEXTROW:
- movdqa xmm0, [rax]
- movdqa xmm1, [rax+16]
- movdqa xmm2, [rax+32]
- movdqa xmm3, [rax+48]
- pmaddwd xmm0, xmm0
- pmaddwd xmm1, xmm1
- pmaddwd xmm2, xmm2
- pmaddwd xmm3, xmm3
-
- paddd xmm0, xmm1
- paddd xmm2, xmm3
- paddd xmm4, xmm0
- paddd xmm4, xmm2
-
- add rax, 0x40
- dec rcx
- ja .NEXTROW
-
- movdqa xmm3,xmm4
- psrldq xmm4,8
- paddd xmm4,xmm3
- movdqa xmm3,xmm4
- psrldq xmm4,4
- paddd xmm4,xmm3
- movq rax,xmm4
-
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;unsigned int vp8_get16x16var_sse2
-;(
-; unsigned char * src_ptr,
-; int source_stride,
-; unsigned char * ref_ptr,
-; int recon_stride,
-; unsigned int * SSE,
-; int * Sum
-;)
-global sym(vp8_get16x16var_sse2)
-sym(vp8_get16x16var_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- SAVE_XMM 7
- push rbx
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(0) ;[src_ptr]
- mov rdi, arg(2) ;[ref_ptr]
-
- movsxd rax, DWORD PTR arg(1) ;[source_stride]
- movsxd rdx, DWORD PTR arg(3) ;[recon_stride]
-
- ; Prefetch data
- lea rcx, [rax+rax*2]
- prefetcht0 [rsi]
- prefetcht0 [rsi+rax]
- prefetcht0 [rsi+rax*2]
- prefetcht0 [rsi+rcx]
- lea rbx, [rsi+rax*4]
- prefetcht0 [rbx]
- prefetcht0 [rbx+rax]
- prefetcht0 [rbx+rax*2]
- prefetcht0 [rbx+rcx]
-
- lea rcx, [rdx+rdx*2]
- prefetcht0 [rdi]
- prefetcht0 [rdi+rdx]
- prefetcht0 [rdi+rdx*2]
- prefetcht0 [rdi+rcx]
- lea rbx, [rdi+rdx*4]
- prefetcht0 [rbx]
- prefetcht0 [rbx+rdx]
- prefetcht0 [rbx+rdx*2]
- prefetcht0 [rbx+rcx]
-
- pxor xmm0, xmm0 ; clear xmm0 for unpack
- pxor xmm7, xmm7 ; clear xmm7 for accumulating diffs
-
- pxor xmm6, xmm6 ; clear xmm6 for accumulating sse
- mov rcx, 16
-
-.var16loop:
- movdqu xmm1, XMMWORD PTR [rsi]
- movdqu xmm2, XMMWORD PTR [rdi]
-
- prefetcht0 [rsi+rax*8]
- prefetcht0 [rdi+rdx*8]
-
- movdqa xmm3, xmm1
- movdqa xmm4, xmm2
-
-
- punpcklbw xmm1, xmm0
- punpckhbw xmm3, xmm0
-
- punpcklbw xmm2, xmm0
- punpckhbw xmm4, xmm0
-
-
- psubw xmm1, xmm2
- psubw xmm3, xmm4
-
- paddw xmm7, xmm1
- pmaddwd xmm1, xmm1
-
- paddw xmm7, xmm3
- pmaddwd xmm3, xmm3
-
- paddd xmm6, xmm1
- paddd xmm6, xmm3
-
- add rsi, rax
- add rdi, rdx
-
- sub rcx, 1
- jnz .var16loop
-
-
- movdqa xmm1, xmm6
- pxor xmm6, xmm6
-
- pxor xmm5, xmm5
- punpcklwd xmm6, xmm7
-
- punpckhwd xmm5, xmm7
- psrad xmm5, 16
-
- psrad xmm6, 16
- paddd xmm6, xmm5
-
- movdqa xmm2, xmm1
- punpckldq xmm1, xmm0
-
- punpckhdq xmm2, xmm0
- movdqa xmm7, xmm6
-
- paddd xmm1, xmm2
- punpckldq xmm6, xmm0
-
- punpckhdq xmm7, xmm0
- paddd xmm6, xmm7
-
- movdqa xmm2, xmm1
- movdqa xmm7, xmm6
-
- psrldq xmm1, 8
- psrldq xmm6, 8
-
- paddd xmm7, xmm6
- paddd xmm1, xmm2
-
- mov rax, arg(5) ;[Sum]
- mov rdi, arg(4) ;[SSE]
-
- movd DWORD PTR [rax], xmm7
- movd DWORD PTR [rdi], xmm1
-
-
- ; begin epilog
- pop rdi
- pop rsi
- pop rbx
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-
-
-;unsigned int vp8_get8x8var_sse2
-;(
-; unsigned char * src_ptr,
-; int source_stride,
-; unsigned char * ref_ptr,
-; int recon_stride,
-; unsigned int * SSE,
-; int * Sum
-;)
-global sym(vp8_get8x8var_sse2)
-sym(vp8_get8x8var_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- sub rsp, 16
- ; end prolog
-
- mov rsi, arg(0) ;[src_ptr]
- mov rdi, arg(2) ;[ref_ptr]
-
- movsxd rax, DWORD PTR arg(1) ;[source_stride]
- movsxd rdx, DWORD PTR arg(3) ;[recon_stride]
-
- pxor xmm0, xmm0 ; clear xmm0 for unpack
- pxor xmm7, xmm7 ; clear xmm7 for accumulating diffs
-
- movq xmm1, QWORD PTR [rsi]
- movq xmm2, QWORD PTR [rdi]
-
- punpcklbw xmm1, xmm0
- punpcklbw xmm2, xmm0
-
- psubsw xmm1, xmm2
- paddw xmm7, xmm1
-
- pmaddwd xmm1, xmm1
-
- movq xmm2, QWORD PTR[rsi + rax]
- movq xmm3, QWORD PTR[rdi + rdx]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
-
- movq xmm2, QWORD PTR[rsi + rax * 2]
- movq xmm3, QWORD PTR[rdi + rdx * 2]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
-
- lea rsi, [rsi + rax * 2]
- lea rdi, [rdi + rdx * 2]
- movq xmm2, QWORD PTR[rsi + rax]
- movq xmm3, QWORD PTR[rdi + rdx]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
- movq xmm2, QWORD PTR[rsi + rax *2]
- movq xmm3, QWORD PTR[rdi + rdx *2]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
-
- lea rsi, [rsi + rax * 2]
- lea rdi, [rdi + rdx * 2]
-
-
- movq xmm2, QWORD PTR[rsi + rax]
- movq xmm3, QWORD PTR[rdi + rdx]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
- movq xmm2, QWORD PTR[rsi + rax *2]
- movq xmm3, QWORD PTR[rdi + rdx *2]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
-
- lea rsi, [rsi + rax * 2]
- lea rdi, [rdi + rdx * 2]
-
- movq xmm2, QWORD PTR[rsi + rax]
- movq xmm3, QWORD PTR[rdi + rdx]
-
- punpcklbw xmm2, xmm0
- punpcklbw xmm3, xmm0
-
- psubsw xmm2, xmm3
- paddw xmm7, xmm2
-
- pmaddwd xmm2, xmm2
- paddd xmm1, xmm2
-
-
- movdqa xmm6, xmm7
- punpcklwd xmm6, xmm0
-
- punpckhwd xmm7, xmm0
- movdqa xmm2, xmm1
-
- paddw xmm6, xmm7
- punpckldq xmm1, xmm0
-
- punpckhdq xmm2, xmm0
- movdqa xmm7, xmm6
-
- paddd xmm1, xmm2
- punpckldq xmm6, xmm0
-
- punpckhdq xmm7, xmm0
- paddw xmm6, xmm7
-
- movdqa xmm2, xmm1
- movdqa xmm7, xmm6
-
- psrldq xmm1, 8
- psrldq xmm6, 8
-
- paddw xmm7, xmm6
- paddd xmm1, xmm2
-
- mov rax, arg(5) ;[Sum]
- mov rdi, arg(4) ;[SSE]
-
- movq rdx, xmm7
- movsx rcx, dx
-
- mov dword ptr [rax], ecx
- movd DWORD PTR [rdi], xmm1
-
- ; begin epilog
- add rsp, 16
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_filter_block2d_bil_var_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int xoffset,
-; int yoffset,
-; int *sum,
-; unsigned int *sumsquared;;
-;
-;)
-global sym(vp8_filter_block2d_bil_var_sse2)
-sym(vp8_filter_block2d_bil_var_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 9
