================
@@ -1472,6 +1472,60 @@ Intrinsic::ID
Intrinsic::getDeinterleaveIntrinsicID(unsigned Factor) {
return InterleaveIntrinsics[Factor - 2].Deinterleave;
}
+LLVM_ABI void Intrinsic::printFPClassMask(raw_ostream &OS,
+ const Constant *ImmArgVal) {
+ if (const auto *CI = dyn_cast<ConstantInt>(ImmArgVal)) {
+ uint64_t Val = CI->getZExtValue();
+ if (Val <= fcAllFlags) {
+ SmallVector<StringRef, 5> Classes;
+ if (Val & fcZero) {
+ if ((Val & fcZero) == fcZero)
+ Classes.push_back("zero");
+ else if (Val & fcPosZero)
+ Classes.push_back("+zero");
+ else
+ Classes.push_back("-zero");
+ }
+ if (Val & fcSubnormal) {
+ if ((Val & fcSubnormal) == fcSubnormal)
+ Classes.push_back("subnormal");
+ else if (Val & fcPosSubnormal)
+ Classes.push_back("+subnormal");
+ else
+ Classes.push_back("-subnormal");
+ }
+ if (Val & fcNormal) {
+ if ((Val & fcNormal) == fcNormal)
+ Classes.push_back("normal");
+ else if (Val & fcPosNormal)
+ Classes.push_back("+normal");
+ else
+ Classes.push_back("-normal");
+ }
+ if (Val & fcInf) {
+ if ((Val & fcInf) == fcInf)
+ Classes.push_back("inf");
+ else if (Val & fcPosInf)
+ Classes.push_back("+inf");
+ else
+ Classes.push_back("-inf");
+ }
+ if (Val & fcNan) {
+ if ((Val & fcNan) == fcNan)
+ Classes.push_back("nan");
+ else if (Val & fcQNan)
+ Classes.push_back("qnan");
+ else
+ Classes.push_back("snan");
+ }
+ OS << llvm::join(Classes, "|");
----------------
spavloff wrote:
May be comma could be used for better reading? `zero,sub,norm` seems better
than `zero|sub|norm` or `zero sub norm`.
https://github.com/llvm/llvm-project/pull/207653
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
