[FFmpeg-devel] [PATCH] convert dsputilenc_mmx.c hadamard functions to yasm

Ronald S. Bultje rsbultje
Fri Sep 17 14:32:27 CEST 2010


Hi,

On Fri, Sep 17, 2010 at 7:00 AM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
> On Thu, Sep 16, 2010 at 11:18 PM, Ramiro Polla <ramiro.polla at gmail.com> wrote:
>> Didn't work =(. Now we get segfault somewhere.
>
> OK, so most likely this is something very, very simple like not
> sign-extending a 32bit integer to 64bit (%ifdef ARCH_X86_64 movsxd r3,
> r3d %endif) or something along those lines. In my tests, all this
> worked. We're almost there. :-).

As I expected. I think this is fixed now (make fate passed all
mpegvideoenc tests, it's still running and I gotta run).

As much as I love and adore the new fate, I would've picked up this
crash earlier if Win64 (or win64/fate) wasn't such a pain to get
useful info out of. For one, "make V=1 fate" does not print out all
commands run on Win64 the way it does on Unixy systems. Also, the
error I got (which made me think I just broke Win64 in general) was
something like:

/home/rsbultje/ffmpeg/svn/tests/data/regression/vsynth1/error: line
134: tests/data/vsynth1.err: No such file or directory

What it should say is something like:

fate failed on vsynth1-error: Segmentation fault (or bus error, or whatever)

Can this be improved?

Ronald

Ronald



More information about the ffmpeg-devel mailing list