[x265] [PATCH 5 of 6] asm: intra_pred_ang4_4 improved by ~44% over SSE4

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1428303677 -19800
#      Mon Apr 06 12:31:17 2015 +0530
# Node ID 604cc2d73e402b9ab3a3edc8ecc34c019239680b
# Parent  6a3e3ace9c9d46eed867a78568bebc56c8310791
asm: intra_pred_ang4_4 improved by ~44% over SSE4

AVX2:
intra_ang_4x4[ 4]       10.62x   94.02           998.80

SSE4:
intra_ang_4x4[ 4]       5.89x    169.02          994.88

diff -r 6a3e3ace9c9d -r 604cc2d73e40 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Mon Apr 06 12:18:54 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Mon Apr 06 12:31:17 2015 +0530
@@ -1701,6 +1701,7 @@
         p.weight_sp = x265_weight_sp_avx2;
 
         // intra_pred functions
+        p.cu[BLOCK_4x4].intra_pred[4] = x265_intra_pred_ang4_4_avx2;
         p.cu[BLOCK_4x4].intra_pred[5] = x265_intra_pred_ang4_5_avx2;
         p.cu[BLOCK_4x4].intra_pred[6] = x265_intra_pred_ang4_6_avx2;
         p.cu[BLOCK_4x4].intra_pred[7] = x265_intra_pred_ang4_7_avx2;
diff -r 6a3e3ace9c9d -r 604cc2d73e40 source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Mon Apr 06 12:18:54 2015 +0530
+++ b/source/common/x86/intrapred.h	Mon Apr 06 12:31:17 2015 +0530
@@ -187,6 +187,7 @@
 DECL_ANG(32, 33, sse4);
 
 #undef DECL_ANG
+void x265_intra_pred_ang4_4_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang4_5_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang4_6_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang4_7_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
diff -r 6a3e3ace9c9d -r 604cc2d73e40 source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Mon Apr 06 12:18:54 2015 +0530
+++ b/source/common/x86/intrapred8.asm	Mon Apr 06 12:31:17 2015 +0530
@@ -515,6 +515,7 @@
 intra_pred4_shuff2:      db 0, 1, 1, 2, 2, 3, 3, 4, 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5
 intra_pred4_shuff31:     db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6
 intra_pred4_shuff33:     db 0, 1, 1, 2, 2, 3, 3, 4, 1, 2, 2, 3, 3, 4, 4, 5, 2, 3, 3, 4, 4, 5, 5, 6, 3, 4, 4, 5, 5, 6, 6, 7
+intra_pred4_shuff4:      db 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14, 11, 12, 12, 13, 13, 14, 14, 15
 intra_pred4_shuff5:      db 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14, 11, 12, 12, 13, 13, 14, 14, 15
 intra_pred4_shuff6:      db 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14, 10, 11, 11, 12, 12, 13, 13, 14
 intra_pred4_shuff7:      db 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 9, 10, 10, 11, 11, 12, 12, 13, 10, 11, 11, 12, 12, 13, 13, 14
@@ -16098,6 +16099,16 @@
     INTRA_PRED_STORE_4x4
     RET
 
+INIT_YMM avx2
+cglobal intra_pred_ang4_4, 3, 3, 1
+    vbroadcasti128    m0, [r2]
+    pshufb            m0, [intra_pred4_shuff5]
+    pmaddubsw         m0, [c_ang4_mode_32]
+    pmulhrsw          m0, [pw_1024]
+    packuswb          m0, m0
+
+    INTRA_PRED_TRANS_STORE_4x4
+    RET
 
 INIT_YMM avx2
 cglobal intra_pred_ang4_5, 3, 3, 1