zlacker

[return to "AMD funded a drop-in CUDA implementation built on ROCm: It's now open-source"]
◧
1. codedo+pF1[view] [source] 2024-02-12 22:09:57
>>mfigui+(OP)
As I understand, Vulkan allows to run custom code on GPU, including the code to multiply matrices. Can one simply use Vulkan and ignore CUDA, PyTorch and ROCm?
◧◩
2. Const-+BZ1[view] [source] 2024-02-13 00:08:55
>>codedo+pF1
I did a few times with Direct3D 11 compute shaders. Here’s an open-source example: https://github.com/Const-me/Cgml

Pretty sure Vulkan gonna work equally well, at the very least there’s an open source DXVK project which implements D3D11 on top of Vulkan.

[go to top]