[PATCH] D69281: [FPEnv][WIP] Constrained FCmp intrinsics
Ulrich Weigand via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 12 09:23:10 PST 2019
uweigand marked an inline comment as done.
uweigand added inline comments.
================
Comment at: llvm/include/llvm/IR/Intrinsics.td:750-791
+ def int_experimental_constrained_fcmpoeq
+ : Intrinsic<[ LLVMScalarOrSameVectorWidth<0, llvm_i1_ty> ],
+ [ llvm_anyfloat_ty, LLVMMatchType<0>, llvm_metadata_ty ]>;
+ def int_experimental_constrained_fcmpogt
+ : Intrinsic<[ LLVMScalarOrSameVectorWidth<0, llvm_i1_ty> ],
+ [ llvm_anyfloat_ty, LLVMMatchType<0>, llvm_metadata_ty ]>;
+ def int_experimental_constrained_fcmpoge
----------------
simoll wrote:
> Out of curiosity: what is your motivation to have one intrinsic per fcmp predicate instead of, say, an `i8 immarg`?
I didn't really make the decision, just took it over from the original patch. But I agree it seems to make sense simply from a perspective of being explicit in the IR: if there's just a random numeric value, it would make the IR harder to read/write.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D69281/new/
https://reviews.llvm.org/D69281
More information about the llvm-commits
mailing list