[compiler-rt] [AMDGPU] Change the representation of double literals in operands (PR #68740)

[Jay Foad via llvm-commits]

================
@@ -4309,7 +4312,18 @@ bool AMDGPUAsmParser::validateVOPLiteral(const MCInst &Inst,
       continue;
 
     if (MO.isImm() && !isInlineConstant(Inst, OpIdx)) {
-      uint32_t Value = static_cast<uint32_t>(MO.getImm());
+      uint64_t Value = static_cast<uint64_t>(MO.getImm());
+      bool IsFP = AMDGPU::isSISrcFPOperand(Desc, OpIdx);
+      bool IsValid32Op = AMDGPU::isValid32BitLiteral(Value, IsFP);
+
+      if (!IsValid32Op && !isInt<32>(Value) && !isUInt<32>(Value)) {
+        Error(getLitLoc(Operands), "invalid operand for instruction");
+        return false;
+      }
+
+      if (IsFP && IsValid32Op)
+        Value = Hi_32(Value);
----------------
jayfoad wrote:

> I also do not like changing isValid32BitLiteral's argument to IsFP64 as it brings even more context than it is needed.

The problem with calling it `IsFP` is that if I write this:
```
  bool IsFP = AMDGPU::isSISrcFPOperand(Desc, OpIdx);
  if (!AMDGPU::isValid32BitLiteral(Value, IsFP))
    Error(getLitLoc(Operands), "invalid operand for instruction");
```
then all 16-bit FP operands are rejected. (Maybe all 32-bit ones too, I didn't check.)

https://github.com/llvm/llvm-project/pull/68740