v8/src/sampler.cc

669 lines
20 KiB
C++
Raw Normal View History

// Copyright 2013 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following
// disclaimer in the documentation and/or other materials provided
// with the distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "sampler.h"
#if V8_OS_POSIX && !V8_OS_CYGWIN
#define USE_SIGNALS
#include <errno.h>
#include <pthread.h>
#include <signal.h>
#include <sys/time.h>
#include <sys/syscall.h>
#if V8_OS_MACOSX
#include <mach/mach.h>
// OpenBSD doesn't have <ucontext.h>. ucontext_t lives in <signal.h>
// and is a typedef for struct sigcontext. There is no uc_mcontext.
#elif(!V8_OS_ANDROID || defined(__BIONIC_HAVE_UCONTEXT_T)) \
&& !V8_OS_OPENBSD
#include <ucontext.h>
#endif
#include <unistd.h>
// GLibc on ARM defines mcontext_t has a typedef for 'struct sigcontext'.
// Old versions of the C library <signal.h> didn't define the type.
#if V8_OS_ANDROID && !defined(__BIONIC_HAVE_UCONTEXT_T) && \
defined(__arm__) && !defined(__BIONIC_HAVE_STRUCT_SIGCONTEXT)
#include <asm/sigcontext.h>
#endif
#elif V8_OS_WIN || V8_OS_CYGWIN
#include "win32-headers.h"
#endif
#include "v8.h"
#include "cpu-profiler-inl.h"
#include "flags.h"
#include "frames-inl.h"
#include "log.h"
#include "platform.h"
#include "simulator.h"
#include "v8threads.h"
#include "vm-state-inl.h"
#if V8_OS_ANDROID && !defined(__BIONIC_HAVE_UCONTEXT_T)
// Not all versions of Android's C library provide ucontext_t.
// Detect this and provide custom but compatible definitions. Note that these
// follow the GLibc naming convention to access register values from
// mcontext_t.
//
// See http://code.google.com/p/android/issues/detail?id=34784
#if defined(__arm__)
typedef struct sigcontext mcontext_t;
typedef struct ucontext {
uint32_t uc_flags;
struct ucontext* uc_link;
stack_t uc_stack;
mcontext_t uc_mcontext;
// Other fields are not used by V8, don't define them here.
} ucontext_t;
#elif defined(__mips__)
// MIPS version of sigcontext, for Android bionic.
typedef struct {
uint32_t regmask;
uint32_t status;
uint64_t pc;
uint64_t gregs[32];
uint64_t fpregs[32];
uint32_t acx;
uint32_t fpc_csr;
uint32_t fpc_eir;
uint32_t used_math;
uint32_t dsp;
uint64_t mdhi;
uint64_t mdlo;
uint32_t hi1;
uint32_t lo1;
uint32_t hi2;
uint32_t lo2;
uint32_t hi3;
uint32_t lo3;
} mcontext_t;
typedef struct ucontext {
uint32_t uc_flags;
struct ucontext* uc_link;
stack_t uc_stack;
mcontext_t uc_mcontext;
// Other fields are not used by V8, don't define them here.
} ucontext_t;
#elif defined(__i386__)
// x86 version for Android.
typedef struct {
uint32_t gregs[19];
void* fpregs;
uint32_t oldmask;
uint32_t cr2;
} mcontext_t;
typedef uint32_t kernel_sigset_t[2]; // x86 kernel uses 64-bit signal masks
typedef struct ucontext {
uint32_t uc_flags;
struct ucontext* uc_link;
stack_t uc_stack;
mcontext_t uc_mcontext;
// Other fields are not used by V8, don't define them here.
