[x265] [PATCH Review Only] asm code for blockcopy_sp, 8x6 partition

[praveen at multicorewareinc.com]

# HG changeset patch
# User Praveen Tiwari
# Date 1383664381 -19800
# Node ID cd6e6dd4443c22bd7b3e1a39542273cc6de816e7
# Parent  053c9cf05f0f4a86855fefb5c9f6c9e6c6450be3
asm code for blockcopy_sp, 8x6 partition

diff -r 053c9cf05f0f -r cd6e6dd4443c source/common/x86/blockcopy8.asm
--- a/source/common/x86/blockcopy8.asm	Tue Nov 05 20:24:16 2013 +0530
+++ b/source/common/x86/blockcopy8.asm	Tue Nov 05 20:43:01 2013 +0530
@@ -848,3 +848,40 @@
 movh       [r4 + r1],          m4
 
 RET
+
+;-----------------------------------------------------------------------------
+; void blockcopy_sp_8x6(pixel *dest, intptr_t destStride, int16_t *src, intptr_t srcStride)
+;-----------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal blockcopy_sp_8x6, 4, 5, 7, dest, destStride, src, srcStride
+
+add         r3,      r3
+
+mova        m0,     [tab_Vm]
+
+movu        m1,     [r2]
+movu        m2,     [r2 + r3]
+movu        m3,     [r2 + 2 * r3]
+lea         r4,     [r2 + 2 * r3]
+movu        m4,     [r4 + r3]
+movu        m5,     [r4 + 2 * r3]
+lea         r4,     [r4 + 2 * r3]
+movu        m6,     [r4 + r3]
+
+pshufb     m1,      m0
+pshufb     m2,      m0
+pshufb     m3,      m0
+pshufb     m4,      m0
+pshufb     m5,      m0
+pshufb     m6,      m0
+
+movh       [r0],            m1
+movh       [r0 + r1],       m2
+movh       [r0 + 2 * r1],   m3
+lea        r4,              [r0 + 2 * r1]
+movh       [r4 + r1],       m4
+movh       [r4 + 2 * r1],   m5
+lea        r4,              [r4 + 2 * r1]
+movh       [r4 + r1],       m6
+
+RET