[x265] [PATCH] asm: pixel_avg_48x64, pixel_avg_8x32

[dnyaneshwar at multicorewareinc.com]

# HG changeset patch
# User Dnyaneshwar Gorade <dnyaneshwar at multicorewareinc.com>
# Date 1384237578 -19800
#      Tue Nov 12 11:56:18 2013 +0530
# Node ID 60303b478851805812f70b06a39c52836c2dffa8
# Parent  1130599ba7772f83bdfdbfafb896110544a0e594
asm: pixel_avg_48x64, pixel_avg_8x32

diff -r 1130599ba777 -r 60303b478851 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Nov 12 11:44:58 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Tue Nov 12 11:56:18 2013 +0530
@@ -121,6 +121,7 @@
     p.pixelavg_pp[LUMA_64x64] = x265_pixel_avg_64x64_ ## cpu; \
     p.pixelavg_pp[LUMA_64x48] = x265_pixel_avg_64x48_ ## cpu; \
     p.pixelavg_pp[LUMA_64x16] = x265_pixel_avg_64x16_ ## cpu; \
+    p.pixelavg_pp[LUMA_48x64] = x265_pixel_avg_48x64_ ## cpu; \
     p.pixelavg_pp[LUMA_32x64] = x265_pixel_avg_32x64_ ## cpu; \
     p.pixelavg_pp[LUMA_32x32] = x265_pixel_avg_32x32_ ## cpu; \
     p.pixelavg_pp[LUMA_32x24] = x265_pixel_avg_32x24_ ## cpu; \
@@ -133,6 +134,7 @@
     p.pixelavg_pp[LUMA_16x12]  = x265_pixel_avg_16x12_ ## cpu; \
     p.pixelavg_pp[LUMA_16x8]  = x265_pixel_avg_16x8_ ## cpu; \
     p.pixelavg_pp[LUMA_16x4]  = x265_pixel_avg_16x4_ ## cpu; \
+    p.pixelavg_pp[LUMA_8x32]  = x265_pixel_avg_8x32_ ## cpu; \
     p.pixelavg_pp[LUMA_8x16]  = x265_pixel_avg_8x16_ ## cpu; \
     p.pixelavg_pp[LUMA_8x8]   = x265_pixel_avg_8x8_ ## cpu; \
     p.pixelavg_pp[LUMA_8x4]   = x265_pixel_avg_8x4_ ## cpu;
diff -r 1130599ba777 -r 60303b478851 source/common/x86/mc-a.asm
--- a/source/common/x86/mc-a.asm	Tue Nov 12 11:44:58 2013 +0530
+++ b/source/common/x86/mc-a.asm	Tue Nov 12 11:56:18 2013 +0530
@@ -239,6 +239,7 @@
 AVG_WEIGHT 32
 AVG_WEIGHT 64
 AVG_WEIGHT 24
+AVG_WEIGHT 48
 %if HIGH_BIT_DEPTH
 INIT_XMM sse2
 AVG_WEIGHT 4,  8
@@ -251,6 +252,7 @@
 AVG_WEIGHT 32, 7
 AVG_WEIGHT 64, 7
 AVG_WEIGHT 24, 7
+AVG_WEIGHT 48, 7
 %define BIWEIGHT BIWEIGHT_SSSE3
 %define BIWEIGHT_START BIWEIGHT_START_SSSE3
 INIT_MMX ssse3
@@ -261,6 +263,7 @@
 AVG_WEIGHT 32, 7
 AVG_WEIGHT 64, 7
 AVG_WEIGHT 24, 7
+AVG_WEIGHT 48, 7
 
 INIT_YMM avx2
 cglobal pixel_avg_weight_w16
@@ -720,6 +723,7 @@
 AVGH 4, 2
 
 AVG_FUNC 8, movq, movq
+AVGH 8, 32
 AVGH 8, 16
 AVGH 8,  8
 AVGH 8,  4
@@ -746,6 +750,9 @@
 AVG_FUNC 24, movq, movq
 AVGH 24, 32
 
+AVG_FUNC 48, movq, movq
+AVGH 48, 64
+
 INIT_XMM sse2
 AVG_FUNC 64, movdqu, movdqa
 AVGH 64, 64
@@ -769,6 +776,11 @@
 AVGH 16, 12
 AVGH 16, 8
 AVGH 16, 4
+
+AVG_FUNC 48, movdqu, movdqa
+AVGH 48, 64
+
+AVGH  8, 32
 AVGH  8, 16
 AVGH  8,  8
 AVGH  8,  4
@@ -791,6 +803,10 @@
 AVGH 16, 12
 AVGH 16, 8
 AVGH 16, 4
+
+AVGH 48, 64
+
+AVGH  8, 32
 AVGH  8, 16
 AVGH  8,  8
 AVGH  8,  4
diff -r 1130599ba777 -r 60303b478851 source/common/x86/pixel.h
--- a/source/common/x86/pixel.h	Tue Nov 12 11:44:58 2013 +0530
+++ b/source/common/x86/pixel.h	Tue Nov 12 11:56:18 2013 +0530
@@ -241,6 +241,7 @@
 DECL_SUF(x265_pixel_avg_64x64, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_64x48, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_64x16, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
+DECL_SUF(x265_pixel_avg_48x64, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_32x64, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_32x32, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_32x24, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
@@ -253,6 +254,7 @@
 DECL_SUF(x265_pixel_avg_16x12, (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_16x8,  (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_16x4,  (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
+DECL_SUF(x265_pixel_avg_8x32,  (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_8x16,  (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_8x8,   (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))
 DECL_SUF(x265_pixel_avg_8x4,   (pixel *, intptr_t, pixel *, intptr_t, pixel *, intptr_t, int))