[FFmpeg-devel] [PATCH] force dnxhd encoder to be independent of qsort internals

[Reimar Döffinger]

New benchmark results:

Command:
./ffmpeg_g -y -flags +bitexact -dct fastint -idct simple -sws_flags +accurate_rnd+bitexact -f image2 -vcodec pgmyuv -i ./tests/vsynth1/%02d.pgm -vcodec dnxhd -flags +ildct -s hd1080 -b 120Mb -pix_fmt yuv422p -an ././tests/data/a-dnxhd-1080i.mov

31715530 dezicycles in qsort, 1 runs, 0 skips
30340880 dezicycles in qsort, 2 runs, 0 skips
29507885 dezicycles in qsort, 4 runs, 0 skips
29701047 dezicycles in qsort, 8 runs, 0 skips
29248313 dezicycles in qsort, 16 runs, 0 skips
29265670 dezicycles in qsort, 32 runs, 0 skips
29055702 dezicycles in qsort, 64 runs, 0 skips

2749630120 dezicycles in encode_fast qsort, 1 runs, 0 skips
2038592080 dezicycles in encode_fast qsort, 2 runs, 0 skips
1669742982 dezicycles in encode_fast qsort, 4 runs, 0 skips
1483996520 dezicycles in encode_fast qsort, 8 runs, 0 skips
1388271382 dezicycles in encode_fast qsort, 16 runs, 0 skips
1344456126 dezicycles in encode_fast qsort, 32 runs, 0 skips
1321935196 dezicycles in encode_fast qsort, 64 runs, 0 skips

21611000 dezicycles in radix_sort, 1 runs, 0 skips
21724385 dezicycles in radix_sort, 2 runs, 0 skips
21598177 dezicycles in radix_sort, 4 runs, 0 skips
21465423 dezicycles in radix_sort, 8 runs, 0 skips
21433041 dezicycles in radix_sort, 16 runs, 0 skips
21450375 dezicycles in radix_sort, 32 runs, 0 skips
21451905 dezicycles in radix_sort, 64 runs, 0 skips

2727159440 dezicycles in encode_fast radix_sort, 1 runs, 0 skips
2018524655 dezicycles in encode_fast radix_sort, 2 runs, 0 skips
1643299490 dezicycles in encode_fast radix_sort, 4 runs, 0 skips
1456110210 dezicycles in encode_fast radix_sort, 8 runs, 0 skips
1362578335 dezicycles in encode_fast radix_sort, 16 runs, 0 skips
1316680249 dezicycles in encode_fast radix_sort, 32 runs, 0 skips
1288873409 dezicycles in encode_fast radix_sort, 64 runs, 0 skips

The results for the whole encode_fast still seem strange, the speedup
should not be that much.