[FFmpeg-devel] [PATCH 02/11] x86: dcadsp: implement int8x8_fmul_int32

James Almer jamrial at gmail.com
Tue Feb 18 07:15:58 CET 2014


On 07/02/14 9:22 PM, Michael Niedermayer wrote:
> On Fri, Feb 07, 2014 at 10:28:19PM +0100, Christophe Gisquet wrote:
>> Hi,
>>
>> here's a new version. It also takes into account changes in the first
>> patch of the series
>>
>> -- 
>> Christophe
> 
>>  dcadsp.asm    |   99 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>  dcadsp_init.c |    6 +++
>>  2 files changed, 105 insertions(+)
>> dada5bbea83c619704000fdb7a4fb9a690ae0645  0004-x86-dcadsp-implement-SSE-lfe_dir.patch
>> From 9a8381fef85b50bb8034ea650f20a3868a146e3c Mon Sep 17 00:00:00 2001
>> From: Christophe Gisquet <christophe.gisquet at gmail.com>
>> Date: Wed, 19 Dec 2012 20:26:05 +0100
>> Subject: [PATCH 04/10] x86: dcadsp: implement SSE lfe_dir
> 
> i think thats not what you wanted to attach
> but the one from the fork seems to have merged without conflict
> so merged
> 
> thanks

Is it just my end, or is the inline version from x86/dca.h always being used no 
matter what? Changes i made to the yasm versions (where i purposely broke stuff 
to test if they were being used) didn't have any effect on the FATE test until i 
changed the line 1384 (current git head) from int8x8_fmul_int32() to 
s->dcadsp.int8x8_fmul_int32().

I checked this on an Win64 build.


More information about the ffmpeg-devel mailing list