AviSynth (and most of the plugins) is not designed to be run on GPU, MP_Pipeline's core is still AviSynth so we have no way to make it run (entirely) on GPU. Some plugins can offload their own computation to GPU and that's the best we can get.

there are this https://github.com/nekopanda/AviSynthPlus

anyway it's kinda selfish to support cuda not opencl since not all people use or can get nvidia

also, I don't think it will be any better to use gpu unless the function is mathematically intense (like NLMeans but there are KNLMeansCL already)

@SAPikachu Ok, I understand, so in the best conditions you get lucky if you have plugins that support GPU computation, so far i tried a lot to figure out the performance issue but nothing seems to get the status in advance. Maybe real.finder know from where i should strike the problem since i'm using his filters (what he assembled together as avisynth filters).

@realfinder i don't know much about OpenCL, if it manages to communicate with GPU libraries, then it's a profit to generalize on all type of GPU (AMD included). also GPU helps to encode very faster regardless of type and intensity of filter/function, in my opinion, every fps boost should be drawn out of the GPU. and btw, correct me if i'm wrong, i assume you're a.k.a AmjadSONY. If that's the case, i'm using Denoising filters (Spatio-Temporal "deen()", and Spatial "eDeen()") that are very heavy for CPU to handle alone, that's what drove me here. more problem details are on doom9 forum StaxRip 2.0 page 111 @Legend Post. I'm using Avisynth+. maybe you know how to solve that, please help ^^.

I did post in doom9

but also I will answer a few things here

1st I didn't see any filter from me, maybe you mean it's my plugins collection pack?

2nd, both "deen()" and Spatial "eDeen()" are very old and close source so you will not get much from them whether quality or speed

and as I and SAPikachu said, GPU is not magic for Performance Boost for everything, It depends on how the filter work, and when the data going from cpu to gpu and gpu to cpu it lost some speed so that why gpu in x264 was not succeed, NVENC is another story since it full hardware encoder, but again when the filter is mathematically intense as I said it will benefit from gpu indeed, also in the case of video decoder it will leave more cpu power for cpu filters (but in your case is not since you didn't use 100% of cpu power in first place)

and if you insists to get GPU in avs, you can try https://github.com/nekopanda/AviSynthPlus/releases as I said with https://github.com/nekopanda/AviSynthCUDAFilters/releases also LSMASHSource has gpu decoder http://avisynth.nl/index.php/LSMASHSource

thanks for replying in doom9, i'm really grateful.

i get good performance, but also some artifacts are showing up for some reason, more details in doom9.

i use these filters Mediafire->Avisynth x64 because they are organised and assembled regularly, and not everyone is doing such thing unless he's passioned ^^

i'm aware that "deen" and "edeen" are very old, but i couldn't get better filters, or i didn't know there are better ones.

For the GPU story i see what you mean, i understand now why it's not working with every filter. now i'm really confused here i don't know what approach i'll engage CPU? or GPU? anyway i will try to make something with CUDA filters using those links you suggested, but please, i need your expertise and i'm really open to your suggestions and ready to consider your adviced filters, especially good ones that you know in Denoising and Sharpening, maybe it will be better if i show you what i'm trying to accomplish, you might know what's better to do.

one last thing! there's also FFmpeg Encoder that supports a lot of libraries, what do you think about it?

thank you very much for everything ^^.

please if you want to contact me, this is my email : boutakbout.yassine@gmail.com

yes, this getting off topic, so it's better to close this ticket

and yes I am the one who made this pack in mediafire, you can find me in discord (with AmjadSONY nickname) since I don't check email every times