[PowerPC] Assume 128bit multiply uses CTR

[TimNN]

Summary

This fixes https://bugs.llvm.org//show_bug.cgi?id=32485

[hfinkel]

A somewhat minified example IR which reproduces the assertion is attached to the linked bug report -- here I would also appreciate some guidance on in what form / where this can be integrated in the regression tests.

Just running llc on the resulting IR in the usual way should be fine.

lib/Target/PowerPC/PPCCTRLoops.cpp

250

It looks like we can have a more-targeted fix than this. The problem seems to be restricted to multiply. See below.

317

Here add:

case Intrinsic::umul_with_overflow: Opcode = ISD::UMULO; break;
case Intrinsic::smul_with_overflow: Opcode = ISD::SMULO; break;

and I think that should take care of it.

[TimNN]

Updated to a more targeted fix and added a test as suggested.

[TimNN]

@hfinkel: Thanks for the feedback!

I have made the changes recommended and uploaded a new revision. (I though this would trigger another mail but apparently it didn't, so I'm commenting again).

[hfinkel]

In D31790#722438, @TimNN wrote:

@hfinkel: Thanks for the feedback!

I have made the changes recommended and uploaded a new revision. (I though this would trigger another mail but apparently it didn't, so I'm commenting again).

Phabricator e-mails were broken for a while.

LGTM.

[TimNN]

I cannot commit this myself, could someone please do it for me?

lib/Target/PowerPC/PPCCTRLoops.cpp

if (CallInst *CI = dyn_cast<CallInst>(J)) {
       const TargetLowering *TLI =
           TM->getSubtargetImpl(*BB->getParent())->getTargetLowering();
 
       if (Function *F = CI->getCalledFunction()) {
         // Most intrinsics don't become function calls, but some might.
         // sin, cos, exp and log are always calls.
         unsigned Opcode = 0;
         if (F->getIntrinsicID() != Intrinsic::not_intrinsic) {
+          // Preemtively assume intrinsics operating on 128 integers use CTR
+          for (unsigned OID = 0; OID < CI->getNumArgOperands(); OID++) {
+            if (isLargeIntegerTy(false, CI->getArgOperand(OID)->getType()->getScalarType())) {
+              return true;
+            }
+          }
+
           switch (F->getIntrinsicID()) {
           default: continue;
           // If we have a call to ppc_is_decremented_ctr_nonzero, or ppc_mtctr
           // we're definitely using CTR.
           case Intrinsic::ppc_is_decremented_ctr_nonzero:
           case Intrinsic::ppc_mtctr:
             return true;
 
#endif
           case Intrinsic::floor:     Opcode = ISD::FFLOOR;     break;
           case Intrinsic::ceil:      Opcode = ISD::FCEIL;      break;
           case Intrinsic::trunc:     Opcode = ISD::FTRUNC;     break;
           case Intrinsic::rint:      Opcode = ISD::FRINT;      break;
           case Intrinsic::nearbyint: Opcode = ISD::FNEARBYINT; break;
           case Intrinsic::round:     Opcode = ISD::FROUND;     break;
           case Intrinsic::minnum:    Opcode = ISD::FMINNUM;    break;
           case Intrinsic::maxnum:    Opcode = ISD::FMAXNUM;    break;
           }
         }
 
         // PowerPC does not use [US]DIVREM or other library calls for
         // operations on regular types which are not otherwise library calls
         // (i.e. soft float or atomics). If adapting for targets that do,
         // additional care is required here.
 
         LibFunc::Func Func;