[PATCH] [mips] Optimize folding of registers derived from the $zero register.

[Vasileios Kalintiris]

Hi dsanders,

Instead of loading the zero value into a new register and then spilling
it into the stack, we can use the $zero register immediatelly.
This patch also changes the semantics of the foldMemoryOperandImpl()
method in order to insert the newly constructed instruction into the
machine basic block.

http://reviews.llvm.org/D8289

Files:
  lib/CodeGen/TargetInstrInfo.cpp
  lib/Target/Mips/Mips16InstrInfo.cpp
  lib/Target/Mips/Mips16InstrInfo.h
  lib/Target/Mips/MipsInstrInfo.h
  lib/Target/Mips/MipsSEInstrInfo.cpp
  lib/Target/Mips/MipsSEInstrInfo.h
  lib/Target/SystemZ/SystemZInstrInfo.cpp
  lib/Target/X86/X86FastISel.cpp
  lib/Target/X86/X86InstrInfo.cpp
  test/CodeGen/Mips/fold-zero-copy.ll

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D8289.21824.patch
Type: text/x-patch
Size: 12958 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150312/4a822b2c/attachment.bin>