[clang] [llvm] [Clang][AMDGPU] Add __builtin_amdgcn_cvt_off_f32_i4 (PR #133741)

[Juan Manuel Martinez Caamaño via llvm-commits]

================
@@ -729,6 +729,25 @@ GCNTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
 
     break;
   }
+  case Intrinsic::amdgcn_cvt_off_f32_i4: {
+    Value* Arg = II.getArgOperand(0);
+    Type *Ty = II.getType();
+
+    if (isa<PoisonValue>(Arg))
+      return IC.replaceInstUsesWith(II, PoisonValue::get(Ty));
+
+    if(IC.getSimplifyQuery().isUndefValue(Arg))
+      return IC.replaceInstUsesWith(II, Constant::getNullValue(Ty));
+
+    ConstantInt *CArg = dyn_cast<ConstantInt>(II.getArgOperand(0));
+    if (!CArg)
+      break;
+
+    int CArg4BitAsInt = CArg->getValue().trunc(4).getSExtValue();
+    float ResVal = 0.0625 * CArg4BitAsInt;
----------------
jmmartinez wrote:

I tabulated the results to avoid any fp operations.

https://github.com/llvm/llvm-project/pull/133741