[llvm] [AMDGPU] Prevent folding of the negative i32 literals as i64 (PR #70274)

[Jay Foad via llvm-commits]

================
@@ -5497,9 +5497,18 @@ bool SIInstrInfo::isOperandLegal(const MachineInstr &MI, unsigned OpIdx,
                      OpInfo.OperandType == AMDGPU::OPERAND_REG_IMM_INT64 ||
                      OpInfo.OperandType == AMDGPU::OPERAND_REG_IMM_V2INT32 ||
                      OpInfo.OperandType == AMDGPU::OPERAND_REG_IMM_V2FP32;
-    if (Is64BitOp && !AMDGPU::isValid32BitLiteral(Imm, Is64BitFPOp) &&
-        !AMDGPU::isInlinableLiteral64(Imm, ST.hasInv2PiInlineImm()))
-      return false;
+    if (Is64BitOp &&
+        !AMDGPU::isInlinableLiteral64(Imm, ST.hasInv2PiInlineImm())) {
+      if (!AMDGPU::isValid32BitLiteral(Imm, Is64BitFPOp))
+        return false;
+
+      // FIXME: We can use sign extended 64-bit literals, but only for signed
+      //        operands. At the moment we do not know if an operand is signed.
+      //        Such operand will be encoded as its low 32 bits and then either
+      //        correctly sign extended or incorrectly zero extended by HW.
+      if (!Is64BitFPOp && (int32_t)Lo_32(Imm) < 0)
----------------
jayfoad wrote:

Nit: `(int32_t)Imm` is simpler.

https://github.com/llvm/llvm-project/pull/70274