ref: 005552639b276709868a4a0f86f27a737c7c6917
parent: bd9cd9a1859aa464b3092f2023b3a4040166572d
author: Yaowu Xu <yaowu@google.com>
date: Tue Mar 12 07:24:04 EDT 2013
removed reference to "LLM" and "x8" The commit changed the name of files and function to remove obselete reference to LLM and x8. Change-Id: I973b20fc1a55149ed68b5408b3874768e6f88516
--- a/test/fdct4x4_test.cc
+++ b/test/fdct4x4_test.cc
@@ -120,7 +120,7 @@
}
// Because the bitstream is not frozen yet, use the idct in the codebase.
- vp9_short_idct4x4llm_c(test_temp_block, test_output_block, pitch);
+ vp9_short_idct4x4_c(test_temp_block, test_output_block, pitch);
for (int j = 0; j < 16; ++j) {const int diff = test_input_block[j] - test_output_block[j];
--- /dev/null
+++ b/test/idct_test.cc
@@ -1,0 +1,118 @@
+/*
+ * 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.
+ */
+
+
+extern "C" {+#include "./vpx_config.h"
+#include "./vp8_rtcd.h"
+}
+#include "test/register_state_check.h"
+#include "third_party/googletest/src/include/gtest/gtest.h"
+
+typedef void (*idct_fn_t)(short *input, unsigned char *pred_ptr,
+ int pred_stride, unsigned char *dst_ptr,
+ int dst_stride);
+namespace {+class IDCTTest : public ::testing::TestWithParam<idct_fn_t> {+ protected:
+ virtual void SetUp() {+ int i;
+
+ UUT = GetParam();
+ memset(input, 0, sizeof(input));
+ /* Set up guard blocks */
+ for (i = 0; i < 256; i++)
+ output[i] = ((i & 0xF) < 4 && (i < 64)) ? 0 : -1;
+ }
+
+ idct_fn_t UUT;
+ short input[16];
+ unsigned char output[256];
+ unsigned char predict[256];
+};
+
+TEST_P(IDCTTest, TestGuardBlocks) {+ int i;
+
+ for (i = 0; i < 256; i++)
+ if ((i & 0xF) < 4 && i < 64)
+ EXPECT_EQ(0, output[i]) << i;
+ else
+ EXPECT_EQ(255, output[i]);
+}
+
+TEST_P(IDCTTest, TestAllZeros) {+ int i;
+
+ REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
+
+ for (i = 0; i < 256; i++)
+ if ((i & 0xF) < 4 && i < 64)
+ EXPECT_EQ(0, output[i]) << "i==" << i;
+ else
+ EXPECT_EQ(255, output[i]) << "i==" << i;
+}
+
+TEST_P(IDCTTest, TestAllOnes) {+ int i;
+
+ input[0] = 4;
+ REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
+
+ for (i = 0; i < 256; i++)
+ if ((i & 0xF) < 4 && i < 64)
+ EXPECT_EQ(1, output[i]) << "i==" << i;
+ else
+ EXPECT_EQ(255, output[i]) << "i==" << i;
+}
+
+TEST_P(IDCTTest, TestAddOne) {+ int i;
+
+ for (i = 0; i < 256; i++)
+ predict[i] = i;
+ input[0] = 4;
+ REGISTER_STATE_CHECK(UUT(input, predict, 16, output, 16));
+
+ for (i = 0; i < 256; i++)
+ if ((i & 0xF) < 4 && i < 64)
+ EXPECT_EQ(i+1, output[i]) << "i==" << i;
+ else
+ EXPECT_EQ(255, output[i]) << "i==" << i;
+}
+
+TEST_P(IDCTTest, TestWithData) {+ int i;
+
+ for (i = 0; i < 16; i++)
+ input[i] = i;
+
+ REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
+
+ for (i = 0; i < 256; i++)
+ if ((i & 0xF) > 3 || i > 63)
+ EXPECT_EQ(255, output[i]) << "i==" << i;
+ else if (i == 0)
+ EXPECT_EQ(11, output[i]) << "i==" << i;
+ else if (i == 34)
+ EXPECT_EQ(1, output[i]) << "i==" << i;
+ else if (i == 2 || i == 17 || i == 32)
+ EXPECT_EQ(3, output[i]) << "i==" << i;
+ else
+ EXPECT_EQ(0, output[i]) << "i==" << i;
+}
+
+INSTANTIATE_TEST_CASE_P(C, IDCTTest,
+ ::testing::Values(vp8_short_idct4x4llm_c));
+#if HAVE_MMX
+INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
+ ::testing::Values(vp8_short_idct4x4llm_mmx));
+#endif
+}
--- a/test/idctllm_test.cc
+++ /dev/null
@@ -1,126 +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.
- */
-
-
-extern "C" {-#include "vpx_config.h"
-#include "vp8_rtcd.h"
-}
-#include "test/register_state_check.h"
-#include "third_party/googletest/src/include/gtest/gtest.h"
-
-typedef void (*idct_fn_t)(short *input, unsigned char *pred_ptr,
- int pred_stride, unsigned char *dst_ptr,
- int dst_stride);
-namespace {-class IDCTTest : public ::testing::TestWithParam<idct_fn_t>
-{- protected:
- virtual void SetUp()
- {- int i;
-
- UUT = GetParam();
- memset(input, 0, sizeof(input));
- /* Set up guard blocks */
- for(i=0; i<256; i++)
- output[i] = ((i&0xF)<4&&(i<64))?0:-1;
- }
-
- idct_fn_t UUT;
- short input[16];
- unsigned char output[256];
- unsigned char predict[256];
-};
-
-TEST_P(IDCTTest, TestGuardBlocks)
-{- int i;
-
- for(i=0; i<256; i++)
- if((i&0xF) < 4 && i<64)
- EXPECT_EQ(0, output[i]) << i;
- else
- EXPECT_EQ(255, output[i]);
-}
-
-TEST_P(IDCTTest, TestAllZeros)
-{- int i;
-
- REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
-
- for(i=0; i<256; i++)
- if((i&0xF) < 4 && i<64)
- EXPECT_EQ(0, output[i]) << "i==" << i;
- else
- EXPECT_EQ(255, output[i]) << "i==" << i;
-}
-
-TEST_P(IDCTTest, TestAllOnes)
-{- int i;
-
- input[0] = 4;
- REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
-
- for(i=0; i<256; i++)
- if((i&0xF) < 4 && i<64)
- EXPECT_EQ(1, output[i]) << "i==" << i;
- else
- EXPECT_EQ(255, output[i]) << "i==" << i;
-}
-
-TEST_P(IDCTTest, TestAddOne)
-{- int i;
-
- for(i=0; i<256; i++)
- predict[i] = i;
-
- input[0] = 4;
- REGISTER_STATE_CHECK(UUT(input, predict, 16, output, 16));
-
- for(i=0; i<256; i++)
- if((i&0xF) < 4 && i<64)
- EXPECT_EQ(i+1, output[i]) << "i==" << i;
- else
- EXPECT_EQ(255, output[i]) << "i==" << i;
-}
-
-TEST_P(IDCTTest, TestWithData)
-{- int i;
-
- for(i=0; i<16; i++)
- input[i] = i;
-
- REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
-
- for(i=0; i<256; i++)
- if((i&0xF) > 3 || i>63)
- EXPECT_EQ(255, output[i]) << "i==" << i;
- else if(i == 0)
- EXPECT_EQ(11, output[i]) << "i==" << i;
- else if(i == 34)
- EXPECT_EQ(1, output[i]) << "i==" << i;
- else if(i == 2 || i == 17 || i == 32)
- EXPECT_EQ(3, output[i]) << "i==" << i;
- else
- EXPECT_EQ(0, output[i]) << "i==" << i;
-}
-
-INSTANTIATE_TEST_CASE_P(C, IDCTTest,
- ::testing::Values(vp8_short_idct4x4llm_c));
-#if HAVE_MMX
-INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
- ::testing::Values(vp8_short_idct4x4llm_mmx));
-#endif
-}
--- a/test/test.mk
+++ b/test/test.mk
@@ -47,7 +47,7 @@
LIBVPX_TEST_SRCS-yes += vp8_boolcoder_test.cc
endif
-LIBVPX_TEST_SRCS-yes += idctllm_test.cc
+LIBVPX_TEST_SRCS-yes += idct_test.cc
LIBVPX_TEST_SRCS-yes += intrapred_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_POSTPROC) += pp_filter_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_ENCODERS) += sad_test.cc
--- /dev/null
+++ b/vp9/common/ppc/vp9_idct_altivec.asm
@@ -1,0 +1,189 @@
+;
+; 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 short_idct4x4_ppc
+
+.macro load_c V, LABEL, OFF, R0, R1
+ lis \R0, \LABEL@ha
+ la \R1, \LABEL@l(\R0)
+ lvx \V, \OFF, \R1
+.endm
+
+;# r3 short *input
+;# r4 short *output
+;# r5 int pitch
+ .align 2
+short_idct4x4_ppc:
+ mfspr r11, 256 ;# get old VRSAVE
+ oris r12, r11, 0xfff8
+ mtspr 256, r12 ;# set VRSAVE
+
+ load_c v8, sinpi8sqrt2, 0, r9, r10
+ load_c v9, cospi8sqrt2minus1, 0, r9, r10
+ load_c v10, hi_hi, 0, r9, r10
+ load_c v11, lo_lo, 0, r9, r10
+ load_c v12, shift_16, 0, r9, r10
+
+ li r10, 16
+ lvx v0, 0, r3 ;# input ip[0], ip[ 4]
+ lvx v1, r10, r3 ;# input ip[8], ip[12]
+
+ ;# first pass
+ vupkhsh v2, v0
+ vupkhsh v3, v1
+ vaddsws v6, v2, v3 ;# a1 = ip[0]+ip[8]
+ vsubsws v7, v2, v3 ;# b1 = ip[0]-ip[8]
+
+ vupklsh v0, v0
+ vmulosh v4, v0, v8
+ vsraw v4, v4, v12
+ vaddsws v4, v4, v0 ;# ip[ 4] * sin(pi/8) * sqrt(2)
+
+ vupklsh v1, v1
+ vmulosh v5, v1, v9
+ vsraw v5, v5, v12 ;# ip[12] * cos(pi/8) * sqrt(2)
+ vaddsws v5, v5, v1
+
+ vsubsws v4, v4, v5 ;# c1
+
+ vmulosh v3, v1, v8
+ vsraw v3, v3, v12
+ vaddsws v3, v3, v1 ;# ip[12] * sin(pi/8) * sqrt(2)
+
+ vmulosh v5, v0, v9
+ vsraw v5, v5, v12 ;# ip[ 4] * cos(pi/8) * sqrt(2)
+ vaddsws v5, v5, v0
+
+ vaddsws v3, v3, v5 ;# d1
+
+ vaddsws v0, v6, v3 ;# a1 + d1
+ vsubsws v3, v6, v3 ;# a1 - d1
+
+ vaddsws v1, v7, v4 ;# b1 + c1
+ vsubsws v2, v7, v4 ;# b1 - c1
+
+ ;# transpose input
+ vmrghw v4, v0, v1 ;# a0 b0 a1 b1
+ vmrghw v5, v2, v3 ;# c0 d0 c1 d1
+
+ vmrglw v6, v0, v1 ;# a2 b2 a3 b3
+ vmrglw v7, v2, v3 ;# c2 d2 c3 d3
+
+ vperm v0, v4, v5, v10 ;# a0 b0 c0 d0
+ vperm v1, v4, v5, v11 ;# a1 b1 c1 d1
+
+ vperm v2, v6, v7, v10 ;# a2 b2 c2 d2
+ vperm v3, v6, v7, v11 ;# a3 b3 c3 d3
+
+ ;# second pass
+ vaddsws v6, v0, v2 ;# a1 = ip[0]+ip[8]
+ vsubsws v7, v0, v2 ;# b1 = ip[0]-ip[8]
+
+ vmulosh v4, v1, v8
+ vsraw v4, v4, v12
+ vaddsws v4, v4, v1 ;# ip[ 4] * sin(pi/8) * sqrt(2)
+
+ vmulosh v5, v3, v9
+ vsraw v5, v5, v12 ;# ip[12] * cos(pi/8) * sqrt(2)
+ vaddsws v5, v5, v3
+
+ vsubsws v4, v4, v5 ;# c1
+
+ vmulosh v2, v3, v8
+ vsraw v2, v2, v12
+ vaddsws v2, v2, v3 ;# ip[12] * sin(pi/8) * sqrt(2)
+
+ vmulosh v5, v1, v9
+ vsraw v5, v5, v12 ;# ip[ 4] * cos(pi/8) * sqrt(2)
+ vaddsws v5, v5, v1
+
+ vaddsws v3, v2, v5 ;# d1
+
+ vaddsws v0, v6, v3 ;# a1 + d1
+ vsubsws v3, v6, v3 ;# a1 - d1
+
+ vaddsws v1, v7, v4 ;# b1 + c1
+ vsubsws v2, v7, v4 ;# b1 - c1
+
+ vspltish v6, 4
+ vspltish v7, 3
+
+ vpkswss v0, v0, v1
+ vpkswss v1, v2, v3
+
+ vaddshs v0, v0, v6
+ vaddshs v1, v1, v6
+
+ vsrah v0, v0, v7
+ vsrah v1, v1, v7
+
+ ;# transpose output
+ vmrghh v2, v0, v1 ;# a0 c0 a1 c1 a2 c2 a3 c3
+ vmrglh v3, v0, v1 ;# b0 d0 b1 d1 b2 d2 b3 d3
+
+ vmrghh v0, v2, v3 ;# a0 b0 c0 d0 a1 b1 c1 d1
+ vmrglh v1, v2, v3 ;# a2 b2 c2 d2 a3 b3 c3 d3
+
+ stwu r1,-416(r1) ;# create space on the stack
+
+ stvx v0, 0, r1
+ lwz r6, 0(r1)
+ stw r6, 0(r4)
+ lwz r6, 4(r1)
+ stw r6, 4(r4)
+
+ add r4, r4, r5
+
+ lwz r6, 8(r1)
+ stw r6, 0(r4)
+ lwz r6, 12(r1)
+ stw r6, 4(r4)
+
+ add r4, r4, r5
+
+ stvx v1, 0, r1
+ lwz r6, 0(r1)
+ stw r6, 0(r4)
+ lwz r6, 4(r1)
+ stw r6, 4(r4)
+
+ add r4, r4, r5
+
+ lwz r6, 8(r1)
+ stw r6, 0(r4)
+ lwz r6, 12(r1)
+ stw r6, 4(r4)
+
+ addi r1, r1, 416 ;# recover stack
+
+ mtspr 256, r11 ;# reset old VRSAVE
+
+ blr
+
+ .align 4
+sinpi8sqrt2:
+ .short 35468, 35468, 35468, 35468, 35468, 35468, 35468, 35468
+
+ .align 4
+cospi8sqrt2minus1:
+ .short 20091, 20091, 20091, 20091, 20091, 20091, 20091, 20091
+
+ .align 4
+shift_16:
+ .long 16, 16, 16, 16
+
+ .align 4
+hi_hi:
+ .byte 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23
+
+ .align 4
+lo_lo:
+ .byte 8, 9, 10, 11, 12, 13, 14, 15, 24, 25, 26, 27, 28, 29, 30, 31
--- a/vp9/common/ppc/vp9_idctllm_altivec.asm
+++ /dev/null
@@ -1,189 +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 short_idct4x4llm_ppc
-
-.macro load_c V, LABEL, OFF, R0, R1
- lis \R0, \LABEL@ha
- la \R1, \LABEL@l(\R0)
- lvx \V, \OFF, \R1
-.endm
-
-;# r3 short *input
-;# r4 short *output
-;# r5 int pitch
- .align 2
-short_idct4x4llm_ppc:
- mfspr r11, 256 ;# get old VRSAVE
- oris r12, r11, 0xfff8
- mtspr 256, r12 ;# set VRSAVE
-
- load_c v8, sinpi8sqrt2, 0, r9, r10
- load_c v9, cospi8sqrt2minus1, 0, r9, r10
- load_c v10, hi_hi, 0, r9, r10
- load_c v11, lo_lo, 0, r9, r10
- load_c v12, shift_16, 0, r9, r10
-
- li r10, 16
- lvx v0, 0, r3 ;# input ip[0], ip[ 4]
- lvx v1, r10, r3 ;# input ip[8], ip[12]
-
- ;# first pass
- vupkhsh v2, v0
- vupkhsh v3, v1
- vaddsws v6, v2, v3 ;# a1 = ip[0]+ip[8]
- vsubsws v7, v2, v3 ;# b1 = ip[0]-ip[8]
-
- vupklsh v0, v0
- vmulosh v4, v0, v8
- vsraw v4, v4, v12
- vaddsws v4, v4, v0 ;# ip[ 4] * sin(pi/8) * sqrt(2)
-
- vupklsh v1, v1
- vmulosh v5, v1, v9
- vsraw v5, v5, v12 ;# ip[12] * cos(pi/8) * sqrt(2)
- vaddsws v5, v5, v1
-
- vsubsws v4, v4, v5 ;# c1
-
- vmulosh v3, v1, v8
- vsraw v3, v3, v12
- vaddsws v3, v3, v1 ;# ip[12] * sin(pi/8) * sqrt(2)
-
- vmulosh v5, v0, v9
- vsraw v5, v5, v12 ;# ip[ 4] * cos(pi/8) * sqrt(2)
- vaddsws v5, v5, v0
-
- vaddsws v3, v3, v5 ;# d1
-
- vaddsws v0, v6, v3 ;# a1 + d1
- vsubsws v3, v6, v3 ;# a1 - d1
-
- vaddsws v1, v7, v4 ;# b1 + c1
- vsubsws v2, v7, v4 ;# b1 - c1
-
- ;# transpose input
- vmrghw v4, v0, v1 ;# a0 b0 a1 b1
- vmrghw v5, v2, v3 ;# c0 d0 c1 d1
-
- vmrglw v6, v0, v1 ;# a2 b2 a3 b3
- vmrglw v7, v2, v3 ;# c2 d2 c3 d3
-
- vperm v0, v4, v5, v10 ;# a0 b0 c0 d0
- vperm v1, v4, v5, v11 ;# a1 b1 c1 d1
-
- vperm v2, v6, v7, v10 ;# a2 b2 c2 d2
- vperm v3, v6, v7, v11 ;# a3 b3 c3 d3
-
- ;# second pass
- vaddsws v6, v0, v2 ;# a1 = ip[0]+ip[8]
- vsubsws v7, v0, v2 ;# b1 = ip[0]-ip[8]
-
- vmulosh v4, v1, v8
- vsraw v4, v4, v12
- vaddsws v4, v4, v1 ;# ip[ 4] * sin(pi/8) * sqrt(2)
-
- vmulosh v5, v3, v9
- vsraw v5, v5, v12 ;# ip[12] * cos(pi/8) * sqrt(2)
- vaddsws v5, v5, v3
-
- vsubsws v4, v4, v5 ;# c1
-
- vmulosh v2, v3, v8
- vsraw v2, v2, v12
- vaddsws v2, v2, v3 ;# ip[12] * sin(pi/8) * sqrt(2)
-
- vmulosh v5, v1, v9
- vsraw v5, v5, v12 ;# ip[ 4] * cos(pi/8) * sqrt(2)
- vaddsws v5, v5, v1
-
- vaddsws v3, v2, v5 ;# d1
-
- vaddsws v0, v6, v3 ;# a1 + d1
- vsubsws v3, v6, v3 ;# a1 - d1
-
- vaddsws v1, v7, v4 ;# b1 + c1
- vsubsws v2, v7, v4 ;# b1 - c1
-
- vspltish v6, 4
- vspltish v7, 3
-
- vpkswss v0, v0, v1
- vpkswss v1, v2, v3
-
- vaddshs v0, v0, v6
- vaddshs v1, v1, v6
-
- vsrah v0, v0, v7
- vsrah v1, v1, v7
-
- ;# transpose output
- vmrghh v2, v0, v1 ;# a0 c0 a1 c1 a2 c2 a3 c3
- vmrglh v3, v0, v1 ;# b0 d0 b1 d1 b2 d2 b3 d3
-
- vmrghh v0, v2, v3 ;# a0 b0 c0 d0 a1 b1 c1 d1
- vmrglh v1, v2, v3 ;# a2 b2 c2 d2 a3 b3 c3 d3
-
- stwu r1,-416(r1) ;# create space on the stack
-
- stvx v0, 0, r1
- lwz r6, 0(r1)
- stw r6, 0(r4)
- lwz r6, 4(r1)
- stw r6, 4(r4)
-
- add r4, r4, r5
-
- lwz r6, 8(r1)
- stw r6, 0(r4)
- lwz r6, 12(r1)
- stw r6, 4(r4)
-
- add r4, r4, r5
-
- stvx v1, 0, r1
- lwz r6, 0(r1)
- stw r6, 0(r4)
- lwz r6, 4(r1)
- stw r6, 4(r4)
-
- add r4, r4, r5
-
- lwz r6, 8(r1)
- stw r6, 0(r4)
- lwz r6, 12(r1)
- stw r6, 4(r4)
-
- addi r1, r1, 416 ;# recover stack
-
- mtspr 256, r11 ;# reset old VRSAVE
-
- blr
-
- .align 4
-sinpi8sqrt2:
- .short 35468, 35468, 35468, 35468, 35468, 35468, 35468, 35468
-
- .align 4
-cospi8sqrt2minus1:
- .short 20091, 20091, 20091, 20091, 20091, 20091, 20091, 20091
-
- .align 4
-shift_16:
- .long 16, 16, 16, 16
-
- .align 4
-hi_hi:
- .byte 0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23
-
- .align 4
-lo_lo:
- .byte 8, 9, 10, 11, 12, 13, 14, 15, 24, 25, 26, 27, 28, 29, 30, 31
--- a/vp9/common/ppc/vp9_systemdependent.c
+++ b/vp9/common/ppc/vp9_systemdependent.c
@@ -63,7 +63,7 @@
void recon2b_ppc(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void recon4b_ppc(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
-extern void short_idct4x4llm_ppc(short *input, short *output, int pitch);
+extern void short_idct4x4_ppc(short *input, short *output, int pitch);
// Generic C
extern subpixel_predict_function vp9_sixtap_predict_c;
@@ -83,8 +83,8 @@
void vp9_recon2b_c(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
void vp9_recon4b_c(short *diff_ptr, unsigned char *pred_ptr, unsigned char *dst_ptr, int stride);
-extern void vp9_short_idct4x4llm_1_c(short *input, short *output, int pitch);
-extern void vp9_short_idct4x4llm_c(short *input, short *output, int pitch);
+extern void vp9_short_idct4x4_1_c(short *input, short *output, int pitch);
+extern void vp9_short_idct4x4_c(short *input, short *output, int pitch);
extern void vp8_dc_only_idct_c(short input_dc, short *output, int pitch);
// PPC
@@ -139,8 +139,8 @@
vp9_sixtap_predict8x4 = sixtap_predict8x4_ppc;
vp9_sixtap_predict = sixtap_predict_ppc;
- vp8_short_idct4x4_1 = vp9_short_idct4x4llm_1_c;
- vp8_short_idct4x4 = short_idct4x4llm_ppc;
+ vp8_short_idct4x4_1 = vp9_short_idct4x4_1_c;
+ vp8_short_idct4x4 = short_idct4x4_ppc;
vp8_dc_only_idct = vp8_dc_only_idct_c;
vp8_lf_mbvfull = loop_filter_mbv_ppc;
--- /dev/null
+++ b/vp9/common/vp9_idct.c
@@ -1,0 +1,1307 @@
+/*
+ * 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 <assert.h>
+#include <math.h>
+
+#include "./vpx_config.h"
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_systemdependent.h"
+#include "vp9/common/vp9_blockd.h"
+#include "vp9/common/vp9_common.h"
+#include "vp9/common/vp9_idct.h"
+
+void vp9_short_iwalsh4x4_c(int16_t *input, int16_t *output, int pitch) {+ int i;
+ int a1, b1, c1, d1;
+ int16_t *ip = input;
+ int16_t *op = output;
+ const int half_pitch = pitch >> 1;
+
+ for (i = 0; i < 4; i++) {+ a1 = (ip[0] + ip[3]) >> WHT_UPSCALE_FACTOR;
+ b1 = (ip[1] + ip[2]) >> WHT_UPSCALE_FACTOR;
+ c1 = (ip[1] - ip[2]) >> WHT_UPSCALE_FACTOR;
+ d1 = (ip[0] - ip[3]) >> WHT_UPSCALE_FACTOR;
+
+ op[0] = (a1 + b1 + 1) >> 1;
+ op[1] = (c1 + d1) >> 1;
+ op[2] = (a1 - b1) >> 1;
+ op[3] = (d1 - c1) >> 1;
+
+ ip += 4;
+ op += half_pitch;
+ }
+
+ ip = output;
+ op = output;
+ for (i = 0; i < 4; i++) {+ a1 = ip[half_pitch * 0] + ip[half_pitch * 3];
+ b1 = ip[half_pitch * 1] + ip[half_pitch * 2];
+ c1 = ip[half_pitch * 1] - ip[half_pitch * 2];
+ d1 = ip[half_pitch * 0] - ip[half_pitch * 3];
+
+
+ op[half_pitch * 0] = (a1 + b1 + 1) >> 1;
+ op[half_pitch * 1] = (c1 + d1) >> 1;
+ op[half_pitch * 2] = (a1 - b1) >> 1;
+ op[half_pitch * 3] = (d1 - c1) >> 1;
+
+ ip++;
+ op++;
+ }
+}
+
+void vp9_short_iwalsh4x4_1_c(int16_t *in, int16_t *out, int pitch) {+ int i;
+ int16_t tmp[4];
+ int16_t *ip = in;
+ int16_t *op = tmp;
+ const int half_pitch = pitch >> 1;
+
+ op[0] = ((ip[0] >> WHT_UPSCALE_FACTOR) + 1) >> 1;
+ op[1] = op[2] = op[3] = (ip[0] >> WHT_UPSCALE_FACTOR) >> 1;
+
+ ip = tmp;
+ op = out;
+ for (i = 0; i < 4; i++) {+ op[half_pitch * 0] = (ip[0] + 1) >> 1;
+ op[half_pitch * 1] = op[half_pitch * 2] = op[half_pitch * 3] = ip[0] >> 1;
+ ip++;
+ op++;
+ }
+}
+
+void vp9_dc_only_inv_walsh_add_c(int input_dc, uint8_t *pred_ptr,
+ uint8_t *dst_ptr,
+ int pitch, int stride) {+ int r, c;
+ int16_t dc = input_dc;
+ int16_t tmp[4 * 4];
+ vp9_short_iwalsh4x4_1_c(&dc, tmp, 4 << 1);
+
+ for (r = 0; r < 4; r++) {+ for (c = 0; c < 4; c++)
+ dst_ptr[c] = clip_pixel(tmp[r * 4 + c] + pred_ptr[c]);
+
+ dst_ptr += stride;
+ pred_ptr += pitch;
+ }
+}
+
+void vp9_idct4_1d_c(int16_t *input, int16_t *output) {+ int16_t step[4];
+ int temp1, temp2;
+ // stage 1
+ temp1 = (input[0] + input[2]) * cospi_16_64;
+ temp2 = (input[0] - input[2]) * cospi_16_64;
