[x265] [PATCH] asm-avx2:intra_pred_ang8_9, 51.13x

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1425560761 -19800
# Node ID 253bff623dea4557596ae6ac141ba6aac6874cbc
# Parent  4deeb4cd656ad7166317f6bd8eeb3fa82154cd39
asm-avx2:intra_pred_ang8_9, 51.13x

intra_ang_8x8[ 9]       51.13x   273.68          13992.92

diff -r 4deeb4cd656a -r 253bff623dea source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Mar 05 18:21:21 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Mar 05 18:36:01 2015 +0530
@@ -1510,6 +1510,7 @@
         p.cu[BLOCK_8x8].intra_pred[29] = x265_intra_pred_ang8_29_avx2;
         p.cu[BLOCK_8x8].intra_pred[8] = x265_intra_pred_ang8_8_avx2;
         p.cu[BLOCK_8x8].intra_pred[28] = x265_intra_pred_ang8_28_avx2;
+        p.cu[BLOCK_8x8].intra_pred[9] = x265_intra_pred_ang8_9_avx2;
         p.cu[BLOCK_8x8].intra_pred[27] = x265_intra_pred_ang8_27_avx2;
 
         // copy_sp primitives
diff -r 4deeb4cd656a -r 253bff623dea source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Thu Mar 05 18:21:21 2015 +0530
+++ b/source/common/x86/intrapred.h	Thu Mar 05 18:36:01 2015 +0530
@@ -173,6 +173,7 @@
 void x265_intra_pred_ang8_29_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_8_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_28_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_9_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_27_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 4deeb4cd656a -r 253bff623dea source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Thu Mar 05 18:21:21 2015 +0530
+++ b/source/common/x86/intrapred8.asm	Thu Mar 05 18:36:01 2015 +0530
@@ -32982,6 +32982,53 @@
     movhps            [r0 + r3], xm2
     RET
 
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_9, 3, 4, 6
+    mova              m3, [pw_1024]
+    vbroadcasti128    m0, [r2 + 17]
+    movu              m5, [c_ang8_src1_9_1_9]
+
+    pshufb            m1, m0, m5
+    pshufb            m2, m0, m5
+    pshufb            m4, m0, m5
+    pshufb            m0, m5
+
+    pmaddubsw         m1, [c_ang8_2_4]
+    pmulhrsw          m1, m3
+    pmaddubsw         m2, [c_ang8_6_8]
+    pmulhrsw          m2, m3
+    pmaddubsw         m4, [c_ang8_10_12]
+    pmulhrsw          m4, m3
+    pmaddubsw         m0, [c_ang8_14_16]
+    pmulhrsw          m0, m3
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    vperm2i128        m2, m1, m4, 00100000b
+    vperm2i128        m1, m1, m4, 00110001b
+    punpcklbw         m4, m2, m1
+    punpckhbw         m2, m1
+    punpcklwd         m1, m4, m2
+    punpckhwd         m4, m2
+    mova              m0, [trans8_shuf]
+    vpermd            m1, m0, m1
+    vpermd            m4, m0, m4
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    movhps            [r0 + r1], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    movhps            [r0 + r1], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + 2 * r1], xm2
+    movhps            [r0 + r3], xm2
+    RET
+
 INIT_YMM avx2
 cglobal intra_pred_ang8_27, 3, 4, 6
     mova              m3, [pw_1024]