Page 2 of 3

Re: CUDA Suggestions and Talk

Posted: Fri Sep 21, 2018 5:07 am
by admin
I'm not seeing anything obviously useful but let's see. One downside of upgrading to later CUDA toolsets is that we have to leave some older cards unsupported. I hate to have to do that.

Re: CUDA Suggestions and Talk

Posted: Wed Oct 31, 2018 5:00 am
by Guest 2
Do you thing is feasible to port a CUDA version of DFTTest and/or FFTFilter? The first above all is really CPU time consuming.

Re: CUDA Suggestions and Talk

Posted: Wed Oct 31, 2018 9:50 am
by admin
I'll have a look as time permits.

Re: CUDA Suggestions and Talk

Posted: Thu Dec 13, 2018 1:22 am
by Chris
Hi Donald,

do I use DGBob correctly so far?

Code: Select all

DGBob(order=-1, mode=0)
And when or why would I want to use mode 1 or 2? I noticed that a 25fps movie is shown as a 50fps movie with mode 1 and as a 25fps movie with mode 2 but with double the runtime.

Also I think since I'm using DGSource I'm good using order -1, right?

As another question: Is there also a resize filter I could use from you that runs on CUDA support? :)

Re: CUDA Suggestions and Talk

Posted: Thu Dec 13, 2018 8:59 am
by admin
Assuming you are using the DGBob() from DGDecodeNV.dll and not the obsolete stand-alone filter, that looks fine.

mode 0 makes a smaller encode but loses motion fluidity
mode 1 makes a larger encode but keeps motion fluidity
mode 2 is to create a slow-motion effect.

Yes, DGBob() can take the field order from DGSource().

DGSource() can resize on the GPU. Use the rw and rh parameters.

Re: CUDA Suggestions and Talk

Posted: Fri Dec 14, 2018 10:54 am
by Chris
Thank you a lot for this package of help! Much appreciated! When using DGBob I assume mode 0 is the one closest to regular yadifmod2?

Re: CUDA Suggestions and Talk

Posted: Fri Dec 14, 2018 10:58 am
by admin
Not sure what you mean by closest. The field->frame algorithm is the same for all modes.

Re: CUDA Suggestions and Talk

Posted: Fri Dec 14, 2018 11:02 am
by Chris
Hehe I knew I should have been more specific :P

What I meant with "the closest" is in terms of quality, accuracy, ..etc as with yadifmod2 I don't get to have any options. Looking at encoding comparisons made I think mode 0 is what compares best with (non-option featured) yadifmod2.

Re: CUDA Suggestions and Talk

Posted: Fri Dec 14, 2018 11:46 am
by admin
As I said it's the same for all modes. Any difference you see will be due to encoding, i.e., your chosen bitrate may be enough for single-rate but not for double-rate.

Re: CUDA Suggestions and Talk

Posted: Sat Dec 15, 2018 12:39 am
by Chris
Alright, thanks. One more question that popped up during browsing forums and finding this: viewtopic.php?f=10&t=107

The thread is quite old so I'm not sure if it's still best to use AviSynths SelectEven or if there is a SelectEven() for CUDA from your end.

Re: CUDA Suggestions and Talk

Posted: Sat Dec 15, 2018 6:18 am
by admin
There is only the Avisynth function. There is no need to try to accelerate this function as it is already virtually instantaneous. It is implemented by simply mapping the requested frame number.

Re: CUDA Suggestions and Talk

Posted: Sat Dec 15, 2018 10:04 am
by Chris
Thanks a lot for your help and assistance!

One more question popped up I want to address: DGIndexNV/EncNV are only able to go with mod4 as lowest, right?
I am asking because sometimes I want to crop/resize by mod2 for where I take off 2 pixels from top (black bar), staxrip set to mod2, applying resize (staxrip) makes the video 720 x 404, removing staxrip resize, adding DGSource resize 720 x 404, makes the video 720 x 402.

Re: CUDA Suggestions and Talk

Posted: Sat Dec 15, 2018 12:30 pm
by admin
It's covered in the DGDecodeNV user manual.

Currently, mod 2 is OK for right and bottom but left and top must be mod 4. It's an nVidia driver issue that will be fixed at some point.

If you must have mod 2 for top or left, then you can use Avisynth cropping.

Re: CUDA Suggestions and Talk

Posted: Wed Aug 07, 2019 5:14 am
by Sharc
SMDegrain.avsi is a powerful but slow avisynth filter. Could similar be done in CUDA? Any thoughts?