+ step[0] = dct_const_round_shift(temp1);
+ step[1] = dct_const_round_shift(temp2);
+ temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
+ temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
+ step[2] = dct_const_round_shift(temp1);
+ step[3] = dct_const_round_shift(temp2);
+
+ // stage 2
+ output[0] = step[0] + step[3];
+ output[1] = step[1] + step[2];
+ output[2] = step[1] - step[2];
+ output[3] = step[0] - step[3];
+}
+
+void vp9_short_idct4x4_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[4 * 4];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[4], temp_out[4];
+
+ // Rows
+ for (i = 0; i < 4; ++i) {+ for (j = 0; j < 4; ++j)
+ temp_in[j] = input[j];
+ vp9_idct4_1d(temp_in, outptr);
+ input += 4;
+ outptr += 4;
+ }
+
+ // Columns
+ for (i = 0; i < 4; ++i) {+ for (j = 0; j < 4; ++j)
+ temp_in[j] = out[j * 4 + i];
+ vp9_idct4_1d(temp_in, temp_out);
+ for (j = 0; j < 4; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 4);
+ }
+}
+
+void vp9_short_idct4x4_1_c(int16_t *input, int16_t *output, int pitch) {+ int i;
+ int a1;
+ int16_t *op = output;
+ const int half_pitch = pitch >> 1;
+ int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ a1 = ROUND_POWER_OF_TWO(out, 4);
+
+ for (i = 0; i < 4; i++) {+ op[0] = op[1] = op[2] = op[3] = a1;
+ op += half_pitch;
+ }
+}
+
+void vp9_dc_only_idct_add_c(int input_dc, uint8_t *pred_ptr,
+ uint8_t *dst_ptr, int pitch, int stride) {+ int a1;
+ int r, c;
+ int16_t out = dct_const_round_shift(input_dc * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ a1 = ROUND_POWER_OF_TWO(out, 4);
+
+ for (r = 0; r < 4; r++) {+ for (c = 0; c < 4; c++)
+ dst_ptr[c] = clip_pixel(a1 + pred_ptr[c]);
+
+ dst_ptr += stride;
+ pred_ptr += pitch;
+ }
+}
+
+static void idct8_1d(int16_t *input, int16_t *output) {+ int16_t step1[8], step2[8];
+ int temp1, temp2;
+ // stage 1
+ step1[0] = input[0];
+ step1[2] = input[4];
+ step1[1] = input[2];
+ step1[3] = input[6];
+ temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
+ temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
+ step1[4] = dct_const_round_shift(temp1);
+ step1[7] = dct_const_round_shift(temp2);
+ temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
+ temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+
+ // stage 2 & stage 3 - even half
+ vp9_idct4_1d(step1, step1);
+
+ // stage 2 - odd half
+ step2[4] = step1[4] + step1[5];
+ step2[5] = step1[4] - step1[5];
+ step2[6] = -step1[6] + step1[7];
+ step2[7] = step1[6] + step1[7];
+
+ // stage 3 -odd half
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+ step1[7] = step2[7];
+
+ // stage 4
+ output[0] = step1[0] + step1[7];
+ output[1] = step1[1] + step1[6];
+ output[2] = step1[2] + step1[5];
+ output[3] = step1[3] + step1[4];
+ output[4] = step1[3] - step1[4];
+ output[5] = step1[2] - step1[5];
+ output[6] = step1[1] - step1[6];
+ output[7] = step1[0] - step1[7];
+}
+
+void vp9_short_idct8x8_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[8 * 8];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[8], temp_out[8];
+
+ // Rows
+ for (i = 0; i < 8; ++i) {+ idct8_1d(input, outptr);
+ input += 8;
+ outptr += 8;
+ }
+
+ // Columns
+ for (i = 0; i < 8; ++i) {+ for (j = 0; j < 8; ++j)
+ temp_in[j] = out[j * 8 + i];
+ idct8_1d(temp_in, temp_out);
+ for (j = 0; j < 8; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
+ }
+}
+
+static void iadst4_1d(int16_t *input, int16_t *output) {+ int s0, s1, s2, s3, s4, s5, s6, s7;
+
+ int x0 = input[0];
+ int x1 = input[1];
+ int x2 = input[2];
+ int x3 = input[3];
+
+ if (!(x0 | x1 | x2 | x3)) {+ output[0] = output[1] = output[2] = output[3] = 0;
+ return;
+ }
+
+ s0 = sinpi_1_9 * x0;
+ s1 = sinpi_2_9 * x0;
+ s2 = sinpi_3_9 * x1;
+ s3 = sinpi_4_9 * x2;
+ s4 = sinpi_1_9 * x2;
+ s5 = sinpi_2_9 * x3;
+ s6 = sinpi_4_9 * x3;
+ s7 = x0 - x2 + x3;
+
+ x0 = s0 + s3 + s5;
+ x1 = s1 - s4 - s6;
+ x2 = sinpi_3_9 * s7;
+ x3 = s2;
+
+ s0 = x0 + x3;
+ s1 = x1 + x3;
+ s2 = x2;
+ s3 = x0 + x1 - x3;
+
+ // 1-D transform scaling factor is sqrt(2).
+ // The overall dynamic range is 14b (input) + 14b (multiplication scaling)
+ // + 1b (addition) = 29b.
+ // Hence the output bit depth is 15b.
+ output[0] = dct_const_round_shift(s0);
+ output[1] = dct_const_round_shift(s1);
+ output[2] = dct_const_round_shift(s2);
+ output[3] = dct_const_round_shift(s3);
+}
+
+void vp9_short_iht4x4_c(int16_t *input, int16_t *output,
+ int pitch, int tx_type) {+ const transform_2d IHT_4[] = {+ { vp9_idct4_1d, vp9_idct4_1d }, // DCT_DCT = 0+ { iadst4_1d, vp9_idct4_1d }, // ADST_DCT = 1+ { vp9_idct4_1d, iadst4_1d }, // DCT_ADST = 2+ { iadst4_1d, iadst4_1d } // ADST_ADST = 3+ };
+
+ int i, j;
+ int16_t out[4 * 4];
+ int16_t *outptr = out;
+ int16_t temp_in[4], temp_out[4];
+
+ // inverse transform row vectors
+ for (i = 0; i < 4; ++i) {+ IHT_4[tx_type].rows(input, outptr);
+ input += 4;
+ outptr += 4;
+ }
+
+ // inverse transform column vectors
+ for (i = 0; i < 4; ++i) {+ for (j = 0; j < 4; ++j)
+ temp_in[j] = out[j * 4 + i];
+ IHT_4[tx_type].cols(temp_in, temp_out);
+ for (j = 0; j < 4; ++j)
+ output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 4);
+ }
+}
+
+static void iadst8_1d(int16_t *input, int16_t *output) {+ int s0, s1, s2, s3, s4, s5, s6, s7;
+
+ int x0 = input[7];
+ int x1 = input[0];
+ int x2 = input[5];
+ int x3 = input[2];
+ int x4 = input[3];
+ int x5 = input[4];
+ int x6 = input[1];
+ int x7 = input[6];
+
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {+ output[0] = output[1] = output[2] = output[3] = output[4]
+ = output[5] = output[6] = output[7] = 0;
+ return;
+ }
+
+ // stage 1
+ s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
+ s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
+ s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
+ s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
+ s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
+ s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
+ s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
+ s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
+
+ x0 = dct_const_round_shift(s0 + s4);
+ x1 = dct_const_round_shift(s1 + s5);
+ x2 = dct_const_round_shift(s2 + s6);
+ x3 = dct_const_round_shift(s3 + s7);
+ x4 = dct_const_round_shift(s0 - s4);
+ x5 = dct_const_round_shift(s1 - s5);
+ x6 = dct_const_round_shift(s2 - s6);
+ x7 = dct_const_round_shift(s3 - s7);
+
+ // stage 2
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
+ s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
+ s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
+ s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
+
+ x0 = s0 + s2;
+ x1 = s1 + s3;
+ x2 = s0 - s2;
+ x3 = s1 - s3;
+ x4 = dct_const_round_shift(s4 + s6);
+ x5 = dct_const_round_shift(s5 + s7);
+ x6 = dct_const_round_shift(s4 - s6);
+ x7 = dct_const_round_shift(s5 - s7);
+
+ // stage 3
+ s2 = cospi_16_64 * (x2 + x3);
+ s3 = cospi_16_64 * (x2 - x3);
+ s6 = cospi_16_64 * (x6 + x7);
+ s7 = cospi_16_64 * (x6 - x7);
+
+ x2 = dct_const_round_shift(s2);
+ x3 = dct_const_round_shift(s3);
+ x6 = dct_const_round_shift(s6);
+ x7 = dct_const_round_shift(s7);
+
+ output[0] = x0;
+ output[1] = -x4;
+ output[2] = x6;
+ output[3] = -x2;
+ output[4] = x3;
+ output[5] = -x7;
+ output[6] = x5;
+ output[7] = -x1;
+}
+
+static const transform_2d IHT_8[] = {+ { idct8_1d, idct8_1d }, // DCT_DCT = 0+ { iadst8_1d, idct8_1d }, // ADST_DCT = 1+ { idct8_1d, iadst8_1d }, // DCT_ADST = 2+ { iadst8_1d, iadst8_1d } // ADST_ADST = 3+};
+
+void vp9_short_iht8x8_c(int16_t *input, int16_t *output,
+ int pitch, int tx_type) {+ int i, j;
+ int16_t out[8 * 8];
+ int16_t *outptr = out;
+ int16_t temp_in[8], temp_out[8];
+ const transform_2d ht = IHT_8[tx_type];
+
+ // inverse transform row vectors
+ for (i = 0; i < 8; ++i) {+ ht.rows(input, outptr);
+ input += 8;
+ outptr += 8;
+ }
+
+ // inverse transform column vectors
+ for (i = 0; i < 8; ++i) {+ for (j = 0; j < 8; ++j)
+ temp_in[j] = out[j * 8 + i];
+ ht.cols(temp_in, temp_out);
+ for (j = 0; j < 8; ++j)
+ output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
+ }
+}
+
+void vp9_short_idct10_8x8_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[8 * 8];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[8], temp_out[8];
+
+ vpx_memset(out, 0, sizeof(out));
+ // First transform rows
+ // only first 4 row has non-zero coefs
+ for (i = 0; i < 4; ++i) {+ idct8_1d(input, outptr);
+ input += 8;
+ outptr += 8;
+ }
+
+ // Then transform columns
+ for (i = 0; i < 8; ++i) {+ for (j = 0; j < 8; ++j)
+ temp_in[j] = out[j * 8 + i];
+ idct8_1d(temp_in, temp_out);
+ for (j = 0; j < 8; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
+ }
+}
+
+void vp9_short_idct1_8x8_c(int16_t *input, int16_t *output) {+ int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ output[0] = ROUND_POWER_OF_TWO(out, 5);
+}
+
+static void idct16_1d(int16_t *input, int16_t *output) {+ int16_t step1[16], step2[16];
+ int temp1, temp2;
+
+ // stage 1
+ step1[0] = input[0/2];
+ step1[1] = input[16/2];
+ step1[2] = input[8/2];
+ step1[3] = input[24/2];
+ step1[4] = input[4/2];
+ step1[5] = input[20/2];
+ step1[6] = input[12/2];
+ step1[7] = input[28/2];
+ step1[8] = input[2/2];
+ step1[9] = input[18/2];
+ step1[10] = input[10/2];
+ step1[11] = input[26/2];
+ step1[12] = input[6/2];
+ step1[13] = input[22/2];
+ step1[14] = input[14/2];
+ step1[15] = input[30/2];
+
+ // stage 2
+ step2[0] = step1[0];
+ step2[1] = step1[1];
+ step2[2] = step1[2];
+ step2[3] = step1[3];
+ step2[4] = step1[4];
+ step2[5] = step1[5];
+ step2[6] = step1[6];
+ step2[7] = step1[7];
+
+ temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
+ temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
+ step2[8] = dct_const_round_shift(temp1);
+ step2[15] = dct_const_round_shift(temp2);
+
+ temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
+ temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
+ step2[9] = dct_const_round_shift(temp1);
+ step2[14] = dct_const_round_shift(temp2);
+
+ temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
+ temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+
+ temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
+ temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
+ step2[11] = dct_const_round_shift(temp1);
+ step2[12] = dct_const_round_shift(temp2);
+
+ // stage 3
+ step1[0] = step2[0];
+ step1[1] = step2[1];
+ step1[2] = step2[2];
+ step1[3] = step2[3];
+
+ temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
+ temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
+ step1[4] = dct_const_round_shift(temp1);
+ step1[7] = dct_const_round_shift(temp2);
+ temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
+ temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+
+ step1[8] = step2[8] + step2[9];
+ step1[9] = step2[8] - step2[9];
+ step1[10] = -step2[10] + step2[11];
+ step1[11] = step2[10] + step2[11];
+ step1[12] = step2[12] + step2[13];
+ step1[13] = step2[12] - step2[13];
+ step1[14] = -step2[14] + step2[15];
+ step1[15] = step2[14] + step2[15];
+
+ temp1 = (step1[0] + step1[1]) * cospi_16_64;
+ temp2 = (step1[0] - step1[1]) * cospi_16_64;
+ step2[0] = dct_const_round_shift(temp1);
+ step2[1] = dct_const_round_shift(temp2);
+ temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
+ temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
+ step2[2] = dct_const_round_shift(temp1);
+ step2[3] = dct_const_round_shift(temp2);
+ step2[4] = step1[4] + step1[5];
+ step2[5] = step1[4] - step1[5];
+ step2[6] = -step1[6] + step1[7];
+ step2[7] = step1[6] + step1[7];
+
+ step2[8] = step1[8];
+ step2[15] = step1[15];
+ temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
+ temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
+ step2[9] = dct_const_round_shift(temp1);
+ step2[14] = dct_const_round_shift(temp2);
+ temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
+ temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+ step2[11] = step1[11];
+ step2[12] = step1[12];
+
+ // stage 5
+ step1[0] = step2[0] + step2[3];
+ step1[1] = step2[1] + step2[2];
+ step1[2] = step2[1] - step2[2];
+ step1[3] = step2[0] - step2[3];
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+ step1[7] = step2[7];
+
+ step1[8] = step2[8] + step2[11];
+ step1[9] = step2[9] + step2[10];
+ step1[10] = step2[9] - step2[10];
+ step1[11] = step2[8] - step2[11];
+ step1[12] = -step2[12] + step2[15];
+ step1[13] = -step2[13] + step2[14];
+ step1[14] = step2[13] + step2[14];
+ step1[15] = step2[12] + step2[15];
+
+ // stage 6
+ step2[0] = step1[0] + step1[7];
+ step2[1] = step1[1] + step1[6];
+ step2[2] = step1[2] + step1[5];
+ step2[3] = step1[3] + step1[4];
+ step2[4] = step1[3] - step1[4];
+ step2[5] = step1[2] - step1[5];
+ step2[6] = step1[1] - step1[6];
+ step2[7] = step1[0] - step1[7];
+ step2[8] = step1[8];
+ step2[9] = step1[9];
+ temp1 = (-step1[10] + step1[13]) * cospi_16_64;
+ temp2 = (step1[10] + step1[13]) * cospi_16_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+ temp1 = (-step1[11] + step1[12]) * cospi_16_64;
+ temp2 = (step1[11] + step1[12]) * cospi_16_64;
+ step2[11] = dct_const_round_shift(temp1);
+ step2[12] = dct_const_round_shift(temp2);
+ step2[14] = step1[14];
+ step2[15] = step1[15];
+
+ // stage 7
+ output[0] = step2[0] + step2[15];
+ output[1] = step2[1] + step2[14];
+ output[2] = step2[2] + step2[13];
+ output[3] = step2[3] + step2[12];
+ output[4] = step2[4] + step2[11];
+ output[5] = step2[5] + step2[10];
+ output[6] = step2[6] + step2[9];
+ output[7] = step2[7] + step2[8];
+ output[8] = step2[7] - step2[8];
+ output[9] = step2[6] - step2[9];
+ output[10] = step2[5] - step2[10];
+ output[11] = step2[4] - step2[11];
+ output[12] = step2[3] - step2[12];
+ output[13] = step2[2] - step2[13];
+ output[14] = step2[1] - step2[14];
+ output[15] = step2[0] - step2[15];
+}
+
+void vp9_short_idct16x16_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[16 * 16];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[16], temp_out[16];
+
+ // First transform rows
+ for (i = 0; i < 16; ++i) {+ idct16_1d(input, outptr);
+ input += 16;
+ outptr += 16;
+ }
+
+ // Then transform columns
+ for (i = 0; i < 16; ++i) {+ for (j = 0; j < 16; ++j)
+ temp_in[j] = out[j * 16 + i];
+ idct16_1d(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
+ }
+}
+
+void iadst16_1d(int16_t *input, int16_t *output) {+ int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15;
+
+ int x0 = input[15];
+ int x1 = input[0];
+ int x2 = input[13];
+ int x3 = input[2];
+ int x4 = input[11];
+ int x5 = input[4];
+ int x6 = input[9];
+ int x7 = input[6];
+ int x8 = input[7];
+ int x9 = input[8];
+ int x10 = input[5];
+ int x11 = input[10];
+ int x12 = input[3];
+ int x13 = input[12];
+ int x14 = input[1];
+ int x15 = input[14];
+
+ if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8
+ | x9 | x10 | x11 | x12 | x13 | x14 | x15)) {+ output[0] = output[1] = output[2] = output[3] = output[4]
+ = output[5] = output[6] = output[7] = output[8]
+ = output[9] = output[10] = output[11] = output[12]
+ = output[13] = output[14] = output[15] = 0;
+ return;
+ }
+
+ // stage 1
+ s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
+ s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
+ s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
+ s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
+ s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
+ s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
+ s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
+ s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
+ s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
+ s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
+ s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
+ s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
+ s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
+ s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
+ s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
+ s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
+
+ x0 = dct_const_round_shift(s0 + s8);
+ x1 = dct_const_round_shift(s1 + s9);
+ x2 = dct_const_round_shift(s2 + s10);
+ x3 = dct_const_round_shift(s3 + s11);
+ x4 = dct_const_round_shift(s4 + s12);
+ x5 = dct_const_round_shift(s5 + s13);
+ x6 = dct_const_round_shift(s6 + s14);
+ x7 = dct_const_round_shift(s7 + s15);
+ x8 = dct_const_round_shift(s0 - s8);
+ x9 = dct_const_round_shift(s1 - s9);
+ x10 = dct_const_round_shift(s2 - s10);
+ x11 = dct_const_round_shift(s3 - s11);
+ x12 = dct_const_round_shift(s4 - s12);
+ x13 = dct_const_round_shift(s5 - s13);
+ x14 = dct_const_round_shift(s6 - s14);
+ x15 = dct_const_round_shift(s7 - s15);
+
+ // stage 2
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = x4;
+ s5 = x5;
+ s6 = x6;
+ s7 = x7;
+ s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
+ s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
+ s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
+ s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
+ s12 = - x12 * cospi_28_64 + x13 * cospi_4_64;
+ s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
+ s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
+ s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
+
+ x0 = s0 + s4;
+ x1 = s1 + s5;
+ x2 = s2 + s6;
+ x3 = s3 + s7;
+ x4 = s0 - s4;
+ x5 = s1 - s5;
+ x6 = s2 - s6;
+ x7 = s3 - s7;
+ x8 = dct_const_round_shift(s8 + s12);
+ x9 = dct_const_round_shift(s9 + s13);
+ x10 = dct_const_round_shift(s10 + s14);
