Update to use rocprofiler-sdk (#55)
- Renames the CMake option "ROCPROFSYS_USE_HIP" to "ROCPROFSYS_USE_ROCM" - Remove the "ROCPROFSYS_USE_ROCM_SMI option. Controlled with the "ROCPROFSYS_USE_ROCM" option, instead. - Runtime configuration can still toggle ROCPROFSYS_USE_ROCM_SMI to disable the sampling. - Rename ROCPROFSYS_HIP_VERSION macro to ROCPROFSYS_ROCM_VERSION and remove blocks for `ROCPROFSYS_ROCM_VERSION < 60000` - Remove ROCPROFSYS_USE_ROCTRACER and ROCPROFSYS_USE_ROCPROFILER - Update test cases - Update docker files and workflows to install cmake 3.21, which is required for the rocprofiler-sdk findPackage script. - Removed rocm-6.2 from workflows due to a rocprofiler-sdk API change.
This commit is contained in:
David Galiffi
GitHub
@@ -25,12 +25,8 @@
|
||||
#include "core/debug.hpp"
|
||||
#include "core/dynamic_library.hpp"
|
||||
#include "core/gpu.hpp"
|
||||
#include "library/components/rocprofiler.hpp"
|
||||
#include "library/components/roctracer.hpp"
|
||||
#include "library/rocm/hsa_rsrc_factory.hpp"
|
||||
#include "library/rocm_smi.hpp"
|
||||
#include "library/rocprofiler.hpp"
|
||||
#include "library/roctracer.hpp"
|
||||
#include "library/rocprofiler-sdk.hpp"
|
||||
#include "library/runtime.hpp"
|
||||
#include "library/thread_data.hpp"
|
||||
#include "library/tracing.hpp"
|
||||
@@ -46,208 +42,18 @@
|
||||
#include <mutex>
|
||||
#include <tuple>
|
||||
|
||||
#if defined(ROCPROFSYS_USE_ROCPROFILER) && ROCPROFSYS_USE_ROCPROFILER > 0
|
||||
# include <rocprofiler.h>
|
||||
#if defined(ROCPROFSYS_USE_ROCM) && ROCPROFSYS_USE_ROCM > 0
|
||||
# include <rocprofiler-sdk/rocprofiler.h>
|
||||
#endif
|
||||
|
||||
using namespace rocprofsys;
|
||||
|
||||
namespace rocprofsys
|
||||
{
|
||||
namespace rocm
|
||||
{
|
||||
std::mutex rocm_mutex = {};
|
||||
bool is_loaded = false;
|
||||
bool on_load_trace = (get_env<int>("ROCP_ONLOAD_TRACE", 0) > 0);
|
||||
std::vector<hardware_counter_info>
|
||||
rocm_events()
|
||||
{
|
||||
return rocprofiler_sdk::get_rocm_events_info();
|
||||
}
|
||||
} // namespace rocm
|
||||
} // namespace rocprofsys
|
||||
|
||||
#if defined(ROCPROFSYS_USE_ROCPROFILER) && ROCPROFSYS_USE_ROCPROFILER > 0
|
||||
std::ostream&
|
||||
operator<<(std::ostream& _os, const rocprofiler_settings_t& _v)
|
||||
{
|
||||
# define ROCPROF_SETTING_FIELD_STR(NAME) JOIN('=', # NAME, _v.NAME)
|
||||
|
||||
_os << JOIN(
|
||||
", ", ROCPROF_SETTING_FIELD_STR(intercept_mode),
|
||||
ROCPROF_SETTING_FIELD_STR(code_obj_tracking),
|
||||
ROCPROF_SETTING_FIELD_STR(memcopy_tracking),
|
||||
ROCPROF_SETTING_FIELD_STR(trace_size), ROCPROF_SETTING_FIELD_STR(trace_local),
|
||||
ROCPROF_SETTING_FIELD_STR(timeout), ROCPROF_SETTING_FIELD_STR(timestamp_on),
|
||||
ROCPROF_SETTING_FIELD_STR(hsa_intercepting),
|
||||
ROCPROF_SETTING_FIELD_STR(k_concurrent), ROCPROF_SETTING_FIELD_STR(opt_mode),
|
||||
ROCPROF_SETTING_FIELD_STR(obj_dumping));
|
||||
return _os;
|
||||
}
|
||||
#endif
|
||||
|
||||
