[FFmpeg-cvslog] r23095 - in trunk/libavcodec: Makefile allcodecs.c mpegaudio.h mpegaudio_tablegen.h mpegaudiodec.c mpegaudiodec_float.c

[Vitor Sessak]

On 06/08/2010 01:41 PM, Carl Eugen Hoyos wrote:
> Carl Eugen Hoyos<cehoyos<at>  ag.or.at>  writes:
>
>> M?ns Rullg?rd<mans<at>  mansr.com>  writes:
>>
>>> The float version is 4x slower on Cortex-A8 not counting post-conversion.
>>> Float and int are about the same speed on Intel i7.
>>
>> mplayer (31175) 100mbDVB.ts -demuxer lavf -novideo -benchmark -ao pcm -ac
>> mp3/ffmp2float/ffmp2
>>
>> P3: Pentium III (Katmai) 500MHz SSE
>> XP: Athlon XP 2600+ Thoroughbred SSE, 3DNow!
>> HP: Intel(R) Pentium(R) M processor 1.86GHz (GenuineIntel 6 13 8) SSE2
>> TP: Intel(R) Pentium(R) M processor 1500MHz (GenuineIntel 6 9 5) SSE2
>>
>> P3: 2.9 / 3.9 / 4.8
>> XP: 0.8 / 1.1 / 1.3
>> HP: 0.5 / 0.6 / 0.9
>> TP: 0.7 / 0.9 / 1.3
>
> Using the same old MPlayer version:
> C2: Core2 Duo E8400
> G4: 800MHz PowerPC G4 (OS X 10.5.8)
>
> C2: 0.17 / 0.22 / 0.27
> G4: 2.6  / 2.7  / 4.9
>
> Latest improvements are hardly measurable on the Core2 Duo, but certainly very
> small compared to the original speed-up.

This is not surprising, that commit should give a very modest speedup. 
Try the attached one on x86 (not for review), should be more noticeable.

-Vitor
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dct32sse.diff
Type: text/x-patch
Size: 8098 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-cvslog/attachments/20100609/8d4e0200/attachment-0001.bin>