2017-12-04 19:03 GMT+01:00 Michael Niedermayer <mich...@niedermayer.cc>:
> On Sun, Dec 03, 2017 at 05:30:32AM +0100, Carl Eugen Hoyos wrote:
>> Hi!
>> Attached patch should fix ticket #6795, only tested with one filter
>> chain (bitexact): testsrc2=r=5:d=10,minterpolate=10
>> Please comment, Carl Eugen
> Does this affect speed ? there are more pointers after the patch

Yes (I had not tested before):
The following filter chain has a higher performance here on x86_64
with vanilla gcc 6.3 with the patch applied:

> also this does not truly fix the issue, it just moves the threshold
> of failure up

The new limit seems more reasonable to me.

Carl Eugen
ffmpeg-devel mailing list

Reply via email to