+ x11 = dct_const_round_shift(s11 + s15);
+ x12 = dct_const_round_shift(s8 - s12);
+ x13 = dct_const_round_shift(s9 - s13);
+ x14 = dct_const_round_shift(s10 - s14);
+ x15 = dct_const_round_shift(s11 - s15);
+
+ // stage 3
+ s0 = x0;
+ s1 = x1;
+ s2 = x2;
+ s3 = x3;
+ s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
+ s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
+ s6 = - x6 * cospi_24_64 + x7 * cospi_8_64;
+ s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
+ s8 = x8;
+ s9 = x9;
+ s10 = x10;
+ s11 = x11;
+ s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
+ s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
+ s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
+ s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
+
+ x0 = s0 + s2;
+ x1 = s1 + s3;
+ x2 = s0 - s2;
+ x3 = s1 - s3;
+ x4 = dct_const_round_shift(s4 + s6);
+ x5 = dct_const_round_shift(s5 + s7);
+ x6 = dct_const_round_shift(s4 - s6);
+ x7 = dct_const_round_shift(s5 - s7);
+ x8 = s8 + s10;
+ x9 = s9 + s11;
+ x10 = s8 - s10;
+ x11 = s9 - s11;
+ x12 = dct_const_round_shift(s12 + s14);
+ x13 = dct_const_round_shift(s13 + s15);
+ x14 = dct_const_round_shift(s12 - s14);
+ x15 = dct_const_round_shift(s13 - s15);
+
+ // stage 4
+ s2 = (- cospi_16_64) * (x2 + x3);
+ s3 = cospi_16_64 * (x2 - x3);
+ s6 = cospi_16_64 * (x6 + x7);
+ s7 = cospi_16_64 * (- x6 + x7);
+ s10 = cospi_16_64 * (x10 + x11);
+ s11 = cospi_16_64 * (- x10 + x11);
+ s14 = (- cospi_16_64) * (x14 + x15);
+ s15 = cospi_16_64 * (x14 - x15);
+
+ x2 = dct_const_round_shift(s2);
+ x3 = dct_const_round_shift(s3);
+ x6 = dct_const_round_shift(s6);
+ x7 = dct_const_round_shift(s7);
+ x10 = dct_const_round_shift(s10);
+ x11 = dct_const_round_shift(s11);
+ x14 = dct_const_round_shift(s14);
+ x15 = dct_const_round_shift(s15);
+
+ output[0] = x0;
+ output[1] = -x8;
+ output[2] = x12;
+ output[3] = -x4;
+ output[4] = x6;
+ output[5] = x14;
+ output[6] = x10;
+ output[7] = x2;
+ output[8] = x3;
+ output[9] = x11;
+ output[10] = x15;
+ output[11] = x7;
+ output[12] = x5;
+ output[13] = -x13;
+ output[14] = x9;
+ output[15] = -x1;
+}
+
+static const transform_2d IHT_16[] = {+ { idct16_1d, idct16_1d }, // DCT_DCT = 0+ { iadst16_1d, idct16_1d }, // ADST_DCT = 1+ { idct16_1d, iadst16_1d }, // DCT_ADST = 2+ { iadst16_1d, iadst16_1d } // ADST_ADST = 3+};
+
+void vp9_short_iht16x16_c(int16_t *input, int16_t *output,
+ int pitch, int tx_type) {+ int i, j;
+ int16_t out[16 * 16];
+ int16_t *outptr = out;
+ int16_t temp_in[16], temp_out[16];
+ const transform_2d ht = IHT_16[tx_type];
+
+ // Rows
+ for (i = 0; i < 16; ++i) {+ ht.rows(input, outptr);
+ input += 16;
+ outptr += 16;
+ }
+
+ // Columns
+ for (i = 0; i < 16; ++i) {+ for (j = 0; j < 16; ++j)
+ temp_in[j] = out[j * 16 + i];
+ ht.cols(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
+ }
+}
+
+void vp9_short_idct10_16x16_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[16 * 16];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[16], temp_out[16];
+
+ /* First transform rows. Since all non-zero dct coefficients are in
+ * upper-left 4x4 area, we only need to calculate first 4 rows here.
+ */
+ vpx_memset(out, 0, sizeof(out));
+ for (i = 0; i < 4; ++i) {+ idct16_1d(input, outptr);
+ input += 16;
+ outptr += 16;
+ }
+
+ // Then transform columns
+ for (i = 0; i < 16; ++i) {+ for (j = 0; j < 16; ++j)
+ temp_in[j] = out[j*16 + i];
+ idct16_1d(temp_in, temp_out);
+ for (j = 0; j < 16; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
+ }
+}
+
+
+void vp9_short_idct1_16x16_c(int16_t *input, int16_t *output) {+ int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ output[0] = ROUND_POWER_OF_TWO(out, 6);
+}
+
+static void idct32_1d(int16_t *input, int16_t *output) {+ int16_t step1[32], step2[32];
+ int temp1, temp2;
+
+ // stage 1
+ step1[0] = input[0];
+ step1[1] = input[16];
+ step1[2] = input[8];
+ step1[3] = input[24];
+ step1[4] = input[4];
+ step1[5] = input[20];
+ step1[6] = input[12];
+ step1[7] = input[28];
+ step1[8] = input[2];
+ step1[9] = input[18];
+ step1[10] = input[10];
+ step1[11] = input[26];
+ step1[12] = input[6];
+ step1[13] = input[22];
+ step1[14] = input[14];
+ step1[15] = input[30];
+
+ temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
+ temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
+ step1[16] = dct_const_round_shift(temp1);
+ step1[31] = dct_const_round_shift(temp2);
+
+ temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
+ temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
+ step1[17] = dct_const_round_shift(temp1);
+ step1[30] = dct_const_round_shift(temp2);
+
+ temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
+ temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
+ step1[18] = dct_const_round_shift(temp1);
+ step1[29] = dct_const_round_shift(temp2);
+
+ temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
+ temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
+ step1[19] = dct_const_round_shift(temp1);
+ step1[28] = dct_const_round_shift(temp2);
+
+ temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
+ temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
+ step1[20] = dct_const_round_shift(temp1);
+ step1[27] = dct_const_round_shift(temp2);
+
+ temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
+ temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
+ step1[21] = dct_const_round_shift(temp1);
+ step1[26] = dct_const_round_shift(temp2);
+
+ temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
+ temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
+ step1[22] = dct_const_round_shift(temp1);
+ step1[25] = dct_const_round_shift(temp2);
+
+ temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
+ temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
+ step1[23] = dct_const_round_shift(temp1);
+ step1[24] = dct_const_round_shift(temp2);
+
+ // stage 2
+ step2[0] = step1[0];
+ step2[1] = step1[1];
+ step2[2] = step1[2];
+ step2[3] = step1[3];
+ step2[4] = step1[4];
+ step2[5] = step1[5];
+ step2[6] = step1[6];
+ step2[7] = step1[7];
+
+ temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
+ temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
+ step2[8] = dct_const_round_shift(temp1);
+ step2[15] = dct_const_round_shift(temp2);
+
+ temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
+ temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
+ step2[9] = dct_const_round_shift(temp1);
+ step2[14] = dct_const_round_shift(temp2);
+
+ temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
+ temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+
+ temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
+ temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
+ step2[11] = dct_const_round_shift(temp1);
+ step2[12] = dct_const_round_shift(temp2);
+
+ step2[16] = step1[16] + step1[17];
+ step2[17] = step1[16] - step1[17];
+ step2[18] = -step1[18] + step1[19];
+ step2[19] = step1[18] + step1[19];
+ step2[20] = step1[20] + step1[21];
+ step2[21] = step1[20] - step1[21];
+ step2[22] = -step1[22] + step1[23];
+ step2[23] = step1[22] + step1[23];
+ step2[24] = step1[24] + step1[25];
+ step2[25] = step1[24] - step1[25];
+ step2[26] = -step1[26] + step1[27];
+ step2[27] = step1[26] + step1[27];
+ step2[28] = step1[28] + step1[29];
+ step2[29] = step1[28] - step1[29];
+ step2[30] = -step1[30] + step1[31];
+ step2[31] = step1[30] + step1[31];
+
+ // stage 3
+ step1[0] = step2[0];
+ step1[1] = step2[1];
+ step1[2] = step2[2];
+ step1[3] = step2[3];
+
+ temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
+ temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
+ step1[4] = dct_const_round_shift(temp1);
+ step1[7] = dct_const_round_shift(temp2);
+ temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
+ temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+
+ step1[8] = step2[8] + step2[9];
+ step1[9] = step2[8] - step2[9];
+ step1[10] = -step2[10] + step2[11];
+ step1[11] = step2[10] + step2[11];
+ step1[12] = step2[12] + step2[13];
+ step1[13] = step2[12] - step2[13];
+ step1[14] = -step2[14] + step2[15];
+ step1[15] = step2[14] + step2[15];
+
+ step1[16] = step2[16];
+ step1[31] = step2[31];
+ temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
+ temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
+ step1[17] = dct_const_round_shift(temp1);
+ step1[30] = dct_const_round_shift(temp2);
+ temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
+ temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
+ step1[18] = dct_const_round_shift(temp1);
+ step1[29] = dct_const_round_shift(temp2);
+ step1[19] = step2[19];
+ step1[20] = step2[20];
+ temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
+ temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
+ step1[21] = dct_const_round_shift(temp1);
+ step1[26] = dct_const_round_shift(temp2);
+ temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
+ temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
+ step1[22] = dct_const_round_shift(temp1);
+ step1[25] = dct_const_round_shift(temp2);
+ step1[23] = step2[23];
+ step1[24] = step2[24];
+ step1[27] = step2[27];
+ step1[28] = step2[28];
+
+ // stage 4
+ temp1 = (step1[0] + step1[1]) * cospi_16_64;
+ temp2 = (step1[0] - step1[1]) * cospi_16_64;
+ step2[0] = dct_const_round_shift(temp1);
+ step2[1] = dct_const_round_shift(temp2);
+ temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
+ temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
+ step2[2] = dct_const_round_shift(temp1);
+ step2[3] = dct_const_round_shift(temp2);
+ step2[4] = step1[4] + step1[5];
+ step2[5] = step1[4] - step1[5];
+ step2[6] = -step1[6] + step1[7];
+ step2[7] = step1[6] + step1[7];
+
+ step2[8] = step1[8];
+ step2[15] = step1[15];
+ temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
+ temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
+ step2[9] = dct_const_round_shift(temp1);
+ step2[14] = dct_const_round_shift(temp2);
+ temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
+ temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+ step2[11] = step1[11];
+ step2[12] = step1[12];
+
+ step2[16] = step1[16] + step1[19];
+ step2[17] = step1[17] + step1[18];
+ step2[18] = step1[17] - step1[18];
+ step2[19] = step1[16] - step1[19];
+ step2[20] = -step1[20] + step1[23];
+ step2[21] = -step1[21] + step1[22];
+ step2[22] = step1[21] + step1[22];
+ step2[23] = step1[20] + step1[23];
+
+ step2[24] = step1[24] + step1[27];
+ step2[25] = step1[25] + step1[26];
+ step2[26] = step1[25] - step1[26];
+ step2[27] = step1[24] - step1[27];
+ step2[28] = -step1[28] + step1[31];
+ step2[29] = -step1[29] + step1[30];
+ step2[30] = step1[29] + step1[30];
+ step2[31] = step1[28] + step1[31];
+
+ // stage 5
+ step1[0] = step2[0] + step2[3];
+ step1[1] = step2[1] + step2[2];
+ step1[2] = step2[1] - step2[2];
+ step1[3] = step2[0] - step2[3];
+ step1[4] = step2[4];
+ temp1 = (step2[6] - step2[5]) * cospi_16_64;
+ temp2 = (step2[5] + step2[6]) * cospi_16_64;
+ step1[5] = dct_const_round_shift(temp1);
+ step1[6] = dct_const_round_shift(temp2);
+ step1[7] = step2[7];
+
+ step1[8] = step2[8] + step2[11];
+ step1[9] = step2[9] + step2[10];
+ step1[10] = step2[9] - step2[10];
+ step1[11] = step2[8] - step2[11];
+ step1[12] = -step2[12] + step2[15];
+ step1[13] = -step2[13] + step2[14];
+ step1[14] = step2[13] + step2[14];
+ step1[15] = step2[12] + step2[15];
+
+ step1[16] = step2[16];
+ step1[17] = step2[17];
+ temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
+ temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
+ step1[18] = dct_const_round_shift(temp1);
+ step1[29] = dct_const_round_shift(temp2);
+ temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
+ temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
+ step1[19] = dct_const_round_shift(temp1);
+ step1[28] = dct_const_round_shift(temp2);
+ temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
+ temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
+ step1[20] = dct_const_round_shift(temp1);
+ step1[27] = dct_const_round_shift(temp2);
+ temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
+ temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
+ step1[21] = dct_const_round_shift(temp1);
+ step1[26] = dct_const_round_shift(temp2);
+ step1[22] = step2[22];
+ step1[23] = step2[23];
+ step1[24] = step2[24];
+ step1[25] = step2[25];
+ step1[30] = step2[30];
+ step1[31] = step2[31];
+
+ // stage 6
+ step2[0] = step1[0] + step1[7];
+ step2[1] = step1[1] + step1[6];
+ step2[2] = step1[2] + step1[5];
+ step2[3] = step1[3] + step1[4];
+ step2[4] = step1[3] - step1[4];
+ step2[5] = step1[2] - step1[5];
+ step2[6] = step1[1] - step1[6];
+ step2[7] = step1[0] - step1[7];
+ step2[8] = step1[8];
+ step2[9] = step1[9];
+ temp1 = (-step1[10] + step1[13]) * cospi_16_64;
+ temp2 = (step1[10] + step1[13]) * cospi_16_64;
+ step2[10] = dct_const_round_shift(temp1);
+ step2[13] = dct_const_round_shift(temp2);
+ temp1 = (-step1[11] + step1[12]) * cospi_16_64;
+ temp2 = (step1[11] + step1[12]) * cospi_16_64;
+ step2[11] = dct_const_round_shift(temp1);
+ step2[12] = dct_const_round_shift(temp2);
+ step2[14] = step1[14];
+ step2[15] = step1[15];
+
+ step2[16] = step1[16] + step1[23];
+ step2[17] = step1[17] + step1[22];
+ step2[18] = step1[18] + step1[21];
+ step2[19] = step1[19] + step1[20];
+ step2[20] = step1[19] - step1[20];
+ step2[21] = step1[18] - step1[21];
+ step2[22] = step1[17] - step1[22];
+ step2[23] = step1[16] - step1[23];
+
+ step2[24] = -step1[24] + step1[31];
+ step2[25] = -step1[25] + step1[30];
+ step2[26] = -step1[26] + step1[29];
+ step2[27] = -step1[27] + step1[28];
+ step2[28] = step1[27] + step1[28];
+ step2[29] = step1[26] + step1[29];
+ step2[30] = step1[25] + step1[30];
+ step2[31] = step1[24] + step1[31];
+
+ // stage 7
+ step1[0] = step2[0] + step2[15];
+ step1[1] = step2[1] + step2[14];
+ step1[2] = step2[2] + step2[13];
+ step1[3] = step2[3] + step2[12];
+ step1[4] = step2[4] + step2[11];
+ step1[5] = step2[5] + step2[10];
+ step1[6] = step2[6] + step2[9];
+ step1[7] = step2[7] + step2[8];
+ step1[8] = step2[7] - step2[8];
+ step1[9] = step2[6] - step2[9];
+ step1[10] = step2[5] - step2[10];
+ step1[11] = step2[4] - step2[11];
+ step1[12] = step2[3] - step2[12];
+ step1[13] = step2[2] - step2[13];
+ step1[14] = step2[1] - step2[14];
+ step1[15] = step2[0] - step2[15];
+
+ step1[16] = step2[16];
+ step1[17] = step2[17];
+ step1[18] = step2[18];
+ step1[19] = step2[19];
+ temp1 = (-step2[20] + step2[27]) * cospi_16_64;
+ temp2 = (step2[20] + step2[27]) * cospi_16_64;
+ step1[20] = dct_const_round_shift(temp1);
+ step1[27] = dct_const_round_shift(temp2);
+ temp1 = (-step2[21] + step2[26]) * cospi_16_64;
+ temp2 = (step2[21] + step2[26]) * cospi_16_64;
+ step1[21] = dct_const_round_shift(temp1);
+ step1[26] = dct_const_round_shift(temp2);
+ temp1 = (-step2[22] + step2[25]) * cospi_16_64;
+ temp2 = (step2[22] + step2[25]) * cospi_16_64;
+ step1[22] = dct_const_round_shift(temp1);
+ step1[25] = dct_const_round_shift(temp2);
+ temp1 = (-step2[23] + step2[24]) * cospi_16_64;
+ temp2 = (step2[23] + step2[24]) * cospi_16_64;
+ step1[23] = dct_const_round_shift(temp1);
+ step1[24] = dct_const_round_shift(temp2);
+ step1[28] = step2[28];
+ step1[29] = step2[29];
+ step1[30] = step2[30];
+ step1[31] = step2[31];
+
+ // final stage
+ output[0] = step1[0] + step1[31];
+ output[1] = step1[1] + step1[30];
+ output[2] = step1[2] + step1[29];
+ output[3] = step1[3] + step1[28];
+ output[4] = step1[4] + step1[27];
+ output[5] = step1[5] + step1[26];
+ output[6] = step1[6] + step1[25];
+ output[7] = step1[7] + step1[24];
+ output[8] = step1[8] + step1[23];
+ output[9] = step1[9] + step1[22];
+ output[10] = step1[10] + step1[21];
+ output[11] = step1[11] + step1[20];
+ output[12] = step1[12] + step1[19];
+ output[13] = step1[13] + step1[18];
+ output[14] = step1[14] + step1[17];
+ output[15] = step1[15] + step1[16];
+ output[16] = step1[15] - step1[16];
+ output[17] = step1[14] - step1[17];
+ output[18] = step1[13] - step1[18];
+ output[19] = step1[12] - step1[19];
+ output[20] = step1[11] - step1[20];
+ output[21] = step1[10] - step1[21];
+ output[22] = step1[9] - step1[22];
+ output[23] = step1[8] - step1[23];
+ output[24] = step1[7] - step1[24];
+ output[25] = step1[6] - step1[25];
+ output[26] = step1[5] - step1[26];
+ output[27] = step1[4] - step1[27];
+ output[28] = step1[3] - step1[28];
+ output[29] = step1[2] - step1[29];
+ output[30] = step1[1] - step1[30];
+ output[31] = step1[0] - step1[31];
+}
+
+void vp9_short_idct32x32_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[32 * 32];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[32], temp_out[32];
+
+ // Rows
+ for (i = 0; i < 32; ++i) {+ idct32_1d(input, outptr);
+ input += 32;
+ outptr += 32;
+ }
+
+ // Columns
+ for (i = 0; i < 32; ++i) {+ for (j = 0; j < 32; ++j)
+ temp_in[j] = out[j * 32 + i];
+ idct32_1d(temp_in, temp_out);
+ for (j = 0; j < 32; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
+ }
+}
+
+void vp9_short_idct1_32x32_c(int16_t *input, int16_t *output) {+ int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ output[0] = ROUND_POWER_OF_TWO(out, 6);
+}
+
+void vp9_short_idct10_32x32_c(int16_t *input, int16_t *output, int pitch) {+ int16_t out[32 * 32];
+ int16_t *outptr = out;
+ const int half_pitch = pitch >> 1;
+ int i, j;
+ int16_t temp_in[32], temp_out[32];
+
+ /* First transform rows. Since all non-zero dct coefficients are in
+ * upper-left 4x4 area, we only need to calculate first 4 rows here.
