Yuqi Dong

icon location Undergraduate in Peking University

Results 5 comments of Yuqi Dong

[QST] possible performance bug due to disabling inlining for ldmatrix

Hi @davidpissarra thank you very much for raising this issue. Here are my experimental results: In my experiments, I did not observe any difference in local memory accesses when switching...

[Feat] Integrate Z3 in TVM Arith Analyzer

This PR brings performance improvements for mha_fwd_bhsd_wgmma_pipelined, as well as other cases such as warp_specialize_gemm_copy1_gemm_0.

[Feature]: Add benchmark scripts for examples

/perf

[Feature]: Add benchmark scripts for examples

/perf

[Feature]: Add benchmark scripts for examples

/perf