X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTransforms%2FIPO%2FPassManagerBuilder.cpp;h=701fb462b4fdad1e6b50bf1bcdee2f8fab2d4d0e;hp=66d979cf2a7e9df5669d23595a64ae54473708cb;hb=d94715e273d879eab53684835375ddbed7eb1eb3;hpb=713aa9431d84805c5b7ddce57159d14ee94cd548 diff --git a/lib/Transforms/IPO/PassManagerBuilder.cpp b/lib/Transforms/IPO/PassManagerBuilder.cpp index 66d979cf2a7..701fb462b4f 100644 --- a/lib/Transforms/IPO/PassManagerBuilder.cpp +++ b/lib/Transforms/IPO/PassManagerBuilder.cpp @@ -14,26 +14,31 @@ #include "llvm/Transforms/IPO/PassManagerBuilder.h" - #include "llvm-c/Transforms/PassManagerBuilder.h" - -#include "llvm/PassManager.h" -#include "llvm/DefaultPasses.h" -#include "llvm/PassManager.h" +#include "llvm/ADT/SmallVector.h" #include "llvm/Analysis/Passes.h" -#include "llvm/Analysis/Verifier.h" +#include "llvm/IR/Verifier.h" +#include "llvm/PassManager.h" #include "llvm/Support/CommandLine.h" +#include "llvm/Support/ManagedStatic.h" #include "llvm/Target/TargetLibraryInfo.h" +#include "llvm/Transforms/IPO.h" #include "llvm/Transforms/Scalar.h" #include "llvm/Transforms/Vectorize.h" -#include "llvm/Transforms/IPO.h" -#include "llvm/ADT/SmallVector.h" -#include "llvm/Support/ManagedStatic.h" using namespace llvm; static cl::opt -RunVectorization("vectorize", cl::desc("Run vectorization passes")); +RunLoopVectorization("vectorize-loops", cl::Hidden, + cl::desc("Run the Loop vectorization passes")); + +static cl::opt +RunSLPVectorization("vectorize-slp", cl::Hidden, + cl::desc("Run the SLP vectorization passes")); + +static cl::opt +RunBBVectorization("vectorize-slp-aggressive", cl::Hidden, + cl::desc("Run the BB vectorization passes")); static cl::opt UseGVNAfterVectorization("use-gvn-after-vectorization", @@ -44,15 +49,27 @@ static cl::opt UseNewSROA("use-new-sroa", cl::init(true), cl::Hidden, cl::desc("Enable the new, experimental SROA pass")); +static cl::opt +RunLoopRerolling("reroll-loops", cl::Hidden, + cl::desc("Run the loop rerolling pass")); + +static cl::opt RunLoadCombine("combine-loads", cl::init(false), + cl::Hidden, + cl::desc("Run the load combining pass")); + PassManagerBuilder::PassManagerBuilder() { OptLevel = 2; SizeLevel = 0; - LibraryInfo = 0; - Inliner = 0; - DisableSimplifyLibCalls = false; + LibraryInfo = nullptr; + Inliner = nullptr; + DisableTailCalls = false; DisableUnitAtATime = false; DisableUnrollLoops = false; - Vectorize = RunVectorization; + BBVectorize = RunBBVectorization; + SLPVectorize = RunSLPVectorization; + LoopVectorize = RunLoopVectorization; + RerollLoops = RunLoopRerolling; + LoadCombine = RunLoadCombine; } PassManagerBuilder::~PassManagerBuilder() { @@ -117,8 +134,16 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) { if (OptLevel == 0) { if (Inliner) { MPM.add(Inliner); - Inliner = 0; + Inliner = nullptr; } + + // FIXME: This is a HACK! The inliner pass above implicitly creates a CGSCC + // pass manager, but we don't want to add extensions into that pass manager. + // To prevent this we must insert a no-op module pass to reset the pass + // manager to get the same behavior as EP_OptimizerLast in non-O0 builds. + if (!GlobalExtensions->empty() || !Extensions.empty()) + MPM.add(createBarrierNoopPass()); + addExtensionsToPM(EP_EnabledOnOptLevel0, MPM); return; } @@ -131,12 +156,13 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) { if (!DisableUnitAtATime) { addExtensionsToPM(EP_ModuleOptimizerEarly, MPM); + MPM.add(createIPSCCPPass()); // IP SCCP MPM.add(createGlobalOptimizerPass()); // Optimize out global vars - MPM.add(createIPSCCPPass()); // IP SCCP MPM.add(createDeadArgEliminationPass()); // Dead argument elimination MPM.add(createInstructionCombiningPass());// Clean up after IPCP & DAE + addExtensionsToPM(EP_Peephole, MPM); MPM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE } @@ -145,7 +171,7 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) { MPM.add(createPruneEHPass()); // Remove dead EH info if (Inliner) { MPM.add(Inliner); - Inliner = 0; + Inliner = nullptr; } if (!DisableUnitAtATime) MPM.add(createFunctionAttrsPass()); // Set readonly/readnone attrs @@ -154,16 +180,19 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) { // Start of function pass. // Break up aggregate allocas, using SSAUpdater. - MPM.add(createScalarReplAggregatesPass(-1, false)); + if (UseNewSROA) + MPM.add(createSROAPass(/*RequiresDomTree*/ false)); + else + MPM.add(createScalarReplAggregatesPass(-1, false)); MPM.add(createEarlyCSEPass()); // Catch trivial redundancies - if (!DisableSimplifyLibCalls) - MPM.add(createSimplifyLibCallsPass()); // Library Call Optimizations MPM.add(createJumpThreadingPass()); // Thread jumps. MPM.add(createCorrelatedValuePropagationPass()); // Propagate conditionals MPM.add(createCFGSimplificationPass()); // Merge & remove BBs MPM.add(createInstructionCombiningPass()); // Combine silly seq's + addExtensionsToPM(EP_Peephole, MPM); - MPM.add(createTailCallEliminationPass()); // Eliminate tail calls + if (!DisableTailCalls) + MPM.add(createTailCallEliminationPass()); // Eliminate tail calls MPM.add(createCFGSimplificationPass()); // Merge & remove BBs MPM.add(createReassociatePass()); // Reassociate expressions MPM.add(createLoopRotatePass()); // Rotate Loop @@ -173,48 +202,82 @@ void PassManagerBuilder::populateModulePassManager(PassManagerBase &MPM) { MPM.add(createIndVarSimplifyPass()); // Canonicalize indvars MPM.add(createLoopIdiomPass()); // Recognize idioms like memset. MPM.add(createLoopDeletionPass()); // Delete dead loops + if (!DisableUnrollLoops) - MPM.add(createLoopUnrollPass()); // Unroll small loops + MPM.add(createSimpleLoopUnrollPass()); // Unroll small loops addExtensionsToPM(EP_LoopOptimizerEnd, MPM); - if (OptLevel > 1) + if (OptLevel > 1) { + MPM.add(createMergedLoadStoreMotionPass()); // Merge load/stores in diamond MPM.add(createGVNPass()); // Remove redundancies + } MPM.add(createMemCpyOptPass()); // Remove memcpy / form memset MPM.add(createSCCPPass()); // Constant prop with SCCP // Run instcombine after redundancy elimination to exploit opportunities // opened up by them. MPM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, MPM); MPM.add(createJumpThreadingPass()); // Thread jumps MPM.add(createCorrelatedValuePropagationPass()); MPM.add(createDeadStoreEliminationPass()); // Delete dead stores addExtensionsToPM(EP_ScalarOptimizerLate, MPM); - if (Vectorize) { + if (RerollLoops) + MPM.add(createLoopRerollPass()); + if (SLPVectorize) + MPM.add(createSLPVectorizerPass()); // Vectorize parallel scalar chains. + + if (BBVectorize) { MPM.add(createBBVectorizePass()); MPM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, MPM); if (OptLevel > 1 && UseGVNAfterVectorization) - MPM.add(createGVNPass()); // Remove redundancies + MPM.add(createGVNPass()); // Remove redundancies else - MPM.add(createEarlyCSEPass()); // Catch trivial redundancies + MPM.add(createEarlyCSEPass()); // Catch trivial redundancies + + // BBVectorize may have significantly shortened a loop body; unroll again. + if (!DisableUnrollLoops) + MPM.add(createLoopUnrollPass()); } + if (LoadCombine) + MPM.add(createLoadCombinePass()); + MPM.add(createAggressiveDCEPass()); // Delete dead instructions - MPM.add(createCFGSimplificationPass()); // Merge & remove BBs + MPM.add(createCFGSimplificationPass()); // Merge & remove BBs MPM.add(createInstructionCombiningPass()); // Clean up after everything. + addExtensionsToPM(EP_Peephole, MPM); + + // FIXME: This is a HACK! The inliner pass above implicitly creates a CGSCC + // pass manager that we are specifically trying to avoid. To prevent this + // we must insert a no-op module pass to reset the pass manager. + MPM.add(createBarrierNoopPass()); + MPM.add(createLoopVectorizePass(DisableUnrollLoops, LoopVectorize)); + // FIXME: Because of #pragma vectorize enable, the passes below are always + // inserted in the pipeline, even when the vectorizer doesn't run (ex. when + // on -O1 and no #pragma is found). Would be good to have these two passes + // as function calls, so that we can only pass them when the vectorizer + // changed the code. + MPM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, MPM); + MPM.add(createCFGSimplificationPass()); + + if (!DisableUnrollLoops) + MPM.add(createLoopUnrollPass()); // Unroll small loops if (!DisableUnitAtATime) { // FIXME: We shouldn't bother with this anymore. MPM.add(createStripDeadPrototypesPass()); // Get rid of dead prototypes - // GlobalOpt already deletes dead functions and globals, at -O3 try a + // GlobalOpt already deletes dead functions and globals, at -O2 try a // late pass of GlobalDCE. It is capable of deleting dead cycles. - if (OptLevel > 2) + if (OptLevel > 1) { MPM.add(createGlobalDCEPass()); // Remove dead fns and globals. - - if (OptLevel > 1) MPM.add(createConstantMergePass()); // Merge dup global constants + } } addExtensionsToPM(EP_OptimizerLast, MPM); } @@ -230,7 +293,7 @@ void PassManagerBuilder::populateLTOPassManager(PassManagerBase &PM, // for a main function. If main is defined, mark all other functions // internal. if (Internalize) - PM.add(createInternalizePass(true)); + PM.add(createInternalizePass("main")); // Propagate constants at call sites into the functions they call. This // opens opportunities for globalopt (and inlining) by substituting function @@ -252,6 +315,7 @@ void PassManagerBuilder::populateLTOPassManager(PassManagerBase &PM, // function pointers. When this happens, we often have to resolve varargs // calls, etc, so let instcombine do this. PM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, PM); // Inline small functions if (RunInliner) @@ -270,7 +334,9 @@ void PassManagerBuilder::populateLTOPassManager(PassManagerBase &PM, // The IPO passes may leave cruft around. Clean up after them. PM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, PM); PM.add(createJumpThreadingPass()); + // Break up allocas if (UseNewSROA) PM.add(createSROAPass()); @@ -282,13 +348,27 @@ void PassManagerBuilder::populateLTOPassManager(PassManagerBase &PM, PM.add(createGlobalsModRefPass()); // IP alias analysis. PM.add(createLICMPass()); // Hoist loop invariants. + PM.add(createMergedLoadStoreMotionPass()); // Merge load/stores in diamonds PM.add(createGVNPass(DisableGVNLoadPRE)); // Remove redundancies. PM.add(createMemCpyOptPass()); // Remove dead memcpys. + // Nuke dead stores. PM.add(createDeadStoreEliminationPass()); + // More loops are countable; try to optimize them. + PM.add(createIndVarSimplifyPass()); + PM.add(createLoopDeletionPass()); + PM.add(createLoopVectorizePass(true, true)); + + // More scalar chains could be vectorized due to more alias information + PM.add(createSLPVectorizerPass()); // Vectorize parallel scalar chains. + + if (LoadCombine) + PM.add(createLoadCombinePass()); + // Cleanup and simplify the code after the scalar optimizations. PM.add(createInstructionCombiningPass()); + addExtensionsToPM(EP_Peephole, PM); PM.add(createJumpThreadingPass()); @@ -299,7 +379,15 @@ void PassManagerBuilder::populateLTOPassManager(PassManagerBase &PM, PM.add(createGlobalDCEPass()); } -LLVMPassManagerBuilderRef LLVMPassManagerBuilderCreate(void) { +inline PassManagerBuilder *unwrap(LLVMPassManagerBuilderRef P) { + return reinterpret_cast(P); +} + +inline LLVMPassManagerBuilderRef wrap(PassManagerBuilder *P) { + return reinterpret_cast(P); +} + +LLVMPassManagerBuilderRef LLVMPassManagerBuilderCreate() { PassManagerBuilder *PMB = new PassManagerBuilder(); return wrap(PMB); } @@ -340,8 +428,7 @@ LLVMPassManagerBuilderSetDisableUnrollLoops(LLVMPassManagerBuilderRef PMB, void LLVMPassManagerBuilderSetDisableSimplifyLibCalls(LLVMPassManagerBuilderRef PMB, LLVMBool Value) { - PassManagerBuilder *Builder = unwrap(PMB); - Builder->DisableSimplifyLibCalls = Value; + // NOTE: The simplify-libcalls pass has been removed. } void @@ -369,9 +456,9 @@ LLVMPassManagerBuilderPopulateModulePassManager(LLVMPassManagerBuilderRef PMB, void LLVMPassManagerBuilderPopulateLTOPassManager(LLVMPassManagerBuilderRef PMB, LLVMPassManagerRef PM, - bool Internalize, - bool RunInliner) { + LLVMBool Internalize, + LLVMBool RunInliner) { PassManagerBuilder *Builder = unwrap(PMB); PassManagerBase *LPM = unwrap(PM); - Builder->populateLTOPassManager(*LPM, Internalize, RunInliner); + Builder->populateLTOPassManager(*LPM, Internalize != 0, RunInliner != 0); }