[x265] [PATCH] asm-intra_pred_ang16_25: improved, 781.13c -> 466.16c

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari <praveen at multicorewareinc.com>
# Date 1426167514 -19800
# Node ID a5648b191717022f24b48b30b6f184d2c46f91f7
# Parent  d012e125bdb1299ba29b9c0680931e148981a42e
asm-intra_pred_ang16_25: improved, 781.13c -> 466.16c

AVX2:
intra_ang_16x16[25]     18.09x   466.16          8434.26

SSE4:
intra_ang_16x16[25]     10.90x   781.13          8511.65

diff -r d012e125bdb1 -r a5648b191717 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Mar 12 18:40:23 2015 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Mar 12 19:08:34 2015 +0530
@@ -1504,6 +1504,7 @@
         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_16x16].intra_pred[25] = x265_intra_pred_ang16_25_avx2;
 
         // copy_sp primitives
         p.cu[BLOCK_16x16].copy_sp = x265_blockcopy_sp_16x16_avx2;
diff -r d012e125bdb1 -r a5648b191717 source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Thu Mar 12 18:40:23 2015 +0530
+++ b/source/common/x86/intrapred.h	Thu Mar 12 19:08:34 2015 +0530
@@ -182,6 +182,7 @@
 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_ang16_25_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);
 void x265_all_angs_pred_16x16_sse4(pixel *dest, pixel *refPix, pixel *filtPix, int bLuma);
diff -r d012e125bdb1 -r a5648b191717 source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Thu Mar 12 18:40:23 2015 +0530
+++ b/source/common/x86/intrapred8.asm	Thu Mar 12 19:08:34 2015 +0530
@@ -113,6 +113,17 @@
                       db 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 25, 7, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2
                       db 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 3, 29, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
 
+ALIGN 32
+c_ang16_mode_25:      db 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28
+                      db 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24
+                      db 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20
+                      db 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+                      db 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 18, 14, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12, 20, 12
+                      db 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 22, 10, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8, 24, 8
+                      db 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 26, 6, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4, 28, 4
+                      db 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 30, 2, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0, 32, 0
+
+ALIGN 32
 ;; (blkSize - 1 - x)
 pw_planar4_0:         dw 3,  2,  1,  0,  3,  2,  1,  0
 pw_planar4_1:         dw 3,  3,  3,  3,  3,  3,  3,  3
@@ -10453,3 +10464,43 @@
     movhps            [r0 + 2 * r1], xm4
     movhps            [r0 + r3], xm2
     RET
+
+%macro INTRA_PRED_ANG16_MC0 3
+    pmaddubsw         m3, m1, [r4 + %3 * mmsize]
+    pmulhrsw          m3, m0
+    pmaddubsw         m4, m2, [r4 + %3 * mmsize]
+    pmulhrsw          m4, m0
+    packuswb          m3, m4
+    movu              [%1], xm3
+    vextracti128      xm4, m3, 1
+    movu              [%2], xm4
+%endmacro
+
+%macro INTRA_PRED_ANG16_25 1
+    INTRA_PRED_ANG16_MC0 r0, r0 + r1, %1
+    INTRA_PRED_ANG16_MC0 r0 + 2 * r1, r0 + r3, (%1 + 1)
+%endmacro
+
+INIT_YMM avx2
+cglobal intra_pred_ang16_25, 3, 5, 5
+    mova              m0, [pw_1024]
+
+    vbroadcasti128    m1, [r2]
+    pshufb            m1, [intra_pred_shuff_0_8]
+    vbroadcasti128    m2, [r2 + 8]
+    pshufb            m2, [intra_pred_shuff_0_8]
+
+    lea               r3, [3 * r1]
+    lea               r4, [c_ang16_mode_25]
+
+    INTRA_PRED_ANG16_25 0
+
+    lea    r0, [r0 + 4 * r1]
+    INTRA_PRED_ANG16_25 2
+
+    lea    r0, [r0 + 4 * r1]
+    INTRA_PRED_ANG16_25 4
+
+    lea    r0, [r0 + 4 * r1]
+    INTRA_PRED_ANG16_25 6
+    RET