Remove critical trace usage in source code

source/bin/omnitrace-causal/impl.cpp

@@ -200,7 +200,6 @@ get_initial_environment()
     update_env(_env, "OMNITRACE_TRACE", false);
     update_env(_env, "OMNITRACE_PROFILE", false);
     update_env(_env, "OMNITRACE_USE_PROCESS_SAMPLING", false);
-    update_env(_env, "OMNITRACE_CRITICAL_TRACE", false);
     update_env(_env, "OMNITRACE_THREAD_POOL_SIZE",
                get_env<int>("OMNITRACE_THREAD_POOL_SIZE", 0));
     update_env(_env, "OMNITRACE_LAUNCHER", "omnitrace-causal");

source/lib/core/argparse.cpp

@@ -1225,7 +1225,6 @@ add_core_arguments(parser_t& _parser, parser_data& _data)
     add_group_arguments(_parser, "perfetto", _data, true);
     add_group_arguments(_parser, "timemory", _data, true);
     add_group_arguments(_parser, "rocm", _data, true);
-    add_group_arguments(_parser, "critical_trace", _data, true);
 
     _parser.start_group("MISCELLANEOUS OPTIONS", "");
 

source/lib/core/categories.hpp

@@ -110,9 +110,6 @@ OMNITRACE_DEFINE_CATEGORY(category, mpi, OMNITRACE_CATEGORY_MPI, "mpi", "MPI reg
 OMNITRACE_DEFINE_CATEGORY(category, ompt, OMNITRACE_CATEGORY_OMPT, "ompt", "OpenMP tools regions")
 OMNITRACE_DEFINE_CATEGORY(category, process_sampling, OMNITRACE_CATEGORY_PROCESS_SAMPLING, "process_sampling", "Process-level data")
 OMNITRACE_DEFINE_CATEGORY(category, comm_data, OMNITRACE_CATEGORY_COMM_DATA, "comm_data", "MPI/RCCL counters for tracking amount of data sent or received")
-OMNITRACE_DEFINE_CATEGORY(category, critical_trace, OMNITRACE_CATEGORY_CRITICAL_TRACE, "critical-trace", "Critical trace data")
-OMNITRACE_DEFINE_CATEGORY(category, host_critical_trace, OMNITRACE_CATEGORY_HOST_CRITICAL_TRACE, "host-critical-trace", "Host-side critical trace data")
-OMNITRACE_DEFINE_CATEGORY(category, device_critical_trace, OMNITRACE_CATEGORY_DEVICE_CRITICAL_TRACE, "device-critical-trace", "Device-side critical trace data")
 OMNITRACE_DEFINE_CATEGORY(category, causal, OMNITRACE_CATEGORY_CAUSAL, "causal", "Causal profiling data")
 OMNITRACE_DEFINE_CATEGORY(category, cpu_freq, OMNITRACE_CATEGORY_CPU_FREQ, "cpu_frequency", "CPU frequency (collected in background thread)")
 OMNITRACE_DEFINE_CATEGORY(category, process_page, OMNITRACE_CATEGORY_PROCESS_PAGE, "process_page_fault", "Memory page faults in process (collected in background thread)")
@@ -174,9 +171,6 @@ using name = perfetto_category<Tp...>;
         OMNITRACE_PERFETTO_CATEGORY(category::sampling),                                 \
         OMNITRACE_PERFETTO_CATEGORY(category::process_sampling),                         \
         OMNITRACE_PERFETTO_CATEGORY(category::comm_data),                                \
-        OMNITRACE_PERFETTO_CATEGORY(category::critical_trace),                           \
-        OMNITRACE_PERFETTO_CATEGORY(category::host_critical_trace),                      \
-        OMNITRACE_PERFETTO_CATEGORY(category::device_critical_trace),                    \
         OMNITRACE_PERFETTO_CATEGORY(category::causal),                                   \
         OMNITRACE_PERFETTO_CATEGORY(category::cpu_freq),                                 \
         OMNITRACE_PERFETTO_CATEGORY(category::process_page),                             \

source/lib/core/config.cpp

@@ -520,10 +520,6 @@ configure_settings(bool _init)
                              _backend, "perfetto")
         ->set_choices({ "inprocess", "system", "all" });
 
-    OMNITRACE_CONFIG_SETTING(bool, "OMNITRACE_CRITICAL_TRACE",
-                             "Enable generation of the critical trace", false, "backend",
-                             "critical_trace");
-
     OMNITRACE_CONFIG_SETTING(bool, "OMNITRACE_TRACE_THREAD_LOCKS",
                              "Enable tracing calls to pthread_mutex_lock, "
                              "pthread_mutex_unlock, pthread_mutex_trylock",
@@ -647,15 +643,6 @@ configure_settings(bool _init)
         "is collected on every available device",
         "", "rocprofiler", "rocm", "hardware_counters");
 
