From 57e1a3e6eed86f689ef8f4bcdbdfda2bd7f52c4d Mon Sep 17 00:00:00 2001
From: Haicheng Wu <haicheng@codeaurora.org>
Date: Wed, 17 Feb 2016 21:00:06 +0000
Subject: [PATCH] [LIR] Avoid turning non-temporal stores into memset

This is to fix PR26645.

llvm-svn: 261149
---
 llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp  |  4 +++
 .../test/Transforms/LoopIdiom/nontemporal_store.ll | 31 ++++++++++++++++++++++
 2 files changed, 35 insertions(+)
 create mode 100644 llvm/test/Transforms/LoopIdiom/nontemporal_store.ll

diff --git a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
index 900cdae..85b185b 100644
--- a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
@@ -314,6 +314,10 @@ bool LoopIdiomRecognize::isLegalStore(StoreInst *SI, bool &ForMemset,
   if (!SI->isSimple())
     return false;
 
+  // Avoid merging nontemporal stores.
+  if (SI->getMetadata(LLVMContext::MD_nontemporal))
+    return false;
+
   Value *StoredVal = SI->getValueOperand();
   Value *StorePtr = SI->getPointerOperand();
 
diff --git a/llvm/test/Transforms/LoopIdiom/nontemporal_store.ll b/llvm/test/Transforms/LoopIdiom/nontemporal_store.ll
new file mode 100644
index 0000000..297f3d3
--- /dev/null
+++ b/llvm/test/Transforms/LoopIdiom/nontemporal_store.ll
@@ -0,0 +1,31 @@
+; RUN: opt -loop-idiom < %s -S | FileCheck %s
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+%struct.bigBlock_t = type { [256 x <4 x float>] }
+
+; CHECK-LABEL: @test(
+; CHECK-NOT: llvm.memset
+define void @test(%struct.bigBlock_t* %p) {
+entry:
+  %0 = getelementptr inbounds %struct.bigBlock_t, %struct.bigBlock_t* %p, i64 0, i32 0, i64 0, i64 0
+  br label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %index.02 = phi i32 [ 0, %entry ], [ %add, %for.body ]
+  %dst.01 = phi float* [ %0, %entry ], [ %add.ptr2, %for.body ]
+  %cast.i5 = bitcast float* %dst.01 to <4 x float>*
+  store <4 x float> zeroinitializer, <4 x float>* %cast.i5, align 16, !nontemporal !0
+  %add.ptr1 = getelementptr inbounds float, float* %dst.01, i64 4
+  %cast.i = bitcast float* %add.ptr1 to <4 x float>*
+  store <4 x float> zeroinitializer, <4 x float>* %cast.i, align 16, !nontemporal !0
+  %add.ptr2 = getelementptr inbounds float, float* %dst.01, i64 8
+  %add = add nuw nsw i32 %index.02, 32
+  %cmp = icmp ult i32 %add, 4096
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:                                          ; preds = %for.body
+  ret void
+}
+
+!0 = !{i32 1}
-- 
2.7.4