[x265] [PATCH 04 of 11] asm: intra_pred_ang4_13_sse2

[dtyx265 at gmail.com]

# HG changeset patch
# User David T Yuen <dtyx265 at gmail.com>
# Date 1427916573 25200
# Node ID 18b2c754beccd1eb2f04e5a36137d1f8a183f599
# Parent  57f560a6292c3f693bf6794a1af3a1246d790aed
asm: intra_pred_ang4_13_sse2

This is backported from sse4 code and replaces c code.

64-bit

./test/TestBench --testbench intrapred | grep "intra_ang_4x4\[13\]"
intra_ang_4x4[13]	3.26x 	 657.49   	 2143.49

32-bit

./test/TestBench --testbench intrapred | grep "intra_ang_4x4\[13\]"
intra_ang_4x4[13]	4.30x 	 715.06   	 3074.11

diff -r 57f560a6292c -r 18b2c754becc source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Wed Apr 01 12:25:28 2015 -0700
+++ b/source/common/x86/asm-primitives.cpp	Wed Apr 01 12:29:33 2015 -0700
@@ -1207,6 +1207,7 @@
         p.cu[BLOCK_4x4].intra_pred[10] = x265_intra_pred_ang4_10_sse2;
         p.cu[BLOCK_4x4].intra_pred[11] = x265_intra_pred_ang4_11_sse2;
         p.cu[BLOCK_4x4].intra_pred[12] = x265_intra_pred_ang4_12_sse2;
+        p.cu[BLOCK_4x4].intra_pred[13] = x265_intra_pred_ang4_13_sse2;
         p.cu[BLOCK_4x4].intra_pred[26] = x265_intra_pred_ang4_26_sse2;
 
         p.cu[BLOCK_4x4].calcresidual = x265_getResidual4_sse2;
diff -r 57f560a6292c -r 18b2c754becc source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Wed Apr 01 12:25:28 2015 -0700
+++ b/source/common/x86/intrapred.h	Wed Apr 01 12:29:33 2015 -0700
@@ -58,6 +58,7 @@
 DECL_ANG(4, 10, sse2);
 DECL_ANG(4, 11, sse2);
 DECL_ANG(4, 12, sse2);
+DECL_ANG(4, 13, sse2);
 DECL_ANG(4, 26, sse2);
 
 DECL_ANG(4, 2, ssse3);
diff -r 57f560a6292c -r 18b2c754becc source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Wed Apr 01 12:25:28 2015 -0700
+++ b/source/common/x86/intrapred8.asm	Wed Apr 01 12:29:33 2015 -0700
@@ -1656,6 +1656,35 @@
     mova        m7, [r3 -  8 * 16]  ; [12]
     jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
 
+cglobal intra_pred_ang4_13, 3,5,8
+    xor         r4d, r4d
+    cmp         r3m, byte 23
+    mov         r3d, 8
+    jz          .next
+    xchg        r3d, r4d
+
+.next:
+    movd        m1, [r2 - 1]            ;[x x A x]
+    movd        m2, [r2 + r4 + 1]       ;[4 3 2 1]
+    movd        m0, [r2 + r3 + 3]       ;[x x B x]
+    punpcklbw   m0, m1                  ;[x x x x A B x x]
+    punpckldq   m0, m2                  ;[4 3 2 1 A B x x]
+    psrldq      m0, 2                   ;[x x 4 3 2 1 A B]
+    punpcklbw   m0, m0                  ;[x x x x 4 4 3 3 2 2 1 1 A A B B]
+    mova        m1, m0
+    psrldq      m0, 3                   ;[x x x x x x x 4 4 3 3 2 2 1 1 A]
+    psrldq      m1, 1                   ;[x x x x x 4 4 3 3 2 2 1 1 A A B]
+    movh        m2, m0
+    punpcklqdq  m0, m0
+    punpcklqdq  m2, m1
+
+    lea         r3, [pw_ang_table + 21 * 16]
+    mova        m4, [r3 +  2 * 16]  ; [23]
+    mova        m5, [r3 -  7 * 16]  ; [14]
+    mova        m6, [r3 - 16 * 16]  ; [ 5]
+    mova        m7, [r3 +  7 * 16]  ; [28]
+    jmp         mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+ .do_filter4x4)
+
 ;---------------------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
 ;---------------------------------------------------------------------------------------------