ARM: align loops to 4 bytes on Cortex-M3 and Cortex-M4.

The Technical Reference Manuals for these two CPUs state that branching
to an unaligned 32-bit instruction incurs an extra pipeline reload
penalty. That's bad.

This also enables the optimization at -Os since it costs on average one
byte per loop in return for 1 cycle per iteration, which is pretty good
going.

llvm-svn: 342127

1 files changed, 7 insertions, 0 deletions

diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 4fed2025942..8b7b6b33193 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -1199,6 +1199,8 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
   // Prefer likely predicted branches to selects on out-of-order cores.
   PredictableSelectIsExpensive = Subtarget->getSchedModel().isOutOfOrder();
 
+  setPrefLoopAlignment(Subtarget->getPrefLoopAlignment());
+
   setMinFunctionAlignment(Subtarget->isThumb() ? 1 : 2);
 }
 
@@ -14695,6 +14697,11 @@ Value *ARMTargetLowering::emitStoreConditional(IRBuilder<> &Builder, Value *Val,
               Addr});
 }
 
+
+bool ARMTargetLowering::alignLoopsWithOptSize() const {
+  return Subtarget->isMClass();
+}
+
 /// A helper function for determining the number of interleaved accesses we
 /// will generate when lowering accesses of the given type.
 unsigned