The Output Type of FPToUIOp and FPToSIOp

qliu93 · November 9, 2020, 11:36am

Hi

I noticed that FPToUIOp::areCastCompatible requires that the output type isSignlessInteger, FPToSIOp the same. This is kind of confusing. What is a “signless integer” and why are signed integer and unsigned integer not allowed for FPToUIOp and FPToSIOp?

github.com

llvm/llvm-project/blob/master/mlir/lib/Dialect/StandardOps/IR/Ops.cpp#L2145


  if (a.isa<FloatType>() && b.isSignlessInteger())
    return true;
  return areVectorCastSimpleCompatible(a, b, areCastCompatible);
}
//===----------------------------------------------------------------------===//
// FPToUIOp
//===----------------------------------------------------------------------===//
bool FPToUIOp::areCastCompatible(Type a, Type b) {
  if (a.isa<FloatType>() && b.isSignlessInteger())
    return true;
  return areVectorCastSimpleCompatible(a, b, areCastCompatible);
}
//===----------------------------------------------------------------------===//
// FPTruncOp
//===----------------------------------------------------------------------===//
bool FPTruncOp::areCastCompatible(Type a, Type b) {
  if (auto fa = a.dyn_cast<FloatType>())

Thanks

qliu93 · November 10, 2020, 2:22am

It seems that these two functions should be implemented like this:

bool FPToUIOp::areCastCompatible(Type a, Type b) {
  if (a.isa<FloatType>() && b.isUnsignedInteger())
    return true;
  return areVectorCastSimpleCompatible(a, b, areCastCompatible);
}

bool FPToSIOp::areCastCompatible(Type a, Type b) {
  if (a.isa<FloatType>() && b.isSignedInteger())
    return true;
  return areVectorCastSimpleCompatible(a, b, areCastCompatible);
}

Looking forward to replies

jpienaar · November 10, 2020, 3:58am

Have you see What are Signless Integers? ?