AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Trace thread config + paranoid level + preload (#176)

Browse Source 
- OMNITRACE_TRACE_THREAD_BARRIERS config option
  - set to OFF to disable wrapping `pthread_barrier`
- OMNITRACE_TRACE_THREAD_JOIN config option
  - set to OFF to disable wrapping `pthread_join`
- allow PAPI with perf_event_paranoid at level 2
- default to no PAPI events
- setenv LD_PRELOAD to not include libomnitrace after preload
  - closes #175 
- bump version to 1.7.1

[ROCm/rocprofiler-systems commit: a3439d5bf2]
This commit is contained in:
Jonathan R. Madsen
2022-10-06 19:11:08 -05:00
committed by GitHub
parent 45e5450bf2
commit 0ec0d18ac8
7 changed files with 94 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-systems/source/lib/omnitrace/library/config.cpp
+42 -12
View File
@@ -459,6 +459,14 @@ configure_settings(bool _init)
"cause deadlocks with MPI distributions.",
false, "backend", "parallelism", "gotcha", "advanced");
OMNITRACE_CONFIG_SETTING(bool, "OMNITRACE_TRACE_THREAD_BARRIERS",
"Enable tracing calls to pthread_barrier functions.", true,
"backend", "parallelism", "gotcha", "advanced");
OMNITRACE_CONFIG_SETTING(bool, "OMNITRACE_TRACE_THREAD_JOIN",
"Enable tracing calls to pthread_join functions.", true,
"backend", "parallelism", "gotcha", "advanced");
OMNITRACE_CONFIG_SETTING(
bool, "OMNITRACE_SAMPLING_KEEP_INTERNAL",
"Configure whether the statistical samples should include call-stack entries "
@@ -601,7 +609,6 @@ configure_settings(bool _init)
_config->get_max_thread_bookmarks() = 1;
_config->get_timing_units() = "sec";
_config->get_memory_units() = "MB";
_config->get_papi_events() = "PAPI_TOT_CYC";
// settings native to timemory but critically and/or extensively used by omnitrace
auto _add_omnitrace_category = [&_config](auto itr) {
@@ -685,21 +692,22 @@ configure_settings(bool _init)
if(_fparanoid) _fparanoid >> _paranoid;
}
if(_paranoid > 1)
if(_paranoid > 2)
{
OMNITRACE_BASIC_VERBOSE(0,
"/proc/sys/kernel/perf_event_paranoid has a value of %i. "
"Disabling PAPI (requires a value <= 1)...\n",
"Disabling PAPI (requires a value <= 2)...\n",
_paranoid);
OMNITRACE_BASIC_VERBOSE(0,
"In order to enable PAPI support, run 'echo N | sudo tee "
"/proc/sys/kernel/perf_event_paranoid' where N is < 2\n");
tim::trait::runtime_enabled<comp::papi_common<void>>::set(false);
tim::trait::runtime_enabled<comp::papi_array_t>::set(false);
tim::trait::runtime_enabled<comp::papi_vector>::set(false);
tim::trait::runtime_enabled<comp::cpu_roofline_flops>::set(false);
tim::trait::runtime_enabled<comp::cpu_roofline_dp_flops>::set(false);
tim::trait::runtime_enabled<comp::cpu_roofline_sp_flops>::set(false);
OMNITRACE_BASIC_VERBOSE(
0, "In order to enable PAPI support, run 'echo N | sudo tee "
"/proc/sys/kernel/perf_event_paranoid' where N is <= 2\n");
trait::runtime_enabled<comp::papi_config>::set(false);
trait::runtime_enabled<comp::papi_common<void>>::set(false);
trait::runtime_enabled<comp::papi_array_t>::set(false);
trait::runtime_enabled<comp::papi_vector>::set(false);
trait::runtime_enabled<comp::cpu_roofline_flops>::set(false);
trait::runtime_enabled<comp::cpu_roofline_dp_flops>::set(false);
trait::runtime_enabled<comp::cpu_roofline_sp_flops>::set(false);
_config->get_papi_events() = std::string{};
}
else
@@ -812,6 +820,14 @@ configure_settings(bool _init)
tim::set_env(std::string{ _dl_verbose->first }, _dl_verbose->second->as_string(),
0);
if(_config->get_papi_events().empty())
{
trait::runtime_enabled<comp::papi_config>::set(false);
trait::runtime_enabled<comp::papi_common<void>>::set(false);
trait::runtime_enabled<comp::papi_array_t>::set(false);
trait::runtime_enabled<comp::papi_vector>::set(false);
}
configure_mode_settings();
configure_signal_handler();
configure_disabled_settings();
@@ -1886,6 +1902,20 @@ get_trace_thread_spin_locks()
return static_cast<tim::tsettings<bool>&>(*_v->second).get();
}
bool
get_trace_thread_barriers()
{
static auto _v = get_config()->find("OMNITRACE_TRACE_THREAD_BARRIERS");
return static_cast<tim::tsettings<bool>&>(*_v->second).get();
}
bool
get_trace_thread_join()
{
static auto _v = get_config()->find("OMNITRACE_TRACE_THREAD_JOIN");
return static_cast<tim::tsettings<bool>&>(*_v->second).get();
}
bool
get_debug_tid()
{