[Power9] Disable removing extra swaps on P9 since it should not be needed.

[stefanp]

Summary

On power 8 we sometimes insert swaps to deal with the difference between Little Endian and Big Endian. The swap removal pass is supposed to clean up these swaps. On power 9 we don't need this pass since we do not need to insert the swaps in the first place.

[nemanjai]

LGTM.

lib/Target/PowerPC/PPCTargetMachine.cpp

369

Please add a comment such as

// Power ISA 3.0 has load/store instructions that preserve
// element order so swaps aren't introduced. The pass to remove
// them is thereby not necessary.

[inouehrs]

lib/Target/PowerPC/PPCTargetMachine.cpp

371

I feel it is better to use Subtarget.needsSwapsForVSXMemOps() to make the meaning of the check clearer.

[nemanjai]

lib/Target/PowerPC/PPCTargetMachine.cpp

371

This is actually a really good point. I forgot that we added that function. Please use that instead.

[jtony]

lib/Target/PowerPC/PPCTargetMachine.cpp

371

I agree with Hiroshi, this is a good way to increase code readability.

[stefanp]

Sorry... We can't do what I wanted to do in the first place because I don't have access to the PPCSubtarget object in the PPCTargetMachine code.
I've moved the guard into the PPCVSXSwapRemoval code.

[echristo]

This is how you should do it. Though I'm not sure it isn't just better to check for "processor >= power9", but *shrug* for now.

I've also marked getSubtargetImpl() as delete so you won't be tempted again. :)

lib/Target/PowerPC/PPCTargetMachine.cpp

#endif
 
   addPass(createPPCVSXCopyPass());
   return false;
 }
 
 void PPCPassConfig::addMachineSSAOptimization() {
   TargetPassConfig::addMachineSSAOptimization();
   // For little endian, remove where possible the vector swap instructions
   // introduced at code generation to normalize vector element order.
   if (TM->getTargetTriple().getArch() == Triple::ppc64le &&
+      !TM->getSubtargetImpl()->isISA3_0() &&
       !DisableVSXSwapRemoval)
     addPass(createPPCVSXSwapRemovalPass());
   // Target-specific peephole cleanups performed after instruction
   // selection.
   if (!DisableMIPeephole) {
     addPass(createPPCMIPeepholePass());
     addPass(&DeadMachineInstructionElimID);
   }