Selur's Little Message Board
[BUG] SSIMD Mitchell ERROR - 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: [BUG] SSIMD Mitchell ERROR (/thread-2162.html)

SSIMD Mitchell ERROR - DevSker - 01.11.2021

Hi Selur,
I made a terrible mistake of using a tempmail and a password I didn't know I wouldn't remember Angel . So I couldn't login to get the dev link.
Since I'm here again, I'll do this right.

So...I've tried to use the SSIMD Resizer from Vapoursynth with MitchellNetravali Kernel and everything I had was a "stats is too small! (byteSize: 0byte)" error. It's not a unique error with Mitchell, using Hermite causes the same problem, probably other kernels as well (Test needed).

Version: Hybrid 2021.09.28.1

RE: SSIMD Mitchell ERROR - Selur - 01.11.2021

Will look into it. Probably a typo somewhere.
kernel: (string) Resample kernel of vszimg/fmtconv.
            Default is 'Bicubic'.
source: https://github.com/WolframRhodium/muvsfunc/blob/f524942ac6cacd937f9479a417eb308a335dfc1d/muvsfunc.py#L3700
-> kernels need to be restricted to
  • point
  • rect
  • linear
  • cubic
  • bicubic
  • lanczos
  • blackman
  • blackmanminlobe
  • spline16
  • spline36
  • spline64
  • gauss
  • sinc
  • impulse <- can't be used, due to missing parameters
see: https://github.com/EleonoreMizo/fmtconv/blob/254ef7f19d53e3dfee6b494d13e6f9fbcbbe4be6/doc/fmtconv.html#L1303
-> will fix
Cu Selur

RE: SSIMD Mitchell ERROR - DevSker - 01.11.2021

(01.11.2021, 09:21)Selur Wrote: Will look into it. Probably a typo somewhere.
kernel: (string) Resample kernel of vszimg/fmtconv.
            Default is 'Bicubic'.
source: https://github.com/WolframRhodium/muvsfunc/blob/f524942ac6cacd937f9479a417eb308a335dfc1d/muvsfunc.py#L3700
-> kernels need to be restricted to
  • point
  • rect
  • linear
  • cubic
  • bicubic
  • lanczos
  • blackman
  • blackmanminlobe
  • spline16
  • spline36
  • spline64
  • gauss
  • sinc
  • impulse <- can't be used, due to missing parameters
see: https://github.com/EleonoreMizo/fmtconv/blob/254ef7f19d53e3dfee6b494d13e6f9fbcbbe4be6/doc/fmtconv.html#L1303
-> will fix
Cu Selur
Hello,
Thank you for the quick fix.
And taking advantage of this thread, to avoid making a new one unnecessarily. Could you let me know if it is possible to use SSIMD using the GPU? I read something about VS Placebo being able to do this, but I didn't get in-depth about it.

RE: SSIMD Mitchell ERROR - Selur - 01.11.2021

placebo.Shader uses the gpu for running GLS shader files, if you find someone who ports the ssim_downscale algorithm to a GLS script which placebo.Shader you could use it.
Looking at the ssim_downscale algorithm is has no option to use the gpu. So either it would need to be ported to GLS or rewritten to use the gpu. Both seem possible, but are nothing I know how to.

Cu Selur