[SCEV] Fix the movement of insertion point in expander. PR35406.

[skatkov]

Summary

We cannot move the insertion point to header if SCEV contains div/rem
operations due to they may go over check for zero denominator.

[sebpop]

LGTM

lib/Analysis/ScalarEvolutionExpander.cpp

1739 ↗

(On Diff #126923)

or SCEVSDivExpr.

[skatkov]

lib/Analysis/ScalarEvolutionExpander.cpp

1739 ↗

(On Diff #126923)

There is no SCEVSDivExpr expression.

llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp

if (SE.isLoopInvariant(S, L)) {
         InsertPt = Preheader->getTerminator();
       else {
         // LSR sets the insertion point for AddRec start/step values to the
         // block start to simplify value reuse, even though it's an invalid
         // position. SCEVExpander must correct for this in all cases.
         InsertPt = &*L->getHeader()->getFirstInsertionPt();
       }
     } else {
+      // We can move insertion point only if there is no div or rem operations
+      // otherwise we are risky to move it over the check for zero denominator.
+      auto SafeToHoist = [](const SCEV *S) {
+        return !SCEVExprContains(S, [](const SCEV *S) {
+                  if (const auto *D = dyn_cast<SCEVUDivExpr>(S)) {
+                    if (const auto *SC = dyn_cast<SCEVConstant>(D->getRHS()))
+                      // Division by non-zero constants can be hoisted.
+                      return SC->getValue()->isZero();
+                    // All other divisions should not be moved as they may be
+                    // divisions by zero and should be kept within the
+                    // conditions of the surrounding loops that guard their
+                    // execution (see PR35406).
+                    return true;
+                  }
+                  return false;
+                });
+      };
       // If the SCEV is computable at this level, insert it into the header
       // after the PHIs (and after any other instructions that we've inserted
       // there) so that it is guaranteed to dominate any user inside the loop.
-      if (L && SE.hasComputableLoopEvolution(S, L) && !PostIncLoops.count(L))
+      if (L && SE.hasComputableLoopEvolution(S, L) && !PostIncLoops.count(L) &&
+          SafeToHoist(S))
         InsertPt = &*L->getHeader()->getFirstInsertionPt();
       while (InsertPt->getIterator() != Builder.GetInsertPoint() &&
              (isInsertedInstruction(InsertPt) ||
               isa<DbgInfoIntrinsic>(InsertPt))) {
         InsertPt = &*std::next(InsertPt->getIterator());
       }
       break;
     }

llvm/trunk/test/Transforms/IndVarSimplify/pr35406.ll

+; RUN: opt -S -indvars %s | FileCheck %s
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128-ni:1"
+target triple = "x86_64-unknown-linux-gnu"
+
+define i32 @testDiv(i8* %p, i64* %p1) {
+; CHECK-LABEL: @testDiv
+entry:
+  br label %loop1
+
+loop1:
+  %local_0_ = phi i32 [ 8, %entry ], [ %9, %loop2.exit ]
+  %local_2_ = phi i32 [ 63864, %entry ], [ %local_2_43, %loop2.exit ]
+  %local_3_ = phi i32 [ 51, %entry ], [ %local_3_44, %loop2.exit ]
+; CHECK-NOT:  udiv
+  %0 = udiv i32 14, %local_0_
+  %1 = icmp ugt i32 %local_0_, 14
+  br i1 %1, label %exit, label %general_case24
+
+; CHECK-LABEL: general_case24
+general_case24:
+  %2 = udiv i32 60392, %0
+  br i1 false, label %loop2, label %loop2.exit
+
+loop2:
+  %local_1_56 = phi i32 [ %2, %general_case24 ], [ %3, %loop2 ]
+  %local_2_57 = phi i32 [ 1, %general_case24 ], [ %7, %loop2 ]
+  %3 = add i32 %local_1_56, -1
+  %4 = load atomic i64, i64* %p1 unordered, align 8
+  %5 = sext i32 %3 to i64
+  %6 = sub i64 %4, %5
+  store atomic i64 %6, i64* %p1 unordered, align 8
+  %7 = add nuw nsw i32 %local_2_57, 1
+  %8 = icmp ugt i32 %local_2_57, 7
+  br i1 %8, label %loop2.exit, label %loop2
+
+loop2.exit:
+  %local_2_43 = phi i32 [ %local_2_, %general_case24 ], [ 9, %loop2 ]
+  %local_3_44 = phi i32 [ %local_3_, %general_case24 ], [ %local_1_56, %loop2 ]
+  %9 = add nuw nsw i32 %local_0_, 1
+  %10 = icmp ugt i32 %local_0_, 129
+  br i1 %10, label %exit, label %loop1
+
+exit:
+  ret i32 0
+}
+
+define i32 @testRem(i8* %p, i64* %p1) {
+; CHECK-LABEL: @testRem
+entry:
+  br label %loop1
+
+loop1:
+  %local_0_ = phi i32 [ 8, %entry ], [ %9, %loop2.exit ]
+  %local_2_ = phi i32 [ 63864, %entry ], [ %local_2_43, %loop2.exit ]
+  %local_3_ = phi i32 [ 51, %entry ], [ %local_3_44, %loop2.exit ]
+; CHECK:  udiv
+; CHECK-NOT:  udiv
+  %0 = udiv i32 14, %local_0_
+  %1 = icmp ugt i32 %local_0_, 14
+  br i1 %1, label %exit, label %general_case24
+
+; CHECK-LABEL: general_case24
+general_case24:
+  %2 = urem i32 60392, %0
+  br i1 false, label %loop2, label %loop2.exit
+
+loop2:
+  %local_1_56 = phi i32 [ %2, %general_case24 ], [ %3, %loop2 ]
+  %local_2_57 = phi i32 [ 1, %general_case24 ], [ %7, %loop2 ]
+  %3 = add i32 %local_1_56, -1
+  %4 = load atomic i64, i64* %p1 unordered, align 8
+  %5 = sext i32 %3 to i64
+  %6 = sub i64 %4, %5
+  store atomic i64 %6, i64* %p1 unordered, align 8
+  %7 = add nuw nsw i32 %local_2_57, 1
+  %8 = icmp ugt i32 %local_2_57, 7
+  br i1 %8, label %loop2.exit, label %loop2
+
+loop2.exit:
+  %local_2_43 = phi i32 [ %local_2_, %general_case24 ], [ 9, %loop2 ]
+  %local_3_44 = phi i32 [ %local_3_, %general_case24 ], [ %local_1_56, %loop2 ]
+  %9 = add nuw nsw i32 %local_0_, 1
+  %10 = icmp ugt i32 %local_0_, 129
+  br i1 %10, label %exit, label %loop1
+
+exit:
+  ret i32 0
+}