} ucontext_t;
enum { REG_EBP = 6, REG_ESP = 7, REG_EIP = 14 };
#endif
#endif // V8_OS_ANDROID && !defined(__BIONIC_HAVE_UCONTEXT_T)
namespace v8 {
namespace internal {
namespace {
class PlatformDataCommon : public Malloced {
public:
PlatformDataCommon() : profiled_thread_id_(ThreadId::Current()) {}
ThreadId profiled_thread_id() { return profiled_thread_id_; }
protected:
~PlatformDataCommon() {}
private:
ThreadId profiled_thread_id_;
};
} // namespace
#if defined(USE_SIGNALS)
class Sampler::PlatformData : public PlatformDataCommon {
public:
PlatformData() : vm_tid_(pthread_self()) {}
pthread_t vm_tid() const { return vm_tid_; }
private:
pthread_t vm_tid_;
};
#elif V8_OS_WIN || V8_OS_CYGWIN
// ----------------------------------------------------------------------------
// Win32 profiler support. On Cygwin we use the same sampler implementation as
// on Win32.
class Sampler::PlatformData : public PlatformDataCommon {
public:
// Get a handle to the calling thread. This is the thread that we are
// going to profile. We need to make a copy of the handle because we are
// going to use it in the sampler thread. Using GetThreadHandle() will
// not work in this case. We're using OpenThread because DuplicateHandle
// for some reason doesn't work in Chrome's sandbox.
PlatformData()
: profiled_thread_(OpenThread(THREAD_GET_CONTEXT |
THREAD_SUSPEND_RESUME |
THREAD_QUERY_INFORMATION,
false,
GetCurrentThreadId())) {}
~PlatformData() {
if (profiled_thread_ != NULL) {
CloseHandle(profiled_thread_);
profiled_thread_ = NULL;
}
}
HANDLE profiled_thread() { return profiled_thread_; }
private:
HANDLE profiled_thread_;
};
#endif
#if defined(USE_SIMULATOR)
class SimulatorHelper {
public:
inline bool Init(Sampler* sampler, Isolate* isolate) {
ThreadId thread_id = sampler->platform_data()->profiled_thread_id();
Isolate::PerIsolateThreadData* per_thread_data = isolate->
FindPerThreadDataForThread(thread_id);
if (!per_thread_data) return false;
simulator_ = per_thread_data->simulator();
// Check if there is active simulator.
return simulator_ != NULL;
}
inline void FillRegisters(RegisterState* state) {
state->pc = reinterpret_cast<Address>(simulator_->get_pc());
state->sp = reinterpret_cast<Address>(simulator_->get_register(
Simulator::sp));
#if V8_TARGET_ARCH_ARM
state->fp = reinterpret_cast<Address>(simulator_->get_register(
Simulator::r11));
#elif V8_TARGET_ARCH_MIPS
state->fp = reinterpret_cast<Address>(simulator_->get_register(
Simulator::fp));
#endif
}
private:
Simulator* simulator_;
};
#endif // USE_SIMULATOR
#if defined(USE_SIGNALS)
class SignalHandler : public AllStatic {
public:
static inline void EnsureInstalled() {
if (signal_handler_installed_) return;
struct sigaction sa;
sa.sa_sigaction = &HandleProfilerSignal;
sigemptyset(&sa.sa_mask);
sa.sa_flags = SA_RESTART | SA_SIGINFO;
signal_handler_installed_ =
(sigaction(SIGPROF, &sa, &old_signal_handler_) == 0);
}
static inline void Restore() {
if (signal_handler_installed_) {
sigaction(SIGPROF, &old_signal_handler_, 0);
signal_handler_installed_ = false;
}
}
static inline bool Installed() {
return signal_handler_installed_;
}
private:
static void HandleProfilerSignal(int signal, siginfo_t* info, void* context);
static bool signal_handler_installed_;
static struct sigaction old_signal_handler_;
};
struct sigaction SignalHandler::old_signal_handler_;
bool SignalHandler::signal_handler_installed_ = false;
void SignalHandler::HandleProfilerSignal(int signal, siginfo_t* info,
void* context) {
#if V8_OS_NACL
// As Native Client does not support signal handling, profiling
// is disabled.
return;
#else
USE(info);
if (signal != SIGPROF) return;
Isolate* isolate = Isolate::UncheckedCurrent();
if (isolate == NULL || !isolate->IsInitialized() || !isolate->IsInUse()) {
// We require a fully initialized and entered isolate.
return;
}
if (v8::Locker::IsActive() &&
!isolate->thread_manager()->IsLockedByCurrentThread()) {
return;
}
Sampler* sampler = isolate->logger()->sampler();
if (sampler == NULL || !sampler->IsActive()) return;
RegisterState state;
#if defined(USE_SIMULATOR)
SimulatorHelper helper;
if (!helper.Init(sampler, isolate)) return;
helper.FillRegisters(&state);
#else
// Extracting the sample from the context is extremely machine dependent.
ucontext_t* ucontext = reinterpret_cast<ucontext_t*>(context);
#if !V8_OS_OPENBSD
mcontext_t& mcontext = ucontext->uc_mcontext;
#endif
#if V8_OS_LINUX
#if V8_HOST_ARCH_IA32
state.pc = reinterpret_cast<Address>(mcontext.gregs[REG_EIP]);
state.sp = reinterpret_cast<Address>(mcontext.gregs[REG_ESP]);
state.fp = reinterpret_cast<Address>(mcontext.gregs[REG_EBP]);
#elif V8_HOST_ARCH_X64
state.pc = reinterpret_cast<Address>(mcontext.gregs[REG_RIP]);
state.sp = reinterpret_cast<Address>(mcontext.gregs[REG_RSP]);
state.fp = reinterpret_cast<Address>(mcontext.gregs[REG_RBP]);
#elif V8_HOST_ARCH_ARM
#if defined(__GLIBC__) && !defined(__UCLIBC__) && \
(__GLIBC__ < 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ <= 3))
// Old GLibc ARM versions used a gregs[] array to access the register
// values from mcontext_t.
state.pc = reinterpret_cast<Address>(mcontext.gregs[R15]);
state.sp = reinterpret_cast<Address>(mcontext.gregs[R13]);
state.fp = reinterpret_cast<Address>(mcontext.gregs[R11]);
#else
state.pc = reinterpret_cast<Address>(mcontext.arm_pc);
state.sp = reinterpret_cast<Address>(mcontext.arm_sp);
state.fp = reinterpret_cast<Address>(mcontext.arm_fp);
#endif // defined(__GLIBC__) && !defined(__UCLIBC__) &&
// (__GLIBC__ < 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ <= 3))
#elif V8_HOST_ARCH_MIPS
state.pc = reinterpret_cast<Address>(mcontext.pc);
state.sp = reinterpret_cast<Address>(mcontext.gregs[29]);
state.fp = reinterpret_cast<Address>(mcontext.gregs[30]);
#endif // V8_HOST_ARCH_*
#elif V8_OS_MACOSX
#if V8_HOST_ARCH_X64
#if __DARWIN_UNIX03
state.pc = reinterpret_cast<Address>(mcontext->__ss.__rip);
state.sp = reinterpret_cast<Address>(mcontext->__ss.__rsp);
state.fp = reinterpret_cast<Address>(mcontext->__ss.__rbp);
#else // !__DARWIN_UNIX03
state.pc = reinterpret_cast<Address>(mcontext->ss.rip);
state.sp = reinterpret_cast<Address>(mcontext->ss.rsp);
state.fp = reinterpret_cast<Address>(mcontext->ss.rbp);
#endif // __DARWIN_UNIX03
#elif V8_HOST_ARCH_IA32
#if __DARWIN_UNIX03
state.pc = reinterpret_cast<Address>(mcontext->__ss.__eip);
state.sp = reinterpret_cast<Address>(mcontext->__ss.__esp);
state.fp = reinterpret_cast<Address>(mcontext->__ss.__ebp);
#else // !__DARWIN_UNIX03
state.pc = reinterpret_cast<Address>(mcontext->ss.eip);
state.sp = reinterpret_cast<Address>(mcontext->ss.esp);
state.fp = reinterpret_cast<Address>(mcontext->ss.ebp);
#endif // __DARWIN_UNIX03
#endif // V8_HOST_ARCH_IA32
#elif V8_OS_FREEBSD
#if V8_HOST_ARCH_IA32
state.pc = reinterpret_cast<Address>(mcontext.mc_eip);
state.sp = reinterpret_cast<Address>(mcontext.mc_esp);
state.fp = reinterpret_cast<Address>(mcontext.mc_ebp);
#elif V8_HOST_ARCH_X64
state.pc = reinterpret_cast<Address>(mcontext.mc_rip);
state.sp = reinterpret_cast<Address>(mcontext.mc_rsp);
state.fp = reinterpret_cast<Address>(mcontext.mc_rbp);
#elif V8_HOST_ARCH_ARM
state.pc = reinterpret_cast<Address>(mcontext.mc_r15);
state.sp = reinterpret_cast<Address>(mcontext.mc_r13);
state.fp = reinterpret_cast<Address>(mcontext.mc_r11);
#endif // V8_HOST_ARCH_*
#elif V8_OS_NETBSD
#if V8_HOST_ARCH_IA32
state.pc = reinterpret_cast<Address>(mcontext.__gregs[_REG_EIP]);
state.sp = reinterpret_cast<Address>(mcontext.__gregs[_REG_ESP]);
state.fp = reinterpret_cast<Address>(mcontext.__gregs[_REG_EBP]);
#elif V8_HOST_ARCH_X64
state.pc = reinterpret_cast<Address>(mcontext.__gregs[_REG_RIP]);
state.sp = reinterpret_cast<Address>(mcontext.__gregs[_REG_RSP]);
state.fp = reinterpret_cast<Address>(mcontext.__gregs[_REG_RBP]);
#endif // V8_HOST_ARCH_*
#elif V8_OS_OPENBSD
#if V8_HOST_ARCH_IA32
state.pc = reinterpret_cast<Address>(ucontext->sc_eip);
state.sp = reinterpret_cast<Address>(ucontext->sc_esp);
state.fp = reinterpret_cast<Address>(ucontext->sc_ebp);
#elif V8_HOST_ARCH_X64
state.pc = reinterpret_cast<Address>(ucontext->sc_rip);
state.sp = reinterpret_cast<Address>(ucontext->sc_rsp);
state.fp = reinterpret_cast<Address>(ucontext->sc_rbp);
#endif // V8_HOST_ARCH_*
#elif V8_OS_SOLARIS
state.pc = reinterpret_cast<Address>(mcontext.gregs[REG_PC]);
state.sp = reinterpret_cast<Address>(mcontext.gregs[REG_SP]);
state.fp = reinterpret_cast<Address>(mcontext.gregs[REG_FP]);
#endif // V8_OS_SOLARIS
#endif // USE_SIMULATOR
sampler->SampleStack(state);
#endif // V8_OS_NACL
}
#endif
class SamplerThread : public Thread {
public:
static const int kSamplerThreadStackSize = 64 * KB;
explicit SamplerThread(int interval)
: Thread(Thread::Options("SamplerThread", kSamplerThreadStackSize)),
interval_(interval) {}
static void SetUp() { if (!mutex_) mutex_ = new Mutex(); }
static void TearDown() { delete mutex_; mutex_ = NULL; }
static void AddActiveSampler(Sampler* sampler) {
bool need_to_start = false;
LockGuard<Mutex> lock_guard(mutex_);
if (instance_ == NULL) {
// Start a thread that will send SIGPROF signal to VM threads,
// when CPU profiling will be enabled.
instance_ = new SamplerThread(sampler->interval());
need_to_start = true;
}
ASSERT(sampler->IsActive());
ASSERT(!instance_->active_samplers_.Contains(sampler));
ASSERT(instance_->interval_ == sampler->interval());
instance_->active_samplers_.Add(sampler);
#if defined(USE_SIGNALS)
SignalHandler::EnsureInstalled();
#endif
if (need_to_start) instance_->StartSynchronously();
}
static void RemoveActiveSampler(Sampler* sampler) {
SamplerThread* instance_to_remove = NULL;
{
LockGuard<Mutex> lock_guard(mutex_);
ASSERT(sampler->IsActive());
bool removed = instance_->active_samplers_.RemoveElement(sampler);
ASSERT(removed);
USE(removed);
// We cannot delete the instance immediately as we need to Join() the
// thread but we are holding mutex_ and the thread may try to acquire it.
