view gen/optimizer.cpp @ 1047:6bb04dbee21f

Some calling convention work for x86-64: - Implement x86-64 extern(C), hopefully correctly. - Tried to be a bit smarter about extern(D) while I was there. Interestingly, this code seems to be generating more efficient code than gcc and llvm-gcc in some edge cases, like returning a `{ [7 x i8] }` loaded from a stack slot from an extern(C) function. (gcc generates 7 1-byte loads, while this code generates a 4-byte, a 2-byte and a 1-byte load) I also added some changes to make sure structs being returned from functions or passed in as parameters are stored in memory where the rest of the backend seems to expect them to be. These should be removed when support for first-class aggregates improves.
author Frits van Bommel <fvbommel wxs.nl>
date Fri, 06 Mar 2009 16:00:47 +0100
parents 6aaa3d3c1183
children e40c65bd8c5d
line wrap: on
line source

#include "llvm/PassManager.h"
#include "llvm/LinkAllPasses.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Target/TargetData.h"

using namespace llvm;

//////////////////////////////////////////////////////////////////////////////////////////

// this function runs some or all of the std-compile-opts passes depending on the
// optimization level given.

void ldc_optimize_module(Module* m, char lvl, bool doinline)
{
    if (!doinline && lvl < 0)
        return;

    PassManager pm;
    pm.add(new TargetData(m));

    // -O0
    if (lvl >= 0)
    {
        //pm.add(createStripDeadPrototypesPass());
        pm.add(createGlobalDCEPass());
    }

    // -O1
    if (lvl >= 1)
    {
        pm.add(createRaiseAllocationsPass());
        pm.add(createCFGSimplificationPass());
        pm.add(createPromoteMemoryToRegisterPass());
        pm.add(createGlobalOptimizerPass());
        pm.add(createGlobalDCEPass());
    }

    // -O2
    if (lvl >= 2)
    {
        pm.add(createIPConstantPropagationPass());
        pm.add(createDeadArgEliminationPass());
        pm.add(createInstructionCombiningPass());
        pm.add(createCFGSimplificationPass());
        pm.add(createPruneEHPass());
    }

    // -inline
    if (doinline) {
        pm.add(createFunctionInliningPass());
    }

    // -O3
    if (lvl >= 3)
    {
        pm.add(createArgumentPromotionPass());
        pm.add(createTailDuplicationPass());
        pm.add(createInstructionCombiningPass());
        pm.add(createCFGSimplificationPass());
        pm.add(createScalarReplAggregatesPass());
        pm.add(createInstructionCombiningPass());
        pm.add(createCondPropagationPass());

        pm.add(createTailCallEliminationPass());
        pm.add(createCFGSimplificationPass());
        pm.add(createReassociatePass());
        pm.add(createLoopRotatePass());
        pm.add(createLICMPass());
        pm.add(createLoopUnswitchPass());
        pm.add(createInstructionCombiningPass());
        pm.add(createIndVarSimplifyPass());
        pm.add(createLoopUnrollPass());
        pm.add(createInstructionCombiningPass());
        pm.add(createGVNPass());
        pm.add(createSCCPPass());

        pm.add(createInstructionCombiningPass());
        pm.add(createCondPropagationPass());

        pm.add(createDeadStoreEliminationPass());
        pm.add(createAggressiveDCEPass());
        pm.add(createCFGSimplificationPass());
        pm.add(createSimplifyLibCallsPass());
        pm.add(createDeadTypeEliminationPass());
        pm.add(createConstantMergePass());
    }

    // level -O4 and -O5 are linktime optimizations

    pm.run(*m);
}