[PATCH] D30333: Split SimplifyCFG to run obscuring switch transforms only during last phase

[Joerg Sonnenberger via Phabricator via llvm-commits]

joerg marked 2 inline comments as done.
joerg added inline comments.


================
Comment at: lib/Transforms/IPO/PassManagerBuilder.cpp:592
     MPM.add(createLoopUnswitchPass(SizeLevel || OptLevel < 3));
     MPM.add(createCFGSimplificationPass());
     addInstructionCombiningPass(MPM);
----------------
mehdi_amini wrote:
> What about this one?
I don't have a strong opinion on whether it should be done before or after SLP handling. I don't think the code is likely to be vectorised.


================
Comment at: lib/Transforms/IPO/PassManagerBuilder.cpp:622
   addExtensionsToPM(EP_Peephole, MPM);
-  MPM.add(createCFGSimplificationPass());
+  MPM.add(createLateCFGSimplificationPass()); // Switches to lookup tables
   addInstructionCombiningPass(MPM);
----------------
mehdi_amini wrote:
> The comment is misleading, it seems that the only reason we run the pass here is for this purpose.
Well, we also want to do the usual transformations again after loop unrolling. But at this point, we are also commiting to the the lookup table conversions.


================
Comment at: lib/Transforms/Scalar/SimplifyCFGPass.cpp:256
+  }
+};
 }
----------------
mehdi_amini wrote:
> Why two passes instead of a parameter?
Two passes makes it easier to test it on the command line.


https://reviews.llvm.org/D30333