[llvm] [AMDGPU] Account for inline asm size in inst_pref_size calculation (PR #192306)

[Jay Foad via llvm-commits]

================
@@ -218,6 +191,22 @@ bool AMDGPUMCExpr::evaluateOccupancy(MCValue &Res,
   return true;
 }
 
+bool AMDGPUMCExpr::evaluateInstPrefSize(MCValue &Res,
+                                        const MCAssembler *Asm) const {
+  assert(Args.size() == 3 &&
+         "AMDGPUMCExpr Argument count incorrect for InstPrefSize");
+
+  SmallVector<uint64_t, 3> Vals;
+  if (!evaluateMCExprs(Args, Asm, Vals))
+    return false;
+  uint64_t CodeSizeInBytes = Vals[0], FieldWidth = Vals[1],
+           CacheLineSize = Vals[2];
----------------
jayfoad wrote:

I don't understand why FieldWidth and CacheLineSize come in via Args. Aren't they just known constants that we can get frmo the Subtarget here?

https://github.com/llvm/llvm-project/pull/192306