LLVM 初探(2) Optimization Pass 编写

Dead code elimination Pass

源码如下，发现llvm cookbook用的api挺老的，把旧的api换成新的了

#include "llvm/Transforms/Scalar.h"
#include "llvm/ADT/DepthFirstIterator.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/CFG.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Function.h"
#include "llvm/Pass.h"
using namespace llvm;

namespace {

struct MYADCE : public FunctionPass {
    static char ID;
    MYADCE() : FunctionPass(ID) {
    }
    bool runOnFunction(Function& F) override;
    void getAnalysisUsage(AnalysisUsage& AU) const override{
        AU.setPreservesCFG();
    }
};
}

char MYADCE::ID = 0;

bool MYADCE::runOnFunction(Function& F) {
    if (skipFunction(F))
        return false;
    SmallPtrSet<Instruction*, 32> Alive;
    SmallVector<Instruction*, 128> Worklist;
    for(Instruction & I:instructions(F)){
        if(isa<TerminatorInst>(I) || isa<DbgInfoIntrinsic>(I)
            || isa<LandingPadInst>(I) || I.mayHaveSideEffects()){
            Alive.insert(&I);
            Worklist.push_back(&I);
        }
    }
    while(!Worklist.empty()){
        Instruction* Curr = Worklist.pop_back_val();
        for( Use &OI : Curr->operands()) {
            if (Instruction* Inst = dyn_cast<Instruction>(OI))
                if(Alive.insert(Inst).second)
                    Worklist.push_back(Inst);
        }
    }
    for(Instruction &I : instructions(F)){
        if(!Alive.count(&I)){
            Worklist.push_back(&I);
            I.dropAllReferences();
        }
    }
    for(Instruction* I : Worklist){
        I->eraseFromParent();
    }
    return !Worklist.empty();
    
    return false;
}
static RegisterPass<MYADCE> X("myadce", "My Aggressive Dead Code Elimination", false, false);

测试用IR如下

declare i32 @strlen(i8*) readonly nounwind
define void @test() {
    call i32 @strlen( i8* null )
    ret void
}

Pass首先是遍历Instruction，寻找这4类Instruction：TerminatorInst，DbgInfoIntrinsic，LandingPadInst，mayHaveSideEffects，也就是最后有意义的指令，然后用bfs反向遍历回去，类似于污点传播的感觉，把所有有意义的指令都插入到Alive里面，最后把不在Alive里面的指令都去除掉

Inlining Transformation Pass

源码如下，因为cookbook的都是旧版本的，所以我下了源码，用lib/Transforms/IPO/InlineSimple.cpp魔改了下cookbook的例子

#include "llvm/Transforms/IPO.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/CallGraph.h"
#include "llvm/Analysis/InlineCost.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/Type.h"
#include "llvm/Transforms/IPO/Inliner.h"

using namespace llvm;
namespace {

class MyInliner : public LegacyInlinerBase {
    
    public:
    MyInliner(): LegacyInlinerBase(ID) {}
    
    static char ID;
    
    InlineCost getInlineCost(CallSite CS) override;
    void getAnalysisUsage(AnalysisUsage &AU) const override;

    bool runOnSCC(CallGraphSCC & SCC) override;
    bool doFinalization(CallGraph &CG) override {
        return removeDeadFunctions(CG,true);
    }
};
}

char MyInliner::ID = 0;
InlineCost MyInliner::getInlineCost(CallSite CS){
    Function* Callee = CS.getCalledFunction();
    if(Callee && !Callee->isDeclaration() && 
        CS.hasFnAttr(Attribute::AlwaysInline))
        return InlineCost::getAlways();
    return InlineCost::getNever();
}

bool MyInliner::runOnSCC(CallGraphSCC & SCC) {
    return LegacyInlinerBase::runOnSCC(SCC);
}

void MyInliner::getAnalysisUsage(AnalysisUsage &AU) const {
    AU.addRequired<TargetTransformInfoWrapperPass>();
    LegacyInlinerBase::getAnalysisUsage(AU);
}

static RegisterPass<MyInliner> X("myinliner", "My Inliner", false, false);

测试IR如下

define i32 @inner1() alwaysinline {
    ret i32 1
}
define i32 @outer1() {
    %r = call i32 @inner1()
    ret i32 %r
}

运行效果

这个Pass是将标志了alwaysinline的函数全部都内联，其他都不内联