Linux Perf Support + Causal Profiling Updates (#276)

* causal backtrace updates

- fix initial causal sampling period value

* causal delay updates

- tweak handling of sleep_for_overhead

* Fix experiment global scaling for prog pts

- results in drastically improved predictions

* pthread_mutex_gotcha updates

- disable all wrappers during causal profiling

* validate-causal-json.py updates

- support decimal stddev
- fix setting stddev from command-line

* causal perform_experiment_impl update

- handle start failing because finalizing

* deprecate causal::component::sample_rate

- appears to not help at all

* Rework sample info

* Increase causal unwind_depth

- use OMNITRACE_MAX_UNWIND_DEPTH

* validate-causal-json updates

- min experiments
  - exclude reporting predictions with less than X experiments at a given speedup
- percent samples
  - only print samples within X% of the peak (default: 95%)

* Update timemory submodule

- extensions to sampling for signals delivered via non-timer method
  - e.g. via HW counter overflow

* dwarf_entry::operator< updates

- sort via file

* causal profiling docs updates

- info about backends
- info about installing/enabling perf

* config updates: causal backend

- CausalBackend enum
- OMNITRACE_CAUSAL_BACKEND: perf, timer, auto
- omnitrace-causal option: --backend

* debug update

- use spin_mutex instead of std::mutex

* address_range::contains update

- range from 0-100 contains range from 10-100 but was returning false because high was == 100 not < 100

* symbol::operator< update

- handle load address differences

* sampling updates (non-causal)

- update get_timer to get_trigger + dynamic_cast

* container::static_vector updates

- support construction from container::c_array
- update_size private member func for handling atomic m_size

* Move perf files

- moved library/causal/perf.{hpp,cpp} to library/perf.{hpp,cpp}

* causal example update

- created impl.hpp (forward decls)
- renamed {cpu,rng}_func_impl to {cpu,rng}_impl_func
- only create two threads which run N iterations instead of two threads each iteration

* Update timemory submodule

- updates to unwind::processed_entry
- updates to procfs::maps

* Updated causal documentation

- fixed line numbers changed by modifications to causal example

* omnitrace-causal exe updates

- set OMNITRACE_THREAD_POOL_SIZE to zero by default

* core/containers updates

- static_vector: provide data() member function
- c_array pop_front() and pop_back() member functions

* core: config and argparse updates + perf

- core/perf.{hpp,cpp}
  - forward decl of enums
  - config-related capabilities
- argparse: --sample-overflow
- renamed some config functions
  - e.g. get_sampling_cpu_freq -> get_sampling_cputime_freq
- added config settings related to overflow sampling via perf
- added timer_sampling and overflow_sampling categories

* Update timemory submodule

- sampling allocator flushing

* binary updates

- lookup_ipaddr_entry
- use bfd_find_nearest_line instead of bfd_find_nearest_line_discriminator
  - discriminators are not used
- explicit instantiations of inlined_symbol::serialize

* Bump VERSION to 1.10.0

* sampling and perf updates

- support overflow sampling via Linux Perf
- update perf namespace
- update perf::perf_event
  - update record ctor: pointer instead of const ref
  - update open member func: return optional string
  - add m_batch_size member variable
- sampling updates
  - support overflow sampling
  - flush allocators
  - increase buffer size from 1024 to 2048
  - restructure post-processing in light of perf overflow supports
  - improve offload memory usage only load buffers for thread
  - load_offload_buffer(tid) uses thread-specific filepos
- component updates
  - backtrace_metrics::operator-=
  - backtrace_metrics::operator-
  - backtrace::sample does not record for overflow signal
  - callchain: perf overflow sample

* core updates

- component::sampling_percent does not report self + uses_percent_units

* causal updates

- tweak get_line_info
- overloads for set_current_selection (uint64_t, c_array, std::array)
- delay
  - use sampling::pause/sampling::resume
- experiment
  - experiment::sample derives from unwind::processed_entry
  - experiment::samples is vector instead of set
  - fixed samples
  - overloads for is_selected (uint64_t, c_array, std::array)
  - scaling factor defaults to 100 instead of 50
  - serialize updates follow change to experiment::sample
  - modify algorithm for increasing/decreasing experiment length
- sample_data
  - use map<uintptr, uint64_t> instead of set<sample_data>
  - get_samples returns vector<sample_data> instead of set<sample_data>
- sampling
  - support overflow via Linux Perf
  - update causal_offload_buffer
  - flush sampling allocator
- backtrace
  - overflow component

