Yaxun (Sam) Liu
f77162dd86
[Clang][Driver] Parallelize device cc1 jobs for offload arches Large offload builds can target many device architectures. A common ROCm build can target around ten `--offload-arch=` values. For a large single translation unit, the offload device cc1 work before clang-linker-wrapper can become a build-time bottleneck. Those per-architecture jobs are independent, but the driver ran them serially, so `--offload-jobs=` only helped the later wrapper work. Borrow the parallel job mechanism used by clang-linker-wrapper for the pre-wrapper device cc1 jobs. Device job construction marks eligible compile and backend jobs with an offload parallel group. The generic executor only consumes that opt-in metadata for adjacent device jobs with distinct bound architectures. The parallel path is disabled for driver-side output modes and callbacks so existing serial output handling is preserved. This is NFC for non-offload compilation.
457 lines
15 KiB
C++
457 lines
15 KiB
C++
//===- Compilation.cpp - Compilation Task Implementation ------------------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "clang/Driver/Compilation.h"
|
|
#include "clang/Basic/LLVM.h"
|
|
#include "clang/Driver/Action.h"
|
|
#include "clang/Driver/CommonArgs.h"
|
|
#include "clang/Driver/Driver.h"
|
|
#include "clang/Driver/Job.h"
|
|
#include "clang/Driver/ToolChain.h"
|
|
#include "clang/Driver/Util.h"
|
|
#include "clang/Options/Options.h"
|
|
#include "llvm/ADT/STLExtras.h"
|
|
#include "llvm/Option/ArgList.h"
|
|
#include "llvm/Option/OptSpecifier.h"
|
|
#include "llvm/Option/Option.h"
|
|
#include "llvm/Support/FileSystem.h"
|
|
#include "llvm/Support/ThreadPool.h"
|
|
#include "llvm/Support/Threading.h"
|
|
#include "llvm/Support/raw_ostream.h"
|
|
#include "llvm/TargetParser/Triple.h"
|
|
#include <algorithm>
|
|
#include <cassert>
|
|
#include <optional>
|
|
#include <string>
|
|
#include <system_error>
|
|
#include <utility>
|
|
|
|
using namespace clang;
|
|
using namespace driver;
|
|
using namespace llvm::opt;
|
|
|
|
Compilation::Compilation(const Driver &D, const ToolChain &_DefaultToolChain,
|
|
InputArgList *_Args, DerivedArgList *_TranslatedArgs,
|
|
bool ContainsError)
|
|
: TheDriver(D), DefaultToolChain(_DefaultToolChain), Args(_Args),
|
|
TranslatedArgs(_TranslatedArgs), ContainsError(ContainsError) {
|
|
// The offloading host toolchain is the default toolchain.
|
|
OrderedOffloadingToolchains.insert(
|
|
std::make_pair(Action::OFK_Host, &DefaultToolChain));
|
|
}
|
|
|
|
Compilation::~Compilation() {
|
|
// Remove temporary files. This must be done before arguments are freed, as
|
|
// the file names might be derived from the input arguments.
|
|
if (!TheDriver.isSaveTempsEnabled() && !ForceKeepTempFiles)
|
|
CleanupFileList(TempFiles);
|
|
|
|
delete TranslatedArgs;
|
|
delete Args;
|
|
|
|
// Free any derived arg lists.
|
|
for (auto Arg : TCArgs)
|
|
if (Arg.second != TranslatedArgs)
|
|
delete Arg.second;
|
|
}
|
|
|
|
const DerivedArgList &
|
|
Compilation::getArgsForToolChain(const ToolChain *TC, BoundArch BA,
|
|
Action::OffloadKind DeviceOffloadKind) {
|
|
if (!TC)
|
|
TC = &DefaultToolChain;
|
|
|
|
DerivedArgList *&Entry = TCArgs[{TC, BA, DeviceOffloadKind}];
|
|
if (!Entry) {
|
|
SmallVector<Arg *, 4> AllocatedArgs;
|
|
DerivedArgList *OpenMPArgs = nullptr;
|
|
// Translate OpenMP toolchain arguments provided via the -Xopenmp-target flags.
