================ @@ -0,0 +1,202 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ---------------- davemgreen wrote:
For reference when they are needed, this was a list of strict-fp ops that we would need to make sure are tested. Hopefully a lot of them are simple expanded into libcalls. ``` llvm.experimental.constrained.acos llvm.experimental.constrained.asin llvm.experimental.constrained.atan2 llvm.experimental.constrained.atan llvm.experimental.constrained.ceil llvm.experimental.constrained.cosh llvm.experimental.constrained.cos llvm.experimental.constrained.exp2 llvm.experimental.constrained.exp llvm.experimental.constrained.fadd llvm.experimental.constrained.fcmp llvm.experimental.constrained.fcmp llvm.experimental.constrained.fdiv llvm.experimental.constrained.floor llvm.experimental.constrained.fma llvm.experimental.constrained.fmuladd llvm.experimental.constrained.fmul llvm.experimental.constrained.fpext llvm.experimental.constrained.fptosi llvm.experimental.constrained.fptoui llvm.experimental.constrained.fptrunc llvm.experimental.constrained.frem llvm.experimental.constrained.fsub llvm.experimental.constrained.ldexp llvm.experimental.constrained.llrint llvm.experimental.constrained.llround llvm.experimental.constrained.log10 llvm.experimental.constrained.log2 llvm.experimental.constrained.log llvm.experimental.constrained.lrint llvm.experimental.constrained.lround llvm.experimental.constrained.maximum llvm.experimental.constrained.maxnum llvm.experimental.constrained.minimum llvm.experimental.constrained.minnum llvm.experimental.constrained.nearbyint llvm.experimental.constrained.powi llvm.experimental.constrained.pow llvm.experimental.constrained.rint llvm.experimental.constrained.roundeven llvm.experimental.constrained.round llvm.experimental.constrained.sinh llvm.experimental.constrained.sin llvm.experimental.constrained.sitofp llvm.experimental.constrained.sqrt llvm.experimental.constrained.tanh llvm.experimental.constrained.tan llvm.experimental.constrained.trunc llvm.experimental.constrained.uitofp ``` There are double. float and fp16 types that are of interest, both soft-fp and hard. I hope the others are handled automatically. I'm not sure what happens with vectors, both neon and mve. The frontend intrinsics I think should be considered as being in the current fp environment, strict-fp and all. Some do not do denormal flushing or set flags though. https://github.com/llvm/llvm-project/pull/137101 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