* libomnitrace-dl updates

- handle dl::InstrumentMode::PythonProfile

* testing updates (causal)

- causal line 155 -> causal line 100
- causal line 165 -> causal line 110

* formatting

* exit_gotcha updates

- exit_info for abort()
- message about non-zero exit code

* testing updates

- fail regex for causal tests
- validate-causal-json: >= min_experiments instead of > min_experiments
- handle OMNITRACE_DEBUG_SETTINGS in omnitrace_write_test_config

* causal sampling updates

- add new lines where appropriate

* causal data updates

- reorder diagnostic info when experiment fails to start

* binary updates

- symbol address range from address to address + symsize + 1
  - add 1 based on debug info

* causal data updates

- sample_selection wait_ns defaults to 1,000 instead of 10,000
- sample_selection wait scaled by iteration number
- save_line_info_impl verbosity
- print latest_eligible_pc when experiment does not start

* causal sampling + component updates

- perf backend disables component::backtrace
- ensure get_sampling_(realtime|cputime|overflow)_signal do not malloc

* causal: remove period stats

* validate-causal-json update

- fix --help

* causal data updates

- improve eligible pc history reporting when experiment fails to start

* causal data updates

- fix compute_eligible_lines_impl
  - eligible address ranges returning too many ranges
  - occasionally, overwrite all *true* eligible address ranges

* causal data updates

- reduce scoped ranges to symbol ranges
- is_eligible_address() returns true contains (not just coarse)
- revert some sample_selection behavior

* binary address_multirange updates

- make coarse_range private
- fix operator+=(pair<coarse, uintptr_t>)

* causal example update

- fix nsync to default to once per iteration

* binary analysis updates

- tweak header file includes

* causal updates

- remove factoring in sleep_for_overhead
- invoke delay::process() even if experiment is not active

* causal data updates

- update latest_eligible_pc structure

* update omnitrace-install.py.in

- fix support for fedora
  - /etc/os-release does not have ID_LIKE
  - fallback to RHEL 8.7 if version not specified

* update omnitrace-install.py.in

- fix support for debian
  - /etc/os-release does not have ID_LIKE
  - version mapping

* Update documentation

- update docs on installation

* causal data and experiment updates

- data: reset_sample_selection

* causal set_current_selection debugging

- debug messages for failed e2e runs

* causal data and backtrace component updates

- data: set_current_selection returns the number of eligible addresses added
- backtrace: if cputime signal has selected zero IPs > 5x, then realtime signal starts contributing call-stacks

* core library updates

- move config::parse_numeric_range to utility namespace
- add core/utility.cpp
- support range:increment, e.g. 5-25:10 expands to '5 15 25' instead of '5 10 15 20 25'

* omnitrace-causal update

- end-to-end expands all speedups
- support range:increment in speedups

* causal backtrace updates

- remove select_ival (realtime signal always contributes when select_count == 0)

* containers: static_vector update

- explicit c_array constructor
- explicit std::array constructor

* causal data updates

- remove set_current_selection(uint64_t)
- remove set_current_selection(std::array)
- sample_selection increase default wait time
- report eligible PC candidates
- move reset_sample_selection to perform_experiment_impl
- decrease latest_eligible_pc array size
- set_current_selection does not guard for experiment::active

* core debug updates

- OMNITRACE_PRINT_COLOR macros

* causal data updates

- tweak to experiment never started message

* causal gotcha updates

- remove unused code

* critical trace updates

- remove unused code

* omnitrace-causal

- OMNITRACE_LAUNCHER

* causal data updates

- don't fail on end-to-end + omnitrace-causal

* causal backtrace updates

- reintroduce select_ival behavior

* causal data updates

- tweak verbose messages about number of PC candidates

* core mproc updates

- utilities for waiting on child PID and diagnosing status
  - omnitrace::mproc::wait_pid
  - omnitrace::mproc::diagnose_status

* omnitrace-run updates

- support --fork argument for executing via fork in current process + execvpe on child instead of execvpe in current process

* omnitrace-causal updates

- wait_pid and diagnose_status just call equivalent functions in omnitrace::mproc

* ubuntu-focal workflow update

- attempt to launch ubuntu-focal-codecov job with CAP_SYS_ADMIN and use perf backend

* tests reorg and updates

- remove binary-rewrite-sampling and runtime-instrument-sampling tests
- rename *-preload tests (which use omnitrace-sample exe) to *-sampling
- split tests/CMakeLists.txt into several tests/omnitrace-<category>-tests.cmake files
- tweak to causal-both-omni-func test
  - add args: -n 2 -b timer

