Open tormento opened 3 months ago
Could you also detect this in earlier versions?
I will try ASAP.
and what about other versions of x265?
What do you mean with "other" versions?
X265 aMod etc.
Please send me the link on GitHub (or elsewhere) of the builds you want me to try.
I don't even know if they provide an AVX build or not.
Please run the test again with the current release. There was an error in the code. I hope the result is now almost identical.
I tested the version that my venerably Sandy Bridge can support:
There must something terribly wrong with gcc13.2.0. Look at speeds:
5,95 11 avx-msvc1937_temp
5,84 11 sandybridge-gcc13.2.0_temp
5,47 11 msvc1937_temp
2,35 11 gcc13.2.0_temp
The graphs look quite good except gcc13.2.0, how are the data of the plain version?
Which is the "plain" version?
Which is the "plain" version?
Sorry, misunderstanding. The values fit in and of themselves.
Have you done some more testing? Are there any new findings?
Let me have some day more.
I have done some tests with the last AVX version and I have found that in most tests, it has different VMAF metrics than plain version:
Sintetic ones such as PSNR and SSIM are almost identical.
I use FFMetrics, it you want to replicate them.