|
|
if (DeviceOffloadKind == Action::OFK_OpenMP) {
|
|
const ToolChain *HostTC = getSingleOffloadToolChain<Action::OFK_Host>();
|
|
bool SameTripleAsHost = (TC->getTriple() == HostTC->getTriple());
|
|
OpenMPArgs = TC->TranslateOpenMPTargetArgs(
|
|
*TranslatedArgs, SameTripleAsHost, AllocatedArgs);
|
|
}
|
|
|
|
DerivedArgList *NewDAL = nullptr;
|
|
if (!OpenMPArgs) {
|
|
NewDAL = TC->TranslateXarchArgs(*TranslatedArgs, BA, DeviceOffloadKind,
|
|
&AllocatedArgs);
|
|
} else {
|
|
NewDAL = TC->TranslateXarchArgs(*OpenMPArgs, BA, DeviceOffloadKind,
|
|
&AllocatedArgs);
|
|
if (!NewDAL)
|
|
NewDAL = OpenMPArgs;
|
|
else
|
|
delete OpenMPArgs;
|
|
}
|
|
|
|
if (!NewDAL) {
|
|
Entry = TC->TranslateArgs(*TranslatedArgs, BA, DeviceOffloadKind);
|
|
if (!Entry)
|
|
Entry = TranslatedArgs;
|
|
} else {
|
|
Entry = TC->TranslateArgs(*NewDAL, BA, DeviceOffloadKind);
|
|
if (!Entry)
|
|
Entry = NewDAL;
|
|
else
|
|
delete NewDAL;
|
|
}
|
|
|
|
// Add allocated arguments to the final DAL.
|
|
for (auto *ArgPtr : AllocatedArgs)
|
|
Entry->AddSynthesizedArg(ArgPtr);
|
|
}
|
|
|
|
return *Entry;
|
|
}
|
|
|
|
bool Compilation::CleanupFile(const char *File, bool IssueErrors) const {
|
|
// FIXME: Why are we trying to remove files that we have not created? For
|
|
// example we should only try to remove a temporary assembly file if
|
|
// "clang -cc1" succeed in writing it. Was this a workaround for when
|
|
// clang was writing directly to a .s file and sometimes leaving it behind
|
|
// during a failure?
|
|
|
|
// FIXME: If this is necessary, we can still try to split
|
|
// llvm::sys::fs::remove into a removeFile and a removeDir and avoid the
|
|
// duplicated stat from is_regular_file.
|
|
|
|
// Don't try to remove files which we don't have write access to (but may be
|
|
// able to remove), or non-regular files. Underlying tools may have
|
|
// intentionally not overwritten them.
|
|
if (!llvm::sys::fs::can_write(File) || !llvm::sys::fs::is_regular_file(File))
|
|
return true;
|
|
|
|
if (std::error_code EC = llvm::sys::fs::remove(File)) {
|
|
// Failure is only failure if the file exists and is "regular". We checked
|
|
// for it being regular before, and llvm::sys::fs::remove ignores ENOENT,
|
|
// so we don't need to check again.
|
|
|
|
if (IssueErrors)
|
|
getDriver().Diag(diag::err_drv_unable_to_remove_file)
|
|
<< EC.message();
|
|
return false;
|
|
}
|
|
return true;
|
|
}
|
|
|
|
bool Compilation::CleanupFileList(const llvm::opt::ArgStringList &Files,
|
|
bool IssueErrors) const {
|
|
bool Success = true;
|
|
for (const auto &File: Files)
|
|
Success &= CleanupFile(File, IssueErrors);
|
|
return Success;
|
|
}
|
|
|
|
bool Compilation::CleanupFileMap(const ArgStringMap &Files,
|
|
const JobAction *JA,
|
|
bool IssueErrors) const {
|
|
bool Success = true;
|
|
for (const auto &File : Files) {
|
|
// If specified, only delete the files associated with the JobAction.
|
|
// Otherwise, delete all files in the map.
|
|
if (JA && File.first != JA)
|
|
continue;
|
|
Success &= CleanupFile(File.second, IssueErrors);
|
|
}
|
|
return Success;
|
|
}
|
|
|
|
int Compilation::ExecuteCommand(const Command &C,
|
|
const Command *&FailingCommand,
|
|
bool LogOnly) const {
|
|
if ((getDriver().CCPrintOptions ||
|
|
getArgs().hasArg(options::OPT_v)) && !getDriver().CCGenDiagnostics) {
|
|
raw_ostream *OS = &llvm::errs();
|
|
std::unique_ptr<llvm::raw_fd_ostream> OwnedStream;
|
|
|
|
// Follow gcc implementation of CC_PRINT_OPTIONS; we could also cache the
|
|
// output stream.