* update validate-causal-json.py

- better reasoning info for adjusting tolerance
- always apply tolerance adjustments in CI mode

* causal e2e tests update

- add label "causal-e2e" label
- tweak params
  - old: 80 12 432525 500000000
  - new: 80 50 432525 100000000
- disable processor affinity for slow-func/line-100 tests
  - artificially inflates some speedups with perf

* unblocking_gotcha updates

- overload operator() according to gotcha function index

* blocking_gotcha updates

- overload operator() according to gotcha function index
- fix bug where potentially post block functors (e.g. pthread_mutex_trylock) throw error if lock is not acquired.

* parse_numeric_range update

- support unordered_set

* config update

- OMNITRACE_DEBUG_{TIDS,PIDS} use parse_numeric_range
This commit is contained in:
Jonathan R. Madsen
2023-04-13 02:14:35 -05:00
committed by GitHub
parent cc14b52584
commit 9de3a6b0b4
96 changed files with 5489 additions and 3013 deletions
@@ -21,6 +21,9 @@
// SOFTWARE.
#include "library/causal/experiment.hpp"
#include "binary/analysis.hpp"
#include "binary/dwarf_entry.hpp"
#include "binary/symbol.hpp"
#include "common/defines.h"
#include "core/config.hpp"
#include "core/debug.hpp"
@@ -29,11 +32,11 @@
#include "library/causal/components/progress_point.hpp"
#include "library/causal/data.hpp"
#include "library/causal/delay.hpp"
#include "library/causal/sample_data.hpp"
#include "library/thread_data.hpp"
#include "library/thread_info.hpp"
#include "library/tracing.hpp"
#include <string>
#include <timemory/components/timing/backends.hpp>
#include <timemory/hash/types.hpp>
#include <timemory/mpl/policy.hpp>
@@ -42,10 +45,12 @@
#include <timemory/tpls/cereal/cereal/archives/json.hpp>
#include <timemory/tpls/cereal/types.hpp>
#include <timemory/units.hpp>
#include <timemory/unwind/dlinfo.hpp>
#include <chrono>
#include <ratio>
#include <regex>
#include <string>
#include <thread>
#include <vector>
@@ -68,26 +73,31 @@ bool use_exp_speedup_scaling =
get_env<bool>("OMNITRACE_CAUSAL_SCALE_EXPERIMENT_TIME_BY_SPEEDUP", false);
} // namespace
experiment::sample::sample(const base_type& _b, uint64_t _c)
: base_type{ _b }
, count{ _c }
{
if(lineinfo)
{
for(const auto& itr : lineinfo.lines)
{
if(itr.inlined)
inlines.emplace_back(
binary::inlined_symbol{ itr.line, itr.location, itr.name });
}
}
}
bool
experiment::sample::operator==(const sample& _v) const
{
return std::tie(address, info.line, info.file, info.func, location) ==
std::tie(_v.address, _v.info.line, _v.info.file, _v.info.func, _v.location);
return base_type::operator==(_v);
}
bool
experiment::sample::operator<(const sample& _v) const
{
if(info.line > 0 && _v.info.line > 0)
{
return std::tie(info.line, info.file) == std::tie(_v.info.line, _v.info.file);
}
else if((info.line + _v.info.line) > 0)
{
return std::tie(info.file, location, info.line) <
std::tie(_v.info.file, _v.location, _v.info.line);
}
return (location < _v.location);
return base_type::operator<(_v);
}
const auto&
@@ -102,8 +112,35 @@ void
experiment::sample::serialize(ArchiveT& ar, const unsigned)
{
namespace cereal = ::tim::cereal;
ar(cereal::make_nvp("location", location), cereal::make_nvp("count", count),
cereal::make_nvp("info", info));
using cereal::make_nvp;
ar(cereal::make_nvp("count", count));
if constexpr(concepts::is_output_archive<ArchiveT>::value)
{
ar(cereal::make_nvp("location", get_identifier()));
}
ar.setNextName("info");
ar.startNode();
ar(make_nvp("address", address), make_nvp("line", lineno), make_nvp("file", location),
make_nvp("func", name));
if constexpr(concepts::is_output_archive<ArchiveT>::value)
{
ar(cereal::make_nvp("dfunc", demangle(name)),
cereal::make_nvp("dwarf_info", std::vector<binary::dwarf_entry>{}));
}
ar(cereal::make_nvp("inlines", inlines));
ar.finishNode();