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- push rbx
- ; end prolog
-
- pxor xmm6, xmm6 ;
- pxor xmm7, xmm7 ;
-
- lea rsi, [GLOBAL(xmm_bi_rd)] ; rounding
- movdqa xmm4, XMMWORD PTR [rsi]
-
- lea rcx, [GLOBAL(vp8_bilinear_filters_sse2)]
- movsxd rax, dword ptr arg(5) ; xoffset
-
- cmp rax, 0 ; skip first_pass filter if xoffset=0
- je filter_block2d_bil_var_sse2_sp_only
-
- shl rax, 5 ; point to filter coeff with xoffset
- lea rax, [rax + rcx] ; HFilter
-
- movsxd rdx, dword ptr arg(6) ; yoffset
-
- cmp rdx, 0 ; skip second_pass filter if yoffset=0
- je filter_block2d_bil_var_sse2_fp_only
-
- shl rdx, 5
- lea rdx, [rdx + rcx] ; VFilter
-
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
-
- pxor xmm0, xmm0 ;
- movq xmm1, QWORD PTR [rsi] ;
- movq xmm3, QWORD PTR [rsi+1] ;
-
- punpcklbw xmm1, xmm0 ;
- pmullw xmm1, [rax] ;
- punpcklbw xmm3, xmm0
- pmullw xmm3, [rax+16] ;
-
- paddw xmm1, xmm3 ;
- paddw xmm1, xmm4 ;
- psraw xmm1, xmm_filter_shift ;
- movdqa xmm5, xmm1
-
- movsxd rbx, dword ptr arg(1) ;ref_pixels_per_line
- lea rsi, [rsi + rbx]
-%if ABI_IS_32BIT=0
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
-filter_block2d_bil_var_sse2_loop:
- movq xmm1, QWORD PTR [rsi] ;
- movq xmm3, QWORD PTR [rsi+1] ;
-
- punpcklbw xmm1, xmm0 ;
- pmullw xmm1, [rax] ;
- punpcklbw xmm3, xmm0 ;
- pmullw xmm3, [rax+16] ;
-
- paddw xmm1, xmm3 ;
- paddw xmm1, xmm4 ;
- psraw xmm1, xmm_filter_shift ;
-
- movdqa xmm3, xmm5 ;
- movdqa xmm5, xmm1 ;
-
- pmullw xmm3, [rdx] ;
- pmullw xmm1, [rdx+16] ;
- paddw xmm1, xmm3 ;
- paddw xmm1, xmm4 ;
- psraw xmm1, xmm_filter_shift ;
-
- movq xmm3, QWORD PTR [rdi] ;
- punpcklbw xmm3, xmm0 ;
-
- psubw xmm1, xmm3 ;
- paddw xmm6, xmm1 ;
-
- pmaddwd xmm1, xmm1 ;
- paddd xmm7, xmm1 ;
-
- lea rsi, [rsi + rbx] ;ref_pixels_per_line
-%if ABI_IS_32BIT
- add rdi, dword ptr arg(3) ;src_pixels_per_line
-%else
- lea rdi, [rdi + r9]
-%endif
-
- sub rcx, 1 ;
- jnz filter_block2d_bil_var_sse2_loop ;
-
- jmp filter_block2d_bil_variance
-
-filter_block2d_bil_var_sse2_sp_only:
- movsxd rdx, dword ptr arg(6) ; yoffset
-
- cmp rdx, 0 ; skip all if both xoffset=0 and yoffset=0
- je filter_block2d_bil_var_sse2_full_pixel
-
- shl rdx, 5
- lea rdx, [rdx + rcx] ; VFilter
-
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
-
- pxor xmm0, xmm0 ;
- movq xmm1, QWORD PTR [rsi] ;
- punpcklbw xmm1, xmm0 ;
-
- movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
- lea rsi, [rsi + rax]
-
-filter_block2d_bil_sp_only_loop:
- movq xmm3, QWORD PTR [rsi] ;
- punpcklbw xmm3, xmm0 ;
- movdqa xmm5, xmm3
-
- pmullw xmm1, [rdx] ;
- pmullw xmm3, [rdx+16] ;
- paddw xmm1, xmm3 ;
- paddw xmm1, xmm4 ;
- psraw xmm1, xmm_filter_shift ;
-
- movq xmm3, QWORD PTR [rdi] ;
- punpcklbw xmm3, xmm0 ;
-
- psubw xmm1, xmm3 ;
- paddw xmm6, xmm1 ;
-
- pmaddwd xmm1, xmm1 ;
- paddd xmm7, xmm1 ;
-
- movdqa xmm1, xmm5 ;
- lea rsi, [rsi + rax] ;ref_pixels_per_line
- lea rdi, [rdi + rbx] ;src_pixels_per_line
-
- sub rcx, 1 ;
- jnz filter_block2d_bil_sp_only_loop ;
-
- jmp filter_block2d_bil_variance
-
-filter_block2d_bil_var_sse2_full_pixel:
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
- movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
- pxor xmm0, xmm0 ;
-
-filter_block2d_bil_full_pixel_loop:
- movq xmm1, QWORD PTR [rsi] ;
- punpcklbw xmm1, xmm0 ;
-
- movq xmm2, QWORD PTR [rdi] ;
- punpcklbw xmm2, xmm0 ;
-
- psubw xmm1, xmm2 ;
- paddw xmm6, xmm1 ;
-
- pmaddwd xmm1, xmm1 ;
- paddd xmm7, xmm1 ;
-
- lea rsi, [rsi + rax] ;ref_pixels_per_line
- lea rdi, [rdi + rbx] ;src_pixels_per_line
-
- sub rcx, 1 ;
- jnz filter_block2d_bil_full_pixel_loop ;
-
- jmp filter_block2d_bil_variance
-
-filter_block2d_bil_var_sse2_fp_only:
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
-
- pxor xmm0, xmm0 ;
- movsxd rbx, dword ptr arg(3) ;src_pixels_per_line
-
-filter_block2d_bil_fp_only_loop:
- movq xmm1, QWORD PTR [rsi] ;
- movq xmm3, QWORD PTR [rsi+1] ;
-
- punpcklbw xmm1, xmm0 ;
- pmullw xmm1, [rax] ;
- punpcklbw xmm3, xmm0 ;
- pmullw xmm3, [rax+16] ;
-
- paddw xmm1, xmm3 ;
- paddw xmm1, xmm4 ;
- psraw xmm1, xmm_filter_shift ;
-
- movq xmm3, QWORD PTR [rdi] ;
- punpcklbw xmm3, xmm0 ;
-
- psubw xmm1, xmm3 ;
- paddw xmm6, xmm1 ;
-
- pmaddwd xmm1, xmm1 ;
- paddd xmm7, xmm1 ;
- lea rsi, [rsi + rdx]
- lea rdi, [rdi + rbx] ;src_pixels_per_line
-
- sub rcx, 1 ;
- jnz filter_block2d_bil_fp_only_loop ;
-
- jmp filter_block2d_bil_variance
-
-filter_block2d_bil_variance:
- movdq2q mm6, xmm6 ;
- movdq2q mm7, xmm7 ;
-
- psrldq xmm6, 8
- psrldq xmm7, 8
-
- movdq2q mm2, xmm6
- movdq2q mm3, xmm7
-
- paddw mm6, mm2
- paddd mm7, mm3
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rsi, arg(7) ; sum
- mov rdi, arg(8) ; sumsquared
-
- movd [rsi], mm2 ; xsum
- movd [rdi], mm4 ; xxsum
-
- ; begin epilog
- pop rbx
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_half_horiz_vert_variance8x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_horiz_vert_variance8x_h_sse2)
-sym(vp8_half_horiz_vert_variance8x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
-%if ABI_IS_32BIT=0
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
-
- pxor xmm0, xmm0 ;
-
- movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
- movq xmm3, QWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s9
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3) horizontal line 1
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line ; next source
-%else
- add rsi, r8
-%endif
-
-vp8_half_horiz_vert_variance8x_h_1:
-
- movq xmm1, QWORD PTR [rsi] ;
