Fix HIP Streams Duplication Error (#313)
* Fix stream duplication and fixed tests * Added comments to explain stream.cpp code, change stream nullptr check to occur in update table to prevent readding null stream, simplified hip-streams bin file code, add destroyStreams to hip-streams bin file code * Removed roctx from CMakeLists.txt * Updated documentation * Fix documentation * Removed update_table for HIP compiler table and updated stream.cpp to remove support for HIP compiler table * Added runtime initialization check for HIP * Changed tool name, working on fixing memory management * Added context for counter collection kernel rename combination * Changed name from map to set and changed description * Fix documentation description for group-by-queue * Merged memory copy and kernel operations onto a single track when on the same stream * Updated perfetto output to remove hardware information from track name to merge all memory copy and kernel operations on the same stream to the same track: * Most pr comments addressed * Added filter for counter collection and removed kernel buffer tracing hack * Added PR comment fixes --------- Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
Šī revīzija ir iekļauta:
1 vecāks
4f03ebc360
revīzija
e626df43eb
26 files changed
+465
-287
No files matched your search
@@ -182,10 +182,9 @@ using rocdecode_buffered_output_t =
|
||||
using rocjpeg_buffered_output_t =
|
||||
buffered_output<rocprofiler_buffer_tracing_rocjpeg_api_record_t, domain_type::ROCJPEG>;
|
||||
using kernel_dispatch_buffered_output_with_stream_t =
|
||||
buffered_output<tool_buffer_tracing_kernel_dispatch_with_stream_record_t,
|
||||
domain_type::KERNEL_DISPATCH>;
|
||||
buffered_output<tool_buffer_tracing_kernel_dispatch_ext_record_t, domain_type::KERNEL_DISPATCH>;
|
||||
using memory_copy_buffered_output_with_stream_t =
|
||||
buffered_output<tool_buffer_tracing_memory_copy_with_stream_record_t, domain_type::MEMORY_COPY>;
|
||||
buffered_output<tool_buffer_tracing_memory_copy_ext_record_t, domain_type::MEMORY_COPY>;
|
||||
using pc_sampling_stochastic_buffered_output_t =
|
||||
buffered_output<rocprofiler::tool::rocprofiler_tool_pc_sampling_stochastic_record_t,
|
||||
domain_type::PC_SAMPLING_STOCHASTIC>;
|
||||
|
||||
@@ -249,10 +249,10 @@ generate_csv(const output_config& cfg,
|
||||
}
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& data,
|
||||
const stats_entry_t& stats)
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& data,
|
||||
const stats_entry_t& stats)
|
||||
{
|
||||
if(data.empty()) return;
|
||||
|
||||
@@ -399,10 +399,10 @@ generate_csv(const output_config& cfg,
|
||||
}
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& data,
|
||||
const stats_entry_t& stats)
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& data,
|
||||
const stats_entry_t& stats)
|
||||
{
|
||||
if(data.empty()) return;
|
||||
|
||||
|
||||
@@ -40,10 +40,10 @@ generate_csv(const output_config& cfg,
|
||||
std::vector<agent_info>& data);
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& data,
|
||||
const stats_entry_t& stats);
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& data,
|
||||
const stats_entry_t& stats);
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
@@ -58,10 +58,10 @@ generate_csv(const output_config& cfg,
|
||||
const stats_entry_t& stats);
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& data,
|
||||
const stats_entry_t& stats);
|
||||
generate_csv(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& data,
|
||||
const stats_entry_t& stats);
|
||||
|
||||
void
|
||||
generate_csv(const output_config& cfg,
|
||||
|
||||
@@ -186,11 +186,11 @@ void
|
||||
write_json(json_output& json_ar,
|
||||
const output_config& /*cfg*/,
|
||||
const metadata& /*tool_metadata*/,
|
||||
const domain_stats_vec_t& domain_stats,
|
||||
generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>&& hip_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_hsa_api_record_t> hsa_api_gen,
