[flang-commits] [flang] [mlir] [MLIR] Add cpow support in ComplexToROCDLLibraryCalls (PR #153183)

[Slava Zakharin via flang-commits]

================
@@ -1625,15 +1647,19 @@ static constexpr MathOperation mathOperations[] = {
      genFuncType<Ty::Real<16>, Ty::Real<16>, Ty::Integer<8>>,
      genMathOp<mlir::math::FPowIOp>},
     {"pow", RTNAME_STRING(cpowi),
----------------
vzakhari wrote:

Sorry, the "runtime" approach that I meant was that we generate the `cpowi` etc. calls here, and then the runtime implementation for AMD GPU device uses the `__ocml_*` intrinsics.  Would that be a viable solution?  I guess the benefit of having the complex operations is some special case handling, like the constant exponent optimizations.

https://github.com/llvm/llvm-project/pull/153183