[llvm] a208623 - Improve module.pcm lock file performance on machines with high core counts
Michael Spencer via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 23 15:00:03 PDT 2020
Author: Ladd Van Tol
Date: 2020-03-23T14:59:39-07:00
New Revision: a20862307fff9385bbf6562de3154d05138f8bec
URL: https://github.com/llvm/llvm-project/commit/a20862307fff9385bbf6562de3154d05138f8bec
DIFF: https://github.com/llvm/llvm-project/commit/a20862307fff9385bbf6562de3154d05138f8bec.diff
LOG: Improve module.pcm lock file performance on machines with high core counts
Summary:
When building a large Xcode project with multiple module dependencies, and mixed Objective-C & Swift, I observed a large number of clang processes stalling at zero CPU for 30+ seconds throughout the build. This was especially prevalent on my 18-core iMac Pro.
After some sampling, the major cause appears to be the lock file implementation for precompiled modules in the module cache. When the lock is heavily contended by multiple clang processes, the exponential backoff runs in lockstep, with some of the processes sleeping for 30+ seconds in order to acquire the file lock.
In the attached patch, I implemented a more aggressive polling mechanism that limits the sleep interval to a max of 500ms, and randomizes the wait time. I preserved a limited form of exponential backoff. I also updated the code to use cross-platform timing, thread sleep, and random number capabilities available in C++11.
On iMac Pro (2.3 GHz Intel Xeon W, 18 core):
Xcode 11.1 bundled clang:
502.2 seconds (average of 5 runs)
Custom clang build with LockFileManager patch applied:
276.6 seconds (average of 5 runs)
This is a 1.82x speedup for this use case.
On MacBook Pro (4 core 3.1GHz Intel i7):
Xcode 11.1 bundled clang:
539.4 seconds (average of 2 runs)
Custom clang build with LockFileManager patch applied:
509.5 seconds (average of 2 runs)
As expected, machines with fewer cores benefit less from this change.
```
Call graph:
2992 Thread_393602 DispatchQueue_1: com.apple.main-thread (serial)
2992 start (in libdyld.dylib) + 1 [0x7fff6a1683d5]
2992 main (in clang) + 297 [0x1097a1059]
2992 driver_main(int, char const**) (in clang) + 2803 [0x1097a5513]
2992 cc1_main(llvm::ArrayRef<char const*>, char const*, void*) (in clang) + 1608 [0x1097a7cc8]
2992 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (in clang) + 3299 [0x1097dace3]
2992 clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (in clang) + 509 [0x1097dcc1d]
2992 clang::FrontendAction::Execute() (in clang) + 42 [0x109818b3a]
2992 clang::ParseAST(clang::Sema&, bool, bool) (in clang) + 185 [0x10981b369]
2992 clang::Parser::ParseFirstTopLevelDecl(clang::OpaquePtr<clang::DeclGroupRef>&) (in clang) + 37 [0x10983e9b5]
2992 clang::Parser::ParseTopLevelDecl(clang::OpaquePtr<clang::DeclGroupRef>&) (in clang) + 141 [0x10983ecfd]
2992 clang::Parser::ParseExternalDeclaration(clang::Parser::ParsedAttributesWithRange&, clang::ParsingDeclSpec*) (in clang) + 695 [0x10983f3b7]
2992 clang::Parser::ParseObjCAtDirectives(clang::Parser::ParsedAttributesWithRange&) (in clang) + 637 [0x10a9be9bd]
2992 clang::Parser::ParseModuleImport(clang::SourceLocation) (in clang) + 170 [0x10c4841ba]
2992 clang::Parser::ParseModuleName(clang::SourceLocation, llvm::SmallVectorImpl<std::__1::pair<clang::IdentifierInfo*, clang::SourceLocation> >&, bool) (in clang) + 503 [0x10c485267]
2992 clang::Preprocessor::Lex(clang::Token&) (in clang) + 316 [0x1098285cc]
2992 clang::Preprocessor::LexAfterModuleImport(clang::Token&) (in clang) + 690 [0x10cc7af62]
2992 clang::CompilerInstance::loadModule(clang::SourceLocation, llvm::ArrayRef<std::__1::pair<clang::IdentifierInfo*, clang::SourceLocation> >, clang::Module::NameVisibilityKind, bool) (in clang) + 7989 [0x10bba6535]
2992 compileAndLoadModule(clang::CompilerInstance&, clang::SourceLocation, clang::SourceLocation, clang::Module*, llvm::StringRef) (in clang) + 296 [0x10bba8318]
2992 llvm::LockFileManager::waitForUnlock() (in clang) + 91 [0x10b6953ab]
2992 nanosleep (in libsystem_c.dylib) + 199 [0x7fff6a22c914]
2992 __semwait_signal (in libsystem_kernel.dylib) + 10 [0x7fff6a2a0f32]
```
Differential Revision: https://reviews.llvm.org/D69575
Added:
Modified:
llvm/include/llvm/Support/LockFileManager.h
llvm/lib/Support/LockFileManager.cpp
Removed:
################################################################################
diff --git a/llvm/include/llvm/Support/LockFileManager.h b/llvm/include/llvm/Support/LockFileManager.h
index 2efeca3b6200..ab66621e6756 100644
--- a/llvm/include/llvm/Support/LockFileManager.h
+++ b/llvm/include/llvm/Support/LockFileManager.h
@@ -78,8 +78,8 @@ class LockFileManager {
/// For a shared lock, wait until the owner releases the lock.
