[x265] [PATCH 2 of 9] asm: intrapred dc8 sse2 high bit

chen chenm003 at 163.com
Fri Mar 6 17:12:33 CET 2015




At 2015-03-06 10:21:33,"Steve Borho" <steve at borho.org> wrote:
>On 03/06, chen wrote:
>>  
>> 
>> At 2015-03-06 08:19:55,dtyx265 at gmail.com wrote:
>> ># HG changeset patch
>> ># User David T Yuen <dtyx265 at gmail.com>
>> ># Date 1425588355 28800
>> ># Node ID 3b2512cba57e1af6ba21c445122a2b5b2efd87dc
>> ># Parent  39111ae2fd72824c60f827023bd6b55b92dc1c41
>> >asm: intrapred dc8 sse2 high bit
>> >
>> >This replaces c code for systems using ssse3 to sse2 processors
>> >The code is backported from intrapred dc8 sse4 high bit
>> >
>> >./test/TestBench --testbench intrapred | grep intra_dc_8x8
>> >intra_dc_8x8[f=0]	1.64x 	 437.56   	 719.58
>> >intra_dc_8x8[f=1]	1.55x 	 747.51   	 1157.56
>> >
>> >diff -r 39111ae2fd72 -r 3b2512cba57e source/common/x86/asm-primitives.cpp
>> >--- a/source/common/x86/asm-primitives.cpp	Thu Mar 05 12:21:23 2015 -0800
>> >+++ b/source/common/x86/asm-primitives.cpp	Thu Mar 05 12:45:55 2015 -0800
>> >@@ -869,6 +869,7 @@
>> >         ALL_LUMA_TU_S(transpose, transpose, sse2);
>> > 
>> >         p.cu[BLOCK_4x4].intra_pred[DC_IDX] = x265_intra_pred_dc4_sse2;
>> >+        p.cu[BLOCK_8x8].intra_pred[DC_IDX] = x265_intra_pred_dc8_sse2;
>> > 
>> >         p.cu[BLOCK_4x4].intra_pred[PLANAR_IDX] = x265_intra_pred_planar4_sse2;
>> > 
>> >diff -r 39111ae2fd72 -r 3b2512cba57e source/common/x86/intrapred16.asm
>> >--- a/source/common/x86/intrapred16.asm	Thu Mar 05 12:21:23 2015 -0800
>> >+++ b/source/common/x86/intrapred16.asm	Thu Mar 05 12:45:55 2015 -0800
>> >@@ -161,6 +161,89 @@
>> > ;-----------------------------------------------------------------------------------
>> > ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel* above, int, int filter)
>> > ;-----------------------------------------------------------------------------------
>> >+INIT_XMM sse2
>> >+cglobal intra_pred_dc8, 5, 8, 2
>> x86 compatible issue
>
>We don't allow 32bit HIGH_BIT_DEPTH builds
>
yes, but we'd better check enviorment here, it reduce works in future (eg: debug in x86+HIGH_BIT or support it in future)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20150307/6c720638/attachment.html>


More information about the x265-devel mailing list