- movq xmm2, QWORD PTR [rsi+1] ;
- pavgb xmm1, xmm2 ; xmm1 = avg(xmm1,xmm3) horizontal line i+1
-
- pavgb xmm5, xmm1 ; xmm = vertical average of the above
- punpcklbw xmm5, xmm0 ; xmm5 = words of above
-
- movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
- punpcklbw xmm3, xmm0 ; xmm3 = words of above
-
- psubw xmm5, xmm3 ; xmm5 -= xmm3
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
-
- movdqa xmm5, xmm1 ; save xmm1 for use on the next row
-
-%if ABI_IS_32BIT
- add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
- add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
-%else
- add rsi, r8
- add rdi, r9
-%endif
-
- sub rcx, 1 ;
- jnz vp8_half_horiz_vert_variance8x_h_1 ;
-
- movdq2q mm6, xmm6 ;
- movdq2q mm7, xmm7 ;
-
- psrldq xmm6, 8
- psrldq xmm7, 8
-
- movdq2q mm2, xmm6
- movdq2q mm3, xmm7
-
- paddw mm6, mm2
- paddd mm7, mm3
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rsi, arg(5) ; sum
- mov rdi, arg(6) ; sumsquared
-
- movd [rsi], mm2 ;
- movd [rdi], mm4 ;
-
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_half_horiz_vert_variance16x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_horiz_vert_variance16x_h_sse2)
-sym(vp8_half_horiz_vert_variance16x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
- movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
-
- pxor xmm0, xmm0 ;
-
- movdqu xmm5, XMMWORD PTR [rsi]
- movdqu xmm3, XMMWORD PTR [rsi+1]
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3) horizontal line 1
-
- lea rsi, [rsi + rax]
-
-vp8_half_horiz_vert_variance16x_h_1:
- movdqu xmm1, XMMWORD PTR [rsi] ;
- movdqu xmm2, XMMWORD PTR [rsi+1] ;
- pavgb xmm1, xmm2 ; xmm1 = avg(xmm1,xmm3) horizontal line i+1
-
- pavgb xmm5, xmm1 ; xmm = vertical average of the above
-
- movdqa xmm4, xmm5
- punpcklbw xmm5, xmm0 ; xmm5 = words of above
- punpckhbw xmm4, xmm0
-
- movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
- punpcklbw xmm3, xmm0 ; xmm3 = words of above
- psubw xmm5, xmm3 ; xmm5 -= xmm3
-
- movq xmm3, QWORD PTR [rdi+8]
- punpcklbw xmm3, xmm0
- psubw xmm4, xmm3
-
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- paddw xmm6, xmm4
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- pmaddwd xmm4, xmm4
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
- paddd xmm7, xmm4
-
- movdqa xmm5, xmm1 ; save xmm1 for use on the next row
-
- lea rsi, [rsi + rax]
- lea rdi, [rdi + rdx]
-
- sub rcx, 1 ;
- jnz vp8_half_horiz_vert_variance16x_h_1 ;
-
- pxor xmm1, xmm1
- pxor xmm5, xmm5
-
- punpcklwd xmm0, xmm6
- punpckhwd xmm1, xmm6
- psrad xmm0, 16
- psrad xmm1, 16
- paddd xmm0, xmm1
- movdqa xmm1, xmm0
-
- movdqa xmm6, xmm7
- punpckldq xmm6, xmm5
- punpckhdq xmm7, xmm5
- paddd xmm6, xmm7
-
- punpckldq xmm0, xmm5
- punpckhdq xmm1, xmm5
- paddd xmm0, xmm1
-
- movdqa xmm7, xmm6
- movdqa xmm1, xmm0
-
- psrldq xmm7, 8
- psrldq xmm1, 8
-
- paddd xmm6, xmm7
- paddd xmm0, xmm1
-
- mov rsi, arg(5) ;[Sum]
- mov rdi, arg(6) ;[SSE]
-
- movd [rsi], xmm0
- movd [rdi], xmm6
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_half_vert_variance8x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_vert_variance8x_h_sse2)
-sym(vp8_half_vert_variance8x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
-%if ABI_IS_32BIT=0
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
-
- pxor xmm0, xmm0 ;
-vp8_half_vert_variance8x_h_1:
- movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
- movq xmm3, QWORD PTR [rsi+rax] ; xmm3 = s1,s2,s3..s9
-
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
- punpcklbw xmm5, xmm0 ; xmm5 = words of above
-
- movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
- punpcklbw xmm3, xmm0 ; xmm3 = words of above
-
- psubw xmm5, xmm3 ; xmm5 -= xmm3
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
-
-%if ABI_IS_32BIT
- add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
- add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
-%else
- add rsi, r8
- add rdi, r9
-%endif
-
- sub rcx, 1 ;
- jnz vp8_half_vert_variance8x_h_1 ;
-
- movdq2q mm6, xmm6 ;
- movdq2q mm7, xmm7 ;
-
- psrldq xmm6, 8
- psrldq xmm7, 8
-
- movdq2q mm2, xmm6
- movdq2q mm3, xmm7
-
- paddw mm6, mm2
- paddd mm7, mm3
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rsi, arg(5) ; sum
- mov rdi, arg(6) ; sumsquared
-
- movd [rsi], mm2 ;
- movd [rdi], mm4 ;
-
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_half_vert_variance16x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_vert_variance16x_h_sse2)
-sym(vp8_half_vert_variance16x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr
-
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
- movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
-
- movdqu xmm5, XMMWORD PTR [rsi]
- lea rsi, [rsi + rax ]
- pxor xmm0, xmm0
-
-vp8_half_vert_variance16x_h_1:
- movdqu xmm3, XMMWORD PTR [rsi]
-
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
- movdqa xmm4, xmm5
- punpcklbw xmm5, xmm0
- punpckhbw xmm4, xmm0
-
- movq xmm2, QWORD PTR [rdi]
- punpcklbw xmm2, xmm0
- psubw xmm5, xmm2
- movq xmm2, QWORD PTR [rdi+8]
- punpcklbw xmm2, xmm0
- psubw xmm4, xmm2
-
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- paddw xmm6, xmm4
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- pmaddwd xmm4, xmm4
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
- paddd xmm7, xmm4
-
- movdqa xmm5, xmm3
-
- lea rsi, [rsi + rax]
- lea rdi, [rdi + rdx]
-
- sub rcx, 1
- jnz vp8_half_vert_variance16x_h_1
-
- pxor xmm1, xmm1
- pxor xmm5, xmm5
-
- punpcklwd xmm0, xmm6
- punpckhwd xmm1, xmm6
- psrad xmm0, 16
- psrad xmm1, 16
- paddd xmm0, xmm1
- movdqa xmm1, xmm0
