[FFmpeg-devel] [PATCH] x86/swr: add SSE2/AVX pack_8ch functions

Ronald S. Bultje rsbultje at gmail.com
Wed Dec 31 02:51:30 CET 2014


Hi,

On Tue, Dec 30, 2014 at 7:58 PM, James Almer <jamrial at gmail.com> wrote:

> On 30/12/14 7:57 PM, Michael Niedermayer wrote:
> > On Tue, Dec 30, 2014 at 06:20:15PM -0300, James Almer wrote:
> >> Signed-off-by: James Almer <jamrial at gmail.com>
> >> ---
> >>  libavcodec/x86/flacdsp.asm             |  13 ---
> >>  libavutil/x86/x86util.asm              |  37 +++++++++
> >>  libswresample/x86/audio_convert.asm    | 145
> +++++++++++++++++++++++++++++++++
> >>  libswresample/x86/audio_convert_init.c |  17 ++++
> >>  4 files changed, 199 insertions(+), 13 deletions(-)
> >
> > looks ok from a quick look and fate passes on linux & mingw 32&64
>
> Does any fate test deal with 8ch fltp/s32p -> flt/s32 conversion? I don't
> think I've
> seen anything beyond 6ch.
>
> I'll wait a bit to see if i can get a more thorough review before pushing.
> I'm mainly
> interested in having the non-transpose-related 32bits stack usage looked
> at.


Looks ok.

More broadly speaking, I honestly do not tend to go out of my way to get
every last cycle out of 32bit asm anymore nowadays, I mean, if people are
using 32bit, it's their own fault. If they want fast software, they should
get 64bit :)

Ronald


More information about the ffmpeg-devel mailing list