Allow merging multiple store sequences on the same chain.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@169111 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 5659069178b7dacb8eb4c143809fa56c372d4b0b..438e23da34b3c8a4310a90339634e28c030de691 100644 (file)
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -8157,8 +8157,21 @@ SDValue DAGCombiner::visitSTORE(SDNode *N) {
 
   // Only perform this optimization before the types are legal, because we
   // don't want to perform this optimization on every DAGCombine invocation.
-  if (!LegalTypes && MergeConsecutiveStores(ST))
-    return SDValue(N, 0);
+  if (!LegalTypes) {
+    bool EverChanged = false;
+
+    do {
+      // There can be multiple store sequences on the same chain.
+      // Keep trying to merge store sequences until we are unable to do so
+      // or until we merge the last store on the chain.
+      bool Changed = MergeConsecutiveStores(ST);
+      EverChanged |= Changed;
+      if (!Changed) break;
+    } while (ST->getOpcode() != ISD::DELETED_NODE);
+
+    if (EverChanged)
+      return SDValue(N, 0);
+  }
 
   return ReduceLoadOpStoreWidth(N);
 }

diff --git a/test/CodeGen/X86/2012-12-1-merge-multiple.ll b/test/CodeGen/X86/2012-12-1-merge-multiple.ll
new file mode 100644 (file)
index 0000000..5931c3d
--- /dev/null
+++ b/test/CodeGen/X86/2012-12-1-merge-multiple.ll
@@ -0,0 +1,31 @@
+; RUN: llc < %s -march=x86-64 -mcpu=corei7 -mtriple=x86_64-pc-win64 | FileCheck %s
+
+; CHECK: multiple_stores_on_chain
+; CHECK: movabsq
+; CHECK: movq
+; CHECK: movabsq
+; CHECK: movq
+; CHECK: ret
+define void @multiple_stores_on_chain(i16 * %A) {
+entry:
+  %a0 = getelementptr inbounds i16* %A, i64 0
+  %a1 = getelementptr inbounds i16* %A, i64 1
+  %a2 = getelementptr inbounds i16* %A, i64 2
+  %a3 = getelementptr inbounds i16* %A, i64 3
+  %a4 = getelementptr inbounds i16* %A, i64 4
+  %a5 = getelementptr inbounds i16* %A, i64 5
+  %a6 = getelementptr inbounds i16* %A, i64 6
+  %a7 = getelementptr inbounds i16* %A, i64 7
+
+  store i16 0, i16* %a0
+  store i16 1, i16* %a1
+  store i16 2, i16* %a2
+  store i16 3, i16* %a3
+  store i16 4, i16* %a4
+  store i16 5, i16* %a5
+  store i16 6, i16* %a6
+  store i16 7, i16* %a7
+
+  ret void
+}
+