[LICM] Small update to note changes made in hoistRegion

[andrew.w.kaylor]

Summary

I'm not sure how important this is, but I think the code was technically incorrect without it.

I was looking at LICM to see if I could improve compilation time. The easiest thing I found turned out to have been fixed last week with r290726. This one liner was all that was left of a patch I was working on. :-)

[mkuper]

Yes, this should definitely be there. LGTM.
(Out of curiosity - you mean r290726 actually has an observable effect on compile time?)

[andrew.w.kaylor]

In D28363#637147, @mkuper wrote:

(Out of curiosity - you mean r290726 actually has an observable effect on compile time?)

I hadn't gotten as far as taking measurements. I did find a file in a spec test where the LCSSA form was being unnecessarily recomputed 43 times, but it was probably relatively quick since nothing needed to change.

LICM was showing up as a time consuming pass on a couple of measurements that were sent my way, so my first step was just reading through the code and looking for anything obvious that could be fixed easily. The two things I saw were the issue you addressed in r290726 and the funclet coloring thing I e-mailed you about.

llvm/trunk/lib/Transforms/Scalar/LICM.cpp

for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;) {
               &I, I.getModule()->getDataLayout(), TLI)) {
         DEBUG(dbgs() << "LICM folding inst: " << I << "  --> " << *C << '\n');
         CurAST->copyValue(&I, C);
         I.replaceAllUsesWith(C);
         if (isInstructionTriviallyDead(&I, TLI)) {
           CurAST->deleteValue(&I);
           I.eraseFromParent();
         }
+        Changed = true;
         continue;
       }
 
       // Try hoisting the instruction out to the preheader.  We can only do this
       // if all of the operands of the instruction are loop invariant and if it
       // is safe to hoist the instruction.
       //
       if (CurLoop->hasLoopInvariantOperands(&I) &&