[Mlir-commits] [mlir] [mlir][emitc] Arith to EmitC: handle floating-point<->integer conversions (PR #87614)

[Simon Camphausen]

================
@@ -201,6 +201,96 @@ class SelectOpConversion : public OpConversionPattern<arith::SelectOp> {
   }
 };
 
+// Floating-point to integer conversions.
+template <typename CastOp>
+class FtoICastOpConversion : public OpConversionPattern<CastOp> {
+public:
+  FtoICastOpConversion(const TypeConverter &typeConverter, MLIRContext *context)
+      : OpConversionPattern<CastOp>(typeConverter, context) {}
+
+  LogicalResult
+  matchAndRewrite(CastOp castOp, typename CastOp::Adaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+
+    Type operandType = adaptor.getIn().getType();
+    if (!emitc::isSupportedFloatType(operandType))
+      return rewriter.notifyMatchFailure(castOp,
+                                         "unsupported cast source type");
+
+    Type dstType = this->getTypeConverter()->convertType(castOp.getType());
+    if (!dstType)
+      return rewriter.notifyMatchFailure(castOp, "type conversion failed");
+
+    // Float-to-bool casts are not supported: any non-zero value must be mapped
+    // to true, which does not hold with rounding for values <= 0.5.
----------------
simon-camp wrote:

I think the comment can be made a bit clearer in what the expected result in the arith dialect is and what the actual behavior in C is.

```suggestion
    // Float-to-i1 casts are not supported: any value with 0 < value < 1 must be
    // truncated to 0, whereas a boolean conversion would return true.
```

https://github.com/llvm/llvm-project/pull/87614