[Mlir-commits] [mlir] Introduce `arith.scaling_extf` and `arith.scaling_truncf` (PR #141965)
Krzysztof Drewniak
llvmlistbot at llvm.org
Thu May 29 10:29:33 PDT 2025
================
@@ -409,6 +421,112 @@ struct F8E8M0TruncFOpConverter : public OpRewritePattern<arith::TruncFOp> {
}
};
+struct ScalingExtFOpConverter : public OpRewritePattern<arith::ScalingExtFOp> {
+ using OpRewritePattern::OpRewritePattern;
+ LogicalResult matchAndRewrite(arith::ScalingExtFOp op,
+ PatternRewriter &rewriter) const final {
+ ImplicitLocOpBuilder b(op.getLoc(), rewriter);
+ auto inputOperand = op.getIn();
+ auto scaleOperand = op.getScale();
+ if (!llvm::isa<Float8E8M0FNUType>(getElementTypeOrSelf(scaleOperand))) {
----------------
krzysz00 wrote:
Can we make this "truncate to f8E8M0" instead?
https://github.com/llvm/llvm-project/pull/141965
More information about the Mlir-commits
mailing list