|
|
if (getDriver().CCPrintOptions &&
|
|
!getDriver().CCPrintOptionsFilename.empty()) {
|
|
std::error_code EC;
|
|
OwnedStream.reset(new llvm::raw_fd_ostream(
|
|
getDriver().CCPrintOptionsFilename, EC,
|
|
llvm::sys::fs::OF_Append | llvm::sys::fs::OF_TextWithCRLF));
|
|
if (EC) {
|
|
getDriver().Diag(diag::err_drv_cc_print_options_failure)
|
|
<< EC.message();
|
|
FailingCommand = &C;
|
|
return 1;
|
|
}
|
|
OS = OwnedStream.get();
|
|
}
|
|
|
|
if (getDriver().CCPrintOptions)
|
|
*OS << "[Logging clang options]\n";
|
|
|
|
C.Print(*OS, "\n", /*Quote=*/getDriver().CCPrintOptions);
|
|
}
|
|
|
|
if (LogOnly)
|
|
return 0;
|
|
|
|
std::string Error;
|
|
bool ExecutionFailed;
|
|
int Res = C.Execute(Redirects, &Error, &ExecutionFailed);
|
|
if (PostCallback)
|
|
PostCallback(C, Res);
|
|
if (!Error.empty()) {
|
|
assert(Res && "Error string set with 0 result code!");
|
|
getDriver().Diag(diag::err_drv_command_failure) << Error;
|
|
}
|
|
|
|
if (Res)
|
|
FailingCommand = &C;
|
|
|
|
return ExecutionFailed ? 1 : Res;
|
|
}
|
|
|
|
using FailingCommandList = SmallVectorImpl<std::pair<int, const Command *>>;
|
|
|
|
static bool ActionFailed(const Action *A,
|
|
const FailingCommandList &FailingCommands) {
|
|
if (FailingCommands.empty())
|
|
return false;
|
|
|
|
// CUDA/HIP/SYCL can have the same input source code compiled multiple times
|
|
// so do not compile again if there are already failures. It is OK to abort
|
|
// the CUDA/HIP/SYCL pipeline on errors.
|
|
if (A->isOffloading(Action::OFK_Cuda) || A->isOffloading(Action::OFK_HIP) ||
|
|
A->isOffloading(Action::OFK_SYCL))
|
|
return true;
|
|
|
|
for (const auto &CI : FailingCommands)
|
|
if (A == &(CI.second->getSource()))
|
|
return true;
|
|
|
|
for (const auto *AI : A->inputs())
|
|
if (ActionFailed(AI, FailingCommands))
|
|
return true;
|
|
|
|
return false;
|
|
}
|
|
|
|
static bool ActionDependsOn(const Action *A, const Action *Other) {
|
|
return A == Other || llvm::any_of(A->inputs(), [&](const Action *Input) {
|
|
return ActionDependsOn(Input, Other);
|
|
});
|
|
}
|
|
|
|
static bool ActionsAreIndependent(const Action *A, const Action *B) {
|
|
return !ActionDependsOn(A, B) && !ActionDependsOn(B, A);
|
|
}
|
|
|
|
static bool CanRunInParallelOffloadJobGroup(const Command &Job) {
|
|
return !Job.InProcess && !Job.PrintInputFilenames &&
|
|
!Job.getBoundArch().empty() &&
|
|
!Job.getOffloadDeviceParallelJobGroup().empty();
|
|
}
|
|
|
|
static bool SameParallelOffloadJobGroup(const Command &A, const Command &B) {
|
|
return A.getOffloadDeviceParallelJobGroup() ==
|
|
B.getOffloadDeviceParallelJobGroup();
|
|
}
|
|
|
|
static bool HasDistinctBoundArch(const Command &Candidate,
|
|
ArrayRef<const Command *> Jobs) {
|
|
BoundArch CandidateArch = Candidate.getBoundArch();
|
|
return llvm::none_of(Jobs, [&](const Command *Job) {
|
|
return Job->getBoundArch() == CandidateArch;
|
|
});
|
|
}
|
|
|
|
static std::optional<llvm::ThreadPoolStrategy>
|
|
getParallelOffloadJobsStrategy(const ArgList &Args, unsigned NumJobs) {
|
|
if (NumJobs < 2)
|
|