-    OMNITRACE_CONFIG_SETTING(bool, "OMNITRACE_CRITICAL_TRACE_DEBUG",
-                             "Enable debugging for critical trace", _omnitrace_debug,
-                             "debugging", "critical_trace", "advanced");
-
-    OMNITRACE_CONFIG_SETTING(
-        bool, "OMNITRACE_CRITICAL_TRACE_SERIALIZE_NAMES",
-        "Include names in serialization of critical trace (mainly for debugging)",
-        _omnitrace_debug, "debugging", "critical_trace", "advanced");
-
     OMNITRACE_CONFIG_SETTING(size_t, "OMNITRACE_PERFETTO_SHMEM_SIZE_HINT_KB",
                              "Hint for shared-memory buffer size in perfetto (in KB)",
                              size_t{ 4096 }, "perfetto", "data", "advanced");
@@ -721,21 +708,6 @@ configure_settings(bool _init)
                            1),
         "parallelism", "advanced");
 
-    OMNITRACE_CONFIG_EXT_SETTING(int64_t, "OMNITRACE_CRITICAL_TRACE_COUNT",
-                                 "Number of critical trace to export (0 == all)",
-                                 int64_t{ 0 }, "critical_trace",
-                                 "omnitrace-critical-trace", "advanced");
-
-    OMNITRACE_CONFIG_SETTING(uint64_t, "OMNITRACE_CRITICAL_TRACE_BUFFER_COUNT",
-                             "Number of critical trace records to store in thread-local "
-                             "memory before submitting to shared buffer",
-                             uint64_t{ 2000 }, "critical_trace", "advanced");
-
-    OMNITRACE_CONFIG_EXT_SETTING(
-        int64_t, "OMNITRACE_CRITICAL_TRACE_PER_ROW",
-        "How many critical traces per row in perfetto (0 == all in one row)",
-        int64_t{ 0 }, "critical_trace", "omnitrace-critical-trace", "advanced");
-
     OMNITRACE_CONFIG_SETTING(
         std::string, "OMNITRACE_TIMEMORY_COMPONENTS",
         "List of components to collect via timemory (see `omnitrace-avail -C`)",
@@ -1157,14 +1129,12 @@ configure_mode_settings(const std::shared_ptr<settings>& _config)
         _set("OMNITRACE_USE_OMPT", false);
         _set("OMNITRACE_USE_SAMPLING", false);
         _set("OMNITRACE_USE_PROCESS_SAMPLING", false);
-        _set("OMNITRACE_CRITICAL_TRACE", false);
     }
     else if(get_mode() == Mode::Causal)
     {
         _set("OMNITRACE_USE_CAUSAL", true);
         _set("OMNITRACE_TRACE", false);
         _set("OMNITRACE_PROFILE", false);
-        _set("OMNITRACE_CRITICAL_TRACE", false);
         _set("OMNITRACE_USE_SAMPLING", false);
         _set("OMNITRACE_USE_PROCESS_SAMPLING", false);
     }
@@ -1223,7 +1193,6 @@ configure_mode_settings(const std::shared_ptr<settings>& _config)
         _set("OMNITRACE_USE_SAMPLING", false);
         _set("OMNITRACE_USE_PROCESS_SAMPLING", false);
         _set("OMNITRACE_USE_CODE_COVERAGE", false);
-        _set("OMNITRACE_CRITICAL_TRACE", false);
         set_setting_value("OMNITRACE_TIMEMORY_COMPONENTS", std::string{});
         set_setting_value("OMNITRACE_PAPI_EVENTS", std::string{});
     }
@@ -1404,7 +1373,6 @@ configure_disabled_settings(const std::shared_ptr<settings>& _config)
     _handle_use_option("OMNITRACE_USE_ROCM_SMI", "rocm_smi");
     _handle_use_option("OMNITRACE_USE_ROCTRACER", "roctracer");
     _handle_use_option("OMNITRACE_USE_ROCPROFILER", "rocprofiler");
-    _handle_use_option("OMNITRACE_CRITICAL_TRACE", "critical_trace");
 
 #if !defined(OMNITRACE_USE_ROCTRACER) || OMNITRACE_USE_ROCTRACER == 0
     _config->find("OMNITRACE_USE_ROCTRACER")->second->set_hidden(true);
@@ -1971,13 +1939,6 @@ get_use_mpip()
     return static_cast<tim::tsettings<bool>&>(*_v->second).get();
 }
 
-bool&
-get_use_critical_trace()
-{
-    static auto _v = get_config()->find("OMNITRACE_CRITICAL_TRACE");
-    return static_cast<tim::tsettings<bool>&>(*_v->second).get();
-}
-
 bool
 get_use_kokkosp()
 {
@@ -2024,20 +1985,6 @@ get_num_threads_hint()
     return static_cast<tim::tsettings<size_t>&>(*_v->second).get();
 }
 