// HSA-runtime tool on-load method
|
||||
extern "C"
|
||||
{
|
||||
#if defined(ROCPROFSYS_USE_ROCPROFILER) && ROCPROFSYS_USE_ROCPROFILER > 0
|
||||
void OnUnloadTool()
|
||||
{
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Unloading...\n");
|
||||
|
||||
rocm::lock_t _lk{ rocm::rocm_mutex, std::defer_lock };
|
||||
if(!_lk.owns_lock()) _lk.lock();
|
||||
|
||||
if(!rocm::is_loaded)
|
||||
{
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(1 || rocm::on_load_trace,
|
||||
"rocprofiler is not loaded\n");
|
||||
return;
|
||||
}
|
||||
rocm::is_loaded = false;
|
||||
|
||||
_lk.unlock();
|
||||
|
||||
// stop_top_level_timer_if_necessary();
|
||||
// Final resources cleanup
|
||||
rocprofsys::rocprofiler::rocm_cleanup();
|
||||
}
|
||||
|
||||
void OnLoadToolProp(rocprofiler_settings_t* settings)
|
||||
{
|
||||
using ::rocprofiler::util::HsaRsrcFactory;
|
||||
|
||||
if(!config::get_use_rocprofiler() || config::get_rocm_events().empty()) return;
|
||||
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Loading...\n");
|
||||
|
||||
rocm::lock_t _lk{ rocm::rocm_mutex, std::defer_lock };
|
||||
if(!_lk.owns_lock()) _lk.lock();
|
||||
|
||||
if(rocm::is_loaded)
|
||||
{
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(1 || rocm::on_load_trace,
|
||||
"rocprofiler is already loaded\n");
|
||||
return;
|
||||
}
|
||||
rocm::is_loaded = true;
|
||||
|
||||
_lk.unlock();
|
||||
|
||||
// Enable timestamping
|
||||
settings->timestamp_on = 1;
|
||||
settings->intercept_mode = 1;
|
||||
settings->hsa_intercepting = 1;
|
||||
settings->k_concurrent = 0;
|
||||
settings->obj_dumping = 0;
|
||||
// settings->code_obj_tracking = 0;
|
||||
// settings->memcopy_tracking = 0;
|
||||
// settings->trace_local = 1;
|
||||
// settings->opt_mode = 1;
|
||||
// settings->trace_size = 0;
|
||||
// settings->timeout = 0;
|
||||
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(1 || rocm::on_load_trace, "rocprofiler settings: %s\n",
|
||||
JOIN("", *settings).c_str());
|
||||
|
||||
// Initialize profiling
|
||||
rocprofsys::rocprofiler::rocm_initialize();
|
||||
HsaRsrcFactory::Instance().PrintGpuAgents("ROCm");
|
||||
}
|
||||
#endif
|
||||
|
||||
bool OnLoad(HsaApiTable* table, uint64_t runtime_version, uint64_t failed_tool_count,
|
||||
const char* const* failed_tool_names)
|
||||
{
|
||||
tim::consume_parameters(table, runtime_version, failed_tool_count,
|
||||
failed_tool_names);
|
||||
|
||||
static bool _once = false;
|
||||
if(_once) return true;
|
||||
_once = true;
|
||||
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Loading...\n");
|
||||
ROCPROFSYS_SCOPED_SAMPLING_ON_CHILD_THREADS(false);
|
||||
|
||||
if(!tim::get_env("ROCPROFSYS_INIT_TOOLING", true)) return true;
|
||||
if(!tim::settings::enabled()) return true;
|
||||
|
||||
roctracer_is_init() = true;
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(1 || rocm::on_load_trace, "Loading ROCm tooling...\n");
|
||||
|
||||
if(!config::settings_are_configured() && get_state() < State::Active)