Re: CUDA Suggestions and Talk

Posted: Wed Aug 07, 2019 7:58 am
by admin
Guest 2 just suggested that as well. I'll take a look into feasibility.

Re: CUDA Suggestions and Talk

Posted: Wed Aug 07, 2019 11:42 am
by Sharc
Thanks.
Will be a big challenge I think because it uses mvtools and masktools. :scratch:

Re: CUDA Suggestions and Talk

Posted: Wed Aug 07, 2019 11:57 am
by admin
Sharc wrote:
Wed Aug 07, 2019 11:42 am
Will be a big challenge
For sure.

Re: CUDA Suggestions and Talk

Posted: Wed Aug 21, 2019 6:34 am
by DJATOM
I'm not sure if it's possible to reuse Opticflow vectors data in mvtools2, but if it's compatible, that should provide nice speedup in scripts.

Re: CUDA Suggestions and Talk

Posted: Mon Sep 23, 2019 5:41 am
by Boulder
Having a high quality resizer like ResampleHQ would be a nice addition to the toolset.

https://gist.github.com/4re/64642122e359c37543fe

I personally use it all the time for downsizing with Bicubic, but I don't know if the GF hardware supports that one (particularly the b and c parameters that I tune for each source with a Zopti run).

Re: CUDA Suggestions and Talk

Posted: Wed Sep 25, 2019 8:15 pm
by Bullwinkle
+2 on that, Boulder, the bold one haha. Meese and squirrels are are worth 2, hoomans 1. Get over it.

Re: CUDA Suggestions and Talk

Posted: Wed Feb 19, 2020 10:52 pm
by JoyBell
Ever since this summer I was wondering if these modern open shader sharpeners could be added as a avisynth script.
Sorry if a bunch of people have brought this up already. They have good enough results to upsample 1440p to 2160p with very good results all while only dropping real time FPS a few percentages. This has to be much faster than my software cutting me from 10fps to 4 fps encoding.

Overview and results
https://www.techspot.com/review/1884-am ... s-reshade/
Image

I don't know how much harder it is to port than from reshade or come to think of it if I am muddling CUDA and shader scripts in my head.
https://www.youtube.com/watch?v=8PON7Vtdbk0

FidelityFX CAS (Contrast Adaptive Sharpening) v1.0
https://github.com/GPUOpen-Effects/Fide ... er/FFX_CAS

Re: CUDA Suggestions and Talk

Posted: Fri Apr 10, 2020 11:44 pm
by hydra3333
Goodness me, the youtube video examples look very nice !

So, are you asking for the code to somehow be incorporated into cudasynth
(yay, although that wonderful initiative seems to have stalled, or worse)
or perhaps into DGdecode ... I could live with that :)
(although I'd guess the author here may suggest it does not align with the prime purpose of dgdecode :) ) ?

In any case it would be nice to hear DG's thoughts.

edit: ah, I just saw this thread https://forum.doom9.org/showthread.php? ... Sharpening

Re: CUDA Suggestions and Talk

Posted: Sat Apr 11, 2020 4:37 am
by Rocky
Ha, and I hadn't noticed this thread before!

Never say never. I will look into it. Something to do over the next (minimum of) 20 days of lockdown. I live alone so I'm going crazy.

Re: CUDA Suggestions and Talk

Posted: Sun Apr 12, 2020 12:00 pm
by hydra3333
Rocky wrote:
Sat Apr 11, 2020 4:37 am
I live alone so I'm going crazy.
I don't and I'm going crazier ;) :D

Our Prime Minister has said for some time "at least six months" (in some form). Many official (or respected ;)) medical officers say even then at-risk people until a vaccine or mitigating medicine appears, if you don't want risk dying. (I read up on what it does, and it is a horrible way to die).

PS, good luck with it ... the code, the crazy, and the other.

Re: CUDA Suggestions and Talk

Posted: Sun Apr 12, 2020 3:34 pm
by Rocky
hydra3333 wrote:
Sun Apr 12, 2020 12:00 pm
I don't and I'm going crazier
LOL. Cat or wife?
Our Prime Minister has said for some time "at least six months" (in some form). Many official (or respected ;)) medical officers say even then at-risk people until a vaccine or mitigating medicine appears
Can't comment on that or admin will bust me for political discussion. Suffice to say, what a load of bollocks. We've been had.

Should we make a new forum for CCP virus talk?