From 353476cbbb80881a754a90abc13851a884a4e26c Mon Sep 17 00:00:00 2001
From: Nadav Rotem <nrotem@apple.com>
Date: Mon, 12 Aug 2013 17:46:44 +0000
Subject: [PATCH] Fix PR16797 - Support PHINodes with multiple inputs from the
 same basic block. Do not generate new vector values for the same entries
 because we know that the incoming values from the same block must be
 identical.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188185 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Vectorize/SLPVectorizer.cpp    | 11 +++++
 .../X86/crash_netbsd_decompress.ll            | 41 +++++++++++++++++++
 2 files changed, 52 insertions(+)
 create mode 100644 test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll

diff --git a/lib/Transforms/Vectorize/SLPVectorizer.cpp b/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 9312b4bb4b0..22a2519d67f 100644
--- a/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -1187,10 +1187,21 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E) {
       PHINode *NewPhi = Builder.CreatePHI(VecTy, PH->getNumIncomingValues());
       E->VectorizedValue = NewPhi;
 
+      // PHINodes may have multiple entries from the same block. We want to
+      // visit every block once.
+      SmallSet<BasicBlock*, 4> VisitedBBs;
+
       for (unsigned i = 0, e = PH->getNumIncomingValues(); i < e; ++i) {
         ValueList Operands;
         BasicBlock *IBB = PH->getIncomingBlock(i);
 
+        if (VisitedBBs.count(IBB)) {
+          NewPhi->addIncoming(NewPhi->getIncomingValueForBlock(IBB), IBB);
+          continue;
+        }
+
+        VisitedBBs.insert(IBB);
+
         // Prepare the operand vector.
         for (unsigned j = 0; j < E->Scalars.size(); ++j)
           Operands.push_back(cast<PHINode>(E->Scalars[j])->
diff --git a/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll b/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll
new file mode 100644
index 00000000000..8da3c34a027
--- /dev/null
+++ b/test/Transforms/SLPVectorizer/X86/crash_netbsd_decompress.ll
@@ -0,0 +1,41 @@
+; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+%struct.DState = type { i32, i32 }
+
+@b = common global %struct.DState zeroinitializer, align 4
+@d = common global i32 0, align 4
+@c = common global i32 0, align 4
+@a = common global i32 0, align 4
+@e = common global i32 0, align 4
+
+define i32 @fn1() {
+entry:
+  %0 = load i32* getelementptr inbounds (%struct.DState* @b, i32 0, i32 0), align 4
+  %1 = load i32* getelementptr inbounds (%struct.DState* @b, i32 0, i32 1), align 4
+  %2 = load i32* @d, align 4
+  %cond = icmp eq i32 %2, 0
+  br i1 %cond, label %sw.bb, label %save_state_and_return
+
+sw.bb:                                            ; preds = %entry
+  %3 = load i32* @c, align 4
+  %and = and i32 %3, 7
+  store i32 %and, i32* @a, align 4
+  switch i32 %and, label %if.end [
+    i32 7, label %save_state_and_return
+    i32 0, label %save_state_and_return
+  ]
+
+if.end:                                           ; preds = %sw.bb
+  br label %save_state_and_return
+
+save_state_and_return:                            ; preds = %sw.bb, %sw.bb, %if.end, %entry
+  %t.0 = phi i32 [ 0, %if.end ], [ %0, %entry ], [ %0, %sw.bb ], [ %0, %sw.bb ]
+  %f.0 = phi i32 [ 0, %if.end ], [ %1, %entry ], [ 0, %sw.bb ], [ 0, %sw.bb ]
+  store i32 %t.0, i32* getelementptr inbounds (%struct.DState* @b, i32 0, i32 0), align 4
+  store i32 %f.0, i32* getelementptr inbounds (%struct.DState* @b, i32 0, i32 1), align 4
+  ret i32 undef
+}
+