[Mlir-commits] [mlir] [MLIR][ARITH] Adds missing foldings for truncf (PR #128096)

[Thomas Raoux]

================
@@ -1517,6 +1517,22 @@ LogicalResult arith::TruncIOp::verify() {
 /// Perform safe const propagation for truncf, i.e., only propagate if FP value
 /// can be represented without precision loss.
 OpFoldResult arith::TruncFOp::fold(FoldAdaptor adaptor) {
+  if (auto extOp = getOperand().getDefiningOp<arith::ExtFOp>()) {
----------------
ThomasRaoux wrote:

llvm does optimize y = trunc(ext(bf16 x to f32) to bf16)) on ToT: https://godbolt.org/z/Grbqo4v3o

https://github.com/llvm/llvm-project/pull/128096