[Metarenamer] Distinguish instructions by opcodes

[mkazantsev]

Summary

This should improve readability of metarenamed code.

[nikic]

llvm/lib/Transforms/Utils/MetaRenamer.cpp

119

What do you think about using I.getOpcodeName() in general here?

[aeubanks]

this deserves a test

[mkazantsev]

llvm/lib/Transforms/Utils/MetaRenamer.cpp

119

Interesting idea.

[mkazantsev]

In D145896#4189775, @aeubanks wrote:

this deserves a test

Ok, will add.

[mkazantsev]

Distinguished all instructions as Nikita has proposed, added test.

[nikic]

LGTM

llvm/test/Transforms/MetaRenamer/opcodes.ll

16 ↗

(On Diff #505360)

In the future, we might want to special-case some of these to more typical shorter names, e.g. getelementptr -> gep.

[mkazantsev]

llvm/test/Transforms/MetaRenamer/opcodes.ll

16 ↗

(On Diff #505360)

I was also thinking about block differentiantion basing on their role, e.g. entry, header, preheader etc. But let's move in small steps.

llvm/lib/Transforms/Utils/MetaRenamer.cpp

 #include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/IR/Argument.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/DerivedTypes.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/GlobalAlias.h"
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/Instruction.h"
+#include "llvm/IR/Instructions.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/PassManager.h"
 #include "llvm/IR/Type.h"
 #include "llvm/IR/TypeFinder.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Transforms/Utils.h"
for (Argument &Arg : F.args())
     if (!Arg.getType()->isVoidTy())
       Arg.setName("arg");
 
   for (auto &BB : F) {
     BB.setName("bb");
 
     for (auto &I : BB)
       if (!I.getType()->isVoidTy())
-        I.setName("inst");
+        I.setName(isa<PHINode>(I) ? "phi" : "inst");
   }
 }
 
 void MetaRename(Module &M,
                 function_ref<TargetLibraryInfo &(Function &)> GetTLI) {
   // Seed our PRNG with simple additive sum of ModuleID. We're looking to
   // simply avoid always having the same function names, and we need to
   // remain deterministic.