llvm-6502/lib/CodeGen/ShrinkWrap.cpp

389 lines
14 KiB
C++
Raw Normal View History

[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
//===-- ShrinkWrap.cpp - Compute safe point for prolog/epilog insertion ---===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass looks for safe point where the prologue and epilogue can be
// inserted.
// The safe point for the prologue (resp. epilogue) is called Save
// (resp. Restore).
// A point is safe for prologue (resp. epilogue) if and only if
// it 1) dominates (resp. post-dominates) all the frame related operations and
// between 2) two executions of the Save (resp. Restore) point there is an
// execution of the Restore (resp. Save) point.
//
// For instance, the following points are safe:
// for (int i = 0; i < 10; ++i) {
// Save
// ...
// Restore
// }
// Indeed, the execution looks like Save -> Restore -> Save -> Restore ...
// And the following points are not:
// for (int i = 0; i < 10; ++i) {
// Save
// ...
// }
// for (int i = 0; i < 10; ++i) {
// ...
// Restore
// }
// Indeed, the execution looks like Save -> Save -> ... -> Restore -> Restore.
//
// This pass also ensures that the safe points are 3) cheaper than the regular
// entry and exits blocks.
//
// Property #1 is ensured via the use of MachineDominatorTree and
// MachinePostDominatorTree.
// Property #2 is ensured via property #1 and MachineLoopInfo, i.e., both
// points must be in the same loop.
// Property #3 is ensured via the MachineBlockFrequencyInfo.
//
// If this pass found points matching all this properties, then
// MachineFrameInfo is updated this that information.
//===----------------------------------------------------------------------===//
#include "llvm/ADT/Statistic.h"
// To check for profitability.
#include "llvm/CodeGen/MachineBlockFrequencyInfo.h"
// For property #1 for Save.
#include "llvm/CodeGen/MachineDominators.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
// To record the result of the analysis.
#include "llvm/CodeGen/MachineFrameInfo.h"
// For property #2.
#include "llvm/CodeGen/MachineLoopInfo.h"
// For property #1 for Restore.
#include "llvm/CodeGen/MachinePostDominators.h"
#include "llvm/CodeGen/Passes.h"
// To know about callee-saved.
#include "llvm/CodeGen/RegisterClassInfo.h"
#include "llvm/Support/Debug.h"
// To query the target about frame lowering.
#include "llvm/Target/TargetFrameLowering.h"
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
// To know about frame setup operation.
#include "llvm/Target/TargetInstrInfo.h"
// To access TargetInstrInfo.
#include "llvm/Target/TargetSubtargetInfo.h"
#define DEBUG_TYPE "shrink-wrap"
using namespace llvm;
STATISTIC(NumFunc, "Number of functions");
STATISTIC(NumCandidates, "Number of shrink-wrapping candidates");
STATISTIC(NumCandidatesDropped,
"Number of shrink-wrapping candidates dropped because of frequency");
namespace {
/// \brief Class to determine where the safe point to insert the
/// prologue and epilogue are.
/// Unlike the paper from Fred C. Chow, PLDI'88, that introduces the
/// shrink-wrapping term for prologue/epilogue placement, this pass
/// does not rely on expensive data-flow analysis. Instead we use the
/// dominance properties and loop information to decide which point
/// are safe for such insertion.
class ShrinkWrap : public MachineFunctionPass {
/// Hold callee-saved information.
RegisterClassInfo RCI;
MachineDominatorTree *MDT;
MachinePostDominatorTree *MPDT;
/// Current safe point found for the prologue.
/// The prologue will be inserted before the first instruction
/// in this basic block.
MachineBasicBlock *Save;
/// Current safe point found for the epilogue.
/// The epilogue will be inserted before the first terminator instruction
/// in this basic block.
MachineBasicBlock *Restore;
/// Hold the information of the basic block frequency.
/// Use to check the profitability of the new points.
MachineBlockFrequencyInfo *MBFI;
/// Hold the loop information. Used to determine if Save and Restore
/// are in the same loop.
MachineLoopInfo *MLI;
/// Frequency of the Entry block.
uint64_t EntryFreq;
/// Current opcode for frame setup.
unsigned FrameSetupOpcode;
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
/// Current opcode for frame destroy.
unsigned FrameDestroyOpcode;
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
/// Entry block.
const MachineBasicBlock *Entry;
/// \brief Check if \p MI uses or defines a callee-saved register or
/// a frame index. If this is the case, this means \p MI must happen
/// after Save and before Restore.
bool useOrDefCSROrFI(const MachineInstr &MI) const;
/// \brief Update the Save and Restore points such that \p MBB is in
/// the region that is dominated by Save and post-dominated by Restore
/// and Save and Restore still match the safe point definition.
