[x265] [PATCH 10 of 18] asm: intra_pred_ang4_26_sse2 16-bit

[dtyx265 at gmail.com]

# HG changeset patch
# User David T Yuen <dtyx265 at gmail.com>
# Date 1428076210 25200
# Node ID 89cf1dc260ad91eba0f4edcacc784be0959636b5
# Parent  95c50ede466fa24a1ff13e5203305b35392a5f64
asm: intra_pred_ang4_26_sse2 16-bit

This is backported from sse4 code and replaces c code.

./test/TestBench --testbench intrapred | grep "intra_ang_4x4\[26\]"
intra_ang_4x4[26]	3.76x 	 200.35   	 754.13

diff -r 95c50ede466f -r 89cf1dc260ad source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Fri Apr 03 08:46:08 2015 -0700
+++ b/source/common/x86/asm-primitives.cpp	Fri Apr 03 08:50:10 2015 -0700
@@ -891,6 +891,7 @@
         p.cu[BLOCK_4x4].intra_pred[8] = x265_intra_pred_ang4_8_sse2;
         p.cu[BLOCK_4x4].intra_pred[9] = x265_intra_pred_ang4_9_sse2;
         p.cu[BLOCK_4x4].intra_pred[10] = x265_intra_pred_ang4_10_sse2;
+        p.cu[BLOCK_4x4].intra_pred[26] = x265_intra_pred_ang4_26_sse2;
         p.cu[BLOCK_4x4].intra_pred[27] = x265_intra_pred_ang4_9_sse2;
         p.cu[BLOCK_4x4].intra_pred[28] = x265_intra_pred_ang4_8_sse2;
         p.cu[BLOCK_4x4].intra_pred[29] = x265_intra_pred_ang4_7_sse2;
diff -r 95c50ede466f -r 89cf1dc260ad source/common/x86/intrapred16.asm
--- a/source/common/x86/intrapred16.asm	Fri Apr 03 08:46:08 2015 -0700
+++ b/source/common/x86/intrapred16.asm	Fri Apr 03 08:50:10 2015 -0700
@@ -946,6 +946,42 @@
     movh        [r0],           m0
     RET
 
+cglobal intra_pred_ang4_26, 3,3,3
+    movh        m0,             [r2 + 2]            ; [8 7 6 5 4 3 2 1]
+    add         r1,             r1
+    ; store
+    movh        [r0],           m0
+    movh        [r0 + r1],      m0
+    movh        [r0 + r1 * 2],  m0
+    lea         r3,             [r1 * 3]
+    movh        [r0 + r3],      m0
+
+    ; filter
+    cmp         r4m,            byte 0
+    jz         .quit
+
+    pshuflw     m0,             m0, 0x00
+    movd        m2,             [r2]
+    pshuflw     m2,             m2, 0x00
+    movh        m1,             [r2 + 18]
+    psubw       m1,             m2
+    psraw       m1,             1
+    paddw       m0,             m1
+    pxor        m1,             m1
+    pmaxsw      m0,             m1
+    pminsw      m0,             [pw_1023]
+
+    movh        r2,             m0
+    mov         [r0],           r2w
+    shr         r2,             16
+    mov         [r0 + r1],      r2w
+    shr         r2,             16
+    mov         [r0 + r1 * 2],  r2w
+    shr         r2,             16
+    mov         [r0 + r3],      r2w
+.quit:
+    RET
+
 ;-----------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel* above, int, int filter)
 ;-----------------------------------------------------------------------------------