From c7eb5df9be401ee84f6ce04691f72e9ff97a2075 Mon Sep 17 00:00:00 2001
From: Michael Kuperstein <mkuper@google.com>
Date: Tue, 10 Jan 2017 19:32:30 +0000
Subject: [PATCH] [LV] Don't panic when encountering the IV of an outer loop.

Bail out instead of asserting when we encounter this situation,
which can actually happen.

The reason the test uses the new PM is that the "bad" phi, incidentally, gets
cleaned up by LoopSimplify. But LICM can create this kind of phi and preserve
loop simplify form, so the cleanup has no chance to run.

This fixes PR31190.
We may want to solve this in a less conservative manner, since this phi is
actually uniform within the inner loop (or we may want LICM to output a cleaner
promotion to begin with).

Differential Revision: https://reviews.llvm.org/D28490


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@291589 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Utils/LoopUtils.cpp       |  9 +++-
 test/Transforms/LoopVectorize/pr31190.ll | 64 ++++++++++++++++++++++++
 2 files changed, 71 insertions(+), 2 deletions(-)
 create mode 100644 test/Transforms/LoopVectorize/pr31190.ll

diff --git a/lib/Transforms/Utils/LoopUtils.cpp b/lib/Transforms/Utils/LoopUtils.cpp
index 09e9f1ddc7f..c8efa9efc7f 100644
--- a/lib/Transforms/Utils/LoopUtils.cpp
+++ b/lib/Transforms/Utils/LoopUtils.cpp
@@ -869,8 +869,13 @@ bool InductionDescriptor::isInductionPHI(PHINode *Phi, const Loop *TheLoop,
     return false;
   }
 
-  assert(TheLoop->getHeader() == Phi->getParent() &&
-         "PHI is an AddRec for a different loop?!");
+  if (AR->getLoop() != TheLoop) {
+    // FIXME: We should treat this as a uniform. Unfortunately, we
+    // don't currently know how to handled uniform PHIs.
+    DEBUG(dbgs() << "LV: PHI is a recurrence with respect to an outer loop.\n");
+    return false;    
+  }
+
   Value *StartValue =
     Phi->getIncomingValueForBlock(AR->getLoop()->getLoopPreheader());
   const SCEV *Step = AR->getStepRecurrence(*SE);
diff --git a/test/Transforms/LoopVectorize/pr31190.ll b/test/Transforms/LoopVectorize/pr31190.ll
new file mode 100644
index 00000000000..afb1754983c
--- /dev/null
+++ b/test/Transforms/LoopVectorize/pr31190.ll
@@ -0,0 +1,64 @@
+; RUN: opt -passes='loop-vectorize' -debug -S < %s 2>&1 | FileCheck %s
+; REQUIRES: asserts
+
+; This checks we don't crash when the inner loop we're trying to vectorize
+; is a SCEV AddRec with respect to an outer loop.
+
+; In this case, the problematic PHI is:
+; %0 = phi i32 [ undef, %for.cond1.preheader ], [ %inc54, %for.body3 ]
+; Since %inc54 is the IV of the outer loop, and %0 equivalent to it,
+; we get the situation described above.
+
+; This test uses the new PM, because with the old PM, running loop-vectorize
+; would explicitly run loop-simplify. Even though this loop is already in
+; simplified form, loop-simplify would still clean up the phi.
+; The reason this matters is that in a real optimizer pipeline, LICM can create
+; such PHIs, and since it preserves loop simplified form, the cleanup has
+; no chance to run.
+
+; Code that leads to this situation can look something like:
+;
+; int a, b[1], c;
+; void fn1 ()
+; {
+;  for (; c; c++)
+;    for (a = 0; a; a++)
+;      b[c] = 4;
+; }
+;
+; The PHI is an artifact of the register promotion of c.
+
+@c = external global i32, align 4
+@a = external global i32, align 4
+@b = external global [1 x i32], align 4
+
+; CHECK: LV: PHI is a recurrence with respect to an outer loop.
+; CHECK: LV: Not vectorizing: Cannot prove legality.
+; CHECK-LABEL: @test
+define void @test() {
+entry:
+  %a.promoted2 = load i32, i32* @a, align 1
+  %c.promoted = load i32, i32* @c, align 1
+  br label %for.cond1.preheader
+
+for.cond1.preheader:                              ; preds = %for.cond1.for.inc4_crit_edge, %entry
+  %inc54 = phi i32 [ %inc5, %for.cond1.for.inc4_crit_edge ], [ %c.promoted, %entry ]
+  %inc.lcssa3 = phi i32 [ %inc.lcssa, %for.cond1.for.inc4_crit_edge ], [ %a.promoted2, %entry ]
+  br label %for.body3
+
+for.body3:                                        ; preds = %for.body3, %for.cond1.preheader
+  %inc1 = phi i32 [ %inc.lcssa3, %for.cond1.preheader ], [ %inc, %for.body3 ]
+  %0 = phi i32 [ undef, %for.cond1.preheader ], [ %inc54, %for.body3 ]
+  %idxprom = sext i32 %0 to i64
+  %arrayidx = getelementptr inbounds [1 x i32], [1 x i32]* @b, i64 0, i64 %idxprom
+  store i32 4, i32* %arrayidx, align 4
+  %inc = add nsw i32 %inc1, 1
+  %tobool2 = icmp eq i32 %inc, 0
+  br i1 %tobool2, label %for.cond1.for.inc4_crit_edge, label %for.body3
+
+for.cond1.for.inc4_crit_edge:                     ; preds = %for.body3
+  %inc.lcssa = phi i32 [ %inc, %for.body3 ]
+  %.lcssa = phi i32 [ %inc54, %for.body3 ]
+  %inc5 = add nsw i32 %.lcssa, 1
+  br label %for.cond1.preheader
+}
-- 
2.40.0