On Tuesday, April 24, 2012, Sean McGovern <gsean...@gmail.com> wrote: > > > On Tuesday, April 24, 2012, Måns Rullgård <m...@mansr.com> wrote: >> Sean McGovern <gsean...@gmail.com> writes: >> >>> Valgrind had a field day when I left it enabled -- it will need >>> further investigation as it must mean AltiVec is implemented >>> differently on POWER7. >> >> I thought the Power variants were generally backwards compatible. How >> is it failing? Does it pass tests without valgrind? > > I'll have to try that manually -- note that Mike Kostylev's ppc64 FATE builds don't use -mcpu=power7 (even though they give --cpu=power7 to 'configure') so they aren't a good base to say why this fails. > >> Actually, this going to break on everything else. Something has to >> detect if vsx is really supported on the selected CPU and disable it if >> not. >> > > I can write a test for that, no problem. >
The results aren't promising. At a quick glance, there are over 90+ failures with '-mcpu=power7 -maltivec -mabi=altivec -mvsx' (FATE run without valgrind). -- Sean McG.
_______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel