Lc0 is useless on M1 because there's no way for it to use the hardware.
dangi12012 wrote:No one wants to touch anything you have posted. That proves you now have negative reputations since everyone knows already you are a forum troll.
Maybe you copied your stockfish commits from someone else too?
I will look into that.
Compile Stockfish from source. It is pretty trivial. Whatever brew serves up is way slower, I don't know how they managed to mess this up.
lc0 is a problem. The latest net architectures (78xxxx 79xxxx 80xxxx) have features that are only supported in the latest CUDA backends. The GPU opencl backend will work (slowly, compared to nVidia) with older net architectures. The blas CPU backend (using Apple Accelerate framework) will work very slowly with 79xxxx and 80xxxx nets. You will need to learn how to compile from source. I use lc0 remotely from my MBA on an Windows RTX3080.
Also, I would recommend Hiarcs Chess Explorer Pro as the best GUI.
Sopel wrote: ↑Mon Aug 01, 2022 12:31 pm
Lc0 is useless on M1 because there's no way for it to use the hardware.
Feel free to contact the LC0 developer.
why would I do that?
dangi12012 wrote:No one wants to touch anything you have posted. That proves you now have negative reputations since everyone knows already you are a forum troll.
Maybe you copied your stockfish commits from someone else too?
I will look into that.
Sopel wrote: ↑Mon Aug 01, 2022 12:31 pm
Lc0 is useless on M1 because there's no way for it to use the hardware.
Feel free to contact the LC0 developer.
Maybe Magnum could start to reach out to Apple about OpenCL support and performance issues in general and post their response here?
I doubt unpaid developers find it worth to focus on proprietary stuff as long as it a) isn't significantly better perfromance than alternatives or b) gets lackluster support from the vendor itself to the open-source community.
This is all new territory to me.