|
||||
generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t> kernel_dispatch_gen,
|
||||
generator<tool_buffer_tracing_memory_copy_with_stream_record_t> memory_copy_gen,
|
||||
const domain_stats_vec_t& domain_stats,
|
||||
generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>&& hip_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_hsa_api_record_t> hsa_api_gen,
|
||||
generator<tool_buffer_tracing_kernel_dispatch_ext_record_t> kernel_dispatch_gen,
|
||||
generator<tool_buffer_tracing_memory_copy_ext_record_t> memory_copy_gen,
|
||||
generator<tool_counter_record_t> counter_collection_gen,
|
||||
generator<rocprofiler_buffer_tracing_marker_api_record_t> marker_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_scratch_memory_record_t> scratch_memory_gen,
|
||||
|
||||
@@ -81,14 +81,14 @@ void
|
||||
write_json(json_output&, const output_config& cfg, const metadata& tool_metadata, uint64_t pid);
|
||||
|
||||
void
|
||||
write_json(json_output& json_ar,
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const domain_stats_vec_t& domain_stats,
|
||||
generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>&& hip_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_hsa_api_record_t> hsa_api_gen,
|
||||
generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t> kernel_dispatch_gen,
|
||||
generator<tool_buffer_tracing_memory_copy_with_stream_record_t> memory_copy_gen,
|
||||
write_json(json_output& json_ar,
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const domain_stats_vec_t& domain_stats,
|
||||
generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>&& hip_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_hsa_api_record_t> hsa_api_gen,
|
||||
generator<tool_buffer_tracing_kernel_dispatch_ext_record_t> kernel_dispatch_gen,
|
||||
generator<tool_buffer_tracing_memory_copy_ext_record_t> memory_copy_gen,
|
||||
generator<tool_counter_record_t> counter_collection_gen,
|
||||
generator<rocprofiler_buffer_tracing_marker_api_record_t> marker_api_gen,
|
||||
generator<rocprofiler_buffer_tracing_scratch_memory_record_t> scratch_memory_gen,
|
||||
|
||||
@@ -356,15 +356,15 @@ create_attribute_list()
|
||||
|
||||
void
|
||||
write_otf2(
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
uint64_t pid,
|
||||
const std::vector<agent_info>& agent_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hip_api_ext_record_t>* hip_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hsa_api_record_t>* hsa_api_data,
|
||||
std::deque<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>* kernel_dispatch_data,
|
||||
std::deque<tool_buffer_tracing_memory_copy_with_stream_record_t>* memory_copy_data,
|
||||
std::deque<rocprofiler_buffer_tracing_marker_api_record_t>* marker_api_data,
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
uint64_t pid,
|
||||
const std::vector<agent_info>& agent_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hip_api_ext_record_t>* hip_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hsa_api_record_t>* hsa_api_data,
|
||||
std::deque<tool_buffer_tracing_kernel_dispatch_ext_record_t>* kernel_dispatch_data,
|
||||
std::deque<tool_buffer_tracing_memory_copy_ext_record_t>* memory_copy_data,
|
||||
std::deque<rocprofiler_buffer_tracing_marker_api_record_t>* marker_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_scratch_memory_record_t>* /*scratch_memory_data*/,
|
||||
std::deque<rocprofiler_buffer_tracing_rccl_api_record_t>* rccl_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_memory_allocation_record_t>* memory_allocation_data,
|
||||
|
||||
@@ -36,19 +36,19 @@ namespace tool
|
||||
{
|
||||
void
|
||||
write_otf2(
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
uint64_t pid,
|
||||
const std::vector<agent_info>& agent_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hip_api_ext_record_t>* hip_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hsa_api_record_t>* hsa_api_data,
|
||||