return std::nullopt;
|
|
|
|
auto OffloadJobs = tools::parseOffloadJobs(Args);
|
|
if (!OffloadJobs.isValid())
|
|
return std::nullopt;
|
|
|
|
if (OffloadJobs.K == tools::OffloadJobsOpt::Kind::Jobserver)
|
|
return llvm::jobserver_concurrency();
|
|
|
|
if (OffloadJobs.NumThreads < 2)
|
|
return std::nullopt;
|
|
|
|
return llvm::hardware_concurrency(std::min(OffloadJobs.NumThreads, NumJobs));
|
|
}
|
|
|
|
struct ParallelJobResult {
|
|
int Res = 0;
|
|
bool ExecutionFailed = false;
|
|
std::string Error;
|
|
};
|
|
|
|
struct ParallelOffloadJobGroupResult {
|
|
size_t NumJobs = 0;
|
|
};
|
|
|
|
static std::optional<ParallelOffloadJobGroupResult>
|
|
tryExecuteParallelOffloadJobGroup(const Driver &D, const ArgList &Args,
|
|
ArrayRef<std::optional<StringRef>> Redirects,
|
|
const JobList::list_type &JobStorage,
|
|
size_t StartIndex,
|
|
FailingCommandList &FailingCommands) {
|
|
const Command &Job = *JobStorage[StartIndex];
|
|
if (!CanRunInParallelOffloadJobGroup(Job))
|
|
return std::nullopt;
|
|
|
|
SmallVector<const Command *, 4> ParallelJobs;
|
|
for (size_t I = StartIndex; I < JobStorage.size(); ++I) {
|
|
const Command &Candidate = *JobStorage[I];
|
|
if (ActionFailed(&Candidate.getSource(), FailingCommands))
|
|
break;
|
|
|
|
if (!CanRunInParallelOffloadJobGroup(Candidate))
|
|
break;
|
|
|
|
if (!SameParallelOffloadJobGroup(Job, Candidate))
|
|
break;
|
|
|
|
if (!HasDistinctBoundArch(Candidate, ParallelJobs))
|
|
break;
|
|
|
|
if (!llvm::all_of(ParallelJobs, [&](const Command *Other) {
|
|
return ActionsAreIndependent(&Candidate.getSource(),
|
|
&Other->getSource());
|
|
}))
|
|
break;
|
|
|
|
ParallelJobs.push_back(&Candidate);
|
|
}
|
|
|
|
std::optional<llvm::ThreadPoolStrategy> Strategy =
|
|
getParallelOffloadJobsStrategy(Args, ParallelJobs.size());
|
|
if (!Strategy)
|
|
return std::nullopt;
|
|
|
|
SmallVector<ParallelJobResult, 4> Results(ParallelJobs.size());
|
|
llvm::DefaultThreadPool Pool(*Strategy);
|
|
for (auto IndexedJob : llvm::enumerate(ParallelJobs)) {
|
|
size_t Index = IndexedJob.index();
|
|
const Command *ParallelJob = IndexedJob.value();
|
|
Pool.async([&, Index, ParallelJob] {
|
|
Results[Index].Res = ParallelJob->Execute(
|
|
Redirects, &Results[Index].Error, &Results[Index].ExecutionFailed);
|
|
});
|
|
}
|
|
Pool.wait();
|
|
|
|
for (auto [Index, ParallelJob] : llvm::enumerate(ParallelJobs)) {
|
|
ParallelJobResult &Result = Results[Index];
|
|
if (!Result.Error.empty()) {
|
|
assert(Result.Res && "Error string set with 0 result code!");
|
|
D.Diag(diag::err_drv_command_failure) << Result.Error;
|
|
}
|
|
|
|
if (Result.Res) {
|
|
FailingCommands.push_back(
|
|
std::make_pair(Result.ExecutionFailed ? 1 : Result.Res, ParallelJob));
|
|
}
|
|
}
|
|
|
|
return ParallelOffloadJobGroupResult{ParallelJobs.size()};
|
|
}
|
|
|
|
void Compilation::ExecuteJobs(const JobList &Jobs,
|
|
FailingCommandList &FailingCommands,
|
|
bool LogOnly) const {
|
|
// According to UNIX standard, driver need to continue compiling all the
|
|
// inputs on the command line even one of them failed.