+ */
+ vpx_memset(out, 0, sizeof(out));
+ for (i = 0; i < 4; ++i) {+ idct32_1d(input, outptr);
+ input += 32;
+ outptr += 32;
+ }
+
+ // Columns
+ for (i = 0; i < 32; ++i) {+ for (j = 0; j < 32; ++j)
+ temp_in[j] = out[j * 32 + i];
+ idct32_1d(temp_in, temp_out);
+ for (j = 0; j < 32; ++j)
+ output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
+ }
+}
--- a/vp9/common/vp9_idctllm.c
+++ /dev/null
@@ -1,1321 +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.
- */
-
-
-/****************************************************************************
- * Notes:
- *
- * This implementation makes use of 16 bit fixed point verio of two multiply
- * constants:
- * 1. sqrt(2) * cos (pi/8)
- * 2. sqrt(2) * sin (pi/8)
- * Becuase the first constant is bigger than 1, to maintain the same 16 bit
- * fixed point precision as the second one, we use a trick of
- * x * a = x + x*(a-1)
- * so
- * x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1).
- **************************************************************************/
-#include <assert.h>
-#include <math.h>
-
-#include "./vpx_config.h"
-#include "./vp9_rtcd.h"
-#include "vp9/common/vp9_systemdependent.h"
-#include "vp9/common/vp9_blockd.h"
-#include "vp9/common/vp9_common.h"
-#include "vp9/common/vp9_idct.h"
-
-void vp9_short_inv_walsh4x4_x8_c(int16_t *input, int16_t *output, int pitch) {- int i;
- int a1, b1, c1, d1;
- int16_t *ip = input;
- int16_t *op = output;
- const int half_pitch = pitch >> 1;
-
- for (i = 0; i < 4; i++) {- a1 = (ip[0] + ip[3]) >> WHT_UPSCALE_FACTOR;
- b1 = (ip[1] + ip[2]) >> WHT_UPSCALE_FACTOR;
- c1 = (ip[1] - ip[2]) >> WHT_UPSCALE_FACTOR;
- d1 = (ip[0] - ip[3]) >> WHT_UPSCALE_FACTOR;
-
- op[0] = (a1 + b1 + 1) >> 1;
- op[1] = (c1 + d1) >> 1;
- op[2] = (a1 - b1) >> 1;
- op[3] = (d1 - c1) >> 1;
-
- ip += 4;
- op += half_pitch;
- }
-
- ip = output;
- op = output;
- for (i = 0; i < 4; i++) {- a1 = ip[half_pitch * 0] + ip[half_pitch * 3];
- b1 = ip[half_pitch * 1] + ip[half_pitch * 2];
- c1 = ip[half_pitch * 1] - ip[half_pitch * 2];
- d1 = ip[half_pitch * 0] - ip[half_pitch * 3];
-
-
- op[half_pitch * 0] = (a1 + b1 + 1) >> 1;
- op[half_pitch * 1] = (c1 + d1) >> 1;
- op[half_pitch * 2] = (a1 - b1) >> 1;
- op[half_pitch * 3] = (d1 - c1) >> 1;
-
- ip++;
- op++;
- }
-}
-
-void vp9_short_inv_walsh4x4_1_x8_c(int16_t *in, int16_t *out, int pitch) {- int i;
- int16_t tmp[4];
- int16_t *ip = in;
- int16_t *op = tmp;
- const int half_pitch = pitch >> 1;
-
- op[0] = ((ip[0] >> WHT_UPSCALE_FACTOR) + 1) >> 1;
- op[1] = op[2] = op[3] = (ip[0] >> WHT_UPSCALE_FACTOR) >> 1;
-
- ip = tmp;
- op = out;
- for (i = 0; i < 4; i++) {- op[half_pitch * 0] = (ip[0] + 1) >> 1;
- op[half_pitch * 1] = op[half_pitch * 2] = op[half_pitch * 3] = ip[0] >> 1;
- ip++;
- op++;
- }
-}
-
-void vp9_dc_only_inv_walsh_add_c(int input_dc, uint8_t *pred_ptr,
- uint8_t *dst_ptr,
- int pitch, int stride) {- int r, c;
- int16_t dc = input_dc;
- int16_t tmp[4 * 4];
- vp9_short_inv_walsh4x4_1_x8_c(&dc, tmp, 4 << 1);
-
- for (r = 0; r < 4; r++) {- for (c = 0; c < 4; c++)
- dst_ptr[c] = clip_pixel(tmp[r * 4 + c] + pred_ptr[c]);
-
- dst_ptr += stride;
- pred_ptr += pitch;
- }
-}
-
-void vp9_idct4_1d_c(int16_t *input, int16_t *output) {- int16_t step[4];
- int temp1, temp2;
- // stage 1
- temp1 = (input[0] + input[2]) * cospi_16_64;
- temp2 = (input[0] - input[2]) * cospi_16_64;
- step[0] = dct_const_round_shift(temp1);
- step[1] = dct_const_round_shift(temp2);
- temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
- temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
- step[2] = dct_const_round_shift(temp1);
- step[3] = dct_const_round_shift(temp2);
-
- // stage 2
- output[0] = step[0] + step[3];
- output[1] = step[1] + step[2];
- output[2] = step[1] - step[2];
- output[3] = step[0] - step[3];
-}
-
-void vp9_short_idct4x4llm_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[4 * 4];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[4], temp_out[4];
-
- // Rows
- for (i = 0; i < 4; ++i) {- for (j = 0; j < 4; ++j)
- temp_in[j] = input[j];
- vp9_idct4_1d(temp_in, outptr);
- input += 4;
- outptr += 4;
- }
-
- // Columns
- for (i = 0; i < 4; ++i) {- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
- vp9_idct4_1d(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 4);
- }
-}
-
-void vp9_short_idct4x4llm_1_c(int16_t *input, int16_t *output, int pitch) {- int i;
- int a1;
- int16_t *op = output;
- const int half_pitch = pitch >> 1;
- int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- a1 = ROUND_POWER_OF_TWO(out, 4);
-
- for (i = 0; i < 4; i++) {- op[0] = op[1] = op[2] = op[3] = a1;
- op += half_pitch;
- }
-}
-
-void vp9_dc_only_idct_add_c(int input_dc, uint8_t *pred_ptr,
- uint8_t *dst_ptr, int pitch, int stride) {- int a1;
- int r, c;
- int16_t out = dct_const_round_shift(input_dc * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- a1 = ROUND_POWER_OF_TWO(out, 4);
-
- for (r = 0; r < 4; r++) {- for (c = 0; c < 4; c++)
- dst_ptr[c] = clip_pixel(a1 + pred_ptr[c]);
-
- dst_ptr += stride;
- pred_ptr += pitch;
- }
-}
-
-static void idct8_1d(int16_t *input, int16_t *output) {- int16_t step1[8], step2[8];
- int temp1, temp2;
- // stage 1
- step1[0] = input[0];
- step1[2] = input[4];
- step1[1] = input[2];
- step1[3] = input[6];
- temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
- temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
- temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
- temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
-
- // stage 2 & stage 3 - even half
- vp9_idct4_1d(step1, step1);
-
- // stage 2 - odd half
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
-
- // stage 3 -odd half
- step1[4] = step2[4];
- temp1 = (step2[6] - step2[5]) * cospi_16_64;
- temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
- step1[7] = step2[7];
-
- // stage 4
- output[0] = step1[0] + step1[7];
- output[1] = step1[1] + step1[6];
- output[2] = step1[2] + step1[5];
- output[3] = step1[3] + step1[4];
- output[4] = step1[3] - step1[4];
- output[5] = step1[2] - step1[5];
- output[6] = step1[1] - step1[6];
- output[7] = step1[0] - step1[7];
-}
-
-void vp9_short_idct8x8_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[8 * 8];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[8], temp_out[8];
-
- // Rows
- for (i = 0; i < 8; ++i) {- idct8_1d(input, outptr);
- input += 8;
- outptr += 8;
- }
-
- // Columns
- for (i = 0; i < 8; ++i) {- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
- idct8_1d(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
- }
-}
-
-static void iadst4_1d(int16_t *input, int16_t *output) {- int s0, s1, s2, s3, s4, s5, s6, s7;
-
- int x0 = input[0];
- int x1 = input[1];
- int x2 = input[2];
- int x3 = input[3];
-
- if (!(x0 | x1 | x2 | x3)) {- output[0] = output[1] = output[2] = output[3] = 0;
- return;
- }
-
- s0 = sinpi_1_9 * x0;
- s1 = sinpi_2_9 * x0;
- s2 = sinpi_3_9 * x1;
- s3 = sinpi_4_9 * x2;
- s4 = sinpi_1_9 * x2;
- s5 = sinpi_2_9 * x3;
- s6 = sinpi_4_9 * x3;
- s7 = x0 - x2 + x3;
-
- x0 = s0 + s3 + s5;
- x1 = s1 - s4 - s6;
- x2 = sinpi_3_9 * s7;
- x3 = s2;
-
- s0 = x0 + x3;
- s1 = x1 + x3;
- s2 = x2;
- s3 = x0 + x1 - x3;
-
- // 1-D transform scaling factor is sqrt(2).
- // The overall dynamic range is 14b (input) + 14b (multiplication scaling)
- // + 1b (addition) = 29b.
- // Hence the output bit depth is 15b.
- output[0] = dct_const_round_shift(s0);
- output[1] = dct_const_round_shift(s1);
- output[2] = dct_const_round_shift(s2);
- output[3] = dct_const_round_shift(s3);
-}
-
-void vp9_short_iht4x4_c(int16_t *input, int16_t *output,
- int pitch, int tx_type) {- const transform_2d IHT_4[] = {- { vp9_idct4_1d, vp9_idct4_1d }, // DCT_DCT = 0- { iadst4_1d, vp9_idct4_1d }, // ADST_DCT = 1- { vp9_idct4_1d, iadst4_1d }, // DCT_ADST = 2- { iadst4_1d, iadst4_1d } // ADST_ADST = 3- };
-
- int i, j;
- int16_t out[4 * 4];
- int16_t *outptr = out;
- int16_t temp_in[4], temp_out[4];
-
- // inverse transform row vectors
- for (i = 0; i < 4; ++i) {- IHT_4[tx_type].rows(input, outptr);
- input += 4;
- outptr += 4;
- }
-
- // inverse transform column vectors
- for (i = 0; i < 4; ++i) {- for (j = 0; j < 4; ++j)
- temp_in[j] = out[j * 4 + i];
- IHT_4[tx_type].cols(temp_in, temp_out);
- for (j = 0; j < 4; ++j)
- output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 4);
- }
-}
-
-static void iadst8_1d(int16_t *input, int16_t *output) {- int s0, s1, s2, s3, s4, s5, s6, s7;
-
- int x0 = input[7];
- int x1 = input[0];
- int x2 = input[5];
- int x3 = input[2];
- int x4 = input[3];
- int x5 = input[4];
- int x6 = input[1];
- int x7 = input[6];
-
- if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {- output[0] = output[1] = output[2] = output[3] = output[4]
- = output[5] = output[6] = output[7] = 0;
- return;
- }
-
- // stage 1
- s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
- s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
- s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
- s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
- s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
- s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
- s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
- s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
-
- x0 = dct_const_round_shift(s0 + s4);
- x1 = dct_const_round_shift(s1 + s5);
- x2 = dct_const_round_shift(s2 + s6);
- x3 = dct_const_round_shift(s3 + s7);
- x4 = dct_const_round_shift(s0 - s4);
- x5 = dct_const_round_shift(s1 - s5);
- x6 = dct_const_round_shift(s2 - s6);
- x7 = dct_const_round_shift(s3 - s7);
-
- // stage 2
- s0 = x0;
- s1 = x1;
- s2 = x2;
- s3 = x3;
- s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
- s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
- s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
- s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
-
- x0 = s0 + s2;
- x1 = s1 + s3;
- x2 = s0 - s2;
- x3 = s1 - s3;
- x4 = dct_const_round_shift(s4 + s6);
- x5 = dct_const_round_shift(s5 + s7);
- x6 = dct_const_round_shift(s4 - s6);
- x7 = dct_const_round_shift(s5 - s7);
-
- // stage 3
- s2 = cospi_16_64 * (x2 + x3);
- s3 = cospi_16_64 * (x2 - x3);
- s6 = cospi_16_64 * (x6 + x7);
- s7 = cospi_16_64 * (x6 - x7);
-
- x2 = dct_const_round_shift(s2);
- x3 = dct_const_round_shift(s3);
- x6 = dct_const_round_shift(s6);
- x7 = dct_const_round_shift(s7);
-
- output[0] = x0;
- output[1] = -x4;
- output[2] = x6;
- output[3] = -x2;
- output[4] = x3;
- output[5] = -x7;
- output[6] = x5;
- output[7] = -x1;
-}
-
-static const transform_2d IHT_8[] = {- { idct8_1d, idct8_1d }, // DCT_DCT = 0- { iadst8_1d, idct8_1d }, // ADST_DCT = 1- { idct8_1d, iadst8_1d }, // DCT_ADST = 2- { iadst8_1d, iadst8_1d } // ADST_ADST = 3-};
-
-void vp9_short_iht8x8_c(int16_t *input, int16_t *output,
- int pitch, int tx_type) {- int i, j;
- int16_t out[8 * 8];
- int16_t *outptr = out;
- int16_t temp_in[8], temp_out[8];
- const transform_2d ht = IHT_8[tx_type];
-
- // inverse transform row vectors
- for (i = 0; i < 8; ++i) {- ht.rows(input, outptr);
- input += 8;
- outptr += 8;
- }
-
- // inverse transform column vectors
- for (i = 0; i < 8; ++i) {- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
- ht.cols(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
- }
-}
-
-void vp9_short_idct10_8x8_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[8 * 8];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[8], temp_out[8];
-
- vpx_memset(out, 0, sizeof(out));
- // First transform rows
- // only first 4 row has non-zero coefs
- for (i = 0; i < 4; ++i) {- idct8_1d(input, outptr);
- input += 8;
- outptr += 8;
- }
-
- // Then transform columns
- for (i = 0; i < 8; ++i) {- for (j = 0; j < 8; ++j)
- temp_in[j] = out[j * 8 + i];
- idct8_1d(temp_in, temp_out);
- for (j = 0; j < 8; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 5);
- }
-}
-
-void vp9_short_idct1_8x8_c(int16_t *input, int16_t *output) {- int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- output[0] = ROUND_POWER_OF_TWO(out, 5);
-}
-
-static void idct16_1d(int16_t *input, int16_t *output) {- int16_t step1[16], step2[16];
- int temp1, temp2;
-
- // stage 1
- step1[0] = input[0/2];
- step1[1] = input[16/2];
- step1[2] = input[8/2];
- step1[3] = input[24/2];
- step1[4] = input[4/2];
- step1[5] = input[20/2];
- step1[6] = input[12/2];
- step1[7] = input[28/2];
- step1[8] = input[2/2];
- step1[9] = input[18/2];
- step1[10] = input[10/2];
- step1[11] = input[26/2];
- step1[12] = input[6/2];
- step1[13] = input[22/2];
- step1[14] = input[14/2];
- step1[15] = input[30/2];
-
- // stage 2
- step2[0] = step1[0];
- step2[1] = step1[1];
- step2[2] = step1[2];
- step2[3] = step1[3];
- step2[4] = step1[4];
- step2[5] = step1[5];
- step2[6] = step1[6];
- step2[7] = step1[7];
-
- temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
- temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = dct_const_round_shift(temp1);
- step2[15] = dct_const_round_shift(temp2);
-
- temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
- temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
-
- temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
- temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
-
- temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
- temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
-
- // stage 3
- step1[0] = step2[0];
- step1[1] = step2[1];
- step1[2] = step2[2];
- step1[3] = step2[3];
-
- temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
- temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
- temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
- temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
-
- step1[8] = step2[8] + step2[9];
- step1[9] = step2[8] - step2[9];
- step1[10] = -step2[10] + step2[11];
- step1[11] = step2[10] + step2[11];
- step1[12] = step2[12] + step2[13];
- step1[13] = step2[12] - step2[13];
- step1[14] = -step2[14] + step2[15];
- step1[15] = step2[14] + step2[15];
-
- temp1 = (step1[0] + step1[1]) * cospi_16_64;
- temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = dct_const_round_shift(temp1);
- step2[1] = dct_const_round_shift(temp2);
- temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
- temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = dct_const_round_shift(temp1);
- step2[3] = dct_const_round_shift(temp2);
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
-
- step2[8] = step1[8];
- step2[15] = step1[15];
- temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
- temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
- temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
- temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
- step2[11] = step1[11];
- step2[12] = step1[12];
-
- // stage 5
- step1[0] = step2[0] + step2[3];
- step1[1] = step2[1] + step2[2];
- step1[2] = step2[1] - step2[2];
- step1[3] = step2[0] - step2[3];
- step1[4] = step2[4];
- temp1 = (step2[6] - step2[5]) * cospi_16_64;
- temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
- step1[7] = step2[7];
-
- step1[8] = step2[8] + step2[11];
- step1[9] = step2[9] + step2[10];
- step1[10] = step2[9] - step2[10];
- step1[11] = step2[8] - step2[11];
- step1[12] = -step2[12] + step2[15];
- step1[13] = -step2[13] + step2[14];
- step1[14] = step2[13] + step2[14];
- step1[15] = step2[12] + step2[15];
-
- // stage 6
- step2[0] = step1[0] + step1[7];
- step2[1] = step1[1] + step1[6];
- step2[2] = step1[2] + step1[5];
- step2[3] = step1[3] + step1[4];
- step2[4] = step1[3] - step1[4];
- step2[5] = step1[2] - step1[5];
- step2[6] = step1[1] - step1[6];
- step2[7] = step1[0] - step1[7];
- step2[8] = step1[8];
- step2[9] = step1[9];
- temp1 = (-step1[10] + step1[13]) * cospi_16_64;
- temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
- temp1 = (-step1[11] + step1[12]) * cospi_16_64;
- temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
- step2[14] = step1[14];
- step2[15] = step1[15];
-
- // stage 7
- output[0] = step2[0] + step2[15];
- output[1] = step2[1] + step2[14];
- output[2] = step2[2] + step2[13];
- output[3] = step2[3] + step2[12];
- output[4] = step2[4] + step2[11];
- output[5] = step2[5] + step2[10];
- output[6] = step2[6] + step2[9];
- output[7] = step2[7] + step2[8];
- output[8] = step2[7] - step2[8];
- output[9] = step2[6] - step2[9];
- output[10] = step2[5] - step2[10];
- output[11] = step2[4] - step2[11];
- output[12] = step2[3] - step2[12];
- output[13] = step2[2] - step2[13];
- output[14] = step2[1] - step2[14];
- output[15] = step2[0] - step2[15];
-}
-
-void vp9_short_idct16x16_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[16 * 16];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[16], temp_out[16];
-
- // First transform rows
- for (i = 0; i < 16; ++i) {- idct16_1d(input, outptr);
- input += 16;
- outptr += 16;
- }
-
- // Then transform columns
- for (i = 0; i < 16; ++i) {- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
- idct16_1d(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
- }
-}
-
-void iadst16_1d(int16_t *input, int16_t *output) {- int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15;
-
- int x0 = input[15];
- int x1 = input[0];
- int x2 = input[13];
- int x3 = input[2];
- int x4 = input[11];
- int x5 = input[4];
- int x6 = input[9];
- int x7 = input[6];
- int x8 = input[7];
- int x9 = input[8];
- int x10 = input[5];
- int x11 = input[10];
- int x12 = input[3];
- int x13 = input[12];
- int x14 = input[1];
- int x15 = input[14];
-
- if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8
- | x9 | x10 | x11 | x12 | x13 | x14 | x15)) {- output[0] = output[1] = output[2] = output[3] = output[4]
- = output[5] = output[6] = output[7] = output[8]
- = output[9] = output[10] = output[11] = output[12]
- = output[13] = output[14] = output[15] = 0;
- return;
- }
-
- // stage 1
- s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
- s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
- s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
- s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
- s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
- s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
- s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
- s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
- s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
- s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
- s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
- s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
- s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
- s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
- s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
- s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
-
- x0 = dct_const_round_shift(s0 + s8);
- x1 = dct_const_round_shift(s1 + s9);
- x2 = dct_const_round_shift(s2 + s10);
- x3 = dct_const_round_shift(s3 + s11);
- x4 = dct_const_round_shift(s4 + s12);
- x5 = dct_const_round_shift(s5 + s13);
- x6 = dct_const_round_shift(s6 + s14);
- x7 = dct_const_round_shift(s7 + s15);
- x8 = dct_const_round_shift(s0 - s8);
- x9 = dct_const_round_shift(s1 - s9);
- x10 = dct_const_round_shift(s2 - s10);
- x11 = dct_const_round_shift(s3 - s11);
- x12 = dct_const_round_shift(s4 - s12);
- x13 = dct_const_round_shift(s5 - s13);
- x14 = dct_const_round_shift(s6 - s14);
- x15 = dct_const_round_shift(s7 - s15);
-
- // stage 2
- s0 = x0;
- s1 = x1;
- s2 = x2;
- s3 = x3;
- s4 = x4;
- s5 = x5;
- s6 = x6;
- s7 = x7;
- s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
- s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
- s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
- s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
- s12 = - x12 * cospi_28_64 + x13 * cospi_4_64;
- s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
- s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
- s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
-
- x0 = s0 + s4;
- x1 = s1 + s5;
- x2 = s2 + s6;
- x3 = s3 + s7;
- x4 = s0 - s4;
- x5 = s1 - s5;
- x6 = s2 - s6;
- x7 = s3 - s7;
- x8 = dct_const_round_shift(s8 + s12);
- x9 = dct_const_round_shift(s9 + s13);
- x10 = dct_const_round_shift(s10 + s14);
- x11 = dct_const_round_shift(s11 + s15);
- x12 = dct_const_round_shift(s8 - s12);
- x13 = dct_const_round_shift(s9 - s13);
- x14 = dct_const_round_shift(s10 - s14);
- x15 = dct_const_round_shift(s11 - s15);
-
- // stage 3
- s0 = x0;
- s1 = x1;
- s2 = x2;
- s3 = x3;
- s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
- s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
- s6 = - x6 * cospi_24_64 + x7 * cospi_8_64;
- s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
- s8 = x8;
- s9 = x9;
- s10 = x10;
- s11 = x11;
- s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
- s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
- s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
- s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
-
- x0 = s0 + s2;
- x1 = s1 + s3;
- x2 = s0 - s2;
- x3 = s1 - s3;
- x4 = dct_const_round_shift(s4 + s6);
- x5 = dct_const_round_shift(s5 + s7);
- x6 = dct_const_round_shift(s4 - s6);
- x7 = dct_const_round_shift(s5 - s7);
- x8 = s8 + s10;
- x9 = s9 + s11;
- x10 = s8 - s10;
- x11 = s9 - s11;
- x12 = dct_const_round_shift(s12 + s14);
- x13 = dct_const_round_shift(s13 + s15);
- x14 = dct_const_round_shift(s12 - s14);
- x15 = dct_const_round_shift(s13 - s15);
-
- // stage 4
- s2 = (- cospi_16_64) * (x2 + x3);
- s3 = cospi_16_64 * (x2 - x3);
- s6 = cospi_16_64 * (x6 + x7);
- s7 = cospi_16_64 * (- x6 + x7);
- s10 = cospi_16_64 * (x10 + x11);
- s11 = cospi_16_64 * (- x10 + x11);
- s14 = (- cospi_16_64) * (x14 + x15);
- s15 = cospi_16_64 * (x14 - x15);
-
- x2 = dct_const_round_shift(s2);
- x3 = dct_const_round_shift(s3);
- x6 = dct_const_round_shift(s6);
- x7 = dct_const_round_shift(s7);
- x10 = dct_const_round_shift(s10);
- x11 = dct_const_round_shift(s11);
- x14 = dct_const_round_shift(s14);
- x15 = dct_const_round_shift(s15);
-
- output[0] = x0;
- output[1] = -x8;
- output[2] = x12;
- output[3] = -x4;
- output[4] = x6;
- output[5] = x14;
- output[6] = x10;
- output[7] = x2;
- output[8] = x3;
- output[9] = x11;
- output[10] = x15;
- output[11] = x7;
- output[12] = x5;
- output[13] = -x13;
- output[14] = x9;
- output[15] = -x1;
-}
-
-static const transform_2d IHT_16[] = {- { idct16_1d, idct16_1d }, // DCT_DCT = 0- { iadst16_1d, idct16_1d }, // ADST_DCT = 1- { idct16_1d, iadst16_1d }, // DCT_ADST = 2- { iadst16_1d, iadst16_1d } // ADST_ADST = 3-};
-
-void vp9_short_iht16x16_c(int16_t *input, int16_t *output,
- int pitch, int tx_type) {- int i, j;
- int16_t out[16 * 16];
- int16_t *outptr = out;
- int16_t temp_in[16], temp_out[16];
- const transform_2d ht = IHT_16[tx_type];
-
- // Rows
- for (i = 0; i < 16; ++i) {- ht.rows(input, outptr);
- input += 16;
- outptr += 16;
- }
-
- // Columns
- for (i = 0; i < 16; ++i) {- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j * 16 + i];
- ht.cols(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- output[j * pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
- }
-}
-
-void vp9_short_idct10_16x16_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[16 * 16];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[16], temp_out[16];
-
- /* First transform rows. Since all non-zero dct coefficients are in
- * upper-left 4x4 area, we only need to calculate first 4 rows here.