std::deque<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>* kernel_dispatch_data,
|
||||
std::deque<tool_buffer_tracing_memory_copy_with_stream_record_t>* memory_copy_data,
|
||||
std::deque<rocprofiler_buffer_tracing_marker_api_record_t>* marker_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_scratch_memory_record_t>* scratch_memory_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rccl_api_record_t>* rccl_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_memory_allocation_record_t>* memory_allocation_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rocdecode_api_ext_record_t>* rocdecode_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rocjpeg_api_record_t>* rocjpeg_api_data);
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
uint64_t pid,
|
||||
const std::vector<agent_info>& agent_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hip_api_ext_record_t>* hip_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_hsa_api_record_t>* hsa_api_data,
|
||||
std::deque<tool_buffer_tracing_kernel_dispatch_ext_record_t>* kernel_dispatch_data,
|
||||
std::deque<tool_buffer_tracing_memory_copy_ext_record_t>* memory_copy_data,
|
||||
std::deque<rocprofiler_buffer_tracing_marker_api_record_t>* marker_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_scratch_memory_record_t>* scratch_memory_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rccl_api_record_t>* rccl_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_memory_allocation_record_t>* memory_allocation_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rocdecode_api_ext_record_t>* rocdecode_api_data,
|
||||
std::deque<rocprofiler_buffer_tracing_rocjpeg_api_record_t>* rocjpeg_api_data);
|
||||
} // namespace tool
|
||||
} // namespace rocprofiler
|
||||
@@ -65,15 +65,15 @@ get_hash_id(Tp&& _val)
|
||||
|
||||
void
|
||||
write_perfetto(
|
||||
const output_config& ocfg,
|
||||
const metadata& tool_metadata,
|
||||
std::vector<agent_info> agent_data,
|
||||
const generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>& hip_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_hsa_api_record_t>& hsa_api_gen,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& kernel_dispatch_gen,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& memory_copy_gen,
|
||||
const generator<tool_counter_record_t>& counter_collection_gen,
|
||||
const generator<rocprofiler_buffer_tracing_marker_api_record_t>& marker_api_gen,
|
||||
const output_config& ocfg,
|
||||
const metadata& tool_metadata,
|
||||
std::vector<agent_info> agent_data,
|
||||
const generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>& hip_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_hsa_api_record_t>& hsa_api_gen,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& kernel_dispatch_gen,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& memory_copy_gen,
|
||||
const generator<tool_counter_record_t>& counter_collection_gen,
|
||||
const generator<rocprofiler_buffer_tracing_marker_api_record_t>& marker_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_scratch_memory_record_t>& /*scratch_memory_gen*/,
|
||||
const generator<rocprofiler_buffer_tracing_rccl_api_record_t>& rccl_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_memory_allocation_record_t>& memory_allocation_gen,
|
||||
@@ -133,10 +133,12 @@ write_perfetto(
|
||||
|
||||
tracing_session->Setup(cfg);
|
||||
tracing_session->StartBlocking();
|
||||
|
||||
auto tids = std::set<rocprofiler_thread_id_t>{};
|
||||
auto demangled = std::unordered_map<std::string_view, std::string>{};
|
||||
auto agent_thread_ids = std::unordered_map<rocprofiler_agent_id_t, std::set<uint64_t>>{};
|
||||
const auto is_hip_initialized =
|
||||
tool_metadata.is_runtime_initialized(ROCPROFILER_RUNTIME_INITIALIZATION_HIP);
|
||||
const auto group_by_queue = ocfg.group_by_queue || !is_hip_initialized;
|
||||
auto tids = std::set<rocprofiler_thread_id_t>{};
|
||||
auto demangled = std::unordered_map<std::string_view, std::string>{};
