Extend r148086 to check for [r +/- reg] address mode. This fixes queens performance regression (due to increased register pressure from overly aggressive pre-inc formation).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@152162 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 15 insertions, 8 deletions

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index b55d84cab8..2532c123ba 100644
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -6142,8 +6142,7 @@ SDValue DAGCombiner::visitBR_CC(SDNode *N) {
 
 /// canFoldInAddressingMode - Return true if 'Use' is a load or a store that
 /// uses N as its base pointer and that N may be folded in the load / store
-/// addressing mode. FIXME: This currently only looks for folding of
-/// [reg +/- imm] addressing modes.
+/// addressing mode.
 static bool canFoldInAddressingMode(SDNode *N, SDNode *Use,
                                     SelectionDAG &DAG,
                                     const TargetLowering &TLI) {
@@ -6163,15 +6162,19 @@ static bool canFoldInAddressingMode(SDNode *N, SDNode *Use,
   if (N->getOpcode() == ISD::ADD) {
     ConstantSDNode *Offset = dyn_cast<ConstantSDNode>(N->getOperand(1));
     if (Offset)
+      // [reg +/- imm]
       AM.BaseOffs = Offset->getSExtValue();
     else
-      return false;
+      // [reg +/- reg]
+      AM.Scale = 1;
   } else if (N->getOpcode() == ISD::SUB) {
     ConstantSDNode *Offset = dyn_cast<ConstantSDNode>(N->getOperand(1));
     if (Offset)
+      // [reg +/- imm]
       AM.BaseOffs = -Offset->getSExtValue();
     else
-      return false;
+      // [reg +/- reg]
+      AM.Scale = 1;
   } else
     return false;
 
diff --git a/test/CodeGen/ARM/shifter_operand.ll b/test/CodeGen/ARM/shifter_operand.ll
index 521ffa1c9e..eb971ff72e 100644
--- a/test/CodeGen/ARM/shifter_operand.ll
+++ b/test/CodeGen/ARM/shifter_operand.ll
@@ -54,12 +54,16 @@ declare i8* @malloc(...)
 define fastcc void @test4(i16 %addr) nounwind {
 entry:
 ; A8: test4:
-; A8: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]!
-; A8: str [[REG]], [r0]
+; A8: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]
+; A8-NOT: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]!
+; A8: str [[REG]], [r0, r1, lsl #2]
+; A8-NOT: str [[REG]], [r0]
 
 ; A9: test4:
-; A9: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]!
-; A9: str [[REG]], [r0]
+; A9: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]
+; A9-NOT: ldr [[REG:r[0-9]+]], [r0, r1, lsl #2]!
+; A9: str [[REG]], [r0, r1, lsl #2]
+; A9-NOT: str [[REG]], [r0]
   %0 = tail call i8* (...)* @malloc(i32 undef) nounwind
   %1 = bitcast i8* %0 to i32*
   %2 = sext i16 %addr to i32