[RFC] Proposal for complex math intrinsic lowering through the MLIR complex dialect

In continuation of @szakharin’s work on using MLIR math dialect operations for Fortran math intrinsic functions, @DavidTruby is proposing to use the Complex dialect for Fortran complex math functions. See the RFC in the following phabricator review.
https://reviews.llvm.org/D134364

1 Like

I have made this change in rG3681a7dd7818, so we now lower through calls to libm for complex numbers unless -fapprox-func is passed (and by implication -ffast-math).