/// Total timeout for the file to appear is ~1.5 minutes.
- /// \param MaxSeconds the maximum wait time per iteration in seconds.
- WaitForUnlockResult waitForUnlock(const unsigned MaxSeconds = 40);
+ /// \param MaxSeconds the maximum total wait time in seconds.
+ WaitForUnlockResult waitForUnlock(const unsigned MaxSeconds = 90);
/// Remove the lock file. This may delete a
diff erent lock file than
/// the one previously read if there is a race.
diff --git a/llvm/lib/Support/LockFileManager.cpp b/llvm/lib/Support/LockFileManager.cpp
index a4793aa7c04f..88489a658953 100644
--- a/llvm/lib/Support/LockFileManager.cpp
+++ b/llvm/lib/Support/LockFileManager.cpp
@@ -17,12 +17,16 @@
#include "llvm/Support/Signals.h"
#include "llvm/Support/raw_ostream.h"
#include <cerrno>
+#include <chrono>
#include <ctime>
#include <memory>
+#include <random>
#include <sys/stat.h>
#include <sys/types.h>
#include <system_error>
+#include <thread>
#include <tuple>
+
#ifdef _WIN32
#include <windows.h>
#endif
@@ -295,23 +299,29 @@ LockFileManager::waitForUnlock(const unsigned MaxSeconds) {
if (getState() != LFS_Shared)
return Res_Success;
-#ifdef _WIN32
- unsigned long Interval = 1;
-#else
- struct timespec Interval;
- Interval.tv_sec = 0;
- Interval.tv_nsec = 1000000;
-#endif
+ // Since we don't yet have an event-based method to wait for the lock file,
+ // implement randomized exponential backoff, similar to Ethernet collision
+ // algorithm. This improves performance on machines with high core counts
+ // when the file lock is heavily contended by multiple clang processes
+ const unsigned long MinWaitDurationMS = 10;
+ const unsigned long MaxWaitMultiplier = 50; // 500ms max wait
+ unsigned long WaitMultiplier = 1;
+ unsigned long ElapsedTimeSeconds = 0;
+
+ std::random_device Device;
+ std::default_random_engine Engine(Device());
+
+ auto StartTime = std::chrono::steady_clock::now();
+
do {
+ // FIXME: implement event-based waiting
+
// Sleep for the designated interval, to allow the owning process time to
// finish up and remove the lock file.
- // FIXME: Should we hook in to system APIs to get a notification when the
- // lock file is deleted?
-#ifdef _WIN32
- Sleep(Interval);
-#else
- nanosleep(&Interval, nullptr);
-#endif
+ std::uniform_int_distribution<unsigned long> Distribution(1,
+ WaitMultiplier);
+ unsigned long WaitDurationMS = MinWaitDurationMS * Distribution(Engine);
+ std::this_thread::sleep_for(std::chrono::milliseconds(WaitDurationMS));
if (sys::fs::access(LockFileName.c_str(), sys::fs::AccessMode::Exist) ==
errc::no_such_file_or_directory) {
@@ -325,24 +335,16 @@ LockFileManager::waitForUnlock(const unsigned MaxSeconds) {
if (!processStillExecuting((*Owner).first, (*Owner).second))
return Res_OwnerDied;
- // Exponentially increase the time we wait for the lock to be removed.
-#ifdef _WIN32
- Interval *= 2;
-#else
- Interval.tv_sec *= 2;
- Interval.tv_nsec *= 2;
- if (Interval.tv_nsec >= 1000000000) {
- ++Interval.tv_sec;
- Interval.tv_nsec -= 1000000000;
+ WaitMultiplier *= 2;
+ if (WaitMultiplier > MaxWaitMultiplier) {
+ WaitMultiplier = MaxWaitMultiplier;
}
-#endif
- } while (
-#ifdef _WIN32
- Interval < MaxSeconds * 1000
-#else
- Interval.tv_sec < (time_t)MaxSeconds
-#endif
- );
+
+ ElapsedTimeSeconds = std::chrono::duration_cast<std::chrono::seconds>(
+ std::chrono::steady_clock::now() - StartTime)
+ .count();
+
+ } while (ElapsedTimeSeconds < MaxSeconds);
// Give up.
return Res_Timeout;
More information about the llvm-commits
mailing list