if (instance_->active_samplers_.is_empty()) {
instance_to_remove = instance_;
instance_ = NULL;
#if defined(USE_SIGNALS)
SignalHandler::Restore();
#endif
}
}
if (!instance_to_remove) return;
instance_to_remove->Join();
delete instance_to_remove;
}
// Implement Thread::Run().
virtual void Run() {
while (true) {
{
LockGuard<Mutex> lock_guard(mutex_);
if (active_samplers_.is_empty()) break;
// When CPU profiling is enabled both JavaScript and C++ code is
// profiled. We must not suspend.
for (int i = 0; i < active_samplers_.length(); ++i) {
Sampler* sampler = active_samplers_.at(i);
if (!sampler->isolate()->IsInitialized()) continue;
if (!sampler->IsProfiling()) continue;
sampler->DoSample();
}
}
OS::Sleep(interval_);
}
}
private:
// Protects the process wide state below.
static Mutex* mutex_;
static SamplerThread* instance_;
const int interval_;
List<Sampler*> active_samplers_;
DISALLOW_COPY_AND_ASSIGN(SamplerThread);
};
Mutex* SamplerThread::mutex_ = NULL;
SamplerThread* SamplerThread::instance_ = NULL;
//
// StackTracer implementation
//
DISABLE_ASAN void TickSample::Init(Isolate* isolate,
const RegisterState& regs) {
ASSERT(isolate->IsInitialized());
pc = regs.pc;
state = isolate->current_vm_state();
// Avoid collecting traces while doing GC.
if (state == GC) return;
Address js_entry_sp = isolate->js_entry_sp();
if (js_entry_sp == 0) {
// Not executing JS now.
return;
}
ExternalCallbackScope* scope = isolate->external_callback_scope();
Address handler = Isolate::handler(isolate->thread_local_top());
// If there is a handler on top of the external callback scope then
// we have already entrered JavaScript again and the external callback
// is not the top function.
if (scope && scope->scope_address() < handler) {
external_callback = scope->callback();
has_external_callback = true;
} else {
// Sample potential return address value for frameless invocation of
// stubs (we'll figure out later, if this value makes sense).
tos = Memory::Address_at(regs.sp);
has_external_callback = false;
}
SafeStackFrameIterator it(isolate, regs.fp, regs.sp, js_entry_sp);
top_frame_type = it.top_frame_type();
int i = 0;
while (!it.done() && i < TickSample::kMaxFramesCount) {
stack[i++] = it.frame()->pc();
it.Advance();
}
frames_count = i;
}
void Sampler::SetUp() {
SamplerThread::SetUp();
}
void Sampler::TearDown() {
SamplerThread::TearDown();
}
Sampler::Sampler(Isolate* isolate, int interval)
: isolate_(isolate),
interval_(interval),
profiling_(false),
Support higher CPU profiler sampling rate on posix systems New flag is added that allows to specify CPU profiler sampling rate in microseconds as command line argument. It was tested to work fine with 100us interval(currently it is 1ms). Default values are kept the same as in the current implementation. The new implementation is enabled only on POSIX platforms which use signals to collect samples. Other platforms that pause thread being sampled are to follow. SIGPROF signals are now sent on the profiler event processor thread to make sure that the processing thread does fall far behind the sampling. The patch is based on the previous one that was rolled out in r13851. The main difference is that the circular queue is not modified for now. On Linux sampling for CPU profiler is initiated on the profiler event processor thread, other platforms to follow. CPU profiler continues to use SamplingCircularQueue, we will probably replace it with a single sample buffer when Mac and Win ports support profiling on the event processing thread. When --prof option is specified profiling is initiated either on the profiler event processor thread if CPU profiler is on or on the SignalSender thread as it used to be if no CPU profiles are being collected. ProfilerEventsProcessor::ProcessEventsAndDoSample now waits in a tight loop, processing collected samples until sampling interval expires. To save CPU resources I'm planning to change that to use nanosleep as only one sample is expected in the queue at any point. BUG=v8:2814 R=bmeurer@chromium.org Review URL: https://codereview.chromium.org/21101002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@16310 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2013-08-26 07:17:12 +00:00
has_processing_thread_(false),
active_(false),
is_counting_samples_(false),
js_and_external_sample_count_(0) {
data_ = new PlatformData;
}
Sampler::~Sampler() {
ASSERT(!IsActive());
delete data_;
}
void Sampler::Start() {
ASSERT(!IsActive());
SetActive(true);
SamplerThread::AddActiveSampler(this);
