[x265] [PATCH] asm: intra_pred_ang16_2

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1425996387 -19800
# Node ID c9b41ce268860151664a4d20336c048b0e71af24
# Parent  b5cc1a011beb7eb1baf6d22d2c0e45730f8bfe9e
asm: intra_pred_ang16_2

diff -r b5cc1a011beb -r c9b41ce26886 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Tue Mar 10 19:31:30 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Tue Mar 10 19:36:27 2015 +0530
@@ -1501,6 +1501,7 @@
         p.cu[BLOCK_8x8].intra_pred[24] = x265_intra_pred_ang8_24_avx2;
         p.cu[BLOCK_8x8].intra_pred[22] = x265_intra_pred_ang8_22_avx2;
         p.cu[BLOCK_8x8].intra_pred[14] = x265_intra_pred_ang8_14_avx2;
+        p.cu[BLOCK_16x16].intra_pred[2] = x265_intra_pred_ang16_2_avx2;
         p.cu[BLOCK_16x16].intra_pred[34] = x265_intra_pred_ang16_34_avx2;
 
         // copy_sp primitives
diff -r b5cc1a011beb -r c9b41ce26886 source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Tue Mar 10 19:31:30 2015 +0530
+++ b/source/common/x86/intrapred.h	Tue Mar 10 19:36:27 2015 +0530
@@ -184,6 +184,7 @@
 void x265_intra_pred_ang8_24_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_22_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_14_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang16_2_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_34_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_all_angs_pred_4x4_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
 void x265_all_angs_pred_8x8_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
diff -r b5cc1a011beb -r c9b41ce26886 source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Tue Mar 10 19:31:30 2015 +0530
+++ b/source/common/x86/intrapred8.asm	Tue Mar 10 19:36:27 2015 +0530
@@ -10619,3 +10619,44 @@
     palignr           xm2, xm1, xm0, 15
     movu              [r0 + r3], xm2
     RET
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_2, 3, 4, 3
+    lea               r3, [3 * r1]
+    movu              m0, [r2 + 32 + 2]
+    movu              [r0], xm0
+    vextracti128      xm1, m0, 1
+    palignr           xm2, xm1, xm0, 1
+    movu              [r0 + r1], xm2
+    palignr           xm2, xm1, xm0, 2
+    movu              [r0 + 2 * r1], xm2
+    palignr           xm2, xm1, xm0, 3
+    movu              [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    palignr           xm2, xm1, xm0, 4
+    movu              [r0], xm2
+    palignr           xm2, xm1, xm0, 5
+    movu              [r0 + r1], xm2
+    palignr           xm2, xm1, xm0, 6
+    movu              [r0 + 2 * r1], xm2
+    palignr           xm2, xm1, xm0, 7
+    movu              [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    palignr           xm2, xm1, xm0, 8
+    movu              [r0], xm2
+    palignr           xm2, xm1, xm0, 9
+    movu              [r0 + r1], xm2
+    palignr           xm2, xm1, xm0, 10
+    movu              [r0 + 2 * r1], xm2
+    palignr           xm2, xm1, xm0, 11
+    movu              [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    palignr           xm2, xm1, xm0, 12
+    movu              [r0], xm2
+    palignr           xm2, xm1, xm0, 13
+    movu              [r0 + r1], xm2
+    palignr           xm2, xm1, xm0, 14
+    movu              [r0 + 2 * r1], xm2
+    palignr           xm2, xm1, xm0, 15
+    movu              [r0 + r3], xm2
+    RET