On Sun, Apr 14, 2013 at 10:51:24PM +0200, Diego Biurrun wrote: > On Sun, Apr 14, 2013 at 10:07:35PM +0200, Nicolas BERTRAND wrote: > > Le 14/04/2013 20:02, Diego Biurrun a écrit : > > >>I'd blame floating-point IDWT. I wonder if it will produce the same result > > >>>with "-mfpmath=sse -msse2" on x86. If yes then push and wait for someone > > >>>to > > >>>rewrite IDWT in integer mode and speed it up for j-b's delight. > > >Indeed, recompiling with those flags in CFLAGS/LDFLAGS fixes FATE on > > >x86_32. > > > > So what would be the solution ? > > Add those flags? > > Fix the code? :)
a) make FATE test account for fuzziness of the result b) add lots of round() and forget about performance in principle c) add an integer version of IDWT (even if only for bitexact tests) d) replace IDWT with an integer version for all instances ö) do nothing and pretend it's OK _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
