2012/2/23 Måns Rullgård <[email protected]>: >> Aren't the timings reported in the patch commit message interesting? >> Because I meant those (for x86 using SSE FP math): >> - 32bits: 102c -> 82c >> - 64bits: 82c -> 69c > > I missed that bit. If you add -funsafe-math-optimizations to CFLAGS, > does gcc unroll this automatically?
No, neither for a 64 nor a 32 bits target. It also does a poor job of interleaving instructions even in the unrolled case, but that's another matter. Note: I'm evaluating with a mingw gcc 4.6.1, which may be poorer at this job. Christophe _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
