[x265] [PATCH 1 of 2] asm: intra_pred_ang8_20 improved by ~4% over SSE4

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1428647479 -19800
#      Fri Apr 10 12:01:19 2015 +0530
# Node ID c25751eb5a1ab4b19a171e9f35f149b408c4c19e
# Parent  984e254f93f7cedc5a9b00851d2e14b49dc94e91
asm: intra_pred_ang8_20 improved by ~4% over SSE4

AVX2:
intra_ang_8x8[20]       7.98x    256.94          2050.52

SSE4:
intra_ang_8x8[20]       7.59x    267.77          2031.49

diff -r 984e254f93f7 -r c25751eb5a1a source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Apr 09 11:48:08 2015 -0500
+++ b/source/common/x86/asm-primitives.cpp	Fri Apr 10 12:01:19 2015 +0530
@@ -1769,6 +1769,7 @@
         p.cu[BLOCK_8x8].intra_pred[24] = x265_intra_pred_ang8_24_avx2;
         p.cu[BLOCK_8x8].intra_pred[11] = x265_intra_pred_ang8_11_avx2;
         p.cu[BLOCK_8x8].intra_pred[13] = x265_intra_pred_ang8_13_avx2;
+        p.cu[BLOCK_8x8].intra_pred[20] = x265_intra_pred_ang8_20_avx2;
         p.cu[BLOCK_8x8].intra_pred[21] = x265_intra_pred_ang8_21_avx2;
         p.cu[BLOCK_8x8].intra_pred[22] = x265_intra_pred_ang8_22_avx2;
         p.cu[BLOCK_8x8].intra_pred[23] = x265_intra_pred_ang8_23_avx2;
diff -r 984e254f93f7 -r c25751eb5a1a source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Thu Apr 09 11:48:08 2015 -0500
+++ b/source/common/x86/intrapred.h	Fri Apr 10 12:01:19 2015 +0530
@@ -236,6 +236,7 @@
 void x265_intra_pred_ang8_13_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_ang8_15_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_20_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_21_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_23_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
diff -r 984e254f93f7 -r c25751eb5a1a source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Thu Apr 09 11:48:08 2015 -0500
+++ b/source/common/x86/intrapred8.asm	Fri Apr 10 12:01:19 2015 +0530
@@ -690,6 +690,12 @@
                       db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26
                       db 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
 
+ALIGN 32
+c_ang8_mode_20:       db 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 21, 11, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
+                      db 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+
 const ang_table
 %assign x 0
 %rep 32
@@ -11945,6 +11951,61 @@
     movhps            [r0 + r3], xm2
     RET
 
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_20, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2]
+    lea               r5, [intra_pred_shuff_0_8]
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 2 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_20]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 3 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 5 + 16], 0
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 6 + 16], 0
+    mova              xm0, xm5
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 8 + 16], 0
+    vinserti128       m0, m0, xm5, 1
+    pshufb            m0, [r5]
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    pmulhrsw          m0, m3
+
+    packuswb          m1, m2
+    packuswb          m4, m0
+
+    lea               r3, [3 * r1]
+    movq              [r0], xm1
+    vextracti128      xm2, m1, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm1
+    movhps            [r0 + r3], xm2
+    lea               r0, [r0 + 4 * r1]
+    movq              [r0], xm4
+    vextracti128      xm2, m4, 1
+    movq              [r0 + r1], xm2
+    movhps            [r0 + 2 * r1], xm4
+    movhps            [r0 + r3], xm2
+    RET
+
 INIT_YMM avx2
 cglobal intra_pred_ang8_21, 3, 6, 6
     mova              m3, [pw_1024]