4 ms·

Do it in Go and I’m interested

• jrhey 14d ago
  Go is simply not feasible for CUDA work mainly because of the go runtime that manages GC, allocation, scheduling etc GPU kernels want explicit memory control and as little go-runtime like overhead as possible