[HELP] Vapoursynth is terribly slow.

[HELP] Vapoursynth is terribly slow. - Printable Version

+- Selur's Little Message Board (https://forum.selur.net)
+-- Forum: Hybrid - Support (https://forum.selur.net/forum-1.html)
+--- Forum: Problems & Questions (https://forum.selur.net/forum-3.html)
+--- Thread: [HELP] Vapoursynth is terribly slow. (/thread-1146.html)

Vapoursynth is terribly slow. - Kylenam - 31.08.2019

Source : 1080p BluRay 8bit Remux
Output : x265 10bit
Same Settings : crop, SMDegrain, LSFmod

MySystem:
HexaCore AMD Ryzen 5 1600, 3700 MHz (37 x 100)
memory 16GB
GeForce GTX 1050
SSD 128G

Vapoursynth:
[Image: v.jpg]

avisynth:

Encoding up to 10 percent is the same.

Is it usually this slow?

RE: Vapoursynth is terribly slow. - Selur - 31.08.2019

I noticed that you used in Avisynth:

# grain handling

SMDegrain(tr=3,thSAD=400,RefineMotion=true,interlaced=false)

# sharpening

LSFmod(Smode=5,Smethod=3,Lmode=4,soft=-2,edgemaskHQ=true)

and in Vapoursynth:

clip = havsfunc.SMDegrain(input=clip, tr=3, thSAD=400, RefineMotion=True, interlaced=False)

clip = havsfunc.LSFmod(input=clip, Smode=2, Smethod=3, secure=True, Lmode=22, preblur=False, soft=-2, edgemaskHQ=True)

so at least the LSFmode differ quite a bit.

On a Ryzen 7 1800x using x265 with the default settings I get:

~7 fps with the Vapoursynth script you posted.
~6.6 with the Avisynth script you posted

One of the advantages of Vapoursynth is the better scalability in regard of cpu cores, so if you run on a system with less then 4 logical cores it might be slower than Avisynth.

-> so depending on your system it is expected that Vapoursynth is slower, but no clue why Vapoursynth should be that slow on your system.
I would have expected that the Vapoursynth script is a bit faster than the Avisynth script.

Cu Selur

Ps.: Noticed that I uploaded an old havsfunc script with the version, using opencl=True and the script from https://github.com/Selur/VapoursynthScriptsInHybrid/blob/master/havsfunc.py fixes helps (otherwise opencl=True doesn't work, with opencl enabled I get ~9.7fps on my system, 30%+ gain) This shows that the main bottleneck is SMDegrain in your script and using opencl does help since this causes Hybrid to use nnedi3cl instead if zneedi3.