Faster Compiling kernels

Discussion related to the LuxCore functionality, implementations and API.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Faster Compiling kernels

Post by Dade »

TAO wrote: Thu Apr 29, 2021 11:19 pm What about the minimum requirements to use LuxCore?
It requires CUDA 10 support at least.
Support LuxCoreRender project with salts and bounties
User avatar
Continuum
Posts: 71
Joined: Mon Dec 02, 2019 11:16 pm
Location: Zambia
Contact:

Re: Faster Compiling kernels

Post by Continuum »

Opencl 3.0 seems to have fixed the issues Nvidia had with it. There are already drivers apparently. I guess someone with Nvidia hardware should confirm this.

https://developer.nvidia.com/blog/nvidi ... onformant/

Personally this is excellent news, Opencl development skills are still very relevant, for the present and future it would seem.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Faster Compiling kernels

Post by Dade »

Continuum wrote: Sat May 01, 2021 3:55 pm Opencl 3.0 seems to have fixed the issues Nvidia had with it.
Where do you read it ? I bet the compilation will be as slow as usual, CUDA is slow too. OpenCL v3.0 is the biggest joke ever made: it is just OpenCL v1.2 under a different name.
Support LuxCoreRender project with salts and bounties
User avatar
TAO
Developer
Developer
Posts: 850
Joined: Sun Mar 24, 2019 4:49 pm
Location: France
Contact:

Re: Faster Compiling kernels

Post by TAO »

Continuum wrote: Sat May 01, 2021 3:55 pm Opencl 3.0 seems to have fixed the issues Nvidia had with it. There are already drivers apparently. I guess someone with Nvidia hardware should confirm this.

https://developer.nvidia.com/blog/nvidi ... onformant/

Personally this is excellent news, Opencl development skills are still very relevant, for the present and future it would seem.
Yes, exactly that was the reason i reported they release the new version of OpenCL and that was a piece of good news because, Nvidia out of nowhere decide to add it in new driver packs, exactly the day when we started this conversation. I'm pretty sure there is a good reason behind that and i think Nvidia does not want to get involved in a GPU fight with AMD at the moment especially after what happens to the Intel share market with AMD and Apple new chip otherwise there is no reason at all why they decide to support OpenCL after all these years and in just 2 days.
I personally have a good feeling about that.
User avatar
TAO
Developer
Developer
Posts: 850
Joined: Sun Mar 24, 2019 4:49 pm
Location: France
Contact:

Re: Faster Compiling kernels

Post by TAO »

With the recently released R465 display driver, NVIDIA is now officially OpenCL 3.0 conformant on both Windows and Linux.
Screenshot 2021-05-01 191311.jpg
User avatar
TAO
Developer
Developer
Posts: 850
Joined: Sun Mar 24, 2019 4:49 pm
Location: France
Contact:

Re: Faster Compiling kernels

Post by TAO »

I already install and test it. it is really OpenCL 3.
Screenshot 2021-05-01 193255.jpg
I check every part of it, not just the file version and the result is the same.
User avatar
Continuum
Posts: 71
Joined: Mon Dec 02, 2019 11:16 pm
Location: Zambia
Contact:

Re: Faster Compiling kernels

Post by Continuum »

Dade wrote: Sat May 01, 2021 5:07 pm
Continuum wrote: Sat May 01, 2021 3:55 pm Opencl 3.0 seems to have fixed the issues Nvidia had with it.
Where do you read it ? I bet the compilation will be as slow as usual, CUDA is slow too. OpenCL v3.0 is the biggest joke ever made: it is just OpenCL v1.2 under a different name.
It wasn't a performance issue that was preventing Opencl adoption I think. I can't find the reference now, but Opencl 2.0 introduced features which would have affected Nvidia drivers negatively. Opencl 3.0 changed that I believe.

I don't know about compilation, the issue can always be shared with the developers. If Vulkan & Opencl become compatible, then precompiled code may be possible (spir).

From what I can see Nvidia supports quite a few technologies that they do not own. They have a fluids simulation engine written in Direct X, it's quite good.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: Faster Compiling kernels

Post by Dade »

You are talking of OpenCL v3.0 like if it is something new: it isn't, it is OpenCL v1.2. Plain nothing has changed. Still everything stuck to OpenCL v1.2 (now known as v3.0).
Support LuxCoreRender project with salts and bounties
User avatar
TAO
Developer
Developer
Posts: 850
Joined: Sun Mar 24, 2019 4:49 pm
Location: France
Contact:

Re: Faster Compiling kernels

Post by TAO »

Honestly, you know OpenCL better than anyone and it is sad to see nothing change in it.
At least we can leave it as it is or just update to the new version for now.
Last edited by TAO on Sat May 01, 2021 7:15 pm, edited 1 time in total.
User avatar
Continuum
Posts: 71
Joined: Mon Dec 02, 2019 11:16 pm
Location: Zambia
Contact:

Re: Faster Compiling kernels

Post by Continuum »

Dade wrote: Sat May 01, 2021 6:41 pm You are talking of OpenCL v3.0 like if it is something new: it isn't, it is OpenCL v1.2. Plain nothing has changed. Still everything stuck to OpenCL v1.2 (now known as v3.0).
Not completely true, if you go through the driver release notes, it's show features beyond v1.2 have been included, including experimental features from v2.0. Link below.

https://us.download.nvidia.com/Windows/ ... -notes.pdf
Post Reply