-
- movdqa xmm6, xmm7
- punpckldq xmm6, xmm5
- punpckhdq xmm7, xmm5
- paddd xmm6, xmm7
-
- punpckldq xmm0, xmm5
- punpckhdq xmm1, xmm5
- paddd xmm0, xmm1
-
- movdqa xmm7, xmm6
- movdqa xmm1, xmm0
-
- psrldq xmm7, 8
- psrldq xmm1, 8
-
- paddd xmm6, xmm7
- paddd xmm0, xmm1
-
- mov rsi, arg(5) ;[Sum]
- mov rdi, arg(6) ;[SSE]
-
- movd [rsi], xmm0
- movd [rdi], xmm6
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-;void vp8_half_horiz_variance8x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_horiz_variance8x_h_sse2)
-sym(vp8_half_horiz_variance8x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
-%if ABI_IS_32BIT=0
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
-
- pxor xmm0, xmm0 ;
-vp8_half_horiz_variance8x_h_1:
- movq xmm5, QWORD PTR [rsi] ; xmm5 = s0,s1,s2..s8
- movq xmm3, QWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s9
-
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
- punpcklbw xmm5, xmm0 ; xmm5 = words of above
-
- movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d8
- punpcklbw xmm3, xmm0 ; xmm3 = words of above
-
- psubw xmm5, xmm3 ; xmm5 -= xmm3
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
-
-%if ABI_IS_32BIT
- add esi, dword ptr arg(1) ;ref_pixels_per_line ; next source
- add edi, dword ptr arg(3) ;src_pixels_per_line ; next destination
-%else
- add rsi, r8
- add rdi, r9
-%endif
- sub rcx, 1 ;
- jnz vp8_half_horiz_variance8x_h_1 ;
-
- movdq2q mm6, xmm6 ;
- movdq2q mm7, xmm7 ;
-
- psrldq xmm6, 8
- psrldq xmm7, 8
-
- movdq2q mm2, xmm6
- movdq2q mm3, xmm7
-
- paddw mm6, mm2
- paddd mm7, mm3
-
- pxor mm3, mm3 ;
- pxor mm2, mm2 ;
-
- punpcklwd mm2, mm6 ;
- punpckhwd mm3, mm6 ;
-
- paddd mm2, mm3 ;
- movq mm6, mm2 ;
-
- psrlq mm6, 32 ;
- paddd mm2, mm6 ;
-
- psrad mm2, 16 ;
- movq mm4, mm7 ;
-
- psrlq mm4, 32 ;
- paddd mm4, mm7 ;
-
- mov rsi, arg(5) ; sum
- mov rdi, arg(6) ; sumsquared
-
- movd [rsi], mm2 ;
- movd [rdi], mm4 ;
-
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_half_horiz_variance16x_h_sse2
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int *sum,
-; unsigned int *sumsquared
-;)
-global sym(vp8_half_horiz_variance16x_h_sse2)
-sym(vp8_half_horiz_variance16x_h_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- pxor xmm6, xmm6 ; error accumulator
- pxor xmm7, xmm7 ; sse eaccumulator
- mov rsi, arg(0) ;ref_ptr ;
-
- mov rdi, arg(2) ;src_ptr ;
- movsxd rcx, dword ptr arg(4) ;Height ;
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
- movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
-
- pxor xmm0, xmm0 ;
-
-vp8_half_horiz_variance16x_h_1:
- movdqu xmm5, XMMWORD PTR [rsi] ; xmm5 = s0,s1,s2..s15
- movdqu xmm3, XMMWORD PTR [rsi+1] ; xmm3 = s1,s2,s3..s16
-
- pavgb xmm5, xmm3 ; xmm5 = avg(xmm1,xmm3)
- movdqa xmm1, xmm5
- punpcklbw xmm5, xmm0 ; xmm5 = words of above
- punpckhbw xmm1, xmm0
-
- movq xmm3, QWORD PTR [rdi] ; xmm3 = d0,d1,d2..d7
- punpcklbw xmm3, xmm0 ; xmm3 = words of above
- movq xmm2, QWORD PTR [rdi+8]
- punpcklbw xmm2, xmm0
-
- psubw xmm5, xmm3 ; xmm5 -= xmm3
- psubw xmm1, xmm2
- paddw xmm6, xmm5 ; xmm6 += accumulated column differences
- paddw xmm6, xmm1
- pmaddwd xmm5, xmm5 ; xmm5 *= xmm5
- pmaddwd xmm1, xmm1
- paddd xmm7, xmm5 ; xmm7 += accumulated square column differences
- paddd xmm7, xmm1
-
- lea rsi, [rsi + rax]
- lea rdi, [rdi + rdx]
-
- sub rcx, 1 ;
- jnz vp8_half_horiz_variance16x_h_1 ;
-
- pxor xmm1, xmm1
- pxor xmm5, xmm5
-
- punpcklwd xmm0, xmm6
- punpckhwd xmm1, xmm6
- psrad xmm0, 16
- psrad xmm1, 16
- paddd xmm0, xmm1
- movdqa xmm1, xmm0
-
- movdqa xmm6, xmm7
- punpckldq xmm6, xmm5
- punpckhdq xmm7, xmm5
- paddd xmm6, xmm7
-
- punpckldq xmm0, xmm5
- punpckhdq xmm1, xmm5
- paddd xmm0, xmm1
-
- movdqa xmm7, xmm6
- movdqa xmm1, xmm0
-
- psrldq xmm7, 8
- psrldq xmm1, 8
-
- paddd xmm6, xmm7
- paddd xmm0, xmm1
-
- mov rsi, arg(5) ;[Sum]
- mov rdi, arg(6) ;[SSE]
-
- movd [rsi], xmm0
- movd [rdi], xmm6
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-SECTION_RODATA
-; short xmm_bi_rd[8] = { 64, 64, 64, 64,64, 64, 64, 64};-align 16
-xmm_bi_rd:
- times 8 dw 64
-align 16
-vp8_bilinear_filters_sse2:
- dw 128, 128, 128, 128, 128, 128, 128, 128, 0, 0, 0, 0, 0, 0, 0, 0
- dw 112, 112, 112, 112, 112, 112, 112, 112, 16, 16, 16, 16, 16, 16, 16, 16
- dw 96, 96, 96, 96, 96, 96, 96, 96, 32, 32, 32, 32, 32, 32, 32, 32
- dw 80, 80, 80, 80, 80, 80, 80, 80, 48, 48, 48, 48, 48, 48, 48, 48
- dw 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
- dw 48, 48, 48, 48, 48, 48, 48, 48, 80, 80, 80, 80, 80, 80, 80, 80
- dw 32, 32, 32, 32, 32, 32, 32, 32, 96, 96, 96, 96, 96, 96, 96, 96
- dw 16, 16, 16, 16, 16, 16, 16, 16, 112, 112, 112, 112, 112, 112, 112, 112
--- a/vp8/encoder/x86/variance_impl_ssse3.asm
+++ /dev/null
@@ -1,364 +1,0 @@
-;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-; Use of this source code is governed by a BSD-style license
-; that can be found in the LICENSE file in the root of the source
-; tree. An additional intellectual property rights grant can be found
-; in the file PATENTS. All contributing project authors may
-; be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-
-%define xmm_filter_shift 7
-
-
-;void vp8_filter_block2d_bil_var_ssse3
-;(
-; unsigned char *ref_ptr,
-; int ref_pixels_per_line,
-; unsigned char *src_ptr,
-; int src_pixels_per_line,
-; unsigned int Height,
-; int xoffset,
-; int yoffset,
-; int *sum,
-; unsigned int *sumsquared;;
-;
-;)
-;Note: The filter coefficient at offset=0 is 128. Since the second register
-;for Pmaddubsw is signed bytes, we must calculate zero offset seperately.