|
||||
auto agent_thread_ids = std::unordered_map<rocprofiler_agent_id_t, std::set<uint64_t>>{};
|
||||
auto agent_thread_ids_alloc = std::unordered_map<rocprofiler_agent_id_t, std::set<uint64_t>>{};
|
||||
auto agent_queue_ids =
|
||||
std::unordered_map<rocprofiler_agent_id_t, std::unordered_set<rocprofiler_queue_id_t>>{};
|
||||
@@ -154,12 +156,7 @@ write_perfetto(
|
||||
auto agent_queue_tracks =
|
||||
std::unordered_map<rocprofiler_agent_id_t,
|
||||
std::unordered_map<rocprofiler_queue_id_t, ::perfetto::Track>>{};
|
||||
auto agent_stream_compute_tracks =
|
||||
std::unordered_map<rocprofiler_agent_id_t,
|
||||
std::unordered_map<rocprofiler_stream_id_t, ::perfetto::Track>>{};
|
||||
auto agent_stream_copy_tracks =
|
||||
std::unordered_map<rocprofiler_agent_id_t,
|
||||
std::unordered_map<rocprofiler_stream_id_t, ::perfetto::Track>>{};
|
||||
auto stream_tracks = std::unordered_map<rocprofiler_stream_id_t, ::perfetto::Track>{};
|
||||
|
||||
auto _get_agent = [&agent_data](rocprofiler_agent_id_t _id) -> const rocprofiler_agent_t* {
|
||||
for(const auto& itr : agent_data)
|
||||
@@ -194,7 +191,7 @@ write_perfetto(
|
||||
{
|
||||
tids.emplace(itr.thread_id);
|
||||
agent_stream_ids[itr.dst_agent_id].emplace(itr.stream_id);
|
||||
if(ocfg.group_by_queue)
|
||||
if(group_by_queue)
|
||||
{
|
||||
agent_thread_ids[itr.dst_agent_id].emplace(itr.thread_id);
|
||||
}
|
||||
@@ -212,7 +209,7 @@ write_perfetto(
|
||||
{
|
||||
tids.emplace(itr.thread_id);
|
||||
agent_stream_ids[itr.dispatch_info.agent_id].emplace(itr.stream_id);
|
||||
if(ocfg.group_by_queue)
|
||||
if(group_by_queue)
|
||||
{
|
||||
agent_queue_ids[itr.dispatch_info.agent_id].emplace(itr.dispatch_info.queue_id);
|
||||
}
|
||||
@@ -297,20 +294,11 @@ write_perfetto(
|
||||
{
|
||||
for(auto sitr : aitr.second)
|
||||
{
|
||||
const auto* _agent = _get_agent(aitr.first);
|
||||
const auto stream_id = sitr.handle;
|
||||
const auto stream_id = sitr.handle;
|
||||
|
||||
{
|
||||
auto _namess = std::stringstream{};
|
||||
_namess << "COMPUTE AGENT [" << _agent->logical_node_id << "] STREAM [" << stream_id
|
||||
<< "] ";
|
||||
|
||||
if(_agent->type == ROCPROFILER_AGENT_TYPE_CPU)
|
||||
_namess << "(CPU)";
|
||||
else if(_agent->type == ROCPROFILER_AGENT_TYPE_GPU)
|
||||
_namess << "(GPU)";
|
||||
else
|
||||
_namess << "(UNK)";
|
||||
_namess << fmt::format("STREAM [\" {} \"] ", stream_id);
|
||||
|
||||
auto _track = ::perfetto::Track{get_hash_id(_namess.str())};
|
||||
auto _desc = _track.Serialize();
|
||||
@@ -318,28 +306,7 @@ write_perfetto(
|
||||
|
||||
perfetto::TrackEvent::SetTrackDescriptor(_track, _desc);
|
||||
|
||||
agent_stream_compute_tracks[aitr.first].emplace(sitr, _track);
|
||||
}
|
||||
|
||||
{
|
||||
auto _namess = std::stringstream{};
|
||||
_namess << "COPY to AGENT [" << _agent->logical_node_id << "] STREAM [" << stream_id
|
||||
<< "] ";
|
||||
|
||||
if(_agent->type == ROCPROFILER_AGENT_TYPE_CPU)
|
||||
_namess << "(CPU)";
|
||||
else if(_agent->type == ROCPROFILER_AGENT_TYPE_GPU)
|
||||
_namess << "(GPU)";
|
||||
else
|
||||
_namess << "(UNK)";
|
||||
|
||||
auto _track = ::perfetto::Track{get_hash_id(_namess.str())};
|
||||
auto _desc = _track.Serialize();
|
||||
_desc.set_name(_namess.str());
|
||||
|
||||
perfetto::TrackEvent::SetTrackDescriptor(_track, _desc);
|
||||
|
||||
agent_stream_copy_tracks[aitr.first].emplace(sitr, _track);
|
||||
stream_tracks.emplace(sitr, _track);
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -570,13 +537,13 @@ write_perfetto(
|
||||
auto name = buffer_names.at(itr.kind, itr.operation);
|
||||
|
||||
::perfetto::Track* _track = nullptr;
|
||||
if(ocfg.group_by_queue)
|
||||
if(group_by_queue)
|
||||
{
|
||||
_track = &agent_thread_tracks.at(itr.dst_agent_id).at(itr.thread_id);
|
||||
}
|
||||
else
|
||||
{
|
||||
_track = &agent_stream_copy_tracks.at(itr.dst_agent_id).at(itr.stream_id);
|
||||
_track = &stream_tracks.at(itr.stream_id);
|
||||
}
|
||||
|
||||
