[flang-commits] [flang] [llvm] [flang] Round COMPLEX(4)**INTEGER to single precision once (PR #230257)

[Kiran Chandramohan via flang-commits]

kiranchandramohan wrote:

Thanks for this accuracy improvement.

Do we see any change in performance on Spec, elefunt and blas/lapack? Can this change be tied to the fast-math or its component flags?

https://github.com/llvm/llvm-project/pull/230257