[x265] [PATCH] asm: 10bpp support to sse_ss for 8xN, 16xN and 32xN blocks

[murugan at multicorewareinc.com]

# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1385980817 -19800
#      Mon Dec 02 16:10:17 2013 +0530
# Node ID 507f68d447760113bad13eaaacd91ce08f1a5cd2
# Parent  df0b4f81609e611989c5b1743e7729adeb51cb01
asm: 10bpp support to sse_ss for 8xN, 16xN and 32xN blocks

diff -r df0b4f81609e -r 507f68d44776 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Mon Dec 02 12:19:34 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Mon Dec 02 16:10:17 2013 +0530
@@ -505,8 +505,18 @@
         p.sse_ss[LUMA_8x4] = x265_pixel_ssd_ss_8x4_sse2;
         p.sse_ss[LUMA_8x8] = x265_pixel_ssd_ss_8x8_sse2;
         p.sse_ss[LUMA_8x16] = x265_pixel_ssd_ss_8x16_sse2;
+        p.sse_ss[LUMA_8x32] = x265_pixel_ssd_ss_8x32_sse2;
+        p.sse_ss[LUMA_16x4] = x265_pixel_ssd_ss_16x4_sse2;
         p.sse_ss[LUMA_16x8] = x265_pixel_ssd_ss_16x8_sse2;
+        p.sse_ss[LUMA_16x12] = x265_pixel_ssd_ss_16x12_sse2;
         p.sse_ss[LUMA_16x16] = x265_pixel_ssd_ss_16x16_sse2;
+        p.sse_ss[LUMA_16x32] = x265_pixel_ssd_ss_16x32_sse2;
+        p.sse_ss[LUMA_16x64] = x265_pixel_ssd_ss_16x64_sse2;
+        p.sse_ss[LUMA_32x8] = x265_pixel_ssd_ss_32x8_sse2;
+        p.sse_ss[LUMA_32x16] = x265_pixel_ssd_ss_32x16_sse2;
+        p.sse_ss[LUMA_32x24] = x265_pixel_ssd_ss_32x24_sse2;
+        p.sse_ss[LUMA_32x32] = x265_pixel_ssd_ss_32x32_sse2;
+        p.sse_ss[LUMA_32x64] = x265_pixel_ssd_ss_32x64_sse2;
 
         p.ssim_4x4x2_core = x265_pixel_ssim_4x4x2_core_sse2;
         PIXEL_AVG_W4(sse2);
diff -r df0b4f81609e -r 507f68d44776 source/common/x86/pixel-a.asm
--- a/source/common/x86/pixel-a.asm	Mon Dec 02 12:19:34 2013 +0530
+++ b/source/common/x86/pixel-a.asm	Mon Dec 02 16:10:17 2013 +0530
@@ -188,8 +188,18 @@
 SSD_ONE     8,  4
 SSD_ONE     8,  8
 SSD_ONE     8, 16
+SSD_ONE     8, 32
+SSD_ONE    16,  4
 SSD_ONE    16,  8
+SSD_ONE    16, 12
 SSD_ONE    16, 16
+SSD_ONE    16, 32
+SSD_ONE    16, 64
+SSD_ONE    32,  8
+SSD_ONE    32, 16
+SSD_ONE    32, 24
+SSD_ONE    32, 32
+SSD_ONE    32, 64
 INIT_YMM avx2
 SSD_ONE    16,  8
 SSD_ONE    16, 16