From 28dad2a5caccf579f7430acd3af8fa6e6f8b575e Mon Sep 17 00:00:00 2001
From: Evan Cheng <evan.cheng@apple.com>
Date: Tue, 18 May 2010 21:31:17 +0000
Subject: Sink dag combine's post index load / store code that swap base ptr
 and index into the target hook. Only the target knows whether the swap is
 safe. In Thumb2 mode, the offset must be an immediate. rdar://7998649

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@104060 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/CodeGen/SelectionDAG/DAGCombiner.cpp    |  4 ----
 lib/Target/ARM/ARMISelLowering.cpp          | 16 +++++++++++++++-
 test/CodeGen/ARM/2010-05-18-PostIndexBug.ll | 25 +++++++++++++++++++++++++
 3 files changed, 40 insertions(+), 5 deletions(-)
 create mode 100644 test/CodeGen/ARM/2010-05-18-PostIndexBug.ll

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index d6e761f..bfa302c 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -5284,10 +5284,6 @@ bool DAGCombiner::CombineToPostIndexedLoadStore(SDNode *N) {
     SDValue Offset;
     ISD::MemIndexedMode AM = ISD::UNINDEXED;
     if (TLI.getPostIndexedAddressParts(N, Op, BasePtr, Offset, AM, DAG)) {
-      if (Ptr == Offset && Op->getOpcode() == ISD::ADD)
-        std::swap(BasePtr, Offset);
-      if (Ptr != BasePtr)
-        continue;
       // Don't create a indexed load / store with zero offset.
       if (isa<ConstantSDNode>(Offset) &&
           cast<ConstantSDNode>(Offset)->isNullValue())
diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index c284a7e..f5373cb 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -4407,9 +4407,11 @@ bool ARMTargetLowering::getPostIndexedAddressParts(SDNode *N, SDNode *Op,
   bool isSEXTLoad = false;
   if (LoadSDNode *LD = dyn_cast<LoadSDNode>(N)) {
     VT  = LD->getMemoryVT();
+    Ptr = LD->getBasePtr();
     isSEXTLoad = LD->getExtensionType() == ISD::SEXTLOAD;
   } else if (StoreSDNode *ST = dyn_cast<StoreSDNode>(N)) {
     VT  = ST->getMemoryVT();
+    Ptr = ST->getBasePtr();
   } else
     return false;
 
@@ -4417,13 +4419,25 @@ bool ARMTargetLowering::getPostIndexedAddressParts(SDNode *N, SDNode *Op,
   bool isLegal = false;
   if (Subtarget->isThumb2())
     isLegal = getT2IndexedAddressParts(Op, VT, isSEXTLoad, Base, Offset,
-                                        isInc, DAG);
+                                       isInc, DAG);
   else
     isLegal = getARMIndexedAddressParts(Op, VT, isSEXTLoad, Base, Offset,
                                         isInc, DAG);
   if (!isLegal)
     return false;
 
+  if (Ptr != Base) {
+    // Swap base ptr and offset to catch more post-index load / store when
+    // it's legal. In Thumb2 mode, offset must be an immediate.
+    if (Ptr == Offset && Op->getOpcode() == ISD::ADD &&
+        !Subtarget->isThumb2())
+      std::swap(Base, Offset);
+
+    // Post-indexed load / store update the base pointer.
+    if (Ptr != Base)
+      return false;
+  }
+
   AM = isInc ? ISD::POST_INC : ISD::POST_DEC;
   return true;
 }
diff --git a/test/CodeGen/ARM/2010-05-18-PostIndexBug.ll b/test/CodeGen/ARM/2010-05-18-PostIndexBug.ll
new file mode 100644
index 0000000..9907228
--- /dev/null
+++ b/test/CodeGen/ARM/2010-05-18-PostIndexBug.ll
@@ -0,0 +1,25 @@
+; RUN: llc < %s -mtriple=armv7-apple-darwin   | FileCheck %s -check-prefix=ARM
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin | FileCheck %s -check-prefix=THUMB
+; rdar://7998649
+
+%struct.foo = type { i64, i64 }
+
+define arm_apcscc zeroext i8 @t(%struct.foo* %this) noreturn optsize {
+entry:
+; ARM:       t:
+; ARM:       str r0, [r1], r0
+
+; THUMB:     t:
+; THUMB-NOT: str r0, [r1], r0
+; THUMB:     str r0, [r1]
+  %0 = getelementptr inbounds %struct.foo* %this, i32 0, i32 1 ; <i64*> [#uses=1]
+  store i32 undef, i32* inttoptr (i32 8 to i32*), align 8
+  br i1 undef, label %bb.nph96, label %bb3
+
+bb3:                                              ; preds = %entry
+  %1 = load i64* %0, align 4                      ; <i64> [#uses=0]
+  unreachable
+
+bb.nph96:                                         ; preds = %entry
+  unreachable
+}
-- 
cgit v1.1