[llvm] e4d1d0c - [SCEV] Fix isImpliedViaMerge() with values from previous iteration (PR56242)

[Nikita Popov via llvm-commits]

Author: Nikita Popov
Date: 2022-07-05T15:31:23+02:00
New Revision: e4d1d0cc2c9ca38f98bc9b70c3e3db3a18f1e06e

URL: https://github.com/llvm/llvm-project/commit/e4d1d0cc2c9ca38f98bc9b70c3e3db3a18f1e06e
DIFF: https://github.com/llvm/llvm-project/commit/e4d1d0cc2c9ca38f98bc9b70c3e3db3a18f1e06e.diff

LOG: [SCEV] Fix isImpliedViaMerge() with values from previous iteration (PR56242)

When trying to prove an implied condition on a phi by proving it
for all incoming values, we need to be careful about values coming
from a backedge, as these may refer to a previous loop iteration.
A variant of this issue was fixed in D101829, but the dominance
condition used there isn't quite right: It checks that the value
dominates the incoming block, which doesn't exclude backedges
(values defined in a loop will usually dominate the loop latch,
which is the incoming block of the backedge).

Instead, we should be checking for domination of the phi block.
Any values defined inside the loop will not dominate the loop
header phi.

Fixes https://github.com/llvm/llvm-project/issues/56242.

Differential Revision: https://reviews.llvm.org/D128640

Added: 
    

Modified: 
    llvm/lib/Analysis/ScalarEvolution.cpp
    llvm/test/Transforms/IRCE/decrementing-loop.ll
    llvm/test/Transforms/IndVarSimplify/pr56242.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index aa349f9435ce9..968665c4d3876 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -11812,7 +11812,7 @@ bool ScalarEvolution::isImpliedViaMerge(ICmpInst::Predicate Pred,
       const SCEV *L = getSCEV(LPhi->getIncomingValueForBlock(IncBB));
       // Make sure L does not refer to a value from a potentially previous
       // iteration of a loop.
-      if (!properlyDominates(L, IncBB))
+      if (!properlyDominates(L, LBB))
         return false;
       if (!ProvedEasily(L, RHS))
         return false;

diff  --git a/llvm/test/Transforms/IRCE/decrementing-loop.ll b/llvm/test/Transforms/IRCE/decrementing-loop.ll
index 5ae4412b34bd7..462607162bcbe 100644
--- a/llvm/test/Transforms/IRCE/decrementing-loop.ll
+++ b/llvm/test/Transforms/IRCE/decrementing-loop.ll
@@ -210,17 +210,16 @@ exit:
   ret void
 }
 
-; TODO: we need to be more careful when trying to look through phi nodes in
-; cycles, because the condition to prove may reference the previous value of
-; the phi. So we currently fail to optimize this case.
 ; Check that we can figure out that IV is non-negative via implication through
 ; two Phi nodes, one being AddRec.
 define void @test_05(i32* %a, i32* %a_len_ptr, i1 %cond) {
 
 ; CHECK-LABEL: test_05
-; CHECK: entry:
-; CHECK:   br label %merge
-; CHECK-NOT: mainloop
+; CHECK:       mainloop:
+; CHECK-NEXT:    br label %loop
+; CHECK:       loop:
+; CHECK:         br i1 true, label %in.bounds, label %out.of.bounds
+; CHECK:       loop.preloop:
 
  entry:
   %len.a = load i32, i32* %a_len_ptr, !range !0

diff  --git a/llvm/test/Transforms/IndVarSimplify/pr56242.ll b/llvm/test/Transforms/IndVarSimplify/pr56242.ll
index 82e1d2252e760..6afed4177e1f7 100644
--- a/llvm/test/Transforms/IndVarSimplify/pr56242.ll
+++ b/llvm/test/Transforms/IndVarSimplify/pr56242.ll
@@ -9,12 +9,14 @@ define void @test(ptr %arr) {
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_INC:%.*]], [[LOOP_LATCH:%.*]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[PREV:%.*]] = phi i32 [ [[V:%.*]], [[LOOP_LATCH]] ], [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    [[PTR:%.*]] = getelementptr inbounds i32, ptr [[ARR:%.*]], i64 [[IV]]
-; CHECK-NEXT:    [[V:%.*]] = load i32, ptr [[PTR]], align 4
+; CHECK-NEXT:    [[V]] = load i32, ptr [[PTR]], align 4
 ; CHECK-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[V]], 0
 ; CHECK-NEXT:    br i1 [[CMP1]], label [[IF:%.*]], label [[LOOP_LATCH]]
 ; CHECK:       if:
-; CHECK-NEXT:    call void @use(i1 false)
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp slt i32 [[PREV]], 0
+; CHECK-NEXT:    call void @use(i1 [[CMP2]])
 ; CHECK-NEXT:    br label [[LOOP_LATCH]]
 ; CHECK:       loop.latch:
 ; CHECK-NEXT:    [[IV_INC]] = add nuw nsw i64 [[IV]], 1