[llvm-bugs] [Bug 45290] New: Infinite loop in SplitCriticalSideEffectEdges

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=45290

            Bug ID: 45290
           Summary: Infinite loop in SplitCriticalSideEffectEdges
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Common Code Generator Code
          Assignee: unassignedbugs at nondot.org
          Reporter: mikael.holmen at ericsson.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 23276
  --> https://bugs.llvm.org/attachment.cgi?id=23276&action=edit
bbi-40381_x86.ll reproducer

Reproduce with:
 llc -O0 -mtriple=x86_64-unknown-linux-gnu -o - bbi-40381_x86.ll

With -debug -debug-Pass=Executions the last thing we see is

[2020-03-24 13:58:41.006703848] 0x60aee00     Executing Pass 'X86 DAG->DAG
Instruction Selection' on Function 'f26'...



=== f26

Running in the debugger we see that we're stuck in the loop in
SplitCriticalSideEffectEdges in SelectionDAGISel.cpp:

static void SplitCriticalSideEffectEdges(Function &Fn, DominatorTree *DT,
                                         LoopInfo *LI) {
  // Loop for blocks with phi nodes.
  for (BasicBlock &BB : Fn) {
    PHINode *PN = dyn_cast<PHINode>(BB.begin());
    if (!PN) continue;

  ReprocessBlock:
    // For each block with a PHI node, check to see if any of the input values
    // are potentially trapping constant expressions.  Constant expressions are
    // the only potentially trapping value that can occur as the argument to a
    // PHI.
    for (BasicBlock::iterator I = BB.begin(); (PN = dyn_cast<PHINode>(I)); ++I)
      for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
        ConstantExpr *CE = dyn_cast<ConstantExpr>(PN->getIncomingValue(i));
        if (!CE || !CE->canTrap()) continue;

        // The only case we have to worry about is when the edge is critical.
        // Since this block has a PHI Node, we assume it has multiple input
        // edges: check to see if the pred has multiple successors.
        BasicBlock *Pred = PN->getIncomingBlock(i);
        if (Pred->getTerminator()->getNumSuccessors() == 1)
          continue;

        // Okay, we have to split this edge.
        SplitCriticalEdge(
            Pred->getTerminator(), GetSuccessorNumber(Pred, &BB),
            CriticalEdgeSplittingOptions(DT, LI).setMergeIdenticalEdges());
        goto ReprocessBlock;
      }
  }
}

The problem in this case is because BB has only one predecessor, so
SplitCriticalEdge bails out early and does nothing and we're stuck.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200324/6211ece1/attachment.html>