-bool
-get_critical_trace_debug()
-{
-    static auto _v = get_config()->find("OMNITRACE_CRITICAL_TRACE_DEBUG");
-    return static_cast<tim::tsettings<bool>&>(*_v->second).get();
-}
-
-bool
-get_critical_trace_serialize_names()
-{
-    static auto _v = get_config()->find("OMNITRACE_CRITICAL_TRACE_SERIALIZE_NAMES");
-    return static_cast<tim::tsettings<bool>&>(*_v->second).get();
-}
-
 bool
 get_sampling_keep_internal()
 {
@@ -2094,13 +2041,6 @@ get_trace_hsa_activity()
     return static_cast<tim::tsettings<bool>&>(*_v->second).get();
 }
 
-int64_t
-get_critical_trace_per_row()
-{
-    static auto _v = get_config()->find("OMNITRACE_CRITICAL_TRACE_PER_ROW");
-    return static_cast<tim::tsettings<int64_t>&>(*_v->second).get();
-}
-
 size_t
 get_perfetto_shmem_size_hint()
 {
@@ -2210,14 +2150,6 @@ get_perfetto_annotations()
     return static_cast<tim::tsettings<bool>&>(*_v->second).get();
 }
 
-uint64_t
-get_critical_trace_update_freq()
-{
-    static uint64_t _v =
-        get_config()->get<uint64_t>("OMNITRACE_CRITICAL_TRACE_BUFFER_COUNT");
-    return _v;
-}
-
 uint64_t
 get_thread_pool_size()
 {
@@ -2389,13 +2321,6 @@ get_sampling_allocator_size()
     return std::max<size_t>(static_cast<tim::tsettings<size_t>&>(*_v->second).get(), 1);
 }
 
-int64_t
-get_critical_trace_count()
-{
-    static auto _v = get_config()->find("OMNITRACE_CRITICAL_TRACE_COUNT");
-    return static_cast<tim::tsettings<int64_t>&>(*_v->second).get();
-}
-
 double
 get_process_sampling_freq()
 {

source/lib/core/config.hpp

@@ -218,9 +218,6 @@ get_use_pid();
 bool&
 get_use_mpip();
 
-bool&
-get_use_critical_trace() OMNITRACE_HOT;
-
 bool
 get_use_kokkosp();
 
@@ -251,12 +248,6 @@ get_trace_hsa_api();
 bool
 get_trace_hsa_activity();
 
-bool
-get_critical_trace_debug();
-
-bool
-get_critical_trace_serialize_names();
-
 size_t
 get_perfetto_shmem_size_hint();
 
@@ -278,9 +269,6 @@ get_disabled_categories();
 bool
 get_perfetto_annotations() OMNITRACE_HOT;
 
-uint64_t
-get_critical_trace_update_freq();
-
 uint64_t
 get_thread_pool_size();
 
@@ -297,9 +285,6 @@ get_perfetto_output_filename();
 bool
 get_perfetto_roctracer_per_stream() OMNITRACE_HOT;
 
-int64_t
-get_critical_trace_count();
-
 double
 get_trace_delay();
 
@@ -360,9 +345,6 @@ get_process_sampling_duration();
 std::string
 get_sampling_gpus();
 
-int64_t
-get_critical_trace_per_row();
-
 bool
 get_trace_thread_locks();
 

source/lib/core/debug.hpp

@@ -67,9 +67,6 @@ get_debug_tid() OMNITRACE_HOT;
 
 bool
 get_debug_pid() OMNITRACE_HOT;
-
-bool
-get_critical_trace_debug() OMNITRACE_HOT;
 }  // namespace config
 
 namespace debug
@@ -560,12 +557,6 @@ as_hex<void*>(void*, size_t);
 #define OMNITRACE_BASIC_DEBUG_F(...)                                                     \
     OMNITRACE_CONDITIONAL_BASIC_PRINT_F(::omnitrace::get_debug_env(), __VA_ARGS__)
 