-global sym(vp8_filter_block2d_bil_var_ssse3)
-sym(vp8_filter_block2d_bil_var_ssse3):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 9
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- pxor xmm6, xmm6
- pxor xmm7, xmm7
-
- lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
- movsxd rax, dword ptr arg(5) ; xoffset
-
- cmp rax, 0 ; skip first_pass filter if xoffset=0
- je .filter_block2d_bil_var_ssse3_sp_only
-
- shl rax, 4 ; point to filter coeff with xoffset
- lea rax, [rax + rcx] ; HFilter
-
- movsxd rdx, dword ptr arg(6) ; yoffset
-
- cmp rdx, 0 ; skip second_pass filter if yoffset=0
- je .filter_block2d_bil_var_ssse3_fp_only
-
- shl rdx, 4
- lea rdx, [rdx + rcx] ; VFilter
-
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
-
- movdqu xmm0, XMMWORD PTR [rsi]
- movdqu xmm1, XMMWORD PTR [rsi+1]
- movdqa xmm2, xmm0
-
- punpcklbw xmm0, xmm1
- punpckhbw xmm2, xmm1
- pmaddubsw xmm0, [rax]
- pmaddubsw xmm2, [rax]
-
- paddw xmm0, [GLOBAL(xmm_bi_rd)]
- paddw xmm2, [GLOBAL(xmm_bi_rd)]
- psraw xmm0, xmm_filter_shift
- psraw xmm2, xmm_filter_shift
-
- packuswb xmm0, xmm2
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line
-%else
- movsxd r8, dword ptr arg(1) ;ref_pixels_per_line
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
- lea rsi, [rsi + r8]
-%endif
-
-.filter_block2d_bil_var_ssse3_loop:
- movdqu xmm1, XMMWORD PTR [rsi]
- movdqu xmm2, XMMWORD PTR [rsi+1]
- movdqa xmm3, xmm1
-
- punpcklbw xmm1, xmm2
- punpckhbw xmm3, xmm2
- pmaddubsw xmm1, [rax]
- pmaddubsw xmm3, [rax]
-
- paddw xmm1, [GLOBAL(xmm_bi_rd)]
- paddw xmm3, [GLOBAL(xmm_bi_rd)]
- psraw xmm1, xmm_filter_shift
- psraw xmm3, xmm_filter_shift
- packuswb xmm1, xmm3
-
- movdqa xmm2, xmm0
- movdqa xmm0, xmm1
- movdqa xmm3, xmm2
-
- punpcklbw xmm2, xmm1
- punpckhbw xmm3, xmm1
- pmaddubsw xmm2, [rdx]
- pmaddubsw xmm3, [rdx]
-
- paddw xmm2, [GLOBAL(xmm_bi_rd)]
- paddw xmm3, [GLOBAL(xmm_bi_rd)]
- psraw xmm2, xmm_filter_shift
- psraw xmm3, xmm_filter_shift
-
- movq xmm1, QWORD PTR [rdi]
- pxor xmm4, xmm4
- punpcklbw xmm1, xmm4
- movq xmm5, QWORD PTR [rdi+8]
- punpcklbw xmm5, xmm4
-
- psubw xmm2, xmm1
- psubw xmm3, xmm5
- paddw xmm6, xmm2
- paddw xmm6, xmm3
- pmaddwd xmm2, xmm2
- pmaddwd xmm3, xmm3
- paddd xmm7, xmm2
- paddd xmm7, xmm3
-
-%if ABI_IS_32BIT
- add rsi, dword ptr arg(1) ;ref_pixels_per_line
- add rdi, dword ptr arg(3) ;src_pixels_per_line
-%else
- lea rsi, [rsi + r8]
- lea rdi, [rdi + r9]
-%endif
-
- sub rcx, 1
- jnz .filter_block2d_bil_var_ssse3_loop
-
- jmp .filter_block2d_bil_variance
-
-.filter_block2d_bil_var_ssse3_sp_only:
- movsxd rdx, dword ptr arg(6) ; yoffset
-
- cmp rdx, 0 ; Both xoffset =0 and yoffset=0
- je .filter_block2d_bil_var_ssse3_full_pixel
-
- shl rdx, 4
- lea rdx, [rdx + rcx] ; VFilter
-
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
-
- movdqu xmm1, XMMWORD PTR [rsi]
- movdqa xmm0, xmm1
-
-%if ABI_IS_32BIT=0
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
- lea rsi, [rsi + rax]
-
-.filter_block2d_bil_sp_only_loop:
- movdqu xmm3, XMMWORD PTR [rsi]
- movdqa xmm2, xmm1
- movdqa xmm0, xmm3
-
- punpcklbw xmm1, xmm3
- punpckhbw xmm2, xmm3
- pmaddubsw xmm1, [rdx]
- pmaddubsw xmm2, [rdx]
-
- paddw xmm1, [GLOBAL(xmm_bi_rd)]
- paddw xmm2, [GLOBAL(xmm_bi_rd)]
- psraw xmm1, xmm_filter_shift
- psraw xmm2, xmm_filter_shift
-
- movq xmm3, QWORD PTR [rdi]
- pxor xmm4, xmm4
- punpcklbw xmm3, xmm4
- movq xmm5, QWORD PTR [rdi+8]
- punpcklbw xmm5, xmm4
-
- psubw xmm1, xmm3
- psubw xmm2, xmm5
- paddw xmm6, xmm1
- paddw xmm6, xmm2
- pmaddwd xmm1, xmm1
- pmaddwd xmm2, xmm2
- paddd xmm7, xmm1
- paddd xmm7, xmm2
-
- movdqa xmm1, xmm0
- lea rsi, [rsi + rax] ;ref_pixels_per_line
-
-%if ABI_IS_32BIT
- add rdi, dword ptr arg(3) ;src_pixels_per_line
-%else
- lea rdi, [rdi + r9]
-%endif
-
- sub rcx, 1
- jnz .filter_block2d_bil_sp_only_loop
-
- jmp .filter_block2d_bil_variance
-
-.filter_block2d_bil_var_ssse3_full_pixel:
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rax, dword ptr arg(1) ;ref_pixels_per_line
- movsxd rdx, dword ptr arg(3) ;src_pixels_per_line
- pxor xmm0, xmm0
-
-.filter_block2d_bil_full_pixel_loop:
- movq xmm1, QWORD PTR [rsi]
- punpcklbw xmm1, xmm0
- movq xmm2, QWORD PTR [rsi+8]
- punpcklbw xmm2, xmm0
-
- movq xmm3, QWORD PTR [rdi]
- punpcklbw xmm3, xmm0
- movq xmm4, QWORD PTR [rdi+8]
- punpcklbw xmm4, xmm0
-
- psubw xmm1, xmm3
- psubw xmm2, xmm4
- paddw xmm6, xmm1
- paddw xmm6, xmm2
- pmaddwd xmm1, xmm1
- pmaddwd xmm2, xmm2
- paddd xmm7, xmm1
- paddd xmm7, xmm2
-
- lea rsi, [rsi + rax] ;ref_pixels_per_line
- lea rdi, [rdi + rdx] ;src_pixels_per_line
- sub rcx, 1
- jnz .filter_block2d_bil_full_pixel_loop
-
- jmp .filter_block2d_bil_variance
-
-.filter_block2d_bil_var_ssse3_fp_only:
- mov rsi, arg(0) ;ref_ptr
- mov rdi, arg(2) ;src_ptr
- movsxd rcx, dword ptr arg(4) ;Height
- movsxd rdx, dword ptr arg(1) ;ref_pixels_per_line
-
- pxor xmm0, xmm0
-
-%if ABI_IS_32BIT=0
- movsxd r9, dword ptr arg(3) ;src_pixels_per_line
-%endif
-
-.filter_block2d_bil_fp_only_loop:
- movdqu xmm1, XMMWORD PTR [rsi]
- movdqu xmm2, XMMWORD PTR [rsi+1]
- movdqa xmm3, xmm1
-
- punpcklbw xmm1, xmm2
- punpckhbw xmm3, xmm2
- pmaddubsw xmm1, [rax]
- pmaddubsw xmm3, [rax]
-
- paddw xmm1, [GLOBAL(xmm_bi_rd)]
- paddw xmm3, [GLOBAL(xmm_bi_rd)]
- psraw xmm1, xmm_filter_shift
- psraw xmm3, xmm_filter_shift
-
- movq xmm2, XMMWORD PTR [rdi]
- pxor xmm4, xmm4
- punpcklbw xmm2, xmm4
- movq xmm5, QWORD PTR [rdi+8]
- punpcklbw xmm5, xmm4
-
- psubw xmm1, xmm2
- psubw xmm3, xmm5
- paddw xmm6, xmm1
- paddw xmm6, xmm3
- pmaddwd xmm1, xmm1
- pmaddwd xmm3, xmm3
- paddd xmm7, xmm1
- paddd xmm7, xmm3
-
- lea rsi, [rsi + rdx]
-%if ABI_IS_32BIT
- add rdi, dword ptr arg(3) ;src_pixels_per_line
-%else
- lea rdi, [rdi + r9]
-%endif
-
- sub rcx, 1
- jnz .filter_block2d_bil_fp_only_loop
-
- jmp .filter_block2d_bil_variance
-
-.filter_block2d_bil_variance:
- pxor xmm0, xmm0
- pxor xmm1, xmm1
- pxor xmm5, xmm5
-
- punpcklwd xmm0, xmm6
- punpckhwd xmm1, xmm6
- psrad xmm0, 16
- psrad xmm1, 16
- paddd xmm0, xmm1
- movdqa xmm1, xmm0
-
- movdqa xmm6, xmm7
- punpckldq xmm6, xmm5
- punpckhdq xmm7, xmm5
- paddd xmm6, xmm7
-
- punpckldq xmm0, xmm5
- punpckhdq xmm1, xmm5
- paddd xmm0, xmm1
-
- movdqa xmm7, xmm6
- movdqa xmm1, xmm0
-
- psrldq xmm7, 8
- psrldq xmm1, 8
-
- paddd xmm6, xmm7
- paddd xmm0, xmm1
-
- mov rsi, arg(7) ;[Sum]
- mov rdi, arg(8) ;[SSE]
-
- movd [rsi], xmm0
- movd [rdi], xmm6
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-SECTION_RODATA
-align 16
-xmm_bi_rd:
- times 8 dw 64
-align 16
-vp8_bilinear_filters_ssse3:
- times 8 db 128, 0
- times 8 db 112, 16
- times 8 db 96, 32
- times 8 db 80, 48
- times 8 db 64, 64
- times 8 db 48, 80
- times 8 db 32, 96
- times 8 db 16, 112
--- a/vp8/encoder/x86/variance_mmx.c
+++ /dev/null
@@ -1,398 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-#include "vpx_config.h"
-#include "vp8/encoder/variance.h"
-#include "vp8/common/pragmas.h"
-#include "vpx_ports/mem.h"
-#include "vp8/common/x86/filter_x86.h"
-
-extern void filter_block1d_h6_mmx
-(
- const unsigned char *src_ptr,
- unsigned short *output_ptr,
- unsigned int src_pixels_per_line,
- unsigned int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- short *filter
-);
-extern void filter_block1d_v6_mmx
-(
- const short *src_ptr,
- unsigned char *output_ptr,
- unsigned int pixels_per_line,
- unsigned int pixel_step,
- unsigned int output_height,
- unsigned int output_width,
- short *filter
-);
-
-extern unsigned int vp8_get_mb_ss_mmx(const short *src_ptr);
-extern unsigned int vp8_get8x8var_mmx
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-extern unsigned int vp8_get4x4var_mmx
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-extern void vp8_filter_block2d_bil4x4_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-extern void vp8_filter_block2d_bil_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-
-
-unsigned int vp8_variance4x4_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
- vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 4));
-
-}
-
-unsigned int vp8_variance8x8_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
- vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
- *sse = var;
-
- return (var - ((unsigned int)(avg * avg) >> 6));
-
-}
-
-unsigned int vp8_mse16x16_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, sse2, sse3, var;
- int sum0, sum1, sum2, sum3;
-
-
- vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
- vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
- vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
-
- var = sse0 + sse1 + sse2 + sse3;
- *sse = var;
- return var;
-}
-
-
-unsigned int vp8_variance16x16_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, sse2, sse3, var;
- int sum0, sum1, sum2, sum3, avg;
-
-
- vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
- vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse2, &sum2) ;
- vp8_get8x8var_mmx(src_ptr + 8 * source_stride + 8, source_stride, ref_ptr + 8 * recon_stride + 8, recon_stride, &sse3, &sum3);
-
- var = sse0 + sse1 + sse2 + sse3;
- avg = sum0 + sum1 + sum2 + sum3;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 8));
-}
-
-unsigned int vp8_variance16x8_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, var;
- int sum0, sum1, avg;
-
- vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_mmx(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
-
- var = sse0 + sse1;
- avg = sum0 + sum1;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 7));
-
-}
-
-
-unsigned int vp8_variance8x16_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, var;
- int sum0, sum1, avg;
-
- vp8_get8x8var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_mmx(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
-
- var = sse0 + sse1;
- avg = sum0 + sum1;
- *sse = var;
-
- return (var - ((unsigned int)(avg * avg) >> 7));
-
-}
-
-
-unsigned int vp8_sub_pixel_variance4x4_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil4x4_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 4));
-}
-
-
-unsigned int vp8_sub_pixel_variance8x8_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{-
- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 6));
-}
-
-unsigned int vp8_sub_pixel_variance16x16_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{-
- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum0, &xxsum0
- );
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum1, &xxsum1
- );
-
- xsum0 += xsum1;
- xxsum0 += xxsum1;
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-
-
-}
-
-unsigned int vp8_sub_pixel_mse16x16_mmx(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- vp8_sub_pixel_variance16x16_mmx(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
- return *sse;
-}
-
-unsigned int vp8_sub_pixel_variance16x8_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum0, &xxsum0
- );
-
-
- vp8_filter_block2d_bil_var_mmx(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 8,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum1, &xxsum1
- );
-
- xsum0 += xsum1;
- xxsum0 += xxsum1;
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
-}
-
-unsigned int vp8_sub_pixel_variance8x16_mmx
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 7));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_h_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 0,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_v_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 0, 4,
- ref_ptr, recon_stride, sse);
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_hv_mmx(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- return vp8_sub_pixel_variance16x16_mmx(src_ptr, source_stride, 4, 4,
- ref_ptr, recon_stride, sse);
-}
--- a/vp8/encoder/x86/variance_sse2.c
+++ /dev/null
@@ -1,557 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-#include "vpx_config.h"
-#include "vp8/encoder/variance.h"
-#include "vp8/common/pragmas.h"
-#include "vpx_ports/mem.h"
-#include "vp8/common/x86/filter_x86.h"
-
-extern void filter_block1d_h6_mmx(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d_v6_mmx(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d8_h6_sse2(const unsigned char *src_ptr, unsigned short *output_ptr, unsigned int src_pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-extern void filter_block1d8_v6_sse2(const short *src_ptr, unsigned char *output_ptr, unsigned int pixels_per_line, unsigned int pixel_step, unsigned int output_height, unsigned int output_width, short *filter);
-
-extern void vp8_filter_block2d_bil4x4_var_mmx
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const short *HFilter,
- const short *VFilter,
- int *sum,
- unsigned int *sumsquared
-);
-
-extern unsigned int vp8_get4x4var_mmx
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-
-unsigned int vp8_get_mb_ss_sse2
-(
- const short *src_ptr
-);
-unsigned int vp8_get16x16var_sse2
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-unsigned int vp8_get8x8var_sse2
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-void vp8_filter_block2d_bil_var_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int xoffset,
- int yoffset,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_vert_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_horiz_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_vert_variance8x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-void vp8_half_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-
-unsigned int vp8_variance4x4_wmt(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
- vp8_get4x4var_mmx(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 4));
-
-}
-
-unsigned int vp8_variance8x8_wmt
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int var;
- int avg;
-
- vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &var, &avg) ;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 6));
-
-}
-
-
-unsigned int vp8_variance16x16_wmt
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0;
- int sum0;
-
-
- vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- *sse = sse0;
- return (sse0 - ((unsigned int)(sum0 * sum0) >> 8));
-}
-unsigned int vp8_mse16x16_wmt(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{-
- unsigned int sse0;
- int sum0;
- vp8_get16x16var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- *sse = sse0;
- return sse0;
-
-}
-
-
-unsigned int vp8_variance16x8_wmt
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, var;
- int sum0, sum1, avg;
-
- vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_sse2(src_ptr + 8, source_stride, ref_ptr + 8, recon_stride, &sse1, &sum1);
-
- var = sse0 + sse1;
- avg = sum0 + sum1;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 7));
-
-}
-
-unsigned int vp8_variance8x16_wmt
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *sse)
-{- unsigned int sse0, sse1, var;
- int sum0, sum1, avg;
-
- vp8_get8x8var_sse2(src_ptr, source_stride, ref_ptr, recon_stride, &sse0, &sum0) ;
- vp8_get8x8var_sse2(src_ptr + 8 * source_stride, source_stride, ref_ptr + 8 * recon_stride, recon_stride, &sse1, &sum1) ;
-
- var = sse0 + sse1;
- avg = sum0 + sum1;
- *sse = var;
- return (var - ((unsigned int)(avg * avg) >> 7));
-
-}
-
-unsigned int vp8_sub_pixel_variance4x4_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
- vp8_filter_block2d_bil4x4_var_mmx(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line,
- vp8_bilinear_filters_x86_4[xoffset], vp8_bilinear_filters_x86_4[yoffset],
- &xsum, &xxsum
- );
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 4));
-}
-
-
-unsigned int vp8_sub_pixel_variance8x8_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum, &xxsum);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum, &xxsum);
- }
-
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 6));
-}
-
-unsigned int vp8_sub_pixel_variance16x16_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
-
- // note we could avoid these if statements if the calling function
- // just called the appropriate functions inside.
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum0, &xxsum0
- );
-
- vp8_filter_block2d_bil_var_sse2(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum1, &xxsum1
- );
- xsum0 += xsum1;
- xxsum0 += xxsum1;
- }
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-}
-
-unsigned int vp8_sub_pixel_mse16x16_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- vp8_sub_pixel_variance16x16_wmt(src_ptr, src_pixels_per_line, xoffset, yoffset, dst_ptr, dst_pixels_per_line, sse);
- return *sse;
-}
-
-unsigned int vp8_sub_pixel_variance16x8_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-
-)
-{- int xsum0, xsum1;
- unsigned int xxsum0, xxsum1;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum0, &xxsum0);
-
- vp8_filter_block2d_bil_var_sse2(
- src_ptr + 8, src_pixels_per_line,
- dst_ptr + 8, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum1, &xxsum1);
- xsum0 += xsum1;
- xxsum0 += xxsum1;
- }
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
-}
-
-unsigned int vp8_sub_pixel_variance8x16_wmt
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum;
- unsigned int xxsum;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance8x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum, &xxsum);
- }
- else
- {- vp8_filter_block2d_bil_var_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum, &xxsum);
- }
-
- *sse = xxsum;
- return (xxsum - ((unsigned int)(xsum * xsum) >> 7));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_h_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
-
- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_v_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-}
-
-
-unsigned int vp8_variance_halfpixvar16x16_hv_wmt(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse)
-{- int xsum0;
- unsigned int xxsum0;
-
- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-}
--- a/vp8/encoder/x86/variance_ssse3.c
+++ /dev/null
@@ -1,165 +1,0 @@
-/*
- * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- * Use of this source code is governed by a BSD-style license
- * that can be found in the LICENSE file in the root of the source
- * tree. An additional intellectual property rights grant can be found
- * in the file PATENTS. All contributing project authors may
- * be found in the AUTHORS file in the root of the source tree.
- */
-
-#include "vpx_config.h"
-#include "vp8/encoder/variance.h"
-#include "vp8/common/pragmas.h"
-#include "vpx_ports/mem.h"
-
-extern unsigned int vp8_get16x16var_sse2
-(
- const unsigned char *src_ptr,
- int source_stride,
- const unsigned char *ref_ptr,
- int recon_stride,
- unsigned int *SSE,
- int *Sum
-);
-extern void vp8_half_horiz_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-extern void vp8_half_horiz_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-extern void vp8_half_vert_variance16x_h_sse2
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int *sum,
- unsigned int *sumsquared
-);
-extern void vp8_filter_block2d_bil_var_ssse3
-(
- const unsigned char *ref_ptr,
- int ref_pixels_per_line,
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- unsigned int Height,
- int xoffset,
- int yoffset,
- int *sum,
- unsigned int *sumsquared
-);
-
-unsigned int vp8_sub_pixel_variance16x16_ssse3
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-)
-{- int xsum0;
- unsigned int xxsum0;
-
- // note we could avoid these if statements if the calling function
- // just called the appropriate functions inside.
