From 01b6f1cc9a9fbfc5c8f1d6038c1f371589b606fe Mon Sep 17 00:00:00 2001
From: Arnold Schwaighofer <aschwaighofer@apple.com>
Date: Fri, 15 Nov 2013 23:09:33 +0000
Subject: [PATCH] LoopVectorizer: Use abi alignment for accesses with no
 alignment

When we vectorize a scalar access with no alignment specified, we have to set
the target's abi alignment of the scalar access on the vectorized access.
Using the same alignment of zero would be wrong because most targets will have a
bigger abi alignment for vector types.

This probably fixes PR17878.

llvm-svn: 194876
---
 lib/Transforms/Vectorize/LoopVectorize.cpp |  4 +++
 test/Transforms/LoopVectorize/align.ll     | 33 ++++++++++++++++++++++
 2 files changed, 37 insertions(+)
 create mode 100644 test/Transforms/LoopVectorize/align.ll

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index bc649b3dbbc..e624bb49f2a 100644
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -1205,6 +1205,10 @@ void InnerLoopVectorizer::vectorizeMemoryInstruction(Instruction *Instr,
   Type *DataTy = VectorType::get(ScalarDataTy, VF);
   Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
   unsigned Alignment = LI ? LI->getAlignment() : SI->getAlignment();
+  // An alignment of 0 means target abi alignment. We need to use the scalar's
+  // target abi alignment in such a case.
+  if (!Alignment)
+    Alignment = DL->getABITypeAlignment(ScalarDataTy);
   unsigned AddressSpace = Ptr->getType()->getPointerAddressSpace();
   unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ScalarDataTy);
   unsigned VectorElementSize = DL->getTypeStoreSize(DataTy)/VF;
diff --git a/test/Transforms/LoopVectorize/align.ll b/test/Transforms/LoopVectorize/align.ll
new file mode 100644
index 00000000000..84b03615d37
--- /dev/null
+++ b/test/Transforms/LoopVectorize/align.ll
@@ -0,0 +1,33 @@
+; RUN: opt < %s  -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -S | FileCheck %s
+
+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"
+
+; Make sure we output the abi alignment if no alignment is specified.
+
+;CHECK-LABEL: @align
+;CHECK: load <4 x i32>* {{.*}} align  4
+;CHECK: load <4 x i32>* {{.*}} align  4
+;CHECK: store <4 x i32> {{.*}} align  4
+
+define void @align(i32* %a, i32* %b, i32* %c) nounwind uwtable ssp {
+  br label %1
+
+; <label>:1                                       ; preds = %1, %0
+  %indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
+  %2 = getelementptr inbounds i32* %b, i64 %indvars.iv
+  %3 = load i32* %2
+  %4 = getelementptr inbounds i32* %c, i64 %indvars.iv
+  %5 = load i32* %4
+  %6 = add nsw i32 %5, %3
+  %7 = getelementptr inbounds i32* %a, i64 %indvars.iv
+  store i32 %6, i32* %7
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 128 
+  br i1 %exitcond, label %8, label %1
+
+; <label>:8                                       ; preds = %1
+  ret void
+}
+