TRACE_EVENT_BEGIN(
|
||||
@@ -640,7 +607,7 @@ write_perfetto(
|
||||
rocprofiler_agent_id_t,
|
||||
std::unordered_map<
|
||||
rocprofiler_queue_id_t,
|
||||
std::vector<tool_buffer_tracing_kernel_dispatch_with_stream_record_t*>>>{};
|
||||
std::vector<tool_buffer_tracing_kernel_dispatch_ext_record_t*>>>{};
|
||||
for(auto& itr : generator)
|
||||
{
|
||||
const auto& info = itr.dispatch_info;
|
||||
@@ -671,14 +638,13 @@ write_perfetto(
|
||||
auto name = std::string_view{sym->kernel_name};
|
||||
|
||||
::perfetto::Track* _track = nullptr;
|
||||
if(ocfg.group_by_queue)
|
||||
if(group_by_queue)
|
||||
{
|
||||
_track = &agent_queue_tracks.at(info.agent_id).at(info.queue_id);
|
||||
}
|
||||
else
|
||||
{
|
||||
_track =
|
||||
&agent_stream_compute_tracks.at(info.agent_id).at((*it)->stream_id);
|
||||
_track = &stream_tracks.at((*it)->stream_id);
|
||||
}
|
||||
|
||||
// Temporary fix until timestamp issues are resolved: Set timestamps to be
|
||||
@@ -729,6 +695,11 @@ write_perfetto(
|
||||
.get_agent_index(agents_map.at(info.agent_id).id,
|
||||
ocfg.agent_index_value)
|
||||
.as_string("-"),
|
||||
"agent_type",
|
||||
tool_metadata
|
||||
.get_agent_index(agents_map.at(info.agent_id).id,
|
||||
ocfg.agent_index_value)
|
||||
.type,
|
||||
"corr_id",
|
||||
current.correlation_id.internal,
|
||||
"queue",
|
||||
|
||||
@@ -37,13 +37,13 @@ namespace tool
|
||||
{
|
||||
void
|
||||
write_perfetto(
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
std::vector<agent_info> agent_data,
|
||||
const generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>& hip_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_hsa_api_record_t>& hsa_api_gen,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& kernel_dispatch_gen,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& memory_copy_gen,
|
||||
const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
std::vector<agent_info> agent_data,
|
||||
const generator<rocprofiler_buffer_tracing_hip_api_ext_record_t>& hip_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_hsa_api_record_t>& hsa_api_gen,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& kernel_dispatch_gen,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& memory_copy_gen,
|
||||
const generator<tool_counter_record_t>& counter_collection_gen,
|
||||
const generator<rocprofiler_buffer_tracing_marker_api_record_t>& marker_api_gen,
|
||||
const generator<rocprofiler_buffer_tracing_scratch_memory_record_t>& scratch_memory_gen,
|
||||
|
||||
@@ -63,8 +63,8 @@ get_stats(const stats_map_t& data_v)
|
||||
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& /*cfg*/,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& data)
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& data)
|
||||
{
|
||||
auto kernel_stats = stats_map_t{};
|
||||
for(auto ditr : data)
|
||||
@@ -119,8 +119,8 @@ generate_stats(const output_config& /*cfg*/,
|
||||
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& /*cfg*/,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& data)
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& data)
|
||||
{
|
||||
auto memory_copy_stats = stats_map_t{};
|
||||
for(auto ditr : data)
|
||||
|
||||
@@ -32,9 +32,9 @@ namespace rocprofiler
|
||||
namespace tool
|
||||
{
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_with_stream_record_t>& data);
|
||||
generate_stats(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_kernel_dispatch_ext_record_t>& data);
|
||||
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& cfg,
|
||||
@@ -47,9 +47,9 @@ generate_stats(const output_config& cfg
|
||||
const generator<rocprofiler_buffer_tracing_hsa_api_record_t>& data);
|
||||
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_with_stream_record_t>& data);
|
||||
generate_stats(const output_config& cfg,
|
||||
const metadata& tool_metadata,
|
||||
const generator<tool_buffer_tracing_memory_copy_ext_record_t>& data);
|
||||
|
||||
stats_entry_t
|
||||
