[x265] [PATCH 4 of 9] asm: intra_pred_ang8_13 improved by ~16% over SSE4

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1428575027 -19800
#      Thu Apr 09 15:53:47 2015 +0530
# Node ID 79c8c583603a8dda7fe22973b55b18d9ff08cc64
# Parent  6f9c3e9aec5218f89389c6f1f363b86181fc20cf
asm: intra_pred_ang8_13 improved by ~16% over SSE4

AVX2:
intra_ang_8x8[13]       10.68x   297.95          3183.33

SSE4:
intra_ang_8x8[13]       9.16x    352.32          3225.62

diff -r 6f9c3e9aec52 -r 79c8c583603a source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Apr 09 13:28:56 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Apr 09 15:53:47 2015 +0530
@@ -1764,6 +1764,7 @@
         p.cu[BLOCK_8x8].intra_pred[12] = x265_intra_pred_ang8_12_avx2;
         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_16x16].intra_pred[3] = x265_intra_pred_ang16_3_avx2;
         p.cu[BLOCK_16x16].intra_pred[4] = x265_intra_pred_ang16_4_avx2;
         p.cu[BLOCK_16x16].intra_pred[5] = x265_intra_pred_ang16_5_avx2;
diff -r 6f9c3e9aec52 -r 79c8c583603a source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Thu Apr 09 13:28:56 2015 +0530
+++ b/source/common/x86/intrapred.h	Thu Apr 09 15:53:47 2015 +0530
@@ -233,6 +233,7 @@
 void x265_intra_pred_ang8_12_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_24_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang8_11_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
+void x265_intra_pred_ang8_13_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_3_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_4_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
 void x265_intra_pred_ang16_5_avx2(pixel* dst, intptr_t dstStride, const pixel* srcPix, int dirMode, int bFilter);
diff -r 6f9c3e9aec52 -r 79c8c583603a source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Thu Apr 09 13:28:56 2015 +0530
+++ b/source/common/x86/intrapred8.asm	Thu Apr 09 15:53:47 2015 +0530
@@ -672,6 +672,11 @@
 pw_planar32_L:        dw 31, 30, 29, 28, 27, 26, 25, 24
 pw_planar32_H:        dw 23, 22, 21, 20, 19, 18, 17, 16
 
+ALIGN 32
+c_ang8_mode_13:       db 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 9, 23, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14
+                      db 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 27, 5, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 13, 19, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10
+                      db 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 31, 1, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
 
 const ang_table
 %assign x 0
@@ -11866,6 +11871,61 @@
     RET
 
 
+
+INIT_YMM avx2
+cglobal intra_pred_ang8_13, 3, 6, 6
+    mova              m3, [pw_1024]
+    movu              xm5, [r2 + 16]
+    pinsrb            xm5, [r2], 0
+    lea               r5, [intra_pred_shuff_0_8]
+    vinserti128       m0, m5, xm5, 1
+    pshufb            m0, [r5]
+
+    lea               r4, [c_ang8_mode_13]
+    pmaddubsw         m1, m0, [r4]
+    pmulhrsw          m1, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 4], 0
+    pshufb            xm4, xm5, [r5]
+    vinserti128       m0, m0, xm4, 1
+    pmaddubsw         m2, m0, [r4 + mmsize]
+    pmulhrsw          m2, m3
+    vinserti128       m0, m0, xm4, 0
+    pmaddubsw         m4, m0, [r4 + 2 * mmsize]
+    pmulhrsw          m4, m3
+    pslldq            xm5, 1
+    pinsrb            xm5, [r2 + 7], 0
+    pshufb            xm5, [r5]
+    vinserti128       m0, m0, xm5, 1
+    pmaddubsw         m0, [r4 + 3 * mmsize]
+    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_12, 3, 5, 5
     mova              m3, [pw_1024]