From fc1604ec7274e9b724ff9b2512c288a691167427 Mon Sep 17 00:00:00 2001
From: Arnold Schwaighofer <aschwaighofer@apple.com>
Date: Wed, 16 Oct 2013 17:52:40 +0000
Subject: [PATCH] SLPVectorizer: Don't vectorize volatile memory operations

radar://15231682

Reapply r192799,
  http://lab.llvm.org:8011/builders/lldb-x86_64-debian-clang/builds/8226
showed that the bot is still broken even with this out.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192820 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Vectorize/SLPVectorizer.cpp    | 11 +++--
 test/Transforms/SLPVectorizer/X86/simplebb.ll | 43 +++++++++++++++++++
 2 files changed, 51 insertions(+), 3 deletions(-)

diff --git a/lib/Transforms/Vectorize/SLPVectorizer.cpp b/lib/Transforms/Vectorize/SLPVectorizer.cpp
index af1c0e74236..4d82bc428db 100644
--- a/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -786,13 +786,14 @@ void BoUpSLP::buildTree_rec(ArrayRef<Value *> VL, unsigned Depth) {
     }
     case Instruction::Load: {
       // Check if the loads are consecutive or of we need to swizzle them.
-      for (unsigned i = 0, e = VL.size() - 1; i < e; ++i)
-        if (!isConsecutiveAccess(VL[i], VL[i + 1])) {
+      for (unsigned i = 0, e = VL.size() - 1; i < e; ++i) {
+        LoadInst *L = cast<LoadInst>(VL[i]);
+        if (!L->isSimple() || !isConsecutiveAccess(VL[i], VL[i + 1])) {
           newTreeEntry(VL, false);
           DEBUG(dbgs() << "SLP: Need to swizzle loads.\n");
           return;
         }
-
+      }
       newTreeEntry(VL, true);
       DEBUG(dbgs() << "SLP: added a vector of loads.\n");
       return;
@@ -1911,6 +1912,10 @@ unsigned SLPVectorizer::collectStores(BasicBlock *BB, BoUpSLP &R) {
     if (!SI)
       continue;
 
+    // Don't touch volatile stores.
+    if (!SI->isSimple())
+      continue;
+
     // Check that the pointer points to scalars.
     Type *Ty = SI->getValueOperand()->getType();
     if (Ty->isAggregateType() || Ty->isVectorTy())
diff --git a/test/Transforms/SLPVectorizer/X86/simplebb.ll b/test/Transforms/SLPVectorizer/X86/simplebb.ll
index 0b76bec07bd..7d682e5e467 100644
--- a/test/Transforms/SLPVectorizer/X86/simplebb.ll
+++ b/test/Transforms/SLPVectorizer/X86/simplebb.ll
@@ -44,3 +44,46 @@ entry:
   store double %mul5, double* %arrayidx5, align 8
   ret void
 }
+
+; Don't vectorize volatile loads.
+; CHECK: test_volatile_load
+; CHECK-NOT: load <2 x double>
+; CHECK: store <2 x double>
+; CHECK: ret
+define void @test_volatile_load(double* %a, double* %b, double* %c) {
+entry:
+  %i0 = load volatile double* %a, align 8
+  %i1 = load volatile double* %b, align 8
+  %mul = fmul double %i0, %i1
+  %arrayidx3 = getelementptr inbounds double* %a, i64 1
+  %i3 = load double* %arrayidx3, align 8
+  %arrayidx4 = getelementptr inbounds double* %b, i64 1
+  %i4 = load double* %arrayidx4, align 8
+  %mul5 = fmul double %i3, %i4
+  store double %mul, double* %c, align 8
+  %arrayidx5 = getelementptr inbounds double* %c, i64 1
+  store double %mul5, double* %arrayidx5, align 8
+  ret void
+}
+
+; Don't vectorize volatile stores.
+; CHECK: test_volatile_store
+; CHECK-NOT: store <2 x double>
+; CHECK: ret
+define void @test_volatile_store(double* %a, double* %b, double* %c) {
+entry:
+  %i0 = load double* %a, align 8
+  %i1 = load double* %b, align 8
+  %mul = fmul double %i0, %i1
+  %arrayidx3 = getelementptr inbounds double* %a, i64 1
+  %i3 = load double* %arrayidx3, align 8
+  %arrayidx4 = getelementptr inbounds double* %b, i64 1
+  %i4 = load double* %arrayidx4, align 8
+  %mul5 = fmul double %i3, %i4
+  store volatile double %mul, double* %c, align 8
+  %arrayidx5 = getelementptr inbounds double* %c, i64 1
+  store volatile double %mul5, double* %arrayidx5, align 8
+  ret void
+}
+
+
-- 
2.34.1