Inline memove for small constant sizes and refactor memcpy and memset.

src/IceTargetLoweringX86Base.h

 //===- subzero/src/IceTargetLoweringX86Base.h - x86 lowering ----*- C++ -*-===//
 //
 //                        The Subzero Code Generator
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 ///
 /// \file
@@ skipping 135 matching lines @@
   /// Naive lowering of cmpxchg.
   void lowerAtomicCmpxchg(Variable *DestPrev, Operand *Ptr, Operand *Expected,
                           Operand *Desired);
   /// Attempt a more optimized lowering of cmpxchg. Returns true if optimized.
   bool tryOptimizedCmpxchgCmpBr(Variable *DestPrev, Operand *Ptr,
                                 Operand *Expected, Operand *Desired);
   void lowerAtomicRMW(Variable *Dest, uint32_t Operation, Operand *Ptr,
                       Operand *Val);
   void lowerCountZeros(bool Cttz, Type Ty, Variable *Dest, Operand *FirstVal,
                        Operand *SecondVal);
-  /// Replace a call to memcpy with inline instructions.
+  /// If Overflow is false, find the largest type whose width is less than or
+  /// equal to Size. If Overflow is true, find the smallest type whose width can
+  /// be larger than Size. The maximum width can be limited to LimitWidth bytes.
+  Type typeForSize(uint32_t Size, bool overflow = false,

[John, 2015/08/12 17:44:57]

Can this method be made static?

This comment tells me that you have 2 functions inside this one:

typeForSize(args, Overflow) {
  if Overflow:
    return largestTypeSmallerThanSize(args)
  else:
    return firstTypeThatFits(args)

The fact that, at every call site overflow is always a constant is another
pretty strong indication that you want two different functions.

I would also defined a constant named UnlimitedWidth (or NoWidthLimit) and use
it instead of = 0.

[ascull, 2015/08/17 22:18:52]

Done.

+                   uint32_t LimitWidth = 0);
+  /// Copy memory of given type from Src to Dest using OffsetAmt on both.

[John, 2015/08/12 17:44:57]

the lower "prefix" threw me off here. I though there were a high level CopyMem
instruction. What do you thing about renaming this copyMemory?

[ascull, 2015/08/17 22:18:52]

Done.

+  void lowerCopyMem(Type Ty, Variable *Dest, Variable *Src, uint32_t OffsetAmt);
+  /// Replace some calls to memcpy with inline instructions.
   void lowerMemcpy(Operand *Dest, Operand *Src, Operand *Count);
-  /// Replace a call to memset with inline instructions.
+  /// Replace some calls to memmove with inline instructions.
+  void lowerMemmove(Operand *Dest, Operand *Src, Operand *Count);
+  /// Replace some calls to memset with inline instructions.
   void lowerMemset(Operand *Dest, Operand *Val, Operand *Count);
 
   /// Lower an indirect jump adding sandboxing when needed.
   void lowerIndirectJump(Variable *Target);
 
   /// Check the comparison is in [Min,Max]. The flags register will be modified
   /// with:
   ///   - below equal, if in range
   ///   - above, set if not in range
   /// The index into the range is returned.
@@ skipping 486 matching lines @@
   }
 
   BoolFolding FoldingInfo;
 };
 } // end of namespace X86Internal
 } // end of namespace Ice
 
 #include "IceTargetLoweringX86BaseImpl.h"
 
 #endif // SUBZERO_SRC_ICETARGETLOWERINGX86BASE_H