PR #24564 [CUDA] MatMulNBits benchmark

[CUDA] MatMulNBits benchmark #24564

snnn merged 4 commits into main from tlwu/benchmark_matmul_8bits

Add benchmark script

9308f462

choose unroll kernel

705185e9

Replace unroll with simple loop

492b8dac

refine accumulation

48cb505e

tianleiwu requested a review from

jiafatom 1 year ago

tianleiwu requested a review from

kunal-vaishnavi 1 year ago

tianleiwu added release:1.22.0

kunal-vaishnavi approved these changes on 2025-04-26

snnn merged 1dd9b992 into main 1 year ago

snnn deleted the tlwu/benchmark_matmul_8bits branch 1 year ago

snnn removed release:1.22.0

Reviewers

kunal-vaishnavi

jiafatom

Assignees

No one assigned

Labels

None yet

Milestone

No milestone