ref: 3633ffcbf7b04b15e2dcb6b2834e1685fbbf6bf4
parent: 97826df96b86176d8cd23d398922349950853300
author: hkuang <hkuang@google.com>
date: Tue Jan 21 09:49:15 EST 2014
Add vp9_tm_predictor_16x16 neon implementation which is 3.5 times faster than C. Change-Id: I24439ba7a2971829c11620f34848facf2c916678
--- a/vp9/common/arm/neon/vp9_reconintra_neon.asm
+++ b/vp9/common/arm/neon/vp9_reconintra_neon.asm
@@ -18,6 +18,7 @@
EXPORT |vp9_h_predictor_32x32_neon|
EXPORT |vp9_tm_predictor_4x4_neon|
EXPORT |vp9_tm_predictor_8x8_neon|
+ EXPORT |vp9_tm_predictor_16x16_neon|
ARM
REQUIRE8
PRESERVE8
@@ -402,5 +403,101 @@
vst1.64 {d1}, [r0], r1bx lr
ENDP ; |vp9_tm_predictor_8x8_neon|
+
+;void vp9_tm_predictor_16x16_neon (uint8_t *dst, ptrdiff_t y_stride,
+; const uint8_t *above,
+; const uint8_t *left)
+; r0 uint8_t *dst
+; r1 ptrdiff_t y_stride
+; r2 const uint8_t *above
+; r3 const uint8_t *left
+
+|vp9_tm_predictor_16x16_neon| PROC
+ ; Load ytop_left = above[-1];
+ sub r12, r2, #1
+ ldrb r12, [r12]
+ vdup.u8 q0, r12
+
+ ; Load above 8 pixels
+ vld1.8 q1, [r2]
+
+ ; preload 8 left into r12
+ vld1.8 d18, [r3]!
+
+ ; Compute above - ytop_left
+ vsubl.u8 q2, d2, d0
+ vsubl.u8 q3, d3, d1
+
+ vmovl.u8 q10, d18
+
+ ; Load left row by row and compute left + (above - ytop_left)
+ ; Process 8 rows in each single loop and loop 2 times to process 16 rows.
+ mov r2, #2
+
+loop_16x16_neon
+ ; Process two rows.
+ vdup.16 q0, d20[0]
+ vdup.16 q8, d20[1]
+ vadd.s16 q1, q0, q2
+ vadd.s16 q0, q0, q3
+ vadd.s16 q11, q8, q2
+ vadd.s16 q8, q8, q3
+ vqshrun.s16 d2, q1, #0
+ vqshrun.s16 d3, q0, #0
+ vqshrun.s16 d22, q11, #0
+ vqshrun.s16 d23, q8, #0
+ vdup.16 q0, d20[2] ; proload next 2 rows data
+ vdup.16 q8, d20[3]
+ vst1.64 {d2,d3}, [r0], r1+ vst1.64 {d22,d23}, [r0], r1+
+ ; Process two rows.
+ vadd.s16 q1, q0, q2
+ vadd.s16 q0, q0, q3
+ vadd.s16 q11, q8, q2
+ vadd.s16 q8, q8, q3
+ vqshrun.s16 d2, q1, #0
+ vqshrun.s16 d3, q0, #0
+ vqshrun.s16 d22, q11, #0
+ vqshrun.s16 d23, q8, #0
+ vdup.16 q0, d21[0] ; proload next 2 rows data
+ vdup.16 q8, d21[1]
+ vst1.64 {d2,d3}, [r0], r1+ vst1.64 {d22,d23}, [r0], r1+
+ vadd.s16 q1, q0, q2
+ vadd.s16 q0, q0, q3
+ vadd.s16 q11, q8, q2
+ vadd.s16 q8, q8, q3
+ vqshrun.s16 d2, q1, #0
+ vqshrun.s16 d3, q0, #0
+ vqshrun.s16 d22, q11, #0
+ vqshrun.s16 d23, q8, #0
+ vdup.16 q0, d21[2] ; proload next 2 rows data
+ vdup.16 q8, d21[3]
+ vst1.64 {d2,d3}, [r0], r1+ vst1.64 {d22,d23}, [r0], r1+
+
+ vadd.s16 q1, q0, q2
+ vadd.s16 q0, q0, q3
+ vadd.s16 q11, q8, q2
+ vadd.s16 q8, q8, q3
+ vqshrun.s16 d2, q1, #0
+ vqshrun.s16 d3, q0, #0
+ vqshrun.s16 d22, q11, #0
+ vqshrun.s16 d23, q8, #0
+ vdup.16 q0, d20[2]
+ vdup.16 q8, d20[3]
+ vld1.8 d18, [r3]! ; preload 8 left into r12
+ vmovl.u8 q10, d18
+ vst1.64 {d2,d3}, [r0], r1+ vst1.64 {d22,d23}, [r0], r1+
+ subs r2, r2, #1
+ bgt loop_16x16_neon
+
+ bx lr
+ ENDP ; |vp9_tm_predictor_16x16_neon|
END
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -135,7 +135,7 @@
specialize vp9_v_predictor_16x16 $sse2_x86inc neon
prototype void vp9_tm_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
-specialize vp9_tm_predictor_16x16 $sse2_x86inc
+specialize vp9_tm_predictor_16x16 $sse2_x86inc neon
prototype void vp9_dc_predictor_16x16 "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left"
specialize vp9_dc_predictor_16x16 $sse2_x86inc dspr2
--
⑨