ar(cereal::make_nvp("dlinfo", info));
}
std::string
experiment::sample::get_identifier() const
{
return (lineno > 0 && !location.empty()) ? join(":", location, lineno)
: demangle(name);
}
template <typename ArchiveT>
@@ -119,7 +156,7 @@ experiment::record::serialize(ArchiveT& ar, const unsigned)
{
ar(cereal::make_nvp("samples", _samples));
for(auto& itr : _samples)
samples.emplace(std::move(itr));
samples.emplace_back(std::move(itr));
}
else
{
@@ -171,8 +208,6 @@ experiment::serialize(ArchiveT& ar, const unsigned)
}
ar(cereal::make_nvp("progress_points", _ppts));
}
ar(cereal::make_nvp("period_stats", period_stats));
}
std::string
@@ -203,9 +238,6 @@ experiment::start()
// sampling period in nanoseconds
sampling_period = backtrace_causal::get_period(units::nsec);
// adjust for the real sampling period
period_stats = causal::component::backtrace::get_period_stats();
if(period_stats.get_count() > 10) sampling_period = period_stats.get_mean();
// experiment time is scaled up for longer speedups
index = experiment_history.size() + 1;
@@ -222,10 +254,14 @@ experiment::start()
OMNITRACE_VERBOSE(0, "Starting causal experiment #%-3u: %s\n", index,
as_string().c_str());
current_experiment_value = *this;
current_selected_count.store(0);
current_experiment.store(this);
return true;
if(get_state() < State::Finalized)
{
current_experiment_value = *this;
current_selected_count.store(0);
current_experiment.store(this);
return true;
}
return false;
}
bool
@@ -258,34 +294,46 @@ experiment::stop()
total_delay = (global_delay - total_delay);
duration = (experiment_time > total_delay) ? (experiment_time - total_delay) : 0;
fini_progress = component::progress_point::get_progress_points();
period_stats = causal::component::backtrace::get_period_stats();
// sync data
delay::sync();
// for larger speedups, we increased the experiment time, so we want to artificially
// increase num by the same factor. E.g. 10 throughput points at speedup 50 should
// really look like 15
double _scale_num = 1.0 + ((use_exp_speedup_scaling) ? delay_scaling : 0.0);
auto _prog_stats = tim::statistics<int64_t>{};
auto _prog_stats = tim::statistics<double>{};
auto _prog_vals = std::vector<int64_t>{};
_prog_vals.reserve(fini_progress.size());
for(auto fitr : fini_progress)
{
auto _pt = fitr.second - init_progress[fitr.first];
int64_t _num =
std::max<int64_t>({ _pt.get_laps(), _pt.get_arrival(), _pt.get_departure() });
if(_num > 0) _prog_stats += (_num * _scale_num);
if(_num > 0) _prog_vals.emplace_back(_num);
}
std::sort(_prog_vals.begin(), _prog_vals.end());
for(auto itr : _prog_vals)
_prog_stats += itr;
auto _mean = (_prog_stats.get_count() > 0) ? _prog_stats.get_mean() : 0;
auto _high = (_prog_stats.get_count() > 0) ? _prog_stats.get_max() : 0;
if(_high < 5)
auto _nvals = _prog_vals.size();
auto _medi = (_nvals > 2) ? _prog_vals.at(_nvals / 2) : _prog_vals.front();
auto _mean = (_nvals > 0) ? _prog_stats.get_mean() : 0;
auto _high = (_nvals > 0) ? _prog_stats.get_max() : 0;
auto _lowv = (_nvals > 0) ? _prog_stats.get_min() : 0;
if(_lowv <= 3 && (_mean < 5 || _medi < 5))
{
OMNITRACE_VERBOSE(2,
"[progress points] increasing experiment time :: low: %6.3f, "
"high: %6.3f, mean: %6.3f, median: %zi\n",
_lowv, _high, _mean, _medi);
global_scaling *= 2;
++global_scaling_increments; // keep track of how many successive increments have
// been performed
}
else if(_mean > 10 && global_scaling > 1)
else if(_mean > 10 && _lowv >= 8 && global_scaling > 1)
{
OMNITRACE_VERBOSE(2,
"[progress points] decreasing experiment time :: low: %6.3f, "
"high: %6.3f, mean: %6.3f, median: %zi\n",
_lowv, _high, _mean, _medi);
global_scaling /= 2;
global_scaling_increments = 0;
}
@@ -304,7 +352,9 @@ experiment::stop()
if(_high > 0) experiment_history.emplace_back(*this);
std::this_thread::sleep_for(std::chrono::nanoseconds{ sampling_period * batch_size });
std::this_thread::sleep_for(
std::chrono::nanoseconds{ 5 * sampling_period * batch_size });
return true;
}
@@ -320,7 +370,6 @@ experiment::as_string() const
_ss << ", duration: " << std::setw(5) << std::fixed << std::setprecision(3)
<< _dur << " sec";
_ss << " :: experiment: " << as_hex(selection.address) << " ";
//_ss << " [" << selection.info.ipaddr().as_string() << "]";
if(selection.symbol_address > 0 && selection.address != selection.symbol_address)
_ss << "(symbol@" << as_hex(selection.symbol_address) << ") ";
if(!selection.symbol.file.empty() && selection.symbol.line > 0)
@@ -375,13 +424,30 @@ experiment::is_active()
return (current_experiment.load(std::memory_order_relaxed) != nullptr);
}
bool
experiment::is_selected(uint64_t _addr)
{
return (is_active() && current_experiment_value.selection.contains(_addr));
}
bool
experiment::is_selected(unwind_addr_t _stack)
{
if(is_active())
{
for(auto itr : _stack)
if(current_experiment_value.selection.contains(itr)) return true;
if(itr > 0 && current_experiment_value.selection.contains(itr)) return true;
}
return false;
}
bool
experiment::is_selected(container::c_array<uint64_t> _stack)
{
if(is_active())
{
for(auto itr : _stack)
if(itr > 0 && current_experiment_value.selection.contains(itr)) return true;
}
return false;
}
@@ -413,9 +479,6 @@ experiment::save_experiments(std::string _fname_base, const filename_config_t& _
{
const auto& _info0 = thread_info::get(0, InternalTID);
// if(experiment_history.size() > 1)
// experiment_history.erase(experiment_history.begin());
auto current_record = record{};
current_record.startup = _info0->lifetime.first;
@@ -446,11 +509,7 @@ experiment::save_experiments(std::string _fname_base, const filename_config_t& _
// update sample data
{
auto _add_sample = [&current_record](sample&& _v) {
auto fitr = current_record.samples.find(_v);
if(fitr != current_record.samples.end())
*fitr += _v;
else
current_record.samples.emplace(std::move(_v));
current_record.samples.emplace_back(std::move(_v));
};
auto _total_samples = std::map<uintptr_t, size_t>{};
@@ -462,41 +521,24 @@ experiment::save_experiments(std::string _fname_base, const filename_config_t& _
}
}
OMNITRACE_VERBOSE_F(1, "Processing line info for %zu sampled addresses...\n",
_total_samples.size());
for(const auto& itr : _total_samples)
{
auto _entry = binary::lookup_ipaddr_entry<true>(itr.first);
if(_entry) _add_sample(sample{ *_entry, itr.second });
}
auto _binfo_cfg = settings::compose_filename_config{};
_binfo_cfg.subdirectory = "causal/binary-info";
_binfo_cfg.use_suffix = config::get_use_pid();
save_line_info(_binfo_cfg, config::get_verbose());
for(const auto& itr : _total_samples)
{
auto _addr = itr.first;
auto _count = itr.second;
if(_count > 0)
{
auto _linfo = get_line_info(_addr, true);
for(const auto& iitr : _linfo)
{
auto _name = (iitr.line > 0) ? join(":", iitr.file, iitr.line)
: demangle(iitr.func);
_name = join(" :: ", as_hex(_addr), _name);
_add_sample(sample{ _count, _addr, _name, iitr });
}
if(_linfo.empty() && config::get_debug())
{
_add_sample(
sample{ _count, _addr, as_hex(_addr), sample::line_info{} });
}
}
}
}
bool _causal_output_reset =
config::get_setting_value<bool>("OMNITRACE_CAUSAL_FILE_RESET").value_or(false);
// if(current_record.experiments.empty()) return;
{
auto _saved_experiments = (_causal_output_reset)
? std::vector<experiment::record>{}
@@ -615,7 +657,8 @@ experiment::save_experiments(std::string _fname_base, const filename_config_t& _
for(const auto& itr : current_record.samples)
{
ofs << "samples\tlocation=" << itr.location << "\tcount=" << itr.count;
ofs << "samples\tlocation=" << itr.get_identifier()
<< "\tcount=" << itr.count;
if(config::get_debug()) ofs << "\taddress=" << as_hex(itr.address);
ofs << "\n";
}