Requesting test files for my GPU denoiser

Support for my Avisynth filters
Locked
DAE avatar
sparktank
Posts: 64
Joined: Thu Mar 28, 2013 5:24 pm
Location: Gotham City

Re: Requesting test files for my GPU denoiser

Post by sparktank »

:wow: whaaaa. Late to the party?

I see from the other thread a regression and some fixin' needin'.

I had the first version downloaded, but forgot it when I moved it to a memory card.
With so many updates everywhere, it's hard to keep track of everything.

This should give me time to dig up some noisy blurays.
Nvidia GeForce GT 1030 | GPU Caps Viewer
Windows 10 Annversary (x64) | ASVPlus_x64
Region A/R1
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: Requesting test files for my GPU denoiser

Post by admin »

I'll try to get it back on line for you today.
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: Requesting test files for my GPU denoiser

Post by admin »

OK, I have re-uploaded it with the regression fixed.

http://rationalqm.us/misc/DGDenoise.rar
DAE avatar
sparktank
Posts: 64
Joined: Thu Mar 28, 2013 5:24 pm
Location: Gotham City

Re: Requesting test files for my GPU denoiser

Post by sparktank »

Thanks! I'll give a test run today and see about getting some clips up on MEGA.co.nz for posterity.
Nvidia GeForce GT 1030 | GPU Caps Viewer
Windows 10 Annversary (x64) | ASVPlus_x64
Region A/R1
DAE avatar
Sharc
Posts: 233
Joined: Thu Sep 23, 2010 1:53 pm

Re: Requesting test files for my GPU denoiser

Post by Sharc »

I'm getting from Avisynth: "Could not find usable CUDA device".
My GPU is 9600 GS, Driver Version 9.18.13.4192 on which DGDecNV works flawless (except for HEVC)
User avatar
admin
Posts: 4551
Joined: Thu Sep 09, 2010 3:08 pm

Re: Requesting test files for my GPU denoiser

Post by admin »

IIRC, CUDA toolkit 7.5 dropped support for that old compute capability. I would have to build it with 6.5 instead. For some reason when I converted from the last server version to the current version, I went with 7.5. Anyway, as I mentioned on the other thread I decided not to proceed with this filter because the speed gain it obtains over KNLMeansCL comes at a steep price of some nasty artifacts. I may integrate the KNN kernel (which does not have the artifacts) into DGDecodeNV, and that would be built with toolkit 6.5.

I'll lock this thread now.
Locked