- */
- vpx_memset(out, 0, sizeof(out));
- for (i = 0; i < 4; ++i) {- idct16_1d(input, outptr);
- input += 16;
- outptr += 16;
- }
-
- // Then transform columns
- for (i = 0; i < 16; ++i) {- for (j = 0; j < 16; ++j)
- temp_in[j] = out[j*16 + i];
- idct16_1d(temp_in, temp_out);
- for (j = 0; j < 16; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
- }
-}
-
-
-void vp9_short_idct1_16x16_c(int16_t *input, int16_t *output) {- int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- output[0] = ROUND_POWER_OF_TWO(out, 6);
-}
-
-static void idct32_1d(int16_t *input, int16_t *output) {- int16_t step1[32], step2[32];
- int temp1, temp2;
-
- // stage 1
- step1[0] = input[0];
- step1[1] = input[16];
- step1[2] = input[8];
- step1[3] = input[24];
- step1[4] = input[4];
- step1[5] = input[20];
- step1[6] = input[12];
- step1[7] = input[28];
- step1[8] = input[2];
- step1[9] = input[18];
- step1[10] = input[10];
- step1[11] = input[26];
- step1[12] = input[6];
- step1[13] = input[22];
- step1[14] = input[14];
- step1[15] = input[30];
-
- temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
- temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
- step1[16] = dct_const_round_shift(temp1);
- step1[31] = dct_const_round_shift(temp2);
-
- temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
- temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
- step1[17] = dct_const_round_shift(temp1);
- step1[30] = dct_const_round_shift(temp2);
-
- temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
- temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
-
- temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
- temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
- step1[19] = dct_const_round_shift(temp1);
- step1[28] = dct_const_round_shift(temp2);
-
- temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
- temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
-
- temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
- temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
-
- temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
- temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
-
- temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
- temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
- step1[23] = dct_const_round_shift(temp1);
- step1[24] = dct_const_round_shift(temp2);
-
- // stage 2
- step2[0] = step1[0];
- step2[1] = step1[1];
- step2[2] = step1[2];
- step2[3] = step1[3];
- step2[4] = step1[4];
- step2[5] = step1[5];
- step2[6] = step1[6];
- step2[7] = step1[7];
-
- temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
- temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
- step2[8] = dct_const_round_shift(temp1);
- step2[15] = dct_const_round_shift(temp2);
-
- temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
- temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
-
- temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
- temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
-
- temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
- temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
-
- step2[16] = step1[16] + step1[17];
- step2[17] = step1[16] - step1[17];
- step2[18] = -step1[18] + step1[19];
- step2[19] = step1[18] + step1[19];
- step2[20] = step1[20] + step1[21];
- step2[21] = step1[20] - step1[21];
- step2[22] = -step1[22] + step1[23];
- step2[23] = step1[22] + step1[23];
- step2[24] = step1[24] + step1[25];
- step2[25] = step1[24] - step1[25];
- step2[26] = -step1[26] + step1[27];
- step2[27] = step1[26] + step1[27];
- step2[28] = step1[28] + step1[29];
- step2[29] = step1[28] - step1[29];
- step2[30] = -step1[30] + step1[31];
- step2[31] = step1[30] + step1[31];
-
- // stage 3
- step1[0] = step2[0];
- step1[1] = step2[1];
- step1[2] = step2[2];
- step1[3] = step2[3];
-
- temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
- temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
- step1[4] = dct_const_round_shift(temp1);
- step1[7] = dct_const_round_shift(temp2);
- temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
- temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
-
- step1[8] = step2[8] + step2[9];
- step1[9] = step2[8] - step2[9];
- step1[10] = -step2[10] + step2[11];
- step1[11] = step2[10] + step2[11];
- step1[12] = step2[12] + step2[13];
- step1[13] = step2[12] - step2[13];
- step1[14] = -step2[14] + step2[15];
- step1[15] = step2[14] + step2[15];
-
- step1[16] = step2[16];
- step1[31] = step2[31];
- temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
- temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
- step1[17] = dct_const_round_shift(temp1);
- step1[30] = dct_const_round_shift(temp2);
- temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
- temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
- step1[19] = step2[19];
- step1[20] = step2[20];
- temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
- temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
- temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
- temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
- step1[23] = step2[23];
- step1[24] = step2[24];
- step1[27] = step2[27];
- step1[28] = step2[28];
-
- // stage 4
- temp1 = (step1[0] + step1[1]) * cospi_16_64;
- temp2 = (step1[0] - step1[1]) * cospi_16_64;
- step2[0] = dct_const_round_shift(temp1);
- step2[1] = dct_const_round_shift(temp2);
- temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
- temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
- step2[2] = dct_const_round_shift(temp1);
- step2[3] = dct_const_round_shift(temp2);
- step2[4] = step1[4] + step1[5];
- step2[5] = step1[4] - step1[5];
- step2[6] = -step1[6] + step1[7];
- step2[7] = step1[6] + step1[7];
-
- step2[8] = step1[8];
- step2[15] = step1[15];
- temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
- temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
- step2[9] = dct_const_round_shift(temp1);
- step2[14] = dct_const_round_shift(temp2);
- temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
- temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
- step2[11] = step1[11];
- step2[12] = step1[12];
-
- step2[16] = step1[16] + step1[19];
- step2[17] = step1[17] + step1[18];
- step2[18] = step1[17] - step1[18];
- step2[19] = step1[16] - step1[19];
- step2[20] = -step1[20] + step1[23];
- step2[21] = -step1[21] + step1[22];
- step2[22] = step1[21] + step1[22];
- step2[23] = step1[20] + step1[23];
-
- step2[24] = step1[24] + step1[27];
- step2[25] = step1[25] + step1[26];
- step2[26] = step1[25] - step1[26];
- step2[27] = step1[24] - step1[27];
- step2[28] = -step1[28] + step1[31];
- step2[29] = -step1[29] + step1[30];
- step2[30] = step1[29] + step1[30];
- step2[31] = step1[28] + step1[31];
-
- // stage 5
- step1[0] = step2[0] + step2[3];
- step1[1] = step2[1] + step2[2];
- step1[2] = step2[1] - step2[2];
- step1[3] = step2[0] - step2[3];
- step1[4] = step2[4];
- temp1 = (step2[6] - step2[5]) * cospi_16_64;
- temp2 = (step2[5] + step2[6]) * cospi_16_64;
- step1[5] = dct_const_round_shift(temp1);
- step1[6] = dct_const_round_shift(temp2);
- step1[7] = step2[7];
-
- step1[8] = step2[8] + step2[11];
- step1[9] = step2[9] + step2[10];
- step1[10] = step2[9] - step2[10];
- step1[11] = step2[8] - step2[11];
- step1[12] = -step2[12] + step2[15];
- step1[13] = -step2[13] + step2[14];
- step1[14] = step2[13] + step2[14];
- step1[15] = step2[12] + step2[15];
-
- step1[16] = step2[16];
- step1[17] = step2[17];
- temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
- temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
- step1[18] = dct_const_round_shift(temp1);
- step1[29] = dct_const_round_shift(temp2);
- temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
- temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
- step1[19] = dct_const_round_shift(temp1);
- step1[28] = dct_const_round_shift(temp2);
- temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
- temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
- temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
- temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
- step1[22] = step2[22];
- step1[23] = step2[23];
- step1[24] = step2[24];
- step1[25] = step2[25];
- step1[30] = step2[30];
- step1[31] = step2[31];
-
- // stage 6
- step2[0] = step1[0] + step1[7];
- step2[1] = step1[1] + step1[6];
- step2[2] = step1[2] + step1[5];
- step2[3] = step1[3] + step1[4];
- step2[4] = step1[3] - step1[4];
- step2[5] = step1[2] - step1[5];
- step2[6] = step1[1] - step1[6];
- step2[7] = step1[0] - step1[7];
- step2[8] = step1[8];
- step2[9] = step1[9];
- temp1 = (-step1[10] + step1[13]) * cospi_16_64;
- temp2 = (step1[10] + step1[13]) * cospi_16_64;
- step2[10] = dct_const_round_shift(temp1);
- step2[13] = dct_const_round_shift(temp2);
- temp1 = (-step1[11] + step1[12]) * cospi_16_64;
- temp2 = (step1[11] + step1[12]) * cospi_16_64;
- step2[11] = dct_const_round_shift(temp1);
- step2[12] = dct_const_round_shift(temp2);
- step2[14] = step1[14];
- step2[15] = step1[15];
-
- step2[16] = step1[16] + step1[23];
- step2[17] = step1[17] + step1[22];
- step2[18] = step1[18] + step1[21];
- step2[19] = step1[19] + step1[20];
- step2[20] = step1[19] - step1[20];
- step2[21] = step1[18] - step1[21];
- step2[22] = step1[17] - step1[22];
- step2[23] = step1[16] - step1[23];
-
- step2[24] = -step1[24] + step1[31];
- step2[25] = -step1[25] + step1[30];
- step2[26] = -step1[26] + step1[29];
- step2[27] = -step1[27] + step1[28];
- step2[28] = step1[27] + step1[28];
- step2[29] = step1[26] + step1[29];
- step2[30] = step1[25] + step1[30];
- step2[31] = step1[24] + step1[31];
-
- // stage 7
- step1[0] = step2[0] + step2[15];
- step1[1] = step2[1] + step2[14];
- step1[2] = step2[2] + step2[13];
- step1[3] = step2[3] + step2[12];
- step1[4] = step2[4] + step2[11];
- step1[5] = step2[5] + step2[10];
- step1[6] = step2[6] + step2[9];
- step1[7] = step2[7] + step2[8];
- step1[8] = step2[7] - step2[8];
- step1[9] = step2[6] - step2[9];
- step1[10] = step2[5] - step2[10];
- step1[11] = step2[4] - step2[11];
- step1[12] = step2[3] - step2[12];
- step1[13] = step2[2] - step2[13];
- step1[14] = step2[1] - step2[14];
- step1[15] = step2[0] - step2[15];
-
- step1[16] = step2[16];
- step1[17] = step2[17];
- step1[18] = step2[18];
- step1[19] = step2[19];
- temp1 = (-step2[20] + step2[27]) * cospi_16_64;
- temp2 = (step2[20] + step2[27]) * cospi_16_64;
- step1[20] = dct_const_round_shift(temp1);
- step1[27] = dct_const_round_shift(temp2);
- temp1 = (-step2[21] + step2[26]) * cospi_16_64;
- temp2 = (step2[21] + step2[26]) * cospi_16_64;
- step1[21] = dct_const_round_shift(temp1);
- step1[26] = dct_const_round_shift(temp2);
- temp1 = (-step2[22] + step2[25]) * cospi_16_64;
- temp2 = (step2[22] + step2[25]) * cospi_16_64;
- step1[22] = dct_const_round_shift(temp1);
- step1[25] = dct_const_round_shift(temp2);
- temp1 = (-step2[23] + step2[24]) * cospi_16_64;
- temp2 = (step2[23] + step2[24]) * cospi_16_64;
- step1[23] = dct_const_round_shift(temp1);
- step1[24] = dct_const_round_shift(temp2);
- step1[28] = step2[28];
- step1[29] = step2[29];
- step1[30] = step2[30];
- step1[31] = step2[31];
-
- // final stage
- output[0] = step1[0] + step1[31];
- output[1] = step1[1] + step1[30];
- output[2] = step1[2] + step1[29];
- output[3] = step1[3] + step1[28];
- output[4] = step1[4] + step1[27];
- output[5] = step1[5] + step1[26];
- output[6] = step1[6] + step1[25];
- output[7] = step1[7] + step1[24];
- output[8] = step1[8] + step1[23];
- output[9] = step1[9] + step1[22];
- output[10] = step1[10] + step1[21];
- output[11] = step1[11] + step1[20];
- output[12] = step1[12] + step1[19];
- output[13] = step1[13] + step1[18];
- output[14] = step1[14] + step1[17];
- output[15] = step1[15] + step1[16];
- output[16] = step1[15] - step1[16];
- output[17] = step1[14] - step1[17];
- output[18] = step1[13] - step1[18];
- output[19] = step1[12] - step1[19];
- output[20] = step1[11] - step1[20];
- output[21] = step1[10] - step1[21];
- output[22] = step1[9] - step1[22];
- output[23] = step1[8] - step1[23];
- output[24] = step1[7] - step1[24];
- output[25] = step1[6] - step1[25];
- output[26] = step1[5] - step1[26];
- output[27] = step1[4] - step1[27];
- output[28] = step1[3] - step1[28];
- output[29] = step1[2] - step1[29];
- output[30] = step1[1] - step1[30];
- output[31] = step1[0] - step1[31];
-}
-
-void vp9_short_idct32x32_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[32 * 32];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[32], temp_out[32];
-
- // Rows
- for (i = 0; i < 32; ++i) {- idct32_1d(input, outptr);
- input += 32;
- outptr += 32;
- }
-
- // Columns
- for (i = 0; i < 32; ++i) {- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
- idct32_1d(temp_in, temp_out);
- for (j = 0; j < 32; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
- }
-}
-
-void vp9_short_idct1_32x32_c(int16_t *input, int16_t *output) {- int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- output[0] = ROUND_POWER_OF_TWO(out, 6);
-}
-
-void vp9_short_idct10_32x32_c(int16_t *input, int16_t *output, int pitch) {- int16_t out[32 * 32];
- int16_t *outptr = out;
- const int half_pitch = pitch >> 1;
- int i, j;
- int16_t temp_in[32], temp_out[32];
-
- /* First transform rows. Since all non-zero dct coefficients are in
- * upper-left 4x4 area, we only need to calculate first 4 rows here.
