[FFmpeg-devel] [PATCH 3/3] two functions were declared inlinebut were referenced elsewhere

J. Bohl jbohl at jbohl.de
Wed Jun 13 00:07:01 CEST 2012


Hi,

I published the sources on http://winffmpeg.codeplex.com/ .
Yes, I agree to your objections - and might add that they may very well
introduce bugs (I did not yet thoroughly test every change). For an
example how this looks like see e.g. h264_qpel_mmx.c
(http://winffmpeg.codeplex.com/SourceControl/changeset/view/17734#226890
) or fft_sse.c
(http://winffmpeg.codeplex.com/SourceControl/changeset/view/17734#226876
) - the latter is a simple case (function "ff_imdct_calc_sse").

Regards,

Juergen

-----Original Message-----
From: ffmpeg-devel-bounces at ffmpeg.org
[mailto:ffmpeg-devel-bounces at ffmpeg.org] On Behalf Of Carl Eugen Hoyos
Sent: Monday, June 11, 2012 11:14 PM
To: ffmpeg-devel at ffmpeg.org
Subject: Re: [FFmpeg-devel] [PATCH 3/3] two functions were declared
inlinebut were referenced elsewhere

J. Bohl <jbohl <at> jbohl.de> writes:

> The Intel compiler's support for AT&T-inline-assembly is not really 
> complete compared to GCC, foremost it is lacking the ability to 
> reference external symbols in assembly code. I had to turn all 
> external references into passing the addresses as input registers 
> ("m"). However, this worked out ok and all code compiles flawless now 
> (and works, at least as far as my testing goes).
> I am not sure whether these changes are of interest here.

I have a suspicion that the patch will either introduce a performance
penalty or break compilation on some platforms (or
both) and is therefore unacceptable (but I may be wrong), but
*please* send the patches (or consider publishing a git clone): 
Many people have asked for it (and many have promised patches that never
showed up).

Carl Eugen



More information about the ffmpeg-devel mailing list