[FFmpeg-devel] [PATCH] fix for roundup issue 2127
Sat Jan 1 01:47:23 CET 2011
I have added the changes and ran some more tests. It passes "make fate"
on 32/64-bit, with/without --disable-optimizations. The speed is the
exact same as the original. Are there any comments on the updated patch?
On Fri, Dec 31, 2010 at 11:54 AM, Ronald S. Bultje <rsbultje at gmail.com>wrote:
> On Fri, Dec 31, 2010 at 11:30 AM, Reimar D?ffinger
> <Reimar.Doeffinger at gmx.de> wrote:
> > On Fri, Dec 31, 2010 at 10:59:01AM -0500, Ronald S. Bultje wrote:
> >> On Fri, Dec 31, 2010 at 10:55 AM, Reimar D?ffinger
> >> <Reimar.Doeffinger at gmx.de> wrote:
> >> > which I think makes this unacceptable, at least if we still have to
> >> > policy that fast code in as many cases as possible is more important
> >> > that compilation with options that are only useful for debugging.
> >> The code is faster after his patch; the original code does 6 leas, the
> >> new code uses direct addressing.
> > Well, it still needs gcc to generate the *3 value. By moving that into
> > the asm code (and making the strides inout values) I think it should be
> > possible to make it work with only 4 registers, though it might be
> > slower (instruction scheduling may be a bit problematic there).
> Worth testing. :-).
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 3489 bytes
Desc: not available
More information about the ffmpeg-devel