[llvm-branch-commits] [llvm] [AMDGPU] Allow 16-bit imm folding in real true16 (PR #173318)

[Stanislav Mekhanoshin via llvm-branch-commits]

================
@@ -1335,6 +1336,15 @@ bool SIInstrInfo::getConstValDefinedInReg(const MachineInstr &MI,
 
     return false;
   }
+  case AMDGPU::V_MOV_B16_t16_e64: {
----------------
rampitec wrote:

I do not think we would do it for an immediate, but added check.

https://github.com/llvm/llvm-project/pull/173318