masahi commented on PR #15111: URL: https://github.com/apache/tvm/pull/15111#issuecomment-1709845769
> It seems that the performance of the default cutlass tile does not perform well on compute bound shape gemm How do you know? The kernel uses a runtime heuristic to select a tile config. For now I'm hoping that this heuristic is "good enough". https://github.com/tlc-pack/cutlass_fpA_intB_gemm/blob/main/cutlass_kernels/fpA_intB_gemm/fpA_intB_gemm_template.h#L818-L826 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
