"Ronald S. Bultje" <[email protected]> writes: > Hi, > > 2011/10/3 Måns Rullgård <[email protected]>: >> "Ronald S. Bultje" <[email protected]> writes: >> >>> From: "Ronald S. Bultje" <[email protected]> >>> >>> ~50% decoding speedup (overall) with 4 threads. >> >> That seems a bit low for an intra-only codec. > > Probably shitty laptop with some non-cpu bottleneck. Desktop numbers: > > for n in 1 1 2 2 4 4; do echo -n "$n: "; ( time ./avconv -threads $n > -i ~/Desktop/Sequence\ 1-Apple\ ProRes\ 422.mov -f null -v 0 -vsync 0 > -an - ) 2>&1 | grep ^real; done > > 1: real 0m4.207s > 1: real 0m4.216s > 2: real 0m2.645s > 2: real 0m2.644s > 4: real 0m1.649s > 4: real 0m1.645s > > Seems good enough?
This seem to speed up ~1.6x for each doubling of the cores. That's certainly better. Good enough? Decoding intra frames is in theory infinitely parallelisable, so what's holding it back? -- Måns Rullgård [email protected] _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
