[llvm] [FuncSpec] Only compute Latency bonus when necessary (PR #113159)

[Hari Limaye via llvm-commits]

================
@@ -875,24 +885,23 @@ bool FunctionSpecializer::findSpecializations(Function *F, unsigned FuncSize,
       AllSpecs[Index].CallSites.push_back(&CS);
     } else {
       // Calculate the specialisation gain.
-      Bonus B;
+      Cost CodeSize;
       unsigned Score = 0;
       InstCostVisitor Visitor = getInstCostVisitorFor(F);
       for (ArgInfo &A : S.Args) {
-        B += Visitor.getSpecializationBonus(A.Formal, A.Actual);
+        CodeSize += Visitor.getCodeSizeBonus(A.Formal, A.Actual);
         Score += getInliningBonus(A.Formal, A.Actual);
       }
-      B += Visitor.getBonusFromPendingPHIs();
-
+      CodeSize += Visitor.getCodeSizeBonusFromPendingPHIs();
 
       LLVM_DEBUG(dbgs() << "FnSpecialization: Specialization bonus {CodeSize = "
-                        << B.CodeSize << ", Latency = " << B.Latency
-                        << ", Inlining = " << Score << "}\n");
+                        << CodeSize << ", Inlining = " << Score << "}\n");
 
+      Bonus B = {CodeSize, 0};
----------------
hazzlim wrote:

Fair point - I was keeping it as it seemed like a reasonable way to abstract away the getValue() logic and to pass to the lambda, but I see it's probably cleaner without. Done.

https://github.com/llvm/llvm-project/pull/113159