commit e45ab8a0a90e4f3a59d8c38038ae3e495ee1fef3
author Jim Grosbach <grosbach@apple.com> Tue Jan 19 18:31:11 2010 +0000
committer Jim Grosbach <grosbach@apple.com> Tue Jan 19 18:31:11 2010 +0000
tree c2cb2db266d2fc51246754d560fdf5b1ba792e8d
parent c0404b3715151d020dda074740bbdaa794f9328a

For aligned load/store instructions, it's only required to know whether a
function can support dynamic stack realignment. That's a much easier question
to answer at instruction selection stage than whether the function actually
will have dynamic alignment prologue. This allows the removal of the
stack alignment heuristic pass, and improves code quality for cases where
the heuristic would result in dynamic alignment code being generated when
it was not strictly necessary.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@93885 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/ARM/ARMBaseInstrInfo.cpp

diff --git a/lib/Target/ARM/ARMBaseInstrInfo.cpp b/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 969c4a4..fd46a4a 100644
--- a/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -715,7 +715,7 @@
             RC == ARM::QPR_VFP2RegisterClass) && "Unknown regclass!");
     // FIXME: Neon instructions should support predicates
     if (Align >= 16
-        && (getRegisterInfo().needsStackRealignment(MF))) {
+        && (getRegisterInfo().canRealignStack(MF))) {
       AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::VST1q64))
                      .addFrameIndex(FI).addImm(0).addImm(0).addImm(128)
                      .addMemOperand(MMO)
@@ -760,7 +760,7 @@
             RC == ARM::QPR_VFP2RegisterClass ||
             RC == ARM::QPR_8RegisterClass) && "Unknown regclass!");
     if (Align >= 16
-        && (getRegisterInfo().needsStackRealignment(MF))) {
+        && (getRegisterInfo().canRealignStack(MF))) {
       AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::VLD1q64), DestReg)
                      .addFrameIndex(FI).addImm(0).addImm(0).addImm(128)
                      .addMemOperand(MMO));

lib/Target/ARM/ARMBaseRegisterInfo.cpp

diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.cpp b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
index f1b6e1d..ba9e044 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.cpp
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
@@ -484,6 +484,14 @@
           MFI->isFrameAddressTaken());
 }
 
+bool ARMBaseRegisterInfo::canRealignStack(const MachineFunction &MF) const {
+  const MachineFrameInfo *MFI = MF.getFrameInfo();
+  const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
+  return (RealignStack &&
+          !AFI->isThumb1OnlyFunction() &&
+          !MFI->hasVarSizedObjects());
+}
+
 bool ARMBaseRegisterInfo::
 needsStackRealignment(const MachineFunction &MF) const {
   const MachineFrameInfo *MFI = MF.getFrameInfo();

lib/Target/ARM/ARMBaseRegisterInfo.h

diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.h b/lib/Target/ARM/ARMBaseRegisterInfo.h
index 2788d07..f5ca25c 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.h
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.h
@@ -96,6 +96,7 @@
 
   bool hasFP(const MachineFunction &MF) const;
 
+  bool canRealignStack(const MachineFunction &MF) const;
   bool needsStackRealignment(const MachineFunction &MF) const;
 
   bool cannotEliminateFrame(const MachineFunction &MF) const;

lib/Target/ARM/ARMTargetMachine.cpp

diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 1c6fca7..4d20a5c 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -93,10 +93,6 @@
   if (Subtarget.hasNEON())
     PM.add(createNEONPreAllocPass());
 
-  // Calculate and set max stack object alignment early, so we can decide
-  // whether we will need stack realignment (and thus FP).
-  PM.add(createMaxStackAlignmentCalculatorPass());
-
   // FIXME: temporarily disabling load / store optimization pass for Thumb1.
   if (OptLevel != CodeGenOpt::None && !Subtarget.isThumb1Only())
     PM.add(createARMLoadStoreOptimizationPass(true));