[LLVMbugs] [Bug 1141] NEW: Instcombine makes wrong array size calculation

bugzilla-daemon at cs.uiuc.edu bugzilla-daemon at cs.uiuc.edu
Sun Jan 28 15:35:13 PST 2007


           Summary: Instcombine makes wrong array size calculation
           Product: libraries
           Version: trunk
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: rspencer at x10sys.com

It seems the recent changes to TargetData have affected the computation of array
sizes in PromoteCastOfAllocation. This is the problem that is killing the recent
nightly tests.  I'm filing this because I didn't have time to sufficiently debug
it. The error is somewhere in this code:

  // Get the type really allocated and the type casted to.
  const Type *AllocElTy = AI.getAllocatedType();
  const Type *CastElTy = PTy->getElementType();
  if (!AllocElTy->isSized() || !CastElTy->isSized()) return 0;

  unsigned AllocElTyAlign = TD->getTypeAlignmentABI(AllocElTy);
  unsigned CastElTyAlign = TD->getTypeAlignmentABI(CastElTy);
  if (CastElTyAlign < AllocElTyAlign) return 0;

  // If the allocation has multiple uses, only promote it if we are strictly
  // increasing the alignment of the resultant allocation.  If we keep it the
  // same, we open the door to infinite loops of various kinds.
  if (!AI.hasOneUse() && CastElTyAlign == AllocElTyAlign) return 0;

  uint64_t AllocElTySize = TD->getTypeSize(AllocElTy);
  uint64_t CastElTySize = TD->getTypeSize(CastElTy);
  if (CastElTySize == 0 || AllocElTySize == 0) return 0;

  // See if we can satisfy the modulus by pulling a scale out of the array
  // size argument.
  unsigned ArraySizeScale, ArrayOffset;
  Value *NumElements = // See if the array size is a decomposable linear expr.
    DecomposeSimpleLinearExpr(AI.getOperand(0), ArraySizeScale, ArrayOffset);

  // If we can now satisfy the modulus, by using a non-1 scale, we really can
  // do the xform.
  if ((AllocElTySize*ArraySizeScale) % CastElTySize != 0 ||
      (AllocElTySize*ArrayOffset   ) % CastElTySize != 0) return 0;

  unsigned Scale = (AllocElTySize*ArraySizeScale)/CastElTySize;

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

More information about the llvm-bugs mailing list