-#define OMNITRACE_CT_DEBUG(...)                                                          \
-    OMNITRACE_CONDITIONAL_PRINT(::omnitrace::get_critical_trace_debug(), __VA_ARGS__)
-
-#define OMNITRACE_CT_DEBUG_F(...)                                                        \
-    OMNITRACE_CONDITIONAL_PRINT_F(::omnitrace::get_critical_trace_debug(), __VA_ARGS__)
-
 //--------------------------------------------------------------------------------------//
 //
 //  Verbose macros

source/lib/core/exception.cpp

@@ -50,14 +50,14 @@ template <typename... Args>
 auto
 get_backtrace(Args... _arg)
 {
+    consume_args(_arg...);
     auto _bt = std::stringstream{};
     if constexpr(sizeof...(Args) > 0)
     {
         ((_bt << _arg), ...) << "\n";
     }
     tim::unwind::detailed_backtrace<2>(_bt, true);
     return strdup(_bt.str().c_str());
-    consume_args(_arg...);
 }
 }  // namespace
 

source/lib/omnitrace-user/omnitrace/categories.h

@@ -63,9 +63,6 @@ extern "C"
         OMNITRACE_CATEGORY_OMPT,
         OMNITRACE_CATEGORY_PROCESS_SAMPLING,
         OMNITRACE_CATEGORY_COMM_DATA,
-        OMNITRACE_CATEGORY_CRITICAL_TRACE,
-        OMNITRACE_CATEGORY_HOST_CRITICAL_TRACE,
-        OMNITRACE_CATEGORY_DEVICE_CRITICAL_TRACE,
         OMNITRACE_CATEGORY_CAUSAL,
         OMNITRACE_CATEGORY_CPU_FREQ,
         OMNITRACE_CATEGORY_PROCESS_PAGE,

source/lib/omnitrace/library.cpp

@@ -48,7 +48,6 @@
 #include "library/components/pthread_gotcha.hpp"
 #include "library/components/rocprofiler.hpp"
 #include "library/coverage.hpp"
-#include "library/critical_trace.hpp"
 #include "library/ompt.hpp"
 #include "library/process_sampler.hpp"
 #include "library/ptl.hpp"
@@ -203,9 +202,6 @@ ensure_finalization(bool _static_init = false)
     return scope::destructor{ []() { omnitrace_finalize_hidden(); } };
 }
 
-using Device = critical_trace::Device;
-using Phase  = critical_trace::Phase;
-
 template <typename... Tp>
 struct fini_bundle
 {
@@ -402,11 +398,6 @@ omnitrace_init_library_hidden()
         if(_debug_init) config::set_setting_value("OMNITRACE_DEBUG", _debug_value);
     } };
 
-    // below will effectively do:
-    //      get_cpu_cid_stack(0)->emplace_back(-1);
-    // plus query some env variables
-    add_critical_trace<Device::CPU, Phase::NONE>(0, -1, 0, 0, 0, 0, 0, 0, 0, -1, 0);
-
     tim::trait::runtime_enabled<comp::roctracer>::set(get_use_roctracer());
     tim::trait::runtime_enabled<comp::roctracer_data>::set(get_use_roctracer() &&
                                                            get_use_timemory());
@@ -920,55 +911,12 @@ omnitrace_finalize_hidden(void)
         causal::finish_experimenting();
     }
 
-    if(get_use_critical_trace() || (get_use_rocm_smi() && get_use_roctracer()))
-    {
-        OMNITRACE_VERBOSE_F(1, "Generating the critical trace...\n");
-        for(size_t i = 0; i < thread_info::get_peak_num_threads(); ++i)
-        {
-            using critical_trace_hash_data =
-                thread_data<critical_trace::hash_ids, critical_trace::id>;
-
-            if(i < critical_trace_hash_data::get()->size() &&
-               critical_trace_hash_data::get()->at(i))
-            {
-                OMNITRACE_DEBUG_F("Copying the hash id data for thread %zu...\n", i);
-                critical_trace::add_hash_id(*critical_trace_hash_data::get()->at(i));
-            }
-        }
-
-        for(size_t i = 0; i < thread_info::get_peak_num_threads(); ++i)
-        {
-            using critical_trace_chain_data = thread_data<critical_trace::call_chain>;
-
-            if(i < critical_trace_chain_data::get()->size() &&
-               critical_trace_chain_data::get()->at(i))
-            {
-                OMNITRACE_DEBUG_F(
-                    "Updating the critical trace call-chains for thread %zu...\n", i);
-                critical_trace::update(i);  // launch update task
-            }
-        }
-
-        OMNITRACE_VERBOSE_F(1, "Waiting on critical trace updates...\n");
-        tasking::join();
-    }
-
     if(get_use_process_sampling())
     {
         OMNITRACE_VERBOSE_F(1, "Post-processing the system-level samples...\n");
         process_sampler::post_process();
     }
 
-    if(get_use_critical_trace())
-    {
-        // launch compute task
-        OMNITRACE_VERBOSE_F(1, "Launching critical trace compute task...\n");
-        critical_trace::compute();
-
-        OMNITRACE_VERBOSE_F(1, "Waiting on critical trace computation...\n");
-        tasking::join();
-    }
-
     // shutdown tasking before timemory is finalized, especially the roctracer thread-pool
     OMNITRACE_VERBOSE_F(1, "Shutting down thread-pools...\n");
     tasking::shutdown();