[MLIR] Fix generic assembly syntax for ArrayAttr containing hex float

[joker-eph]

When a float attribute is printed with Hex, we should not elide the type because it is parsed back as i64 otherwise.

[llvmbot]

@llvm/pr-subscribers-mlir

@llvm/pr-subscribers-mlir-core

Author: Mehdi Amini (joker-eph)

Changes

When a float attribute is printed with Hex, we should not elide the type because it is parsed back as i64 otherwise.

---

Full diff: https://github.com/llvm/llvm-project/pull/94583.diff

2 Files Affected:

• (modified) mlir/lib/IR/AsmPrinter.cpp (+6-3)
• (modified) mlir/test/IR/array-of-attr.mlir (+3)

diff --git a/mlir/lib/IR/AsmPrinter.cpp b/mlir/lib/IR/AsmPrinter.cpp
index 29e36210f1270..ba47283e37bee 100644
--- a/mlir/lib/IR/AsmPrinter.cpp
+++ b/mlir/lib/IR/AsmPrinter.cpp
@@ -2061,7 +2061,7 @@ void AsmPrinter::Impl::printLocationInternal(LocationAttr loc, bool pretty,
 
 /// Print a floating point value in a way that the parser will be able to
 /// round-trip losslessly.
-static void printFloatValue(const APFloat &apValue, raw_ostream &os) {
+static void printFloatValue(const APFloat &apValue, raw_ostream &os, bool *printedHex = nullptr) {
   // We would like to output the FP constant value in exponential notation,
   // but we cannot do this if doing so will lose precision.  Check here to
   // make sure that we only output it in exponential format if we can parse
@@ -2102,6 +2102,8 @@ static void printFloatValue(const APFloat &apValue, raw_ostream &os) {
 
   // Print special values in hexadecimal format. The sign bit should be included
   // in the literal.
+  if (printedHex)
+    *printedHex = true;
   SmallVector<char, 16> str;
   APInt apInt = apValue.bitcastToAPInt();
   apInt.toString(str, /*Radix=*/16, /*Signed=*/false,
@@ -2275,10 +2277,11 @@ void AsmPrinter::Impl::printAttributeImpl(Attribute attr,
       return;
 
   } else if (auto floatAttr = llvm::dyn_cast<FloatAttr>(attr)) {
-    printFloatValue(floatAttr.getValue(), os);
+    bool printedHex = false;
+    printFloatValue(floatAttr.getValue(), os, &printedHex);
 
     // FloatAttr elides the type if F64.
-    if (typeElision == AttrTypeElision::May && floatAttr.getType().isF64())
+    if (typeElision == AttrTypeElision::May && floatAttr.getType().isF64() && !printedHex)
       return;
 
   } else if (auto strAttr = llvm::dyn_cast<StringAttr>(attr)) {
diff --git a/mlir/test/IR/array-of-attr.mlir b/mlir/test/IR/array-of-attr.mlir
index 1b6fe55205959..c64a02f230372 100644
--- a/mlir/test/IR/array-of-attr.mlir
+++ b/mlir/test/IR/array-of-attr.mlir
@@ -12,3 +12,6 @@ test.array_of_attr_op
 // CHECK: test.array_of_attr_op
 // CHECK-SAME: a = [], b = [], c = []
 test.array_of_attr_op a = [], b = [], c = []
+
+// CHECK-SAME: 1.000000e+00 : f32, 1.000000e+00, 0x7FF0000000000000 : f64
+"test.test"() {test.float_arr = [1.0 : f32, 1.0 : f64, 0x7FF0000000000000 : f64]} : () -> ()

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.