On 02.11.2016 11:31, Michael Niedermayer wrote: > On Wed, Nov 02, 2016 at 01:40:01AM +0100, Carl Eugen Hoyos wrote: >> 2016-11-01 0:27 GMT+01:00 Andreas Cadhalpun >> <andreas.cadhal...@googlemail.com>: >> >>> -#if HAVE_ALTIVEC >>> +#if HAVE_ALTIVEC && !(ARCH_PPC64 && HAVE_BIGENDIAN) >>> if (!PPC_ALTIVEC(av_get_cpu_flags())) >> >> Do the dnxhd tests work for you on 32bit BE powerpc?
Not when using the altivec functions. (It's not tested on the Debian buildd, since the baseline of Debian's powerpc architecture does not include altivec.) >> (It fails here afaict) > > i find it strange to belive that the code never worked on either HW Your intuition is correct. ;-) > if we assume it did once work for some case, what is differnt now ? > the testcase? It wasn't tested with FATE, so it wasn't noticed, when it broke. > the compiler? No. > the code? Yes, by the following Libav commit: 4c387c7 ppc: dsputil: do unaligned block accesses correctly LOL! I'll send a patch fixing this in a moment. Best regards, Andreas _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