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_ssse3(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 16,
- xoffset, yoffset,
- &xsum0, &xxsum0);
- }
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 8));
-}
-
-unsigned int vp8_sub_pixel_variance16x8_ssse3
-(
- const unsigned char *src_ptr,
- int src_pixels_per_line,
- int xoffset,
- int yoffset,
- const unsigned char *dst_ptr,
- int dst_pixels_per_line,
- unsigned int *sse
-
-)
-{- int xsum0;
- unsigned int xxsum0;
-
- if (xoffset == 4 && yoffset == 0)
- {- vp8_half_horiz_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 0 && yoffset == 4)
- {- vp8_half_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else if (xoffset == 4 && yoffset == 4)
- {- vp8_half_horiz_vert_variance16x_h_sse2(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- &xsum0, &xxsum0);
- }
- else
- {- vp8_filter_block2d_bil_var_ssse3(
- src_ptr, src_pixels_per_line,
- dst_ptr, dst_pixels_per_line, 8,
- xoffset, yoffset,
- &xsum0, &xxsum0);
- }
-
- *sse = xxsum0;
- return (xxsum0 - ((unsigned int)(xsum0 * xsum0) >> 7));
-}
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -64,8 +64,11 @@
VP8_COMMON_SRCS-yes += common/reconinter.c
VP8_COMMON_SRCS-yes += common/reconintra.c
VP8_COMMON_SRCS-yes += common/reconintra4x4.c
+VP8_COMMON_SRCS-yes += common/sad_c.c
VP8_COMMON_SRCS-yes += common/setupintrarecon.c
VP8_COMMON_SRCS-yes += common/swapyv12buffer.c
+VP8_COMMON_SRCS-yes += common/variance_c.c
+VP8_COMMON_SRCS-yes += common/variance.h
@@ -83,22 +86,35 @@
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/idctllm_mmx.asm
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/idctllm_mmx_test.cc
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/iwalsh_mmx.asm
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/loopfilter_mmx.asm
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/recon_mmx.asm
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/sad_mmx.asm
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/subpixel_mmx.asm
-VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/loopfilter_mmx.asm
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/variance_mmx.c
+VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/variance_impl_mmx.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/idct_blk_sse2.c
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/idctllm_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/recon_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/recon_wrapper_sse2.c
+VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/sad_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/subpixel_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/loopfilter_sse2.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/iwalsh_sse2.asm
+VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/variance_sse2.c
+VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/variance_impl_sse2.asm
+VP8_COMMON_SRCS-$(HAVE_SSE3) += common/x86/sad_sse3.asm
+VP8_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/sad_ssse3.asm
VP8_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/subpixel_ssse3.asm
+VP8_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/variance_ssse3.c
+VP8_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/variance_impl_ssse3.asm
+VP8_COMMON_SRCS-$(HAVE_SSE4_1) += common/x86/sad_sse4.asm
+
ifeq ($(CONFIG_POSTPROC),yes)
VP8_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/postproc_x86.c
VP8_COMMON_SRCS-$(HAVE_MMX) += common/x86/postproc_mmx.asm
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/postproc_sse2.asm
endif
+
ifeq ($(ARCH_X86_64),yes)
VP8_COMMON_SRCS-$(HAVE_SSE2) += common/x86/loopfilter_block_sse2.asm
endif
@@ -108,6 +124,7 @@
VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/loopfilter_arm.c
VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/reconintra_arm.c
VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/dequantize_arm.c
+VP8_COMMON_SRCS-$(ARCH_ARM) += common/arm/variance_arm.c
# common (media)
VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/bilinearfilter_arm.c
@@ -127,6 +144,12 @@
VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/dequant_idct_v6$(ASM)
VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/dequantize_v6$(ASM)
VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/idct_blk_v6.c
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_sad16x16_armv6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance8x8_armv6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance16x16_armv6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6$(ASM)
+VP8_COMMON_SRCS-$(HAVE_MEDIA) += common/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
# common (neon)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/bilinearpredict4x4_neon$(ASM)
@@ -143,6 +166,8 @@
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/loopfiltersimpleverticaledge_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/mbloopfilter_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/shortidct4x4llm_neon$(ASM)
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/sad8_neon$(ASM)
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/sad16_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/sixtappredict4x4_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/sixtappredict8x4_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/sixtappredict8x8_neon$(ASM)
@@ -154,3 +179,7 @@
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/idct_dequant_0_2x_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/dequantizeb_neon$(ASM)
VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/idct_blk_neon.c
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/variance_neon$(ASM)
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp8_subpixelvariance8x8_neon$(ASM)
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp8_subpixelvariance16x16_neon$(ASM)
+VP8_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp8_subpixelvariance16x16s_neon$(ASM)
--- a/vp8/vp8cx.mk
+++ b/vp8/vp8cx.mk
@@ -63,7 +63,6 @@
VP8_CX_SRCS-yes += encoder/rdopt.h
VP8_CX_SRCS-yes += encoder/tokenize.h
VP8_CX_SRCS-yes += encoder/treewriter.h
-VP8_CX_SRCS-yes += encoder/variance.h
VP8_CX_SRCS-yes += encoder/mcomp.c
VP8_CX_SRCS-yes += encoder/modecosts.c
VP8_CX_SRCS-yes += encoder/onyx_if.c
@@ -73,13 +72,11 @@
VP8_CX_SRCS-yes += encoder/quantize.c
VP8_CX_SRCS-yes += encoder/ratectrl.c
VP8_CX_SRCS-yes += encoder/rdopt.c
-VP8_CX_SRCS-yes += encoder/sad_c.c
VP8_CX_SRCS-yes += encoder/segmentation.c
VP8_CX_SRCS-yes += encoder/segmentation.h
VP8_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/ssim.c
VP8_CX_SRCS-yes += encoder/tokenize.c
VP8_CX_SRCS-yes += encoder/treewriter.c
-VP8_CX_SRCS-yes += encoder/variance_c.c
VP8_CX_SRCS-$(CONFIG_INTERNAL_STATS) += common/postproc.h
VP8_CX_SRCS-$(CONFIG_INTERNAL_STATS) += common/postproc.c
VP8_CX_SRCS-yes += encoder/temporal_filter.c
@@ -91,27 +88,16 @@
VP8_CX_SRCS_REMOVE-yes += encoder/temporal_filter.c
endif
-VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/variance_mmx.c
-VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/variance_impl_mmx.asm
-VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/sad_mmx.asm
VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/dct_mmx.asm
VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/subtract_mmx.asm
VP8_CX_SRCS-$(HAVE_MMX) += encoder/x86/vp8_enc_stubs_mmx.c
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/dct_sse2.asm
-VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/variance_sse2.c
-VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/variance_impl_sse2.asm
-VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/sad_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/fwalsh_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/quantize_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/subtract_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/temporal_filter_apply_sse2.asm
VP8_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp8_enc_stubs_sse2.c
-VP8_CX_SRCS-$(HAVE_SSE3) += encoder/x86/sad_sse3.asm
-VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/sad_ssse3.asm
-VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/variance_ssse3.c
-VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/variance_impl_ssse3.asm
VP8_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/quantize_ssse3.asm
-VP8_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/sad_sse4.asm
VP8_CX_SRCS-$(HAVE_SSE4_1) += encoder/x86/quantize_sse4.asm
VP8_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/quantize_mmx.asm
VP8_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/encodeopt.asm
--- a/vp8/vp8cx_arm.mk
+++ b/vp8/vp8cx_arm.mk
@@ -15,7 +15,6 @@
# encoder
VP8_CX_SRCS-$(ARCH_ARM) += encoder/arm/dct_arm.c
VP8_CX_SRCS-$(ARCH_ARM) += encoder/arm/quantize_arm.c
-VP8_CX_SRCS-$(ARCH_ARM) += encoder/arm/variance_arm.c
#File list for edsp
# encoder
@@ -31,13 +30,7 @@
VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_subtract_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_short_fdct4x4_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_fast_quantize_b_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_sad16x16_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_variance16x16_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_h_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_mse16x16_armv6$(ASM)
-VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/vp8_variance8x8_armv6$(ASM)
VP8_CX_SRCS-$(HAVE_MEDIA) += encoder/arm/armv6/walsh_v6$(ASM)
#File list for neon
@@ -44,14 +37,8 @@
# encoder
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/fastquantizeb_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/picklpf_arm.c
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/sad8_neon$(ASM)
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/sad16_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/shortfdct_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/subtract_neon$(ASM)
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/variance_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_mse16x16_neon$(ASM)
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_subpixelvariance8x8_neon$(ASM)
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_subpixelvariance16x16_neon$(ASM)
-VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_subpixelvariance16x16s_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_memcpy_neon$(ASM)
VP8_CX_SRCS-$(HAVE_NEON) += encoder/arm/neon/vp8_shortwalsh4x4_neon$(ASM)
--
⑨