llvm-project
5f935e91 - [AArch64] Optimize fp64 <-> fp16 SIMD conversions

Commit

1 year ago

[AArch64] Optimize fp64 <-> fp16 SIMD conversions Legalization would result in needless scalarization. Add some DAGCombines to fix this up.

Author

majnemer

Committer

majnemer

Parents