[Ffmpeg-devel] Re: Intel IPP

[Tomas Christiansson]

Hi,

I think that the DCT functions can be found in the 'ippcv' or 'ippvc'
libraries (I link to both just in case). If you add the ippsZero_16s
(clear_blocks) call you should also add the 'ipps' library.

/tomas

> -----Original Message-----
> From: ffmpeg-devel-bounces at mplayerhq.hu 
> [mailto:ffmpeg-devel-bounces at mplayerhq.hu] On Behalf Of g.
> Sent: den 15 november 2005 17:16
> To: ffmpeg-devel at mplayerhq.hu
> Subject: [Ffmpeg-devel] Re: Intel IPP
> 
> 
> Tomas Christiansson <tomas.christiansson <at> gasoptics.com> writes:
> > 
> > Hi, 
> > 
> > Yes, we tried this (for MPEG2 decompression) last week using IPP 4.0
> > (Linux) and  managed to get an 8-9% speedup of the 
> decompression (gcc
> > 3.3.4).
> 
> Thanks for your quick reply. 
> 
> What CPU was that on?
> 
> > Hopefully we'll be able to get more speedup when we move up to
> > IPP version 5.0. Below, you'll find patches (for
> > 'libavcodec/i386/dsputil_mmx.c' and 'configure') from our tests..
> > 
> > What kind of linking are you using? Shared libraries or 
> static? Which
> > IPP-version are you using?
> 
> I'm using v5.0 (or at least, I'm trying to).
> 
> I'm trying to use shared at the moment. I didn't get any static libs 
> presumably because I simply downloaded the trial version.
> 
> In the makefile of libavcodec I simply added -LIPP_PATH 
> -lippi -lippcore but I 
> keep on getting
> 
> "In function `dsputil_init_mmx':
> dsputil_mmx.c:2608: undefined reference to `ippiDCT8x8Fwd_16s_C1I at 4'
> collect2: ld returned 1 exit status"
> 
> I'm not sure what the "I at 4" suffix refers to, I simply tried to use 
> ippiDCT8x8Fwd_16s_C1
> 
> I used MinGW and the Windows version of IPP.
> 
> Regards
> 
> Graham
> 
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> http://mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>