Support forum for DG Tools

Seeing as how you are working on a native vpy version of DGHDRtoSDR, do you have any plans or intentions of doing the same for the filters such as DGDenoise, DGSharpen etc.
If I missed something really obvious, like it already has a native version, I apologize for my senility

That's right, gonca, the utility filters do not have native Vapoursynth support. It could be added but right now I find that the DGHDRtoSDR Vapoursynth native version runs at half the speed of the Avisynth+ one. Unless I can resolve that and bring those into line, Vapoursynth support seems pointless. You can always run the filters in avscompat mode.

Thanks for the answer.

You can always run the filters in avscompat mode.

I already do, or change the script to an avs one.
I was just checking since I couldn't remember if you had released a native version.
Once again, thanks for the answer

You're welcome.

I'm on a little kick studying performance matters. I discovered something interesting. This script runs fast (182 fps for 3840x2160):

DGSource()

This script runs way slower (36 fps!):

DGSource()
prefetch(6)

A bit surprising to me as DGSource is MT_SERIALIZED.

But if I add DGHDRtoSDR in there then the results are reversed, i.e., the script with prefetch(6) runs faster (102 fps versus 82 fps).

Regarding Vapoursynth, I am currently clueless why it runs at half the speed of Avisynth+. Could be differences in multithreading and/or caching designs. Still investigating it...

Just did a quick test on a 1080p clip
On a DGDource() script Vapoursynth is slightly faster than Avisynth+, and prefetch does make it even slower
This is weird

Hi.

I have found that unless you are using some filters in the AVS file, one might as well leave out Prefetch(x), as that will just slow down things.

If the source one have is pure progressive, and all that needs to be done is cropping and maybe resizing, then leaving out prefetch(x), will be faster than having it in.

Why. I don't know. But its the way it seems to be.

renols