[x265] [PATCH 5 of 5] asm: avx2 code for high_bit_depth satd_48x64, improved over ~50% than previous asm

[dnyaneshwar at multicorewareinc.com]

# HG changeset patch
# User Dnyaneshwar G <dnyaneshwar at multicorewareinc.com>
# Date 1430990919 -19800
#      Thu May 07 14:58:39 2015 +0530
# Node ID e50526d3101e7eecfde8204e0801a7ba335f8f33
# Parent  443ee8d99edf8c3a11f0c1a62f1dd210ee797048
asm: avx2 code for high_bit_depth satd_48x64, improved over ~50% than previous asm

diff -r 443ee8d99edf -r e50526d3101e source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu May 07 14:49:34 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu May 07 14:58:39 2015 +0530
@@ -1181,6 +1181,8 @@
     }
     if (cpuMask & X265_CPU_AVX2)
     {
+        p.pu[LUMA_48x64].satd = x265_pixel_satd_48x64_avx2;
+
         p.pu[LUMA_64x16].satd = x265_pixel_satd_64x16_avx2;
         p.pu[LUMA_64x32].satd = x265_pixel_satd_64x32_avx2;
         p.pu[LUMA_64x48].satd = x265_pixel_satd_64x48_avx2;
diff -r 443ee8d99edf -r e50526d3101e source/common/x86/pixel-a.asm
--- a/source/common/x86/pixel-a.asm	Thu May 07 14:49:34 2015 +0530
+++ b/source/common/x86/pixel-a.asm	Thu May 07 14:58:39 2015 +0530
@@ -11615,6 +11615,58 @@
     movd            eax, xm6
     RET
 
+cglobal pixel_satd_48x64, 4,8,8
+    add             r1d, r1d
+    add             r3d, r3d
+    lea             r4, [3 * r1]
+    lea             r5, [3 * r3]
+    pxor            m6, m6
+    mov             r6, r0
+    mov             r7, r2
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 32]
+    lea             r2, [r7 + 32]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    lea             r0, [r6 + 64]
+    lea             r2, [r7 + 64]
+
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+    call            calc_satd_16x8
+
+    vextracti128    xm7, m6, 1
+    paddd           xm6, xm7
+    pxor            xm7, xm7
+    movhlps         xm7, xm6
+    paddd           xm6, xm7
+    pshufd          xm7, xm6, 1
+    paddd           xm6, xm7
+    movd            eax, xm6
+    RET
+
 cglobal pixel_satd_64x16, 4,8,8
     add             r1d, r1d
     add             r3d, r3d