Make ARM and X86 LowerMEMCPY identical by moving the isThumb check into getMaxInlineS...

and by restructuring the X86 version.

New I just have to move this to a common place :-)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@43554 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index ef1c86d22a8f8f782d3f0e9882c630dff6f6af4c..b7e37660a7b6cb5e988ba50ff294f6f52e12558c 100644 (file)
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -1287,8 +1287,7 @@ static SDOperand LowerSRx(SDOperand Op, SelectionDAG &DAG,
   return DAG.getNode(ISD::BUILD_PAIR, MVT::i64, Lo, Hi);
 }
 
-SDOperand ARMTargetLowering::LowerMEMCPY(SDOperand Op, SelectionDAG &DAG,
-                                         const ARMSubtarget *ST) {
+SDOperand ARMTargetLowering::LowerMEMCPY(SDOperand Op, SelectionDAG &DAG) {
   SDOperand ChainOp = Op.getOperand(0);
   SDOperand DestOp = Op.getOperand(1);
   SDOperand SourceOp = Op.getOperand(2);
@@ -1311,11 +1310,9 @@ SDOperand ARMTargetLowering::LowerMEMCPY(SDOperand Op, SelectionDAG &DAG,
   // The libc version is likely to be faster for the these cases. It can
   // use the address value and run time information about the CPU.
   // With glibc 2.6.1 on a core 2, coping an array of 100M longs was 30% faster
-  // FIXME: For now, we don't lower memcpy's to loads / stores for Thumb. Change
-  // this once Thumb ldmia / stmia support is added.
   unsigned Size = I->getValue();
   if (AlwaysInline ||
-      (!ST->isThumb() && Size <= Subtarget->getMaxInlineSizeThreshold() &&
+      (Size <= Subtarget->getMaxInlineSizeThreshold() &&
        (Align & 3) == 0))
     return LowerMEMCPYInline(ChainOp, DestOp, SourceOp, Size, Align, DAG);
   return LowerMEMCPYCall(ChainOp, DestOp, SourceOp, CountOp, DAG);
@@ -1461,7 +1458,7 @@ SDOperand ARMTargetLowering::LowerOperation(SDOperand Op, SelectionDAG &DAG) {
   case ISD::RETURNADDR:    break;
   case ISD::FRAMEADDR:     break;
   case ISD::GLOBAL_OFFSET_TABLE: return LowerGLOBAL_OFFSET_TABLE(Op, DAG);
-  case ISD::MEMCPY:        return LowerMEMCPY(Op, DAG, Subtarget);
+  case ISD::MEMCPY:        return LowerMEMCPY(Op, DAG);
   }
   return SDOperand();
 }

diff --git a/lib/Target/ARM/ARMISelLowering.h b/lib/Target/ARM/ARMISelLowering.h
index 47cb2a177183c506aaefd2bc9d583da4b4b87765..41045e7569624482c083842da4257580c473fdcb 100644 (file)
--- a/lib/Target/ARM/ARMISelLowering.h
+++ b/lib/Target/ARM/ARMISelLowering.h
@@ -134,8 +134,7 @@ namespace llvm {
     SDOperand LowerGLOBAL_OFFSET_TABLE(SDOperand Op, SelectionDAG &DAG);
     SDOperand LowerFORMAL_ARGUMENTS(SDOperand Op, SelectionDAG &DAG);
     SDOperand LowerBR_JT(SDOperand Op, SelectionDAG &DAG);
-    SDOperand LowerMEMCPY(SDOperand Op, SelectionDAG &DAG,
-                          const ARMSubtarget *ST);
+    SDOperand LowerMEMCPY(SDOperand Op, SelectionDAG &DAG);
     SDOperand LowerMEMCPYCall(SDOperand Chain, SDOperand Dest,
                               SDOperand Source, SDOperand Count,
                               SelectionDAG &DAG);

diff --git a/lib/Target/ARM/ARMSubtarget.h b/lib/Target/ARM/ARMSubtarget.h
index 5b5ee39ced079e9547c2e58a932a54f8dbdb04be..087623dee2f5b948eca1146f9338b552a3f10e10 100644 (file)
--- a/lib/Target/ARM/ARMSubtarget.h
+++ b/lib/Target/ARM/ARMSubtarget.h
@@ -62,7 +62,11 @@ protected:
   ///
   ARMSubtarget(const Module &M, const std::string &FS, bool thumb);
 
-  unsigned getMaxInlineSizeThreshold() const { return 64; }
+  unsigned getMaxInlineSizeThreshold() const {
+    // FIXME: For now, we don't lower memcpy's to loads / stores for Thumb.
+    // Change this once Thumb ldmia / stmia support is added.
+    return isThumb() ? 0 : 64;
+  }
   /// ParseSubtargetFeatures - Parses features string setting specified 
   /// subtarget options.  Definition of function is auto generated by tblgen.
   void ParseSubtargetFeatures(const std::string &FS, const std::string &CPU);

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index f1bf150e1f9e8dcae071a88a58bae448b1bf92ce..3180e647033e3b5f470ee6bf99424bcd1ca664f6 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -4496,24 +4496,17 @@ SDOperand X86TargetLowering::LowerMEMCPY(SDOperand Op, SelectionDAG &DAG) {
     assert(!AlwaysInline && "Cannot inline copy of unknown size");
     return LowerMEMCPYCall(ChainOp, DestOp, SourceOp, CountOp, DAG);
   }
-  unsigned Size = I->getValue();
-
-  if (AlwaysInline)
-    return LowerMEMCPYInline(ChainOp, DestOp, SourceOp, Size, Align, DAG);
 
+  // If not DWORD aligned or if size is more than threshold, then call memcpy.
   // The libc version is likely to be faster for the following cases. It can
   // use the address value and run time information about the CPU.
   // With glibc 2.6.1 on a core 2, coping an array of 100M longs was 30% faster
-
-  // If not DWORD aligned, call memcpy.
-  if ((Align & 3) != 0)
-    return LowerMEMCPYCall(ChainOp, DestOp, SourceOp, CountOp, DAG);
-
-  // If size is more than the threshold, call memcpy.
-  if (Size > Subtarget->getMaxInlineSizeThreshold())
-    return LowerMEMCPYCall(ChainOp, DestOp, SourceOp, CountOp, DAG);
-
-  return LowerMEMCPYInline(ChainOp, DestOp, SourceOp, Size, Align, DAG);
+  unsigned Size = I->getValue();
+  if (AlwaysInline ||
+      (Size <= Subtarget->getMaxInlineSizeThreshold() &&
+       (Align & 3) == 0))
+    return LowerMEMCPYInline(ChainOp, DestOp, SourceOp, Size, Align, DAG);
+  return LowerMEMCPYCall(ChainOp, DestOp, SourceOp, CountOp, DAG);
 }
 
 SDOperand X86TargetLowering::LowerMEMCPYCall(SDOperand Chain,