[PATCH] Add a shrink-wrapping pass to improve the placement of prologue and epilogue.

Wed Apr 22 15:50:39 PDT 2015

Hi grosbach,

Hi,

@Jim, putting you as reviewer as you seemed to have touched PEI more than most people :).

This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function.
The interest is to first safe points that are cheaper than the entry and exits blocks.

** Context **

Currently we insert the prologue and epilogue of the method/function in the entry and exits blocks. Although this is correct, we can do a better job when those are not immediately required and insert them at less frequently executed places.
The job of the shrink-wrapping pass is to identify such places.

** Motivating example **

Let us consider the following function that perform a call only in one branch of a if:
define i32 @f(i32 %a, i32 %b)  {
  %tmp = alloca i32, align 4
  %tmp2 = icmp slt i32 %a, %b
  br i1 %tmp2, label %true, label %false

true:
  store i32 %a, i32* %tmp, align 4
  %tmp4 = call i32 @doSomething(i32 0, i32* %tmp)
  br label %false

false:
  %tmp.0 = phi i32 [ %tmp4, %true ], [ %a, %0 ]
  ret i32 %tmp.0
}

On AArch64 this code generates (removing the cfi directives to ease readabilities):
_f:                                     ; @f
; BB#0:
	stp	x29, x30, [sp, #-16]!
	mov	 x29, sp
	sub	sp, sp, #16             ; =16
	cmp	 w0, w1
	b.ge	LBB0_2
; BB#1:                                 ; %true
	stur	w0, [x29, #-4]
	sub	x1, x29, #4             ; =4
	mov	 w0, wzr
	bl	_doSomething
LBB0_2:                                 ; %false
	mov	 sp, x29
	ldp	x29, x30, [sp], #16
	ret

With shrink-wrapping we could generate:
_f:                                     ; @f
; BB#0:
	cmp	 w0, w1
	b.ge	LBB0_2
; BB#1:                                 ; %true
	stp	x29, x30, [sp, #-16]!
	mov	 x29, sp
	sub	sp, sp, #16             ; =16
	stur	w0, [x29, #-4]
	sub	x1, x29, #4             ; =4
	mov	 w0, wzr
	bl	_doSomething
	add	sp, x29, #16            ; =16
	ldp	x29, x30, [sp], #16
LBB0_2:                                 ; %false
	ret

Therefore, we would pay the overhead of setting up/destroying the frame only if we actually do the call.

** Proposed Solution **

This patch introduces a new machine pass that perform the shrink-wrapping analysis (See the comments at the beginning of ShrinkWrap.cpp for more details). It then stores the safe save and restore point into the MachineFrameInfo attached to the MachineFunction.
This information is then used by the PrologEpilogInserter (PEI) to place the related code at the right place. This pass runs right before the PEI.

Unlike the original paper of Chow from PLDI’88, this implementation of shrink-wrapping does not use expensive data-flow analysis and does not need hack to properly avoid frequently executed point. Instead, it relies on dominance and loop properties.

The pass is off by default and each target can opt-in by setting the EnableShrinkWrap boolean to true in their derived class of TargetPassConfig. This setting can also be overwritten on the command line by using -enable-shrink-wrap.

Before you try out the pass for your target, make sure you properly fix your emitProlog/emitEpilog/adjustForXXX method to cope with basic blocks that are not necessarily the entry block.

** Design Decisions **

1. ShrinkWrap is its own pass right now. It could frankly be merged into PEI but for debugging and clarity I thought it was best to have its own file.
2. Right now, we only support one save point and one restore point. At some point we can expand this to several save point and restore point, the impacted component would then be:
- The pass itself: New algorithm needed.
- MachineFrameInfo: Hold a list or set of Save/Restore point instead of one pointer.
- PEI: Should loop over the save point and restore point.
Anyhow, at least for this first iteration, I do not believe this is interesting to support the complex cases. We should revisit that when we motivating examples.

That being said, the target specific code should not change, which is another point for not blocking the optimization on that :).

** Feedback Needed **

Right now, I haven’t added any new target hook, but I am wondering if some more would make sense.
In particular:
1. Should we have a target hook to be able to fix something on the entry block if this one was not the save point?
2. Same question with all the exit blocks?

For #2, for instance, ARM needs to expand the TCRETURN pseudo-instruction, but this can be done in the expand pseudo pass. Therefore, I do not know if this is actually needed.

** What Is Next? **

I have patches to enable this for AArch64 and ARM and I think I will look into X86 as well. For the record, this implementation of shrink-wrapping applies on about 20% of the function for both O3 and Os with no-regressions and a few improvements.
PGO would certainly helped, but I haven’t tried.

Thanks for your feedbacks,
-Quentin

REPOSITORY
  rL LLVM

http://reviews.llvm.org/D9210

Files:
  include/llvm/CodeGen/MachineFrameInfo.h
  include/llvm/CodeGen/Passes.h
  include/llvm/InitializePasses.h
  include/llvm/Target/TargetFrameLowering.h
  lib/CodeGen/CMakeLists.txt
  lib/CodeGen/CodeGen.cpp
  lib/CodeGen/MachineFunction.cpp
  lib/CodeGen/Passes.cpp
  lib/CodeGen/PrologEpilogInserter.cpp
  lib/CodeGen/ShrinkWrap.cpp
  lib/Target/AArch64/AArch64FrameLowering.cpp
  lib/Target/AArch64/AArch64FrameLowering.h
  lib/Target/ARM/ARMFrameLowering.cpp
  lib/Target/ARM/ARMFrameLowering.h
  lib/Target/ARM/Thumb1FrameLowering.cpp
  lib/Target/ARM/Thumb1FrameLowering.h
  lib/Target/BPF/BPFFrameLowering.cpp
  lib/Target/BPF/BPFFrameLowering.h
  lib/Target/Hexagon/HexagonFrameLowering.cpp
  lib/Target/Hexagon/HexagonFrameLowering.h
  lib/Target/MSP430/MSP430FrameLowering.cpp
  lib/Target/MSP430/MSP430FrameLowering.h
  lib/Target/Mips/Mips16FrameLowering.cpp
  lib/Target/Mips/Mips16FrameLowering.h
  lib/Target/Mips/MipsSEFrameLowering.cpp
  lib/Target/Mips/MipsSEFrameLowering.h
  lib/Target/NVPTX/NVPTXFrameLowering.cpp
  lib/Target/NVPTX/NVPTXFrameLowering.h
  lib/Target/NVPTX/NVPTXPrologEpilogPass.cpp
  lib/Target/PowerPC/PPCFrameLowering.cpp
  lib/Target/PowerPC/PPCFrameLowering.h
  lib/Target/R600/AMDGPUFrameLowering.cpp
  lib/Target/R600/AMDGPUFrameLowering.h
  lib/Target/Sparc/SparcFrameLowering.cpp
  lib/Target/Sparc/SparcFrameLowering.h
  lib/Target/SystemZ/SystemZFrameLowering.cpp
  lib/Target/SystemZ/SystemZFrameLowering.h
  lib/Target/X86/X86FrameLowering.cpp
  lib/Target/X86/X86FrameLowering.h
  lib/Target/XCore/XCoreFrameLowering.cpp
  lib/Target/XCore/XCoreFrameLowering.h

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D9210.24260.patch
Type: text/x-patch
Size: 57901 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150422/b4b50fb3/attachment.bin>