|
||||
rocprofsys_init_tooling_hidden();
|
||||
|
||||
ROCPROFSYS_SCOPED_THREAD_STATE(ThreadState::Internal);
|
||||
|
||||
#if ROCPROFSYS_HIP_VERSION < 50300
|
||||
ROCPROFSYS_VERBOSE_F(1 || rocm::on_load_trace,
|
||||
"Computing the roctracer clock skew...\n");
|
||||
(void) rocprofsys::get_clock_skew();
|
||||
#endif
|
||||
|
||||
if(get_use_process_sampling() && get_use_rocm_smi())
|
||||
{
|
||||
ROCPROFSYS_VERBOSE_F(1 || rocm::on_load_trace,
|
||||
"Setting rocm_smi state to active...\n");
|
||||
rocm_smi::set_state(State::Active);
|
||||
}
|
||||
|
||||
comp::roctracer::setup(static_cast<void*>(table), rocm::on_load_trace);
|
||||
|
||||
#if defined(ROCPROFSYS_USE_ROCPROFILER) && ROCPROFSYS_USE_ROCPROFILER > 0
|
||||
bool _force_rocprofiler_init =
|
||||
tim::get_env("ROCPROFSYS_FORCE_ROCPROFILER_INIT", false, false);
|
||||
#else
|
||||
bool _force_rocprofiler_init = false;
|
||||
#endif
|
||||
|
||||
bool _success = true;
|
||||
bool _is_empty =
|
||||
(config::settings_are_configured() && config::get_rocm_events().empty());
|
||||
if(_force_rocprofiler_init || (get_use_rocprofiler() && !_is_empty))
|
||||
{
|
||||
#if ROCPROFSYS_HIP_VERSION < 50500
|
||||
auto _rocprof = dynamic_library{
|
||||
"ROCPROFSYS_ROCPROFILER_LIBRARY",
|
||||
find_library_path(
|
||||
"librocprofiler64.so", { "ROCPROFSYS_ROCM_PATH", "ROCM_PATH" },
|
||||
{ ROCPROFSYS_DEFAULT_ROCM_PATH },
|
||||
{ "lib", "lib64", "rocprofiler/lib", "rocprofiler/lib64" }),
|
||||
(RTLD_LAZY | RTLD_GLOBAL), false
|
||||
};
|
||||
|
||||
ROCPROFSYS_VERBOSE_F(1 || rocm::on_load_trace,
|
||||
"Loading rocprofiler library (%s=%s)...\n",
|
||||
_rocprof.envname.c_str(), _rocprof.filename.c_str());
|
||||
_rocprof.open();
|
||||
|
||||
on_load_t _rocprof_load = nullptr;
|
||||
_success = _rocprof.invoke("OnLoad", _rocprof_load, table, runtime_version,
|
||||
failed_tool_count, failed_tool_names);
|
||||
ROCPROFSYS_CONDITIONAL_PRINT_F(!_success,
|
||||
"Warning! Invoking rocprofiler's OnLoad "
|
||||
"failed! ROCPROFSYS_ROCPROFILER_LIBRARY=%s\n",
|
||||
_rocprof.filename.c_str());
|
||||
ROCPROFSYS_CI_THROW(!_success,
|
||||
"Warning! Invoking rocprofiler's OnLoad "
|
||||
"failed! ROCPROFSYS_ROCPROFILER_LIBRARY=%s\n",
|
||||
_rocprof.filename.c_str());
|
||||
#endif
|
||||
}
|
||||
else
|
||||
{
|
||||
using ::rocprofiler::util::HsaRsrcFactory;
|
||||
|
||||
HsaRsrcFactory::Instance().PrintGpuAgents("ROCm");
|
||||
}
|
||||
|
||||
gpu::add_hip_device_metadata();
|
||||
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Loading... %s\n",
|
||||
(_success) ? "Done" : "Failed");
|
||||
return _success;
|
||||
}
|
||||
|
||||
// HSA-runtime on-unload method
|
||||
void OnUnload()
|
||||
{
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Unloading...\n");
|
||||
rocprofsys_finalize_hidden();
|
||||
ROCPROFSYS_BASIC_VERBOSE_F(2 || rocm::on_load_trace, "Unloading... Done\n");
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user