}
void Sampler::Stop() {
ASSERT(IsActive());
SamplerThread::RemoveActiveSampler(this);
SetActive(false);
}
void Sampler::SampleStack(const RegisterState& state) {
TickSample* sample = isolate_->cpu_profiler()->StartTickSample();
TickSample sample_obj;
if (sample == NULL) sample = &sample_obj;
sample->Init(isolate_, state);
if (is_counting_samples_) {
if (sample->state == JS || sample->state == EXTERNAL) {
++js_and_external_sample_count_;
}
}
Tick(sample);
if (sample != &sample_obj) {
isolate_->cpu_profiler()->FinishTickSample();
}
}
Support higher CPU profiler sampling rate on posix systems New flag is added that allows to specify CPU profiler sampling rate in microseconds as command line argument. It was tested to work fine with 100us interval(currently it is 1ms). Default values are kept the same as in the current implementation. The new implementation is enabled only on POSIX platforms which use signals to collect samples. Other platforms that pause thread being sampled are to follow. SIGPROF signals are now sent on the profiler event processor thread to make sure that the processing thread does fall far behind the sampling. The patch is based on the previous one that was rolled out in r13851. The main difference is that the circular queue is not modified for now. On Linux sampling for CPU profiler is initiated on the profiler event processor thread, other platforms to follow. CPU profiler continues to use SamplingCircularQueue, we will probably replace it with a single sample buffer when Mac and Win ports support profiling on the event processing thread. When --prof option is specified profiling is initiated either on the profiler event processor thread if CPU profiler is on or on the SignalSender thread as it used to be if no CPU profiles are being collected. ProfilerEventsProcessor::ProcessEventsAndDoSample now waits in a tight loop, processing collected samples until sampling interval expires. To save CPU resources I'm planning to change that to use nanosleep as only one sample is expected in the queue at any point. BUG=v8:2814 R=bmeurer@chromium.org Review URL: https://codereview.chromium.org/21101002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@16310 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2013-08-26 07:17:12 +00:00
bool Sampler::CanSampleOnProfilerEventsProcessorThread() {
#if defined(USE_SIGNALS)
return true;
#elif V8_OS_WIN || V8_OS_CYGWIN
return true;
Support higher CPU profiler sampling rate on posix systems New flag is added that allows to specify CPU profiler sampling rate in microseconds as command line argument. It was tested to work fine with 100us interval(currently it is 1ms). Default values are kept the same as in the current implementation. The new implementation is enabled only on POSIX platforms which use signals to collect samples. Other platforms that pause thread being sampled are to follow. SIGPROF signals are now sent on the profiler event processor thread to make sure that the processing thread does fall far behind the sampling. The patch is based on the previous one that was rolled out in r13851. The main difference is that the circular queue is not modified for now. On Linux sampling for CPU profiler is initiated on the profiler event processor thread, other platforms to follow. CPU profiler continues to use SamplingCircularQueue, we will probably replace it with a single sample buffer when Mac and Win ports support profiling on the event processing thread. When --prof option is specified profiling is initiated either on the profiler event processor thread if CPU profiler is on or on the SignalSender thread as it used to be if no CPU profiles are being collected. ProfilerEventsProcessor::ProcessEventsAndDoSample now waits in a tight loop, processing collected samples until sampling interval expires. To save CPU resources I'm planning to change that to use nanosleep as only one sample is expected in the queue at any point. BUG=v8:2814 R=bmeurer@chromium.org Review URL: https://codereview.chromium.org/21101002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@16310 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2013-08-26 07:17:12 +00:00
#else
return false;
#endif
}
#if defined(USE_SIGNALS)
void Sampler::DoSample() {
if (!SignalHandler::Installed()) return;
pthread_kill(platform_data()->vm_tid(), SIGPROF);
}
#elif V8_OS_WIN || V8_OS_CYGWIN
void Sampler::DoSample() {
HANDLE profiled_thread = platform_data()->profiled_thread();
