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

[Krzysztof Drewniak]

================
@@ -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>()) {
----------------
krzysz00 wrote:

Instruction availability. There's no physical "truncate f16 to f8" function, and folding away the rewrite that works around this would be bad

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