[XviD-devel] recent mmx patch (xvid_bench results)

jmtest at voila.fr jmtest at voila.fr
Thu Apr 15 15:17:10 CEST 2004


Hello,

I'm happy to see it works, this time... Pete, did you check the timing before and after to be sure there's improvement on this computer? You know i can write false code that goes through the bench, so I guess I can write faster code that goes slower...
I did tests on pentium 4 an mobile pentium, it's faster, but  I don't have an old pentium to check...


> Message du 15/04/04 à 11h44
> De : suxen_drol <suxen_drol at hotmail.com>
> A : xvid-devel at xvid.org
> Copie à : 
> Objet : Re: [XviD-devel] recent mmx patch (xvid_bench results)
> 
> On Wed, 14 Apr 2004 21:58:51 +0200 Edouard Gomez <ed.gomez at free.fr> wrote:
> > <jmtest at voila.fr> (jmtest at voila.fr) wrote:
> > > So i did remove the cmov instructions from the mmx code
> > 
> > The patch looked correct, so i applied it on _my_ head branch. It works
> > fine too for me(tm)
> > 
> > I'm now waiting for pete's feedback before applying it to the CVS head.
> 
> i ran two short encodes with quant, and its all good.
> xvid_bench output follows. mingw was used; msvc's gettimeofday()
> exhibits inaccurate timing. 
> 
>  =====  test quant =====
> PLAINC -   quant_mpeg_intra 217.500 usec       crc32=0xfd6a21a4 
> PLAINC -   quant_mpeg_inter 232.773 usec       crc32=0xf6de7757 
> PLAINC - dequant_mpeg_intra 137.695 usec       crc32=0x2def7bc7 
> PLAINC - dequant_mpeg_inter 165.469 usec       crc32=0xd878c722 
> PLAINC -   quant_h263_intra 75.508 usec       crc32=0x2eba9d43 
> PLAINC -   quant_h263_inter 104.453 usec       crc32=0xbd315a7e 
> PLAINC - dequant_h263_intra 88.398 usec       crc32=0x9841212a 
> PLAINC - dequant_h263_inter 101.719 usec       crc32=0xe7df8fba 
>  --- 
> MMX    -   quant_mpeg_intra 21.133 usec       crc32=0xdacabdb6 | ERROR
> MMX    -   quant_mpeg_inter 25.039 usec       crc32=0x72883ab6 | ERROR
> MMX    - dequant_mpeg_intra 22.656 usec       crc32=0x2def7bc7 
> MMX    - dequant_mpeg_inter 25.820 usec       crc32=0xd878c722 
> MMX    -   quant_h263_intra 12.930 usec       crc32=0x2eba9d43 
> MMX    -   quant_h263_inter 12.500 usec       crc32=0xbd315a7e 
> MMX    - dequant_h263_intra 17.227 usec       crc32=0x9841212a 
> MMX    - dequant_h263_inter 17.227 usec       crc32=0xe7df8fba 
>  --- 
> 3DNOW  -   quant_mpeg_intra 217.500 usec       crc32=0xfd6a21a4 
> 3DNOW  -   quant_mpeg_inter 232.773 usec       crc32=0xf6de7757 
> 3DNOW  - dequant_mpeg_intra 137.305 usec       crc32=0x2def7bc7 
> 3DNOW  - dequant_mpeg_inter 165.078 usec       crc32=0xd878c722 
> 3DNOW  -   quant_h263_intra 75.508 usec       crc32=0x2eba9d43 
> 3DNOW  -   quant_h263_inter 104.844 usec       crc32=0xbd315a7e 
> 3DNOW  - dequant_h263_intra 88.008 usec       crc32=0x9841212a 
> 3DNOW  - dequant_h263_inter 102.109 usec       crc32=0xe7df8fba 
> 
> -- pete
> 
> 
> _______________________________________________
> XviD-devel mailing list
> XviD-devel at xvid.org
> http://list.xvid.org/mailman/listinfo/xvid-devel
> 
------------------------------------------

Faites un voeu et puis Voila ! www.voila.fr 




More information about the XviD-devel mailing list