if (profiled_thread == NULL) return;
#if defined(USE_SIMULATOR)
SimulatorHelper helper;
if (!helper.Init(this, isolate())) return;
Support higher CPU profiler sampling rate on posix systems New flag is added that allows to specify CPU profiler sampling rate in microseconds as command line argument. It was tested to work fine with 100us interval(currently it is 1ms). Default values are kept the same as in the current implementation. The new implementation is enabled only on POSIX platforms which use signals to collect samples. Other platforms that pause thread being sampled are to follow. SIGPROF signals are now sent on the profiler event processor thread to make sure that the processing thread does fall far behind the sampling. The patch is based on the previous one that was rolled out in r13851. The main difference is that the circular queue is not modified for now. On Linux sampling for CPU profiler is initiated on the profiler event processor thread, other platforms to follow. CPU profiler continues to use SamplingCircularQueue, we will probably replace it with a single sample buffer when Mac and Win ports support profiling on the event processing thread. When --prof option is specified profiling is initiated either on the profiler event processor thread if CPU profiler is on or on the SignalSender thread as it used to be if no CPU profiles are being collected. ProfilerEventsProcessor::ProcessEventsAndDoSample now waits in a tight loop, processing collected samples until sampling interval expires. To save CPU resources I'm planning to change that to use nanosleep as only one sample is expected in the queue at any point. BUG=v8:2814 R=bmeurer@chromium.org Review URL: https://codereview.chromium.org/21101002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@16310 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2013-08-26 07:17:12 +00:00
#endif
const DWORD kSuspendFailed = static_cast<DWORD>(-1);
if (SuspendThread(profiled_thread) == kSuspendFailed) return;
// Context used for sampling the register state of the profiled thread.
CONTEXT context;
memset(&context, 0, sizeof(context));
context.ContextFlags = CONTEXT_FULL;
if (GetThreadContext(profiled_thread, &context) != 0) {
RegisterState state;
#if defined(USE_SIMULATOR)
helper.FillRegisters(&state);
#else
#if V8_HOST_ARCH_X64
state.pc = reinterpret_cast<Address>(context.Rip);
state.sp = reinterpret_cast<Address>(context.Rsp);
state.fp = reinterpret_cast<Address>(context.Rbp);
#else
state.pc = reinterpret_cast<Address>(context.Eip);
state.sp = reinterpret_cast<Address>(context.Esp);
state.fp = reinterpret_cast<Address>(context.Ebp);
#endif
#endif // USE_SIMULATOR
SampleStack(state);
}
ResumeThread(profiled_thread);
Support higher CPU profiler sampling rate on posix systems New flag is added that allows to specify CPU profiler sampling rate in microseconds as command line argument. It was tested to work fine with 100us interval(currently it is 1ms). Default values are kept the same as in the current implementation. The new implementation is enabled only on POSIX platforms which use signals to collect samples. Other platforms that pause thread being sampled are to follow. SIGPROF signals are now sent on the profiler event processor thread to make sure that the processing thread does fall far behind the sampling. The patch is based on the previous one that was rolled out in r13851. The main difference is that the circular queue is not modified for now. On Linux sampling for CPU profiler is initiated on the profiler event processor thread, other platforms to follow. CPU profiler continues to use SamplingCircularQueue, we will probably replace it with a single sample buffer when Mac and Win ports support profiling on the event processing thread. When --prof option is specified profiling is initiated either on the profiler event processor thread if CPU profiler is on or on the SignalSender thread as it used to be if no CPU profiles are being collected. ProfilerEventsProcessor::ProcessEventsAndDoSample now waits in a tight loop, processing collected samples until sampling interval expires. To save CPU resources I'm planning to change that to use nanosleep as only one sample is expected in the queue at any point. BUG=v8:2814 R=bmeurer@chromium.org Review URL: https://codereview.chromium.org/21101002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@16310 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2013-08-26 07:17:12 +00:00
}
#endif // USE_SIGNALS
} } // namespace v8::internal