/// Such point may not exist and Save and/or Restore may be null after
/// this call.
void updateSaveRestorePoints(MachineBasicBlock &MBB);
/// \brief Initialize the pass for \p MF.
void init(MachineFunction &MF) {
RCI.runOnMachineFunction(MF);
MDT = &getAnalysis<MachineDominatorTree>();
MPDT = &getAnalysis<MachinePostDominatorTree>();
Save = nullptr;
Restore = nullptr;
MBFI = &getAnalysis<MachineBlockFrequencyInfo>();
MLI = &getAnalysis<MachineLoopInfo>();
EntryFreq = MBFI->getEntryFreq();
const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
FrameSetupOpcode = TII.getCallFrameSetupOpcode();
FrameDestroyOpcode = TII.getCallFrameDestroyOpcode();
Entry = &MF.front();
++NumFunc;
}
/// Check whether or not Save and Restore points are still interesting for
/// shrink-wrapping.
bool ArePointsInteresting() const { return Save != Entry && Save && Restore; }
public:
static char ID;
ShrinkWrap() : MachineFunctionPass(ID) {
initializeShrinkWrapPass(*PassRegistry::getPassRegistry());
}
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesAll();
AU.addRequired<MachineBlockFrequencyInfo>();
AU.addRequired<MachineDominatorTree>();
AU.addRequired<MachinePostDominatorTree>();
AU.addRequired<MachineLoopInfo>();
MachineFunctionPass::getAnalysisUsage(AU);
}
const char *getPassName() const override {
return "Shrink Wrapping analysis";
}
/// \brief Perform the shrink-wrapping analysis and update
/// the MachineFrameInfo attached to \p MF with the results.
bool runOnMachineFunction(MachineFunction &MF) override;
};
} // End anonymous namespace.
char ShrinkWrap::ID = 0;
char &llvm::ShrinkWrapID = ShrinkWrap::ID;
INITIALIZE_PASS_BEGIN(ShrinkWrap, "shrink-wrap", "Shrink Wrap Pass", false,
false)
INITIALIZE_PASS_DEPENDENCY(MachineBlockFrequencyInfo)
INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree)
INITIALIZE_PASS_DEPENDENCY(MachinePostDominatorTree)
INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo)
INITIALIZE_PASS_END(ShrinkWrap, "shrink-wrap", "Shrink Wrap Pass", false, false)
bool ShrinkWrap::useOrDefCSROrFI(const MachineInstr &MI) const {
if (MI.getOpcode() == FrameSetupOpcode ||
MI.getOpcode() == FrameDestroyOpcode) {
DEBUG(dbgs() << "Frame instruction: " << MI << '\n');
return true;
}
for (const MachineOperand &MO : MI.operands()) {
bool UseCSR = false;
if (MO.isReg()) {
unsigned PhysReg = MO.getReg();
if (!PhysReg)
continue;
assert(TargetRegisterInfo::isPhysicalRegister(PhysReg) &&
"Unallocated register?!");
UseCSR = RCI.getLastCalleeSavedAlias(PhysReg);
}
// TODO: Handle regmask more accurately.
// For now, be conservative about them.
if (UseCSR || MO.isFI() || MO.isRegMask()) {
DEBUG(dbgs() << "Use or define CSR(" << UseCSR << ") or FI(" << MO.isFI()
<< "): " << MI << '\n');
return true;
}
}
return false;
}
/// \brief Helper function to find the immediate (post) dominator.
template <typename ListOfBBs, typename DominanceAnalysis>
MachineBasicBlock *FindIDom(MachineBasicBlock &Block, ListOfBBs BBs,
DominanceAnalysis &Dom) {
MachineBasicBlock *IDom = &Block;
for (MachineBasicBlock *BB : BBs) {
IDom = Dom.findNearestCommonDominator(IDom, BB);
if (!IDom)
break;
}
return IDom;
}
void ShrinkWrap::updateSaveRestorePoints(MachineBasicBlock &MBB) {
// Get rid of the easy cases first.
if (!Save)
Save = &MBB;
else
Save = MDT->findNearestCommonDominator(Save, &MBB);
if (!Save) {
DEBUG(dbgs() << "Found a block that is not reachable from Entry\n");
return;
}
if (!Restore)
Restore = &MBB;
else
Restore = MPDT->findNearestCommonDominator(Restore, &MBB);
// Make sure we would be able to insert the restore code before the
// terminator.
if (Restore == &MBB) {
for (const MachineInstr &Terminator : MBB.terminators()) {
if (!useOrDefCSROrFI(Terminator))
continue;
// One of the terminator needs to happen before the restore point.
if (MBB.succ_empty()) {
Restore = nullptr;
break;
}
// Look for a restore point that post-dominates all the successors.
// The immediate post-dominator is what we are looking for.
Restore = FindIDom<>(*Restore, Restore->successors(), *MPDT);
break;
}
}
if (!Restore) {
DEBUG(dbgs() << "Restore point needs to be spanned on several blocks\n");
return;
}
// Make sure Save and Restore are suitable for shrink-wrapping:
// 1. all path from Save needs to lead to Restore before exiting.
// 2. all path to Restore needs to go through Save from Entry.
// We achieve that by making sure that:
// A. Save dominates Restore.
// B. Restore post-dominates Save.