- */
- vpx_memset(out, 0, sizeof(out));
- for (i = 0; i < 4; ++i) {- idct32_1d(input, outptr);
- input += 32;
- outptr += 32;
- }
-
- // Columns
- for (i = 0; i < 32; ++i) {- for (j = 0; j < 32; ++j)
- temp_in[j] = out[j * 32 + i];
- idct32_1d(temp_in, temp_out);
- for (j = 0; j < 32; ++j)
- output[j * half_pitch + i] = ROUND_POWER_OF_TWO(temp_out[j], 6);
- }
-}
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -279,11 +279,11 @@
#
# dct
#
-prototype void vp9_short_idct4x4llm_1 "int16_t *input, int16_t *output, int pitch"
-specialize vp9_short_idct4x4llm_1
+prototype void vp9_short_idct4x4_1 "int16_t *input, int16_t *output, int pitch"
+specialize vp9_short_idct4x4_1
-prototype void vp9_short_idct4x4llm "int16_t *input, int16_t *output, int pitch"
-specialize vp9_short_idct4x4llm sse2
+prototype void vp9_short_idct4x4 "int16_t *input, int16_t *output, int pitch"
+specialize vp9_short_idct4x4 sse2
prototype void vp9_short_idct8x8 "int16_t *input, int16_t *output, int pitch"
specialize vp9_short_idct8x8
@@ -330,10 +330,10 @@
prototype void vp9_dc_only_idct_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
specialize vp9_dc_only_idct_add sse2
-prototype void vp9_short_inv_walsh4x4_1_x8 "int16_t *input, int16_t *output, int pitch"
-specialize vp9_short_inv_walsh4x4_1_x8
-prototype void vp9_short_inv_walsh4x4_x8 "int16_t *input, int16_t *output, int pitch"
-specialize vp9_short_inv_walsh4x4_x8
+prototype void vp9_short_iwalsh4x4_1 "int16_t *input, int16_t *output, int pitch"
+specialize vp9_short_iwalsh4x4_1
+prototype void vp9_short_iwalsh4x4 "int16_t *input, int16_t *output, int pitch"
+specialize vp9_short_iwalsh4x4
prototype void vp9_dc_only_inv_walsh_add "int input_dc, uint8_t *pred_ptr, uint8_t *dst_ptr, int pitch, int stride"
specialize vp9_dc_only_inv_walsh_add
@@ -600,11 +600,11 @@
prototype void vp9_short_fdct16x16 "int16_t *InputData, int16_t *OutputData, int pitch"
specialize vp9_short_fdct16x16
-prototype void vp9_short_walsh4x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch"
-specialize vp9_short_walsh4x4_x8
+prototype void vp9_short_walsh4x4 "int16_t *InputData, int16_t *OutputData, int pitch"
+specialize vp9_short_walsh4x4
-prototype void vp9_short_walsh8x4_x8 "int16_t *InputData, int16_t *OutputData, int pitch"
-specialize vp9_short_walsh8x4_x8
+prototype void vp9_short_walsh8x4 "int16_t *InputData, int16_t *OutputData, int pitch"
+specialize vp9_short_walsh8x4
#
# Motion search
--- /dev/null
+++ b/vp9/common/x86/vp9_idct_sse2.asm
@@ -1,0 +1,712 @@
+;
+; 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"
+
+;void vp9_idct_dequant_0_2x_sse2
+; (
+; short *qcoeff - 0
+; short *dequant - 1
+; unsigned char *pre - 2
+; unsigned char *dst - 3
+; int dst_stride - 4
+; int blk_stride - 5
+; )
+
+global sym(vp9_idct_dequant_0_2x_sse2) PRIVATE
+sym(vp9_idct_dequant_0_2x_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 6
+ GET_GOT rbx
+ ; end prolog
+
+ mov rdx, arg(1) ; dequant
+ mov rax, arg(0) ; qcoeff
+
+ movd xmm4, [rax]
+ movd xmm5, [rdx]
+
+ pinsrw xmm4, [rax+32], 4
+ pinsrw xmm5, [rdx], 4
+
+ pmullw xmm4, xmm5
+
+ ; Zero out xmm5, for use unpacking
+ pxor xmm5, xmm5
+
+ ; clear coeffs
+ movd [rax], xmm5
+ movd [rax+32], xmm5
+;pshufb
+ pshuflw xmm4, xmm4, 00000000b
+ pshufhw xmm4, xmm4, 00000000b
+
+ mov rax, arg(2) ; pre
+ paddw xmm4, [GLOBAL(fours)]
+
+ movsxd rcx, dword ptr arg(5) ; blk_stride
+ psraw xmm4, 3
+
+ movq xmm0, [rax]
+ movq xmm1, [rax+rcx]
+ movq xmm2, [rax+2*rcx]
+ lea rcx, [3*rcx]
+ movq xmm3, [rax+rcx]
+
+ punpcklbw xmm0, xmm5
+ punpcklbw xmm1, xmm5
+ punpcklbw xmm2, xmm5
+ punpcklbw xmm3, xmm5
+
+ mov rax, arg(3) ; dst
+ movsxd rdx, dword ptr arg(4) ; dst_stride
+
+ ; Add to predict buffer
+ paddw xmm0, xmm4
+ paddw xmm1, xmm4
+ paddw xmm2, xmm4
+ paddw xmm3, xmm4
+
+ ; pack up before storing
+ packuswb xmm0, xmm5
+ packuswb xmm1, xmm5
+ packuswb xmm2, xmm5
+ packuswb xmm3, xmm5
+
+ ; store blocks back out
+ movq [rax], xmm0
+ movq [rax + rdx], xmm1
+
+ lea rax, [rax + 2*rdx]
+
+ movq [rax], xmm2
+ movq [rax + rdx], xmm3
+
+ ; begin epilog
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+global sym(vp9_idct_dequant_full_2x_sse2) PRIVATE
+sym(vp9_idct_dequant_full_2x_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ ; special case when 2 blocks have 0 or 1 coeffs
+ ; dc is set as first coeff, so no need to load qcoeff
+ mov rax, arg(0) ; qcoeff
+ mov rsi, arg(2) ; pre
+ mov rdi, arg(3) ; dst
+ movsxd rcx, dword ptr arg(5) ; blk_stride
+
+ ; Zero out xmm7, for use unpacking
+ pxor xmm7, xmm7
+
+ mov rdx, arg(1) ; dequant
+
+ ; note the transpose of xmm1 and xmm2, necessary for shuffle
+ ; to spit out sensicle data
+ movdqa xmm0, [rax]
+ movdqa xmm2, [rax+16]
+ movdqa xmm1, [rax+32]
+ movdqa xmm3, [rax+48]
+
+ ; Clear out coeffs
+ movdqa [rax], xmm7
+ movdqa [rax+16], xmm7
+ movdqa [rax+32], xmm7
+ movdqa [rax+48], xmm7
+
+ ; dequantize qcoeff buffer
+ pmullw xmm0, [rdx]
+ pmullw xmm2, [rdx+16]
+ pmullw xmm1, [rdx]
+ pmullw xmm3, [rdx+16]
+
+ ; repack so block 0 row x and block 1 row x are together
+ movdqa xmm4, xmm0
+ punpckldq xmm0, xmm1
+ punpckhdq xmm4, xmm1
+
+ pshufd xmm0, xmm0, 11011000b
+ pshufd xmm1, xmm4, 11011000b
+
+ movdqa xmm4, xmm2
+ punpckldq xmm2, xmm3
+ punpckhdq xmm4, xmm3
+
+ pshufd xmm2, xmm2, 11011000b
+ pshufd xmm3, xmm4, 11011000b
+
+ ; first pass
+ psubw xmm0, xmm2 ; b1 = 0-2
+ paddw xmm2, xmm2 ;
+
+ movdqa xmm5, xmm1
+ paddw xmm2, xmm0 ; a1 = 0+2
+
+ pmulhw xmm5, [GLOBAL(x_s1sqr2)]
+ paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
+
+ movdqa xmm7, xmm3
+ pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
+
+ paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw xmm7, xmm5 ; c1
+
+ movdqa xmm5, xmm1
+ movdqa xmm4, xmm3
+
+ pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
+ paddw xmm5, xmm1
+
+ pmulhw xmm3, [GLOBAL(x_s1sqr2)]
+ paddw xmm3, xmm4
+
+ paddw xmm3, xmm5 ; d1
+ movdqa xmm6, xmm2 ; a1
+
+ movdqa xmm4, xmm0 ; b1
+ paddw xmm2, xmm3 ;0
+
+ paddw xmm4, xmm7 ;1
+ psubw xmm0, xmm7 ;2
+
+ psubw xmm6, xmm3 ;3
+
+ ; transpose for the second pass
+ movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
+ punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
+ punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
+
+ movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
+ punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
+ punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
+
+
+ movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
+ punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
+ punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
+
+ movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
+ punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
+ punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
+
+
+ movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
+ punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
+ punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
+
+ movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
+ punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
+ punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
+
+ pshufd xmm0, xmm2, 11011000b
+ pshufd xmm2, xmm1, 11011000b
+
+ pshufd xmm1, xmm5, 11011000b
+ pshufd xmm3, xmm7, 11011000b
+
+ ; second pass
+ psubw xmm0, xmm2 ; b1 = 0-2
+ paddw xmm2, xmm2
+
+ movdqa xmm5, xmm1
+ paddw xmm2, xmm0 ; a1 = 0+2
+
+ pmulhw xmm5, [GLOBAL(x_s1sqr2)]
+ paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
+
+ movdqa xmm7, xmm3
+ pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
+
+ paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw xmm7, xmm5 ; c1
+
+ movdqa xmm5, xmm1
+ movdqa xmm4, xmm3
+
+ pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
+ paddw xmm5, xmm1
+
+ pmulhw xmm3, [GLOBAL(x_s1sqr2)]
+ paddw xmm3, xmm4
+
+ paddw xmm3, xmm5 ; d1
+ paddw xmm0, [GLOBAL(fours)]
+
+ paddw xmm2, [GLOBAL(fours)]
+ movdqa xmm6, xmm2 ; a1
+
+ movdqa xmm4, xmm0 ; b1
+ paddw xmm2, xmm3 ;0
+
+ paddw xmm4, xmm7 ;1
+ psubw xmm0, xmm7 ;2
+
+ psubw xmm6, xmm3 ;3
+ psraw xmm2, 3
+
+ psraw xmm0, 3
+ psraw xmm4, 3
+
+ psraw xmm6, 3
+
+ ; transpose to save
+ movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
+ punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
+ punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
+
+ movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
+ punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
+ punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
+
+
+ movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
+ punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
+ punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
+
+ movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
+ punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
+ punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
+
+
+ movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
+ punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
+ punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
+
+ movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
+ punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
+ punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
+
+ pshufd xmm0, xmm2, 11011000b
+ pshufd xmm2, xmm1, 11011000b
+
+ pshufd xmm1, xmm5, 11011000b
+ pshufd xmm3, xmm7, 11011000b
+
+ pxor xmm7, xmm7
+
+ ; Load up predict blocks
+ movq xmm4, [rsi]
+ movq xmm5, [rsi+rcx]
+
+ punpcklbw xmm4, xmm7
+ punpcklbw xmm5, xmm7
+
+ paddw xmm0, xmm4
+ paddw xmm1, xmm5
+
+ movq xmm4, [rsi+2*rcx]
+ lea rcx, [3*rcx]
+ movq xmm5, [rsi+rcx]
+
+ punpcklbw xmm4, xmm7
+ punpcklbw xmm5, xmm7
+
+ paddw xmm2, xmm4
+ paddw xmm3, xmm5
+
+.finish:
+
+ ; pack up before storing
+ packuswb xmm0, xmm7
+ packuswb xmm1, xmm7
+ packuswb xmm2, xmm7
+ packuswb xmm3, xmm7
+
+ ; Load destination stride before writing out,
+ ; doesn't need to persist
+ movsxd rdx, dword ptr arg(4) ; dst_stride
+
+ ; store blocks back out
+ movq [rdi], xmm0
+ movq [rdi + rdx], xmm1
+
+ lea rdi, [rdi + 2*rdx]
+
+ movq [rdi], xmm2
+ movq [rdi + rdx], xmm3
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+;void vp9_idct_dequant_dc_0_2x_sse2
+; (
+; short *qcoeff - 0
+; short *dequant - 1
+; unsigned char *pre - 2
+; unsigned char *dst - 3
+; int dst_stride - 4
+; short *dc - 5
+; )
+global sym(vp9_idct_dequant_dc_0_2x_sse2) PRIVATE
+sym(vp9_idct_dequant_dc_0_2x_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ ; special case when 2 blocks have 0 or 1 coeffs
+ ; dc is set as first coeff, so no need to load qcoeff
+ mov rax, arg(0) ; qcoeff
+ mov rsi, arg(2) ; pre
+ mov rdi, arg(3) ; dst
+ mov rdx, arg(5) ; dc
+
+ ; Zero out xmm5, for use unpacking
+ pxor xmm5, xmm5
+
+ ; load up 2 dc words here == 2*16 = doubleword
+ movd xmm4, [rdx]
+
+ ; Load up predict blocks
+ movq xmm0, [rsi]
+ movq xmm1, [rsi+16]
+ movq xmm2, [rsi+32]
+ movq xmm3, [rsi+48]
+
+ ; Duplicate and expand dc across
+ punpcklwd xmm4, xmm4
+ punpckldq xmm4, xmm4
+
+ ; Rounding to dequant and downshift
+ paddw xmm4, [GLOBAL(fours)]
+ psraw xmm4, 3
+
+ ; Predict buffer needs to be expanded from bytes to words
+ punpcklbw xmm0, xmm5
+ punpcklbw xmm1, xmm5
+ punpcklbw xmm2, xmm5
+ punpcklbw xmm3, xmm5
+
+ ; Add to predict buffer
+ paddw xmm0, xmm4
+ paddw xmm1, xmm4
+ paddw xmm2, xmm4
+ paddw xmm3, xmm4
+
+ ; pack up before storing
+ packuswb xmm0, xmm5
+ packuswb xmm1, xmm5
+ packuswb xmm2, xmm5
+ packuswb xmm3, xmm5
+
+ ; Load destination stride before writing out,
+ ; doesn't need to persist
+ movsxd rdx, dword ptr arg(4) ; dst_stride
+
+ ; store blocks back out
+ movq [rdi], xmm0
+ movq [rdi + rdx], xmm1
+
+ lea rdi, [rdi + 2*rdx]
+
+ movq [rdi], xmm2
+ movq [rdi + rdx], xmm3
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+global sym(vp9_idct_dequant_dc_full_2x_sse2) PRIVATE
+sym(vp9_idct_dequant_dc_full_2x_sse2):
+ push rbp
+ mov rbp, rsp
+ SHADOW_ARGS_TO_STACK 7
+ SAVE_XMM 7
+ GET_GOT rbx
+ push rsi
+ push rdi
+ ; end prolog
+
+ ; special case when 2 blocks have 0 or 1 coeffs
+ ; dc is set as first coeff, so no need to load qcoeff
+ mov rax, arg(0) ; qcoeff
+ mov rsi, arg(2) ; pre
+ mov rdi, arg(3) ; dst
+
+ ; Zero out xmm7, for use unpacking
+ pxor xmm7, xmm7
+
+ mov rdx, arg(1) ; dequant
+
+ ; note the transpose of xmm1 and xmm2, necessary for shuffle
+ ; to spit out sensicle data
+ movdqa xmm0, [rax]
+ movdqa xmm2, [rax+16]
+ movdqa xmm1, [rax+32]
+ movdqa xmm3, [rax+48]
+
+ ; Clear out coeffs
+ movdqa [rax], xmm7
+ movdqa [rax+16], xmm7
+ movdqa [rax+32], xmm7
+ movdqa [rax+48], xmm7
+
+ ; dequantize qcoeff buffer
+ pmullw xmm0, [rdx]
+ pmullw xmm2, [rdx+16]
+ pmullw xmm1, [rdx]
+ pmullw xmm3, [rdx+16]
+
+ ; DC component
+ mov rdx, arg(5)
+
+ ; repack so block 0 row x and block 1 row x are together
+ movdqa xmm4, xmm0
+ punpckldq xmm0, xmm1
+ punpckhdq xmm4, xmm1
+
+ pshufd xmm0, xmm0, 11011000b
+ pshufd xmm1, xmm4, 11011000b
+
+ movdqa xmm4, xmm2
+ punpckldq xmm2, xmm3
+ punpckhdq xmm4, xmm3
+
+ pshufd xmm2, xmm2, 11011000b
+ pshufd xmm3, xmm4, 11011000b
+
+ ; insert DC component
+ pinsrw xmm0, [rdx], 0
+ pinsrw xmm0, [rdx+2], 4
+
+ ; first pass
+ psubw xmm0, xmm2 ; b1 = 0-2
+ paddw xmm2, xmm2 ;
+
+ movdqa xmm5, xmm1
+ paddw xmm2, xmm0 ; a1 = 0+2
+
+ pmulhw xmm5, [GLOBAL(x_s1sqr2)]
+ paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
+
+ movdqa xmm7, xmm3
+ pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
+
+ paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw xmm7, xmm5 ; c1
+
+ movdqa xmm5, xmm1
+ movdqa xmm4, xmm3
+
+ pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
+ paddw xmm5, xmm1
+
+ pmulhw xmm3, [GLOBAL(x_s1sqr2)]
+ paddw xmm3, xmm4
+
+ paddw xmm3, xmm5 ; d1
+ movdqa xmm6, xmm2 ; a1
+
+ movdqa xmm4, xmm0 ; b1
+ paddw xmm2, xmm3 ;0
+
+ paddw xmm4, xmm7 ;1
+ psubw xmm0, xmm7 ;2
+
+ psubw xmm6, xmm3 ;3
+
+ ; transpose for the second pass
+ movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
+ punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
+ punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
+
+ movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
+ punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
+ punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
+
+
+ movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
+ punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
+ punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
+
+ movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
+ punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
+ punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
+
+
+ movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
+ punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
+ punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
+
+ movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
+ punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
+ punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
+
+ pshufd xmm0, xmm2, 11011000b
+ pshufd xmm2, xmm1, 11011000b
+
+ pshufd xmm1, xmm5, 11011000b
+ pshufd xmm3, xmm7, 11011000b
+
+ ; second pass
+ psubw xmm0, xmm2 ; b1 = 0-2
+ paddw xmm2, xmm2
+
+ movdqa xmm5, xmm1
+ paddw xmm2, xmm0 ; a1 = 0+2
+
+ pmulhw xmm5, [GLOBAL(x_s1sqr2)]
+ paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
+
+ movdqa xmm7, xmm3
+ pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
+
+ paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw xmm7, xmm5 ; c1
+
+ movdqa xmm5, xmm1
+ movdqa xmm4, xmm3
+
+ pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
+ paddw xmm5, xmm1
+
+ pmulhw xmm3, [GLOBAL(x_s1sqr2)]
+ paddw xmm3, xmm4
+
+ paddw xmm3, xmm5 ; d1
+ paddw xmm0, [GLOBAL(fours)]
+
+ paddw xmm2, [GLOBAL(fours)]
+ movdqa xmm6, xmm2 ; a1
+
+ movdqa xmm4, xmm0 ; b1
+ paddw xmm2, xmm3 ;0
+
+ paddw xmm4, xmm7 ;1
+ psubw xmm0, xmm7 ;2
+
+ psubw xmm6, xmm3 ;3
+ psraw xmm2, 3
+
+ psraw xmm0, 3
+ psraw xmm4, 3
+
+ psraw xmm6, 3
+
+ ; transpose to save
+ movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
+ punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
+ punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
+
+ movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
+ punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
+ punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
+
+
+ movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
+ punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
+ punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
+
+ movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
+ punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
+ punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
+
+
+ movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
+ punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
+ punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
+
+ movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
+ punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
+ punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
+
+ pshufd xmm0, xmm2, 11011000b
+ pshufd xmm2, xmm1, 11011000b
+
+ pshufd xmm1, xmm5, 11011000b
+ pshufd xmm3, xmm7, 11011000b
+
+ pxor xmm7, xmm7
+
+ ; Load up predict blocks
+ movq xmm4, [rsi]
+ movq xmm5, [rsi+16]
+
+ punpcklbw xmm4, xmm7
+ punpcklbw xmm5, xmm7
+
+ paddw xmm0, xmm4
+ paddw xmm1, xmm5
+
+ movq xmm4, [rsi+32]
+ movq xmm5, [rsi+48]
+
+ punpcklbw xmm4, xmm7
+ punpcklbw xmm5, xmm7
+
+ paddw xmm2, xmm4
+ paddw xmm3, xmm5
+
+.finish:
+
+ ; pack up before storing
+ packuswb xmm0, xmm7
+ packuswb xmm1, xmm7
+ packuswb xmm2, xmm7
+ packuswb xmm3, xmm7
+
+ ; Load destination stride before writing out,
+ ; doesn't need to persist
+ movsxd rdx, dword ptr arg(4) ; dst_stride
+
+ ; store blocks back out
+ movq [rdi], xmm0
+ movq [rdi + rdx], xmm1
+
+ lea rdi, [rdi + 2*rdx]
+
+ movq [rdi], xmm2
+ movq [rdi + rdx], xmm3
+
+
+ ; begin epilog
+ pop rdi
+ pop rsi
+ RESTORE_GOT
+ RESTORE_XMM
+ UNSHADOW_ARGS
+ pop rbp
+ ret
+
+SECTION_RODATA
+align 16
+fours:
+ times 8 dw 0x0004
+align 16
+x_s1sqr2:
+ times 8 dw 0x8A8C
+align 16
+x_c1sqr2less1:
+ times 8 dw 0x4E7B
--- /dev/null
+++ b/vp9/common/x86/vp9_idct_x86.c
@@ -1,0 +1,237 @@
+/*
+ * Copyright (c) 2012 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 <assert.h>
+#include <emmintrin.h> // SSE2
+#include "./vpx_config.h"
+#include "vpx/vpx_integer.h"
+#include "vp9/common/vp9_common.h"
+#include "vp9/common/vp9_idct.h"
+
+#if HAVE_SSE2
+// In order to improve performance, clip absolute diff values to [0, 255],
+// which allows to keep the additions/subtractions in 8 bits.
+void vp9_dc_only_idct_add_sse2(int input_dc, uint8_t *pred_ptr,
+ uint8_t *dst_ptr, int pitch, int stride) {+ int a1;
+ int16_t out;
+ uint8_t abs_diff;
+ __m128i p0, p1, p2, p3;
+ unsigned int extended_diff;
+ __m128i diff;
+
+ out = dct_const_round_shift(input_dc * cospi_16_64);
+ out = dct_const_round_shift(out * cospi_16_64);
+ a1 = ROUND_POWER_OF_TWO(out, 4);
+
+ // Read prediction data.
+ p0 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 0 * pitch));
+ p1 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 1 * pitch));
+ p2 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 2 * pitch));
+ p3 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 3 * pitch));
+
+ // Unpack prediction data, and store 4x4 array in 1 XMM register.
+ p0 = _mm_unpacklo_epi32(p0, p1);
+ p2 = _mm_unpacklo_epi32(p2, p3);
+ p0 = _mm_unpacklo_epi64(p0, p2);
+
+ // Clip dc value to [0, 255] range. Then, do addition or subtraction
+ // according to its sign.
+ if (a1 >= 0) {+ abs_diff = (a1 > 255) ? 255 : a1;
+ extended_diff = abs_diff * 0x01010101u;
+ diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
+
+ p1 = _mm_adds_epu8(p0, diff);
+ } else {+ abs_diff = (a1 < -255) ? 255 : -a1;
+ extended_diff = abs_diff * 0x01010101u;
+ diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
+
+ p1 = _mm_subs_epu8(p0, diff);
+ }
+
+ // Store results to dst.