|
|
// In all but CLMode, execute all the jobs unless the necessary inputs for the
|
|
// job is missing due to previous failures.
|
|
bool CanRunJobsInParallel =
|
|
!LogOnly && !getDriver().CCPrintOptions &&
|
|
!getDriver().CCPrintProcessStats && !getDriver().CCGenDiagnostics &&
|
|
!getDriver().IsCLMode() && !getArgs().hasArg(options::OPT_v) &&
|
|
Redirects.empty() && !PostCallback;
|
|
|
|
const auto &JobStorage = Jobs.getJobs();
|
|
for (size_t I = 0; I < JobStorage.size();) {
|
|
const auto &Job = *JobStorage[I];
|
|
if (ActionFailed(&Job.getSource(), FailingCommands)) {
|
|
++I;
|
|
continue;
|
|
}
|
|
|
|
if (CanRunJobsInParallel) {
|
|
if (std::optional<ParallelOffloadJobGroupResult> Result =
|
|
tryExecuteParallelOffloadJobGroup(getDriver(), getArgs(),
|
|
Redirects, JobStorage, I,
|
|
FailingCommands)) {
|
|
I += Result->NumJobs;
|
|
continue;
|
|
}
|
|
}
|
|
|
|
const Command *FailingCommand = nullptr;
|
|
if (int Res = ExecuteCommand(Job, FailingCommand, LogOnly)) {
|
|
FailingCommands.push_back(std::make_pair(Res, FailingCommand));
|
|
// Bail as soon as one command fails in cl driver mode.
|
|
if (TheDriver.IsCLMode())
|
|
return;
|
|
}
|
|
++I;
|
|
}
|
|
}
|
|
|
|
void Compilation::initCompilationForDiagnostics() {
|
|
ForDiagnostics = true;
|
|
|
|
// Free actions and jobs.
|
|
Actions.clear();
|
|
AllActions.clear();
|
|
Jobs.clear();
|
|
|
|
// Remove temporary files.
|
|
if (!TheDriver.isSaveTempsEnabled() && !ForceKeepTempFiles)
|
|
CleanupFileList(TempFiles);
|
|
|
|
// Clear temporary/results file lists.
|
|
TempFiles.clear();
|
|
ResultFiles.clear();
|
|
FailureResultFiles.clear();
|
|
|
|
// Remove any user specified output. Claim any unclaimed arguments, so as
|
|
// to avoid emitting warnings about unused args.
|
|
OptSpecifier OutputOpts[] = {
|
|
options::OPT_o, options::OPT_MD, options::OPT_MMD, options::OPT_M,
|
|
options::OPT_MM, options::OPT_MF, options::OPT_MG, options::OPT_MJ,
|
|
options::OPT_MQ, options::OPT_MT, options::OPT_MV};
|
|
for (const auto &Opt : OutputOpts) {
|
|
if (TranslatedArgs->hasArg(Opt))
|
|
TranslatedArgs->eraseArg(Opt);
|
|
}
|
|
TranslatedArgs->ClaimAllArgs();
|
|
|
|
// Force re-creation of the toolchain Args, otherwise our modifications just
|
|
// above will have no effect.
|
|
for (auto Arg : TCArgs)
|
|
if (Arg.second != TranslatedArgs)
|
|
delete Arg.second;
|
|
TCArgs.clear();
|
|
|
|
// Redirect stdout/stderr to /dev/null.
|
|
Redirects = {std::nullopt, {""}, {""}};
|
|
|
|
// Temporary files added by diagnostics should be kept.
|
|
ForceKeepTempFiles = true;
|
|
}
|
|
|
|
StringRef Compilation::getSysRoot() const {
|
|
return getDriver().SysRoot;
|
|
}
|
|
|
|
void Compilation::Redirect(ArrayRef<std::optional<StringRef>> Redirects) {
|
|
this->Redirects = Redirects;
|
|
}
|