// C. Save and Restore are in the same loop.
bool SaveDominatesRestore = false;
bool RestorePostDominatesSave = false;
while (Save && Restore &&
(!(SaveDominatesRestore = MDT->dominates(Save, Restore)) ||
!(RestorePostDominatesSave = MPDT->dominates(Restore, Save)) ||
MLI->getLoopFor(Save) != MLI->getLoopFor(Restore))) {
// Fix (A).
if (!SaveDominatesRestore) {
Save = MDT->findNearestCommonDominator(Save, Restore);
continue;
}
// Fix (B).
if (!RestorePostDominatesSave)
Restore = MPDT->findNearestCommonDominator(Restore, Save);
// Fix (C).
if (Save && Restore && Save != Restore &&
MLI->getLoopFor(Save) != MLI->getLoopFor(Restore)) {
if (MLI->getLoopDepth(Save) > MLI->getLoopDepth(Restore))
// Push Save outside of this loop.
Save = FindIDom<>(*Save, Save->predecessors(), *MDT);
else
// Push Restore outside of this loop.
Restore = FindIDom<>(*Restore, Restore->successors(), *MPDT);
}
}
}
bool ShrinkWrap::runOnMachineFunction(MachineFunction &MF) {
if (MF.empty())
return false;
DEBUG(dbgs() << "**** Analysing " << MF.getName() << '\n');
init(MF);
for (MachineBasicBlock &MBB : MF) {
DEBUG(dbgs() << "Look into: " << MBB.getNumber() << ' ' << MBB.getName()
<< '\n');
for (const MachineInstr &MI : MBB) {
if (!useOrDefCSROrFI(MI))
continue;
// Save (resp. restore) point must dominate (resp. post dominate)
// MI. Look for the proper basic block for those.
updateSaveRestorePoints(MBB);
// If we are at a point where we cannot improve the placement of
// save/restore instructions, just give up.
if (!ArePointsInteresting()) {
DEBUG(dbgs() << "No Shrink wrap candidate found\n");
return false;
}
// No need to look for other instructions, this basic block
// will already be part of the handled region.
break;
}
}
if (!ArePointsInteresting()) {
// If the points are not interesting at this point, then they must be null
// because it means we did not encounter any frame/CSR related code.
// Otherwise, we would have returned from the previous loop.
assert(!Save && !Restore && "We miss a shrink-wrap opportunity?!");
DEBUG(dbgs() << "Nothing to shrink-wrap\n");
return false;
}
DEBUG(dbgs() << "\n ** Results **\nFrequency of the Entry: " << EntryFreq
<< '\n');
const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering();
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
do {
DEBUG(dbgs() << "Shrink wrap candidates (#, Name, Freq):\nSave: "
<< Save->getNumber() << ' ' << Save->getName() << ' '
<< MBFI->getBlockFreq(Save).getFrequency() << "\nRestore: "
<< Restore->getNumber() << ' ' << Restore->getName() << ' '
<< MBFI->getBlockFreq(Restore).getFrequency() << '\n');
bool IsSaveCheap, TargetCanUseSaveAsPrologue = false;
if (((IsSaveCheap = EntryFreq >= MBFI->getBlockFreq(Save).getFrequency()) &&
EntryFreq >= MBFI->getBlockFreq(Restore).getFrequency()) &&
((TargetCanUseSaveAsPrologue = TFI->canUseAsPrologue(*Save)) &&
TFI->canUseAsEpilogue(*Restore)))
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
break;
DEBUG(dbgs() << "New points are too expensive or invalid for the target\n");
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
MachineBasicBlock *NewBB;
if (!IsSaveCheap || !TargetCanUseSaveAsPrologue) {
[ShrinkWrap] Add (a simplified version) of shrink-wrapping. This patch introduces a new pass that computes the safe point to insert the prologue and epilogue of the function. The interest is to find safe points that are cheaper than the entry and exits blocks. As an example and to avoid regressions to be introduce, this patch also implements the required bits to enable the shrink-wrapping pass for AArch64. ** 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. Differential Revision: http://reviews.llvm.org/D9210 <rdar://problem/3201744> git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@236507 91177308-0d34-0410-b5e6-96231b3b80d8
2015-05-05 17:38:16 +00:00
Save = FindIDom<>(*Save, Save->predecessors(), *MDT);
if (!Save)
break;
NewBB = Save;
} else {
// Restore is expensive.
Restore = FindIDom<>(*Restore, Restore->successors(), *MPDT);
if (!Restore)
break;
NewBB = Restore;
}
updateSaveRestorePoints(*NewBB);
} while (Save && Restore);
if (!ArePointsInteresting()) {
++NumCandidatesDropped;
return false;
}
DEBUG(dbgs() << "Final shrink wrap candidates:\nSave: " << Save->getNumber()
<< ' ' << Save->getName() << "\nRestore: "
<< Restore->getNumber() << ' ' << Restore->getName() << '\n');
MachineFrameInfo *MFI = MF.getFrameInfo();
MFI->setSavePoint(Save);
MFI->setRestorePoint(Restore);
++NumCandidates;
return false;
}