In this method, we invoke `SimplifyICmpOperands` which takes the `Cond` predicate
by reference and may change it along with `LHS` and `RHS` SCEVs. But then we invoke
`computeShiftCompareExitLimit` with Values from which the SCEVs have been derived,
these Values have not been modified while `Cond` could be.
One of possible outcomes of this is that we may falsely prove that an infinite loop ends
within some finite number of iterations.
In this patch, we save the original `Cond` and pass it along with original operands.
This logic may be removed in future once `computeShiftCompareExitLimit` works
with SCEVs instead of value operands.
Reviewed By: sanjoy
Differential Revision: https://reviews.llvm.org/D40953
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@320142
91177308-0d34-0410-b5e6-
96231b3b80d8
Cond = ExitCond->getPredicate();
else
Cond = ExitCond->getInversePredicate();
+ const ICmpInst::Predicate OriginalCond = Cond;
// Handle common loops like: for (X = "string"; *X; ++X)
if (LoadInst *LI = dyn_cast<LoadInst>(ExitCond->getOperand(0)))
return ExhaustiveCount;
return computeShiftCompareExitLimit(ExitCond->getOperand(0),
- ExitCond->getOperand(1), L, Cond);
+ ExitCond->getOperand(1), L, OriginalCond);
}
ScalarEvolution::ExitLimit
ret void
}
+define void @test9() {
+; CHECK-LABEL: Determining loop execution counts for: @test9
+; CHECK: Loop %loop: Unpredictable max backedge-taken count.
+
+; This is an infinite loop, make sure that it recognized as such.
+
+entry:
+ br label %loop
+
+leave:
+ ret void
+
+loop:
+ %iv = phi i32 [ -20, %entry ], [ %iv.shift, %loop ]
+ %iv.shift = ashr i32 %iv, 1
+ %exit.cond = icmp sgt i32 %iv, -1
+ br i1 %exit.cond, label %leave, label %loop
+}
+
!0 = !{i32 0, i32 50000}
!1 = !{i32 -5000, i32 -1}