[llvm] [X86] combineKSHIFT - fold kshiftr(kshiftr/extract_subvector(X, C1), C2) --> kshiftr(X, C1+C2) (PR #115528)

Phoebe Wang via llvm-commits llvm-commits at lists.llvm.org
Fri Nov 8 19:34:40 PST 2024


phoebewang wrote:

> Merge serial KSHIFTR nodes, possibly separated by EXTRACT_SUBVECTOR, to allow mask instructions to be computed in parallel.

I think parallel is not much important for mask operations, becasue they mainly serve for vector instructions. That says, the contiguous mask operations would not be a bottleneck in real workload.

https://github.com/llvm/llvm-project/pull/115528


More information about the llvm-commits mailing list