+ *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
+ dst_ptr += stride;
+
+ p1 = _mm_srli_si128(p1, 4);
+ *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
+ dst_ptr += stride;
+
+ p1 = _mm_srli_si128(p1, 4);
+ *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
+ dst_ptr += stride;
+
+ p1 = _mm_srli_si128(p1, 4);
+ *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
+}
+
+void vp9_short_idct4x4_sse2(int16_t *input, int16_t *output, int pitch) {+ const __m128i zero = _mm_setzero_si128();
+ const __m128i eight = _mm_set1_epi16(8);
+ const __m128i cst = _mm_setr_epi16((int16_t)cospi_16_64, (int16_t)cospi_16_64,
+ (int16_t)cospi_16_64, (int16_t)-cospi_16_64,
+ (int16_t)cospi_24_64, (int16_t)-cospi_8_64,
+ (int16_t)cospi_8_64, (int16_t)cospi_24_64);
+ const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
+ const int half_pitch = pitch >> 1;
+ __m128i input0, input1, input2, input3;
+
+ // Rows
+ input0 = _mm_loadl_epi64((__m128i *)input);
+ input1 = _mm_loadl_epi64((__m128i *)(input + 4));
+ input2 = _mm_loadl_epi64((__m128i *)(input + 8));
+ input3 = _mm_loadl_epi64((__m128i *)(input + 12));
+
+ // Construct i3, i1, i3, i1, i2, i0, i2, i0
+ input0 = _mm_shufflelo_epi16(input0, 0xd8);
+ input1 = _mm_shufflelo_epi16(input1, 0xd8);
+ input2 = _mm_shufflelo_epi16(input2, 0xd8);
+ input3 = _mm_shufflelo_epi16(input3, 0xd8);
+
+ input0 = _mm_unpacklo_epi32(input0, input0);
+ input1 = _mm_unpacklo_epi32(input1, input1);
+ input2 = _mm_unpacklo_epi32(input2, input2);
+ input3 = _mm_unpacklo_epi32(input3, input3);
+
+ // Stage 1
+ input0 = _mm_madd_epi16(input0, cst);
+ input1 = _mm_madd_epi16(input1, cst);
+ input2 = _mm_madd_epi16(input2, cst);
+ input3 = _mm_madd_epi16(input3, cst);
+
+ input0 = _mm_add_epi32(input0, rounding);
+ input1 = _mm_add_epi32(input1, rounding);
+ input2 = _mm_add_epi32(input2, rounding);
+ input3 = _mm_add_epi32(input3, rounding);
+
+ input0 = _mm_srai_epi32(input0, DCT_CONST_BITS);
+ input1 = _mm_srai_epi32(input1, DCT_CONST_BITS);
+ input2 = _mm_srai_epi32(input2, DCT_CONST_BITS);
+ input3 = _mm_srai_epi32(input3, DCT_CONST_BITS);
+
+ // Stage 2
+ input0 = _mm_packs_epi32(input0, zero);
+ input1 = _mm_packs_epi32(input1, zero);
+ input2 = _mm_packs_epi32(input2, zero);
+ input3 = _mm_packs_epi32(input3, zero);
+
+ // Transpose
+ input1 = _mm_unpacklo_epi16(input0, input1);
+ input3 = _mm_unpacklo_epi16(input2, input3);
+ input0 = _mm_unpacklo_epi32(input1, input3);
+ input1 = _mm_unpackhi_epi32(input1, input3);
+
+ // Switch column2, column 3, and then, we got:
+ // input2: column1, column 0; input3: column2, column 3.
+ input1 = _mm_shuffle_epi32(input1, 0x4e);
+ input2 = _mm_add_epi16(input0, input1);
+ input3 = _mm_sub_epi16(input0, input1);
+
+ // Columns
+ // Construct i3, i1, i3, i1, i2, i0, i2, i0
+ input0 = _mm_shufflelo_epi16(input2, 0xd8);
+ input1 = _mm_shufflehi_epi16(input2, 0xd8);
+ input2 = _mm_shufflehi_epi16(input3, 0xd8);
+ input3 = _mm_shufflelo_epi16(input3, 0xd8);
+
+ input0 = _mm_unpacklo_epi32(input0, input0);
+ input1 = _mm_unpackhi_epi32(input1, input1);
+ input2 = _mm_unpackhi_epi32(input2, input2);
+ input3 = _mm_unpacklo_epi32(input3, input3);
+
+ // Stage 1
+ input0 = _mm_madd_epi16(input0, cst);
+ input1 = _mm_madd_epi16(input1, cst);
+ input2 = _mm_madd_epi16(input2, cst);
+ input3 = _mm_madd_epi16(input3, cst);
+
+ input0 = _mm_add_epi32(input0, rounding);
+ input1 = _mm_add_epi32(input1, rounding);
+ input2 = _mm_add_epi32(input2, rounding);
+ input3 = _mm_add_epi32(input3, rounding);
+
+ input0 = _mm_srai_epi32(input0, DCT_CONST_BITS);
+ input1 = _mm_srai_epi32(input1, DCT_CONST_BITS);
+ input2 = _mm_srai_epi32(input2, DCT_CONST_BITS);
+ input3 = _mm_srai_epi32(input3, DCT_CONST_BITS);
+
+ // Stage 2
+ input0 = _mm_packs_epi32(input0, zero);
+ input1 = _mm_packs_epi32(input1, zero);
+ input2 = _mm_packs_epi32(input2, zero);
+ input3 = _mm_packs_epi32(input3, zero);
+
+ // Transpose
+ input1 = _mm_unpacklo_epi16(input0, input1);
+ input3 = _mm_unpacklo_epi16(input2, input3);
+ input0 = _mm_unpacklo_epi32(input1, input3);
+ input1 = _mm_unpackhi_epi32(input1, input3);
+
+ // Switch column2, column 3, and then, we got:
+ // input2: column1, column 0; input3: column2, column 3.
+ input1 = _mm_shuffle_epi32(input1, 0x4e);
+ input2 = _mm_add_epi16(input0, input1);
+ input3 = _mm_sub_epi16(input0, input1);
+
+ // Final round and shift
+ input2 = _mm_add_epi16(input2, eight);
+ input3 = _mm_add_epi16(input3, eight);
+
+ input2 = _mm_srai_epi16(input2, 4);
+ input3 = _mm_srai_epi16(input3, 4);
+
+ // Store results
+ _mm_storel_epi64((__m128i *)output, input2);
+ input2 = _mm_srli_si128(input2, 8);
+ _mm_storel_epi64((__m128i *)(output + half_pitch), input2);
+
+ _mm_storel_epi64((__m128i *)(output + 3 * half_pitch), input3);
+ input3 = _mm_srli_si128(input3, 8);
+ _mm_storel_epi64((__m128i *)(output + 2 * half_pitch), input3);
+}
+
+void vp9_idct4_1d_sse2(int16_t *input, int16_t *output) {+ const __m128i zero = _mm_setzero_si128();
+ const __m128i c1 = _mm_setr_epi16((int16_t)cospi_16_64, (int16_t)cospi_16_64,
+ (int16_t)cospi_16_64, (int16_t)-cospi_16_64,
+ (int16_t)cospi_24_64, (int16_t)-cospi_8_64,
+ (int16_t)cospi_8_64, (int16_t)cospi_24_64);
+ const __m128i c2 = _mm_setr_epi16(1, 1, 1, 1, 1, -1, 1, -1);
+
+ const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
+ __m128i in, temp;
+
+ // Load input data.
+ in = _mm_loadl_epi64((__m128i *)input);
+
+ // Construct i3, i1, i3, i1, i2, i0, i2, i0
+ in = _mm_shufflelo_epi16(in, 0xd8);
+ in = _mm_unpacklo_epi32(in, in);
+
+ // Stage 1
+ in = _mm_madd_epi16(in, c1);
+ in = _mm_add_epi32(in, rounding);
+ in = _mm_srai_epi32(in, DCT_CONST_BITS);
+ in = _mm_packs_epi32(in, zero);
+
+ // Stage 2
+ temp = _mm_shufflelo_epi16(in, 0x9c);
+ in = _mm_shufflelo_epi16(in, 0xc9);
+ in = _mm_unpacklo_epi64(temp, in);
+ in = _mm_madd_epi16(in, c2);
+ in = _mm_packs_epi32(in, zero);
+
+ // Store results
+ _mm_storel_epi64((__m128i *)output, in);
+}
+
+#endif
--- a/vp9/common/x86/vp9_idctllm_sse2.asm
+++ /dev/null
@@ -1,712 +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"
-
-;void vp9_idct_dequant_0_2x_sse2
-; (
-; short *qcoeff - 0
-; short *dequant - 1
-; unsigned char *pre - 2
-; unsigned char *dst - 3
-; int dst_stride - 4
-; int blk_stride - 5
-; )
-
-global sym(vp9_idct_dequant_0_2x_sse2) PRIVATE
-sym(vp9_idct_dequant_0_2x_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- GET_GOT rbx
- ; end prolog
-
- mov rdx, arg(1) ; dequant
- mov rax, arg(0) ; qcoeff
-
- movd xmm4, [rax]
- movd xmm5, [rdx]
-
- pinsrw xmm4, [rax+32], 4
- pinsrw xmm5, [rdx], 4
-
- pmullw xmm4, xmm5
-
- ; Zero out xmm5, for use unpacking
- pxor xmm5, xmm5
-
- ; clear coeffs
- movd [rax], xmm5
- movd [rax+32], xmm5
-;pshufb
- pshuflw xmm4, xmm4, 00000000b
- pshufhw xmm4, xmm4, 00000000b
-
- mov rax, arg(2) ; pre
- paddw xmm4, [GLOBAL(fours)]
-
- movsxd rcx, dword ptr arg(5) ; blk_stride
- psraw xmm4, 3
-
- movq xmm0, [rax]
- movq xmm1, [rax+rcx]
- movq xmm2, [rax+2*rcx]
- lea rcx, [3*rcx]
- movq xmm3, [rax+rcx]
-
- punpcklbw xmm0, xmm5
- punpcklbw xmm1, xmm5
- punpcklbw xmm2, xmm5
- punpcklbw xmm3, xmm5
-
- mov rax, arg(3) ; dst
- movsxd rdx, dword ptr arg(4) ; dst_stride
-
- ; Add to predict buffer
- paddw xmm0, xmm4
- paddw xmm1, xmm4
- paddw xmm2, xmm4
- paddw xmm3, xmm4
-
- ; pack up before storing
- packuswb xmm0, xmm5
- packuswb xmm1, xmm5
- packuswb xmm2, xmm5
- packuswb xmm3, xmm5
-
- ; store blocks back out
- movq [rax], xmm0
- movq [rax + rdx], xmm1
-
- lea rax, [rax + 2*rdx]
-
- movq [rax], xmm2
- movq [rax + rdx], xmm3
-
- ; begin epilog
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-global sym(vp9_idct_dequant_full_2x_sse2) PRIVATE
-sym(vp9_idct_dequant_full_2x_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- ; special case when 2 blocks have 0 or 1 coeffs
- ; dc is set as first coeff, so no need to load qcoeff
- mov rax, arg(0) ; qcoeff
- mov rsi, arg(2) ; pre
- mov rdi, arg(3) ; dst
- movsxd rcx, dword ptr arg(5) ; blk_stride
-
- ; Zero out xmm7, for use unpacking
- pxor xmm7, xmm7
-
- mov rdx, arg(1) ; dequant
-
- ; note the transpose of xmm1 and xmm2, necessary for shuffle
- ; to spit out sensicle data
- movdqa xmm0, [rax]
- movdqa xmm2, [rax+16]
- movdqa xmm1, [rax+32]
- movdqa xmm3, [rax+48]
-
- ; Clear out coeffs
- movdqa [rax], xmm7
- movdqa [rax+16], xmm7
- movdqa [rax+32], xmm7
- movdqa [rax+48], xmm7
-
- ; dequantize qcoeff buffer
- pmullw xmm0, [rdx]
- pmullw xmm2, [rdx+16]
- pmullw xmm1, [rdx]
- pmullw xmm3, [rdx+16]
-
- ; repack so block 0 row x and block 1 row x are together
- movdqa xmm4, xmm0
- punpckldq xmm0, xmm1
- punpckhdq xmm4, xmm1
-
- pshufd xmm0, xmm0, 11011000b
- pshufd xmm1, xmm4, 11011000b
-
- movdqa xmm4, xmm2
- punpckldq xmm2, xmm3
- punpckhdq xmm4, xmm3
-
- pshufd xmm2, xmm2, 11011000b
- pshufd xmm3, xmm4, 11011000b
-
- ; first pass
- psubw xmm0, xmm2 ; b1 = 0-2
- paddw xmm2, xmm2 ;
-
- movdqa xmm5, xmm1
- paddw xmm2, xmm0 ; a1 = 0+2
-
- pmulhw xmm5, [GLOBAL(x_s1sqr2)]
- paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
-
- movdqa xmm7, xmm3
- pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
-
- paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw xmm7, xmm5 ; c1
-
- movdqa xmm5, xmm1
- movdqa xmm4, xmm3
-
- pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
- paddw xmm5, xmm1
-
- pmulhw xmm3, [GLOBAL(x_s1sqr2)]
- paddw xmm3, xmm4
-
- paddw xmm3, xmm5 ; d1
- movdqa xmm6, xmm2 ; a1
-
- movdqa xmm4, xmm0 ; b1
- paddw xmm2, xmm3 ;0
-
- paddw xmm4, xmm7 ;1
- psubw xmm0, xmm7 ;2
-
- psubw xmm6, xmm3 ;3
-
- ; transpose for the second pass
- movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
- punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
- punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
-
- movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
- punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
- punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
-
-
- movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
- punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
- punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
-
- movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
- punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
- punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
-
-
- movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
- punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
- punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
-
- movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
- punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
- punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
-
- pshufd xmm0, xmm2, 11011000b
- pshufd xmm2, xmm1, 11011000b
-
- pshufd xmm1, xmm5, 11011000b
- pshufd xmm3, xmm7, 11011000b
-
- ; second pass
- psubw xmm0, xmm2 ; b1 = 0-2
- paddw xmm2, xmm2
-
- movdqa xmm5, xmm1
- paddw xmm2, xmm0 ; a1 = 0+2
-
- pmulhw xmm5, [GLOBAL(x_s1sqr2)]
- paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
-
- movdqa xmm7, xmm3
- pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
-
- paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw xmm7, xmm5 ; c1
-
- movdqa xmm5, xmm1
- movdqa xmm4, xmm3
-
- pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
- paddw xmm5, xmm1
-
- pmulhw xmm3, [GLOBAL(x_s1sqr2)]
- paddw xmm3, xmm4
-
- paddw xmm3, xmm5 ; d1
- paddw xmm0, [GLOBAL(fours)]
-
- paddw xmm2, [GLOBAL(fours)]
- movdqa xmm6, xmm2 ; a1
-
- movdqa xmm4, xmm0 ; b1
- paddw xmm2, xmm3 ;0
-
- paddw xmm4, xmm7 ;1
- psubw xmm0, xmm7 ;2
-
- psubw xmm6, xmm3 ;3
- psraw xmm2, 3
-
- psraw xmm0, 3
- psraw xmm4, 3
-
- psraw xmm6, 3
-
- ; transpose to save
- movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
- punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
- punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
-
- movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
- punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
- punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
-
-
- movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
- punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
- punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
-
- movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
- punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
- punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
-
-
- movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
- punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
- punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
-
- movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
- punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
- punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
-
- pshufd xmm0, xmm2, 11011000b
- pshufd xmm2, xmm1, 11011000b
-
- pshufd xmm1, xmm5, 11011000b
- pshufd xmm3, xmm7, 11011000b
-
- pxor xmm7, xmm7
-
- ; Load up predict blocks
- movq xmm4, [rsi]
- movq xmm5, [rsi+rcx]
-
- punpcklbw xmm4, xmm7
- punpcklbw xmm5, xmm7
-
- paddw xmm0, xmm4
- paddw xmm1, xmm5
-
- movq xmm4, [rsi+2*rcx]
- lea rcx, [3*rcx]
- movq xmm5, [rsi+rcx]
-
- punpcklbw xmm4, xmm7
- punpcklbw xmm5, xmm7
-
- paddw xmm2, xmm4
- paddw xmm3, xmm5
-
-.finish:
-
- ; pack up before storing
- packuswb xmm0, xmm7
- packuswb xmm1, xmm7
- packuswb xmm2, xmm7
- packuswb xmm3, xmm7
-
- ; Load destination stride before writing out,
- ; doesn't need to persist
- movsxd rdx, dword ptr arg(4) ; dst_stride
-
- ; store blocks back out
- movq [rdi], xmm0
- movq [rdi + rdx], xmm1
-
- lea rdi, [rdi + 2*rdx]
-
- movq [rdi], xmm2
- movq [rdi + rdx], xmm3
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp9_idct_dequant_dc_0_2x_sse2
-; (
-; short *qcoeff - 0
-; short *dequant - 1
-; unsigned char *pre - 2
-; unsigned char *dst - 3
-; int dst_stride - 4
-; short *dc - 5
-; )
-global sym(vp9_idct_dequant_dc_0_2x_sse2) PRIVATE
-sym(vp9_idct_dequant_dc_0_2x_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- ; special case when 2 blocks have 0 or 1 coeffs
- ; dc is set as first coeff, so no need to load qcoeff
- mov rax, arg(0) ; qcoeff
- mov rsi, arg(2) ; pre
- mov rdi, arg(3) ; dst
- mov rdx, arg(5) ; dc
-
- ; Zero out xmm5, for use unpacking
- pxor xmm5, xmm5
-
- ; load up 2 dc words here == 2*16 = doubleword
- movd xmm4, [rdx]
-
- ; Load up predict blocks
- movq xmm0, [rsi]
- movq xmm1, [rsi+16]
- movq xmm2, [rsi+32]
- movq xmm3, [rsi+48]
-
- ; Duplicate and expand dc across
- punpcklwd xmm4, xmm4
- punpckldq xmm4, xmm4
-
- ; Rounding to dequant and downshift
- paddw xmm4, [GLOBAL(fours)]
- psraw xmm4, 3
-
- ; Predict buffer needs to be expanded from bytes to words
- punpcklbw xmm0, xmm5
- punpcklbw xmm1, xmm5
- punpcklbw xmm2, xmm5
- punpcklbw xmm3, xmm5
-
- ; Add to predict buffer
- paddw xmm0, xmm4
- paddw xmm1, xmm4
- paddw xmm2, xmm4
- paddw xmm3, xmm4
-
- ; pack up before storing
- packuswb xmm0, xmm5
- packuswb xmm1, xmm5
- packuswb xmm2, xmm5
- packuswb xmm3, xmm5
-
- ; Load destination stride before writing out,
- ; doesn't need to persist
- movsxd rdx, dword ptr arg(4) ; dst_stride
-
- ; store blocks back out
- movq [rdi], xmm0
- movq [rdi + rdx], xmm1
-
- lea rdi, [rdi + 2*rdx]
-
- movq [rdi], xmm2
- movq [rdi + rdx], xmm3
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-global sym(vp9_idct_dequant_dc_full_2x_sse2) PRIVATE
-sym(vp9_idct_dequant_dc_full_2x_sse2):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- SAVE_XMM 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- ; special case when 2 blocks have 0 or 1 coeffs
- ; dc is set as first coeff, so no need to load qcoeff
- mov rax, arg(0) ; qcoeff
- mov rsi, arg(2) ; pre
- mov rdi, arg(3) ; dst
-
- ; Zero out xmm7, for use unpacking
- pxor xmm7, xmm7
-
- mov rdx, arg(1) ; dequant
-
- ; note the transpose of xmm1 and xmm2, necessary for shuffle
- ; to spit out sensicle data
- movdqa xmm0, [rax]
- movdqa xmm2, [rax+16]
- movdqa xmm1, [rax+32]
- movdqa xmm3, [rax+48]
-
- ; Clear out coeffs
- movdqa [rax], xmm7
- movdqa [rax+16], xmm7
- movdqa [rax+32], xmm7
- movdqa [rax+48], xmm7
-
- ; dequantize qcoeff buffer
- pmullw xmm0, [rdx]
- pmullw xmm2, [rdx+16]
- pmullw xmm1, [rdx]
- pmullw xmm3, [rdx+16]
-
- ; DC component
- mov rdx, arg(5)
-
- ; repack so block 0 row x and block 1 row x are together
- movdqa xmm4, xmm0
- punpckldq xmm0, xmm1
- punpckhdq xmm4, xmm1
-
- pshufd xmm0, xmm0, 11011000b
- pshufd xmm1, xmm4, 11011000b
-
- movdqa xmm4, xmm2
- punpckldq xmm2, xmm3
- punpckhdq xmm4, xmm3
-
- pshufd xmm2, xmm2, 11011000b
- pshufd xmm3, xmm4, 11011000b
-
- ; insert DC component
- pinsrw xmm0, [rdx], 0
- pinsrw xmm0, [rdx+2], 4
-
- ; first pass
- psubw xmm0, xmm2 ; b1 = 0-2
- paddw xmm2, xmm2 ;
-
- movdqa xmm5, xmm1
- paddw xmm2, xmm0 ; a1 = 0+2
-
- pmulhw xmm5, [GLOBAL(x_s1sqr2)]
- paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
-
- movdqa xmm7, xmm3
- pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
-
- paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw xmm7, xmm5 ; c1
-
- movdqa xmm5, xmm1
- movdqa xmm4, xmm3
-
- pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
- paddw xmm5, xmm1
-
- pmulhw xmm3, [GLOBAL(x_s1sqr2)]
- paddw xmm3, xmm4
-
- paddw xmm3, xmm5 ; d1
- movdqa xmm6, xmm2 ; a1
-
- movdqa xmm4, xmm0 ; b1
- paddw xmm2, xmm3 ;0
-
- paddw xmm4, xmm7 ;1
- psubw xmm0, xmm7 ;2
-
- psubw xmm6, xmm3 ;3
-
- ; transpose for the second pass
- movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
- punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
- punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
-
- movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
- punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
- punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
-
-
- movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
- punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
- punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
-
- movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
- punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
- punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
-
-
- movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
- punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
- punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
-
- movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
- punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
- punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
-
- pshufd xmm0, xmm2, 11011000b
- pshufd xmm2, xmm1, 11011000b
-
- pshufd xmm1, xmm5, 11011000b
- pshufd xmm3, xmm7, 11011000b
-
- ; second pass
- psubw xmm0, xmm2 ; b1 = 0-2
- paddw xmm2, xmm2
-
- movdqa xmm5, xmm1
- paddw xmm2, xmm0 ; a1 = 0+2
-
- pmulhw xmm5, [GLOBAL(x_s1sqr2)]
- paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
-
- movdqa xmm7, xmm3
- pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
-
- paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw xmm7, xmm5 ; c1
-
- movdqa xmm5, xmm1
- movdqa xmm4, xmm3
-
- pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
- paddw xmm5, xmm1
-
- pmulhw xmm3, [GLOBAL(x_s1sqr2)]
- paddw xmm3, xmm4
-
- paddw xmm3, xmm5 ; d1
- paddw xmm0, [GLOBAL(fours)]
-
- paddw xmm2, [GLOBAL(fours)]
- movdqa xmm6, xmm2 ; a1
-
- movdqa xmm4, xmm0 ; b1
- paddw xmm2, xmm3 ;0
-
- paddw xmm4, xmm7 ;1
- psubw xmm0, xmm7 ;2
-
- psubw xmm6, xmm3 ;3
- psraw xmm2, 3
-
- psraw xmm0, 3
- psraw xmm4, 3
-
- psraw xmm6, 3
-
- ; transpose to save
- movdqa xmm7, xmm2 ; 103 102 101 100 003 002 001 000
- punpcklwd xmm2, xmm0 ; 007 003 006 002 005 001 004 000
- punpckhwd xmm7, xmm0 ; 107 103 106 102 105 101 104 100
-
- movdqa xmm5, xmm4 ; 111 110 109 108 011 010 009 008
- punpcklwd xmm4, xmm6 ; 015 011 014 010 013 009 012 008
- punpckhwd xmm5, xmm6 ; 115 111 114 110 113 109 112 108
-
-
- movdqa xmm1, xmm2 ; 007 003 006 002 005 001 004 000
- punpckldq xmm2, xmm4 ; 013 009 005 001 012 008 004 000
- punpckhdq xmm1, xmm4 ; 015 011 007 003 014 010 006 002
-
- movdqa xmm6, xmm7 ; 107 103 106 102 105 101 104 100
- punpckldq xmm7, xmm5 ; 113 109 105 101 112 108 104 100
- punpckhdq xmm6, xmm5 ; 115 111 107 103 114 110 106 102
-
-
- movdqa xmm5, xmm2 ; 013 009 005 001 012 008 004 000
- punpckldq xmm2, xmm7 ; 112 108 012 008 104 100 004 000
- punpckhdq xmm5, xmm7 ; 113 109 013 009 105 101 005 001
-
- movdqa xmm7, xmm1 ; 015 011 007 003 014 010 006 002
- punpckldq xmm1, xmm6 ; 114 110 014 010 106 102 006 002
- punpckhdq xmm7, xmm6 ; 115 111 015 011 107 103 007 003
-
- pshufd xmm0, xmm2, 11011000b
- pshufd xmm2, xmm1, 11011000b
-
- pshufd xmm1, xmm5, 11011000b
- pshufd xmm3, xmm7, 11011000b
-
- pxor xmm7, xmm7
-
- ; Load up predict blocks
- movq xmm4, [rsi]
- movq xmm5, [rsi+16]
-
- punpcklbw xmm4, xmm7
- punpcklbw xmm5, xmm7
-
- paddw xmm0, xmm4
- paddw xmm1, xmm5
-
- movq xmm4, [rsi+32]
- movq xmm5, [rsi+48]
-
- punpcklbw xmm4, xmm7
- punpcklbw xmm5, xmm7
-
- paddw xmm2, xmm4
- paddw xmm3, xmm5
-
-.finish:
-
- ; pack up before storing
- packuswb xmm0, xmm7
- packuswb xmm1, xmm7
- packuswb xmm2, xmm7
- packuswb xmm3, xmm7
-
- ; Load destination stride before writing out,
- ; doesn't need to persist
- movsxd rdx, dword ptr arg(4) ; dst_stride
-
- ; store blocks back out
- movq [rdi], xmm0
- movq [rdi + rdx], xmm1
-
- lea rdi, [rdi + 2*rdx]
-
- movq [rdi], xmm2
- movq [rdi + rdx], xmm3
-
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- RESTORE_XMM
- UNSHADOW_ARGS
- pop rbp
- ret
-
-SECTION_RODATA
-align 16
-fours:
- times 8 dw 0x0004
-align 16
-x_s1sqr2:
- times 8 dw 0x8A8C
-align 16
-x_c1sqr2less1:
- times 8 dw 0x4E7B
--- a/vp9/common/x86/vp9_idctllm_x86.c
+++ /dev/null
@@ -1,237 +1,0 @@
-/*
- * Copyright (c) 2012 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 <assert.h>
-#include <emmintrin.h> // SSE2
-#include "./vpx_config.h"
-#include "vpx/vpx_integer.h"
-#include "vp9/common/vp9_common.h"
-#include "vp9/common/vp9_idct.h"
-
-#if HAVE_SSE2
-// In order to improve performance, clip absolute diff values to [0, 255],
-// which allows to keep the additions/subtractions in 8 bits.
