[x265] [PATCH] asm code for blockcopy_ps_6x8

praveen at multicorewareinc.com praveen at multicorewareinc.com
Mon Nov 11 15:54:43 CET 2013


# HG changeset patch
# User Praveen Tiwari
# Date 1384181673 -19800
# Node ID e39f9ba2f96fe065dd676456f07832e8fbe8ae2d
# Parent  184059c875b5f68a68c3fd0a9935fed88cba5e7a
asm code for blockcopy_ps_6x8

diff -r 184059c875b5 -r e39f9ba2f96f source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Mon Nov 11 19:20:41 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Mon Nov 11 20:24:33 2013 +0530
@@ -460,6 +460,7 @@
         p.chroma_copy_ps[CHROMA_4x2] = x265_blockcopy_ps_4x2_sse4;
         p.chroma_copy_ps[CHROMA_4x4] = x265_blockcopy_ps_4x4_sse4;
         p.chroma_copy_ps[CHROMA_4x8] = x265_blockcopy_ps_4x8_sse4;
+        p.chroma_copy_ps[CHROMA_6x8] = x265_blockcopy_ps_6x8_sse4;
         p.chroma_copy_ps[CHROMA_8x2] = x265_blockcopy_ps_8x2_sse4;
         p.chroma_copy_ps[CHROMA_8x4] = x265_blockcopy_ps_8x4_sse4;
         p.chroma_copy_ps[CHROMA_8x6] = x265_blockcopy_ps_8x6_sse4;
diff -r 184059c875b5 -r e39f9ba2f96f source/common/x86/blockcopy8.asm
--- a/source/common/x86/blockcopy8.asm	Mon Nov 11 19:20:41 2013 +0530
+++ b/source/common/x86/blockcopy8.asm	Mon Nov 11 20:24:33 2013 +0530
@@ -1830,6 +1830,52 @@
 
 BLOCKCOPY_PS_W4_H4 4, 8
 
+
+;-----------------------------------------------------------------------------
+; void blockcopy_ps_%1x%2(int16_t *dest, intptr_t destStride, pixel *src, intptr_t srcStride);
+;-----------------------------------------------------------------------------
+%macro BLOCKCOPY_PS_W6_H4 2
+INIT_XMM sse4
+cglobal blockcopy_ps_%1x%2, 4, 5, 1, dest, destStride, src, srcStride
+
+add     r1,      r1
+mov    r4d,      %2/4
+
+.loop
+      movh       m0,                [r2]
+      pmovzxbw   m0,                m0
+      movh       [r0],              m0
+      pextrd     [r0 + 8],          m0,            2
+
+      movh       m0,                [r2 + r3]
+      pmovzxbw   m0,                m0
+      movh       [r0 + r1],         m0
+      pextrd     [r0 + r1 + 8],     m0,            2
+
+      movh       m0,                [r2 + 2 * r3]
+      pmovzxbw   m0,                m0
+      movh       [r0 + 2 * r1],     m0
+      pextrd     [r0 + 2 * r1 + 8], m0,            2
+
+      lea        r2,                [r2 + 2 * r3]
+      lea        r0,                [r0 + 2 * r1]
+
+      movh       m0,                [r2 + r3]
+      pmovzxbw   m0,                m0
+      movh       [r0 + r1],         m0
+      pextrd     [r0 + r1 + 8],     m0,            2
+
+      lea        r0,                [r0 + 2 * r1]
+      lea        r2,                [r2 + 2 * r3]
+
+      dec        r4d
+      jnz        .loop
+
+RET
+%endmacro
+
+BLOCKCOPY_PS_W6_H4 6, 8
+
 ;-----------------------------------------------------------------------------
 ; void blockcopy_ps_8x2(int16_t *dest, intptr_t destStride, pixel *src, intptr_t srcStride);
 ;-----------------------------------------------------------------------------
diff -r 184059c875b5 -r e39f9ba2f96f source/common/x86/blockcopy8.h
--- a/source/common/x86/blockcopy8.h	Mon Nov 11 19:20:41 2013 +0530
+++ b/source/common/x86/blockcopy8.h	Mon Nov 11 20:24:33 2013 +0530
@@ -99,6 +99,7 @@
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(4, 2, cpu); \
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(4, 4, cpu); \
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(4, 8, cpu); \
+    SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(6, 8, cpu); \
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(8, 2, cpu); \
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(8, 4, cpu); \
     SETUP_CHROMA_BLOCKCOPY_FUNC_SSE4(8, 6, cpu); \


More information about the x265-devel mailing list