[FEA] Could developer optimize cute for L2, bank conflict, double buffer and so on? Thank you!!!

[ziyuhuang123]

https://github.com/NVIDIA/cutlass/blob/main/examples/cute/tutorial/sgemm_nt_1.cu

It seems this does not implement L2 cache optimization (swizzle), and shared memory bank conflict, double buffer, optimization? Is it possible to use them? Thank you!!!!

[thakkarV]

the goal of this example was to be a four part series of tutorials. they will be released over time in the future with the part four approaching about the speed of light. That said, all of the optimizations you mention can certainly be implemented in this example as well.

[ziyuhuang123]

Well.... thank you! It is just, I am not that clever to write the code. So I will wait for your update, thanks~

[mnicely]

Closing due to inactivity. Feel free to reopen if needed.