-void vp9_dc_only_idct_add_sse2(int input_dc, uint8_t *pred_ptr,
- uint8_t *dst_ptr, int pitch, int stride) {- int a1;
- int16_t out;
- uint8_t abs_diff;
- __m128i p0, p1, p2, p3;
- unsigned int extended_diff;
- __m128i diff;
-
- out = dct_const_round_shift(input_dc * cospi_16_64);
- out = dct_const_round_shift(out * cospi_16_64);
- a1 = ROUND_POWER_OF_TWO(out, 4);
-
- // Read prediction data.
- p0 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 0 * pitch));
- p1 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 1 * pitch));
- p2 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 2 * pitch));
- p3 = _mm_cvtsi32_si128 (*(const int *)(pred_ptr + 3 * pitch));
-
- // Unpack prediction data, and store 4x4 array in 1 XMM register.
- p0 = _mm_unpacklo_epi32(p0, p1);
- p2 = _mm_unpacklo_epi32(p2, p3);
- p0 = _mm_unpacklo_epi64(p0, p2);
-
- // Clip dc value to [0, 255] range. Then, do addition or subtraction
- // according to its sign.
- if (a1 >= 0) {- abs_diff = (a1 > 255) ? 255 : a1;
- extended_diff = abs_diff * 0x01010101u;
- diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
-
- p1 = _mm_adds_epu8(p0, diff);
- } else {- abs_diff = (a1 < -255) ? 255 : -a1;
- extended_diff = abs_diff * 0x01010101u;
- diff = _mm_shuffle_epi32(_mm_cvtsi32_si128((int)extended_diff), 0);
-
- p1 = _mm_subs_epu8(p0, diff);
- }
-
- // Store results to dst.
- *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
- dst_ptr += stride;
-
- p1 = _mm_srli_si128(p1, 4);
- *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
- dst_ptr += stride;
-
- p1 = _mm_srli_si128(p1, 4);
- *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
- dst_ptr += stride;
-
- p1 = _mm_srli_si128(p1, 4);
- *(int *)dst_ptr = _mm_cvtsi128_si32(p1);
-}
-
-void vp9_short_idct4x4llm_sse2(int16_t *input, int16_t *output, int pitch) {- const __m128i zero = _mm_setzero_si128();
- const __m128i eight = _mm_set1_epi16(8);
- const __m128i cst = _mm_setr_epi16((int16_t)cospi_16_64, (int16_t)cospi_16_64,
- (int16_t)cospi_16_64, (int16_t)-cospi_16_64,
- (int16_t)cospi_24_64, (int16_t)-cospi_8_64,
- (int16_t)cospi_8_64, (int16_t)cospi_24_64);
- const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
- const int half_pitch = pitch >> 1;
- __m128i input0, input1, input2, input3;
-
- // Rows
- input0 = _mm_loadl_epi64((__m128i *)input);
- input1 = _mm_loadl_epi64((__m128i *)(input + 4));
- input2 = _mm_loadl_epi64((__m128i *)(input + 8));
- input3 = _mm_loadl_epi64((__m128i *)(input + 12));
-
- // Construct i3, i1, i3, i1, i2, i0, i2, i0
- input0 = _mm_shufflelo_epi16(input0, 0xd8);
- input1 = _mm_shufflelo_epi16(input1, 0xd8);
- input2 = _mm_shufflelo_epi16(input2, 0xd8);
- input3 = _mm_shufflelo_epi16(input3, 0xd8);
-
- input0 = _mm_unpacklo_epi32(input0, input0);
- input1 = _mm_unpacklo_epi32(input1, input1);
- input2 = _mm_unpacklo_epi32(input2, input2);
- input3 = _mm_unpacklo_epi32(input3, input3);
-
- // Stage 1
- input0 = _mm_madd_epi16(input0, cst);
- input1 = _mm_madd_epi16(input1, cst);
- input2 = _mm_madd_epi16(input2, cst);
- input3 = _mm_madd_epi16(input3, cst);
-
- input0 = _mm_add_epi32(input0, rounding);
- input1 = _mm_add_epi32(input1, rounding);
- input2 = _mm_add_epi32(input2, rounding);
- input3 = _mm_add_epi32(input3, rounding);
-
- input0 = _mm_srai_epi32(input0, DCT_CONST_BITS);
- input1 = _mm_srai_epi32(input1, DCT_CONST_BITS);
- input2 = _mm_srai_epi32(input2, DCT_CONST_BITS);
- input3 = _mm_srai_epi32(input3, DCT_CONST_BITS);
-
- // Stage 2
- input0 = _mm_packs_epi32(input0, zero);
- input1 = _mm_packs_epi32(input1, zero);
- input2 = _mm_packs_epi32(input2, zero);
- input3 = _mm_packs_epi32(input3, zero);
-
- // Transpose
- input1 = _mm_unpacklo_epi16(input0, input1);
- input3 = _mm_unpacklo_epi16(input2, input3);
- input0 = _mm_unpacklo_epi32(input1, input3);
- input1 = _mm_unpackhi_epi32(input1, input3);
-
- // Switch column2, column 3, and then, we got:
- // input2: column1, column 0; input3: column2, column 3.
- input1 = _mm_shuffle_epi32(input1, 0x4e);
- input2 = _mm_add_epi16(input0, input1);
- input3 = _mm_sub_epi16(input0, input1);
-
- // Columns
- // Construct i3, i1, i3, i1, i2, i0, i2, i0
- input0 = _mm_shufflelo_epi16(input2, 0xd8);
- input1 = _mm_shufflehi_epi16(input2, 0xd8);
- input2 = _mm_shufflehi_epi16(input3, 0xd8);
- input3 = _mm_shufflelo_epi16(input3, 0xd8);
-
- input0 = _mm_unpacklo_epi32(input0, input0);
- input1 = _mm_unpackhi_epi32(input1, input1);
- input2 = _mm_unpackhi_epi32(input2, input2);
- input3 = _mm_unpacklo_epi32(input3, input3);
-
- // Stage 1
- input0 = _mm_madd_epi16(input0, cst);
- input1 = _mm_madd_epi16(input1, cst);
- input2 = _mm_madd_epi16(input2, cst);
- input3 = _mm_madd_epi16(input3, cst);
-
- input0 = _mm_add_epi32(input0, rounding);
- input1 = _mm_add_epi32(input1, rounding);
- input2 = _mm_add_epi32(input2, rounding);
- input3 = _mm_add_epi32(input3, rounding);
-
- input0 = _mm_srai_epi32(input0, DCT_CONST_BITS);
- input1 = _mm_srai_epi32(input1, DCT_CONST_BITS);
- input2 = _mm_srai_epi32(input2, DCT_CONST_BITS);
- input3 = _mm_srai_epi32(input3, DCT_CONST_BITS);
-
- // Stage 2
- input0 = _mm_packs_epi32(input0, zero);
- input1 = _mm_packs_epi32(input1, zero);
- input2 = _mm_packs_epi32(input2, zero);
- input3 = _mm_packs_epi32(input3, zero);
-
- // Transpose
- input1 = _mm_unpacklo_epi16(input0, input1);
- input3 = _mm_unpacklo_epi16(input2, input3);
- input0 = _mm_unpacklo_epi32(input1, input3);
- input1 = _mm_unpackhi_epi32(input1, input3);
-
- // Switch column2, column 3, and then, we got:
- // input2: column1, column 0; input3: column2, column 3.
- input1 = _mm_shuffle_epi32(input1, 0x4e);
- input2 = _mm_add_epi16(input0, input1);
- input3 = _mm_sub_epi16(input0, input1);
-
- // Final round and shift
- input2 = _mm_add_epi16(input2, eight);
- input3 = _mm_add_epi16(input3, eight);
-
- input2 = _mm_srai_epi16(input2, 4);
- input3 = _mm_srai_epi16(input3, 4);
-
- // Store results
- _mm_storel_epi64((__m128i *)output, input2);
- input2 = _mm_srli_si128(input2, 8);
- _mm_storel_epi64((__m128i *)(output + half_pitch), input2);
-
- _mm_storel_epi64((__m128i *)(output + 3 * half_pitch), input3);
- input3 = _mm_srli_si128(input3, 8);
- _mm_storel_epi64((__m128i *)(output + 2 * half_pitch), input3);
-}
-
-void vp9_idct4_1d_sse2(int16_t *input, int16_t *output) {- const __m128i zero = _mm_setzero_si128();
- const __m128i c1 = _mm_setr_epi16((int16_t)cospi_16_64, (int16_t)cospi_16_64,
- (int16_t)cospi_16_64, (int16_t)-cospi_16_64,
- (int16_t)cospi_24_64, (int16_t)-cospi_8_64,
- (int16_t)cospi_8_64, (int16_t)cospi_24_64);
- const __m128i c2 = _mm_setr_epi16(1, 1, 1, 1, 1, -1, 1, -1);
-
- const __m128i rounding = _mm_set1_epi32(DCT_CONST_ROUNDING);
- __m128i in, temp;
-
- // Load input data.
- in = _mm_loadl_epi64((__m128i *)input);
-
- // Construct i3, i1, i3, i1, i2, i0, i2, i0
- in = _mm_shufflelo_epi16(in, 0xd8);
- in = _mm_unpacklo_epi32(in, in);
-
- // Stage 1
- in = _mm_madd_epi16(in, c1);
- in = _mm_add_epi32(in, rounding);
- in = _mm_srai_epi32(in, DCT_CONST_BITS);
- in = _mm_packs_epi32(in, zero);
-
- // Stage 2
- temp = _mm_shufflelo_epi16(in, 0x9c);
- in = _mm_shufflelo_epi16(in, 0xc9);
- in = _mm_unpacklo_epi64(temp, in);
- in = _mm_madd_epi16(in, c2);
- in = _mm_packs_epi32(in, zero);
-
- // Store results
- _mm_storel_epi64((__m128i *)output, in);
-}
-
-#endif
--- a/vp9/decoder/vp9_decodframe.c
+++ b/vp9/decoder/vp9_decodframe.c
@@ -124,14 +124,14 @@
if (mb->lossless) {assert(qindex == 0);
- mb->inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8;
- mb->inv_txm4x4 = vp9_short_inv_walsh4x4_x8;
+ mb->inv_txm4x4_1 = vp9_short_iwalsh4x4_1;
+ mb->inv_txm4x4 = vp9_short_iwalsh4x4;
mb->itxm_add = vp9_dequant_idct_add_lossless_c;
mb->itxm_add_y_block = vp9_dequant_idct_add_y_block_lossless_c;
mb->itxm_add_uv_block = vp9_dequant_idct_add_uv_block_lossless_c;
} else {- mb->inv_txm4x4_1 = vp9_short_idct4x4llm_1;
- mb->inv_txm4x4 = vp9_short_idct4x4llm;
+ mb->inv_txm4x4_1 = vp9_short_idct4x4_1;
+ mb->inv_txm4x4 = vp9_short_idct4x4;
mb->itxm_add = vp9_dequant_idct_add;
mb->itxm_add_y_block = vp9_dequant_idct_add_y_block;
mb->itxm_add_uv_block = vp9_dequant_idct_add_uv_block;
--- a/vp9/decoder/vp9_dequantize.c
+++ b/vp9/decoder/vp9_dequantize.c
@@ -126,7 +126,7 @@
input[i] *= dq[i];
// the idct halves ( >> 1) the pitch
- vp9_short_idct4x4llm(input, output, 4 << 1);
+ vp9_short_idct4x4(input, output, 4 << 1);
vpx_memset(input, 0, 32);
@@ -148,7 +148,7 @@
input[i] *= dq[i];
// the idct halves ( >> 1) the pitch
- vp9_short_idct4x4llm(input, output, 4 << 1);
+ vp9_short_idct4x4(input, output, 4 << 1);
vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, pred, pitch, dest, stride);
}
@@ -163,7 +163,7 @@
for (i = 0; i < 16; i++)
input[i] *= dq[i];
- vp9_short_inv_walsh4x4_x8_c(input, output, 4 << 1);
+ vp9_short_iwalsh4x4_c(input, output, 4 << 1);
vpx_memset(input, 0, 32);
@@ -186,7 +186,7 @@
for (i = 1; i < 16; i++)
input[i] *= dq[i];
- vp9_short_inv_walsh4x4_x8_c(input, output, 4 << 1);
+ vp9_short_iwalsh4x4_c(input, output, 4 << 1);
vpx_memset(input, 0, 32);
vp9_add_residual_4x4(output, pred, pitch, dest, stride);
}
--- a/vp9/encoder/vp9_dct.c
+++ b/vp9/encoder/vp9_dct.c
@@ -374,7 +374,7 @@
}
}
-void vp9_short_walsh4x4_x8_c(short *input, short *output, int pitch) {+void vp9_short_walsh4x4_c(short *input, short *output, int pitch) {int i;
int a1, b1, c1, d1;
short *ip = input;
@@ -414,9 +414,9 @@
}
}
-void vp9_short_walsh8x4_x8_c(short *input, short *output, int pitch) {- vp9_short_walsh4x4_x8_c(input, output, pitch);
- vp9_short_walsh4x4_x8_c(input + 4, output + 16, pitch);
+void vp9_short_walsh8x4_c(short *input, short *output, int pitch) {+ vp9_short_walsh4x4_c(input, output, pitch);
+ vp9_short_walsh4x4_c(input + 4, output + 16, pitch);
}
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -1217,10 +1217,10 @@
static void switch_lossless_mode(VP9_COMP *cpi, int lossless) { if (lossless) {- cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4_x8;
- cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4_x8;
- cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8;
- cpi->mb.e_mbd.inv_txm4x4 = vp9_short_inv_walsh4x4_x8;
+ cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4;
+ cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4;
+ cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_iwalsh4x4_1;
+ cpi->mb.e_mbd.inv_txm4x4 = vp9_short_iwalsh4x4;
cpi->mb.optimize = 0;
cpi->common.filter_level = 0;
cpi->zbin_mode_boost_enabled = FALSE;
@@ -1228,8 +1228,8 @@
} else {cpi->mb.fwd_txm8x4 = vp9_short_fdct8x4;
cpi->mb.fwd_txm4x4 = vp9_short_fdct4x4;
- cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_idct4x4llm_1;
- cpi->mb.e_mbd.inv_txm4x4 = vp9_short_idct4x4llm;
+ cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_idct4x4_1;
+ cpi->mb.e_mbd.inv_txm4x4 = vp9_short_idct4x4;
}
}
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -843,8 +843,8 @@
cpi->mb.fwd_txm8x4 = vp9_short_fdct8x4;
cpi->mb.fwd_txm4x4 = vp9_short_fdct4x4;
if (cpi->oxcf.lossless || cpi->mb.e_mbd.lossless) {- cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4_x8;
- cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4_x8;
+ cpi->mb.fwd_txm8x4 = vp9_short_walsh8x4;
+ cpi->mb.fwd_txm4x4 = vp9_short_walsh4x4;
}
cpi->mb.quantize_b_4x4 = vp9_regular_quantize_b_4x4;
@@ -1217,11 +1217,11 @@
cpi->oxcf.lossless = oxcf->lossless;
if (cpi->oxcf.lossless) {- cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8;
- cpi->mb.e_mbd.inv_txm4x4 = vp9_short_inv_walsh4x4_x8;
+ cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_iwalsh4x4_1;
+ cpi->mb.e_mbd.inv_txm4x4 = vp9_short_iwalsh4x4;
} else {- cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_idct4x4llm_1;
- cpi->mb.e_mbd.inv_txm4x4 = vp9_short_idct4x4llm;
+ cpi->mb.e_mbd.inv_txm4x4_1 = vp9_short_idct4x4_1;
+ cpi->mb.e_mbd.inv_txm4x4 = vp9_short_idct4x4;
}
cpi->baseline_gf_interval = DEFAULT_GF_INTERVAL;
--- a/vp9/vp9_common.mk
+++ b/vp9/vp9_common.mk
@@ -28,7 +28,7 @@
VP9_COMMON_SRCS-yes += common/vp9_filter.h
VP9_COMMON_SRCS-yes += common/vp9_findnearmv.c
VP9_COMMON_SRCS-yes += common/generic/vp9_systemdependent.c
-VP9_COMMON_SRCS-yes += common/vp9_idctllm.c
+VP9_COMMON_SRCS-yes += common/vp9_idct.c
VP9_COMMON_SRCS-yes += common/vp9_alloccommon.h
VP9_COMMON_SRCS-yes += common/vp9_blockd.h
VP9_COMMON_SRCS-yes += common/vp9_common.h
@@ -91,7 +91,7 @@
VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_iwalsh_mmx.asm
VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_recon_mmx.asm
VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_loopfilter_mmx.asm
-VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_idctllm_sse2.asm
+VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_idct_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_iwalsh_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_loopfilter_sse2.asm
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_recon_sse2.asm
@@ -110,13 +110,13 @@
VP9_COMMON_SRCS-$(HAVE_SSE3) += common/x86/vp9_mask_sse3.asm
endif
-VP9_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/vp9_idctllm_x86.c
+VP9_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/vp9_idct_x86.c
VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_sadmxn_x86.c
ifeq ($(HAVE_SSE2),yes)
-vp9/common/x86/vp9_idctllm_x86.c.o: CFLAGS += -msse2
+vp9/common/x86/vp9_idct_x86.c.o: CFLAGS += -msse2
vp9/common/x86/vp9_loopfilter_x86.c.o: CFLAGS += -msse2
vp9/common/x86/vp9_sadmxn_x86.c.o: CFLAGS += -msse2
-vp9/common/x86/vp9_idctllm_x86.c.d: CFLAGS += -msse2
+vp9/common/x86/vp9_idct_x86.c.d: CFLAGS += -msse2
vp9/common/x86/vp9_loopfilter_x86.c.d: CFLAGS += -msse2
vp9/common/x86/vp9_sadmxn_x86.c.d: CFLAGS += -msse2
endif