[llvm] [X86] Prefer `lock or` over mfence (PR #106555)

Valentin Churavy via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 27 06:03:39 PDT 2024

Previous message: [llvm] Reland "[NVVM] Upgrade nvvm.ptr.* intrinics to addrspace cast" (PR #110262)
Next message: [compiler-rt] [compiler-rt][profile] Duplicate filename in `parseAndSetFilename` if exists (PR #110264)
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

vchuravy wrote:

On a Genoa machine (AMD EPYC 9384X), a benchmark of mine takes 14.13s to execute with seq_cst defaulting to `mfence` and 9.99s with `lock or`. This is single-threaded...


https://github.com/llvm/llvm-project/pull/106555

Previous message: [llvm] Reland "[NVVM] Upgrade nvvm.ptr.* intrinics to addrspace cast" (PR #110262)
Next message: [compiler-rt] [compiler-rt][profile] Duplicate filename in `parseAndSetFilename` if exists (PR #110264)
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

More information about the llvm-commits mailing list