Arith Dialect Operation Folding

MLIR
1

I have a question about canonicalization in Arith Dialect:
I think MLIR can only fold x + (-0.0) into x, as shown in figure below. Why MLIR cannot fold x + (+0.0) into x?
image
Similarly, why does MLIR only fold x - (+0.0) into x?

2

Blame layer for the code points to a discussion at ⚙ D118244 [MLIR][arith] More float op folders

Actually I think that %x + 0.0 can’t be folded to %x, but %x + -0.0 can!

That’s because -0. + 0. == +0.0.

Seems very relevant to [RFC] FastMath flags support in MLIR (arith dialect) - #2 by jfurtek - MLIR - LLVM Discussion Forums