[BranchFolding] Set correct mem refs (2nd try)

This is a recommit of r257253 which was reverted in r257270.
Previous testcase can make failure on some targets due to using opt with O3 option.

Original Summary:
Merge MBBICommon and MBBI's MMOs.

Differential Revision: http://reviews.llvm.org/D15990

llvm-svn: 257317

2 files changed, 34 insertions, 1 deletions

diff --git a/llvm/lib/CodeGen/BranchFolding.cpp b/llvm/lib/CodeGen/BranchFolding.cpp
index 60803490a5d..df5cac5a9f7 100644
--- a/llvm/lib/CodeGen/BranchFolding.cpp
+++ b/llvm/lib/CodeGen/BranchFolding.cpp
@@ -780,7 +780,7 @@ removeMMOsFromMemoryOperations(MachineBasicBlock::iterator MBBIStartPos,
     assert(MBBICommon->isIdenticalTo(&*MBBI) && "Expected matching MIIs!");
 
     if (MBBICommon->mayLoad() || MBBICommon->mayStore())
-      MBBICommon->setMemRefs(MBBI->mergeMemRefsWith(*MBBI));
+      MBBICommon->setMemRefs(MBBICommon->mergeMemRefsWith(*MBBI));
 
     ++MBBI;
     ++MBBICommon;
diff --git a/llvm/test/CodeGen/AArch64/branch-folder-merge-mmos.ll b/llvm/test/CodeGen/AArch64/branch-folder-merge-mmos.ll
new file mode 100644
index 00000000000..3f9c0239fe4
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/branch-folder-merge-mmos.ll
@@ -0,0 +1,33 @@
+; RUN: llc -march=aarch64 -mtriple=aarch64-none-linux-gnu -stop-after branch-folder -o /dev/null < %s | FileCheck %s
+target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
+
+; Function Attrs: norecurse nounwind
+define void @foo(i32 %a, i32 %b, float* nocapture %foo_arr) #0 {
+; CHECK: (load 4 from %ir.arrayidx1.{{i[1-2]}}), (load 4 from %ir.arrayidx1.{{i[1-2]}})
+entry:
+  %cmp = icmp sgt i32 %a, 0
+  br i1 %cmp, label %if.then, label %if.end
+
+if.then:                                          ; preds = %entry
+  %0 = load float, float* %foo_arr, align 4
+  %arrayidx1.i1 = getelementptr inbounds float, float* %foo_arr, i64 1
+  %1 = load float, float* %arrayidx1.i1, align 4
+  %sub.i = fsub float %0, %1
+  store float %sub.i, float* %foo_arr, align 4
+  br label %if.end3
+
+if.end:                                           ; preds = %entry
+  %cmp1 = icmp sgt i32 %b, 0
+  br i1 %cmp1, label %if.then2, label %if.end3
+
+if.then2:                                         ; preds = %if.end
+  %2 = load float, float* %foo_arr, align 4
+  %arrayidx1.i2 = getelementptr inbounds float, float* %foo_arr, i64 1
+  %3 = load float, float* %arrayidx1.i2, align 4
+  %sub.i3 = fsub float %2, %3
+  store float %sub.i3, float* %foo_arr, align 4
+  br label %if.end3
+
+if.end3:                                          ; preds = %if.then2, %if.end, %if.then
+  ret void
+}