[PATCH] D29004: New OptimizationRemarkEmitter pass for MIR

[Adam Nemet via Phabricator via llvm-commits]

anemet updated this revision to Diff 85668.
anemet marked 7 inline comments as done.
anemet added a comment.

The main change is that we now report spills & reloads per loop as suggested
by Matthias.

For this I had to somehow find the source line info for a MachineLoop (see
MachineLoop::getStartLoc()).  This is a dumbed-down version of the LLVM IR
variant which uses LoopID.  We can explore later if we can somehow find our
way back to the IR LoopID.


https://reviews.llvm.org/D29004

Files:
  include/llvm/Analysis/OptimizationDiagnosticInfo.h
  include/llvm/CodeGen/MachineLoopInfo.h
  include/llvm/CodeGen/MachineOptimizationRemarkEmitter.h
  include/llvm/IR/DiagnosticInfo.h
  include/llvm/InitializePasses.h
  lib/Analysis/OptimizationDiagnosticInfo.cpp
  lib/CodeGen/CMakeLists.txt
  lib/CodeGen/MachineLoopInfo.cpp
  lib/CodeGen/MachineOptimizationRemarkEmitter.cpp
  lib/CodeGen/RegAllocGreedy.cpp
  lib/IR/DiagnosticInfo.cpp
  test/CodeGen/AArch64/arm64-spill-remarks.ll
  tools/llc/llc.cpp

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D29004.85668.patch
Type: text/x-patch
Size: 30610 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170125/d1774181/attachment-0001.bin>