> I haven't benchmarked it at all, but as for Huffman parsing, this is > an example where bitstream reading does matter. > > Irrespective of the hardening/cleaness/... this provides, it would be > interesting to see START/STOP_TIMER results on dnxhd_decode_dct_block > for e.g. arm and x86. >
I benchmarked dnxhd ( https://www.youtube.com/watch?v=9TkHpvaO09c encoded to dnxhd) with perf stat, the result is 3.2% speed up for x86_64. _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