generate_stats(const output_config& cfg,
|
||||
|
||||
@@ -475,6 +475,26 @@ metadata::add_external_correlation_id(uint64_t val)
|
||||
[](auto& _data, uint64_t _val) { return _data.emplace(_val).second; }, val);
|
||||
}
|
||||
|
||||
bool
|
||||
metadata::add_runtime_initialization(rocprofiler_runtime_initialization_operation_t runtime_op)
|
||||
{
|
||||
return runtime_initialization_set.wlock(
|
||||
[](auto& _data, rocprofiler_runtime_initialization_operation_t _runtime_op) {
|
||||
return _data.emplace(_runtime_op).second;
|
||||
},
|
||||
runtime_op);
|
||||
}
|
||||
|
||||
bool
|
||||
metadata::is_runtime_initialized(rocprofiler_runtime_initialization_operation_t runtime_op) const
|
||||
{
|
||||
return runtime_initialization_set.rlock(
|
||||
[](const auto& _data, rocprofiler_runtime_initialization_operation_t _runtime_op) {
|
||||
return _data.count(_runtime_op) > 0;
|
||||
},
|
||||
runtime_op);
|
||||
}
|
||||
|
||||
std::string_view
|
||||
metadata::get_marker_message(uint64_t corr_id) const
|
||||
{
|
||||
|
||||
@@ -90,6 +90,8 @@ using synced_map = common::Synchronized<Tp, true>;
|
||||
template <typename Tp>
|
||||
using synced_obj = common::Synchronized<Tp, true>;
|
||||
using pc_sampling_stats_t = rocprofiler_tool_pc_sampling_stats;
|
||||
using runtime_initialization_set_t =
|
||||
std::unordered_set<rocprofiler_runtime_initialization_operation_t>;
|
||||
|
||||
enum class agent_indexing
|
||||
{
|
||||
@@ -126,17 +128,18 @@ struct metadata
|
||||
agent_counter_info_map_t agent_counter_info = {};
|
||||
agent_pc_sample_config_info_map_t agent_pc_sample_config_info = {};
|
||||
|
||||
sdk::buffer_name_info buffer_names = {};
|
||||
sdk::callback_name_info callback_names = {};
|
||||
synced_map<code_object_data_map_t> code_objects = {};
|
||||
synced_map<kernel_symbol_data_map_t> kernel_symbols = {};
|
||||
synced_map<marker_message_map_t> marker_messages = {};
|
||||
synced_map<string_entry_map_t> string_entries = {};
|
||||
synced_map<external_corr_id_set_t> external_corr_ids = {};
|
||||
synced_map<host_function_info_map_t> host_functions = {};
|
||||
synced_map<code_object_load_info_vec_t> code_object_load = {};
|
||||
att_filenames_map_t att_filenames = {};
|
||||
synced_obj<pc_sampling_stats_t> pc_sampling_stats = {};
|
||||
sdk::buffer_name_info buffer_names = {};
|
||||
sdk::callback_name_info callback_names = {};
|
||||
synced_map<code_object_data_map_t> code_objects = {};
|
||||
synced_map<kernel_symbol_data_map_t> kernel_symbols = {};
|
||||
synced_map<marker_message_map_t> marker_messages = {};
|
||||
synced_map<string_entry_map_t> string_entries = {};
|
||||
synced_map<external_corr_id_set_t> external_corr_ids = {};
|
||||
synced_map<host_function_info_map_t> host_functions = {};
|
||||
synced_map<code_object_load_info_vec_t> code_object_load = {};
|
||||
att_filenames_map_t att_filenames = {};
|
||||
synced_obj<pc_sampling_stats_t> pc_sampling_stats = {};
|
||||
synced_obj<runtime_initialization_set_t> runtime_initialization_set = {};
|
||||
|
||||
metadata() = default;
|
||||
metadata(inprocess);
|
||||
@@ -182,6 +185,7 @@ struct metadata
|
||||
bool add_host_function(host_function_info&& func);
|
||||
bool add_string_entry(size_t key, std::string_view str);
|
||||
bool add_external_correlation_id(uint64_t);
|
||||
bool add_runtime_initialization(rocprofiler_runtime_initialization_operation_t);
|
||||
|
||||
std::string_view get_marker_message(uint64_t corr_id) const;
|
||||
std::string_view get_kernel_name(uint64_t kernel_id, uint64_t rename_id) const;
|
||||
@@ -193,6 +197,7 @@ struct metadata
|
||||
rocprofiler_tracing_operation_t op) const;
|
||||
agent_index get_agent_index(rocprofiler_agent_id_t agent, agent_indexing index) const;
|
||||
const std::string* get_string_entry(size_t key) const;
|
||||
bool is_runtime_initialized(rocprofiler_runtime_initialization_operation_t) const;
|
||||
|
||||
private:
|
||||
bool inprocess_init = false;
|
||||
|
||||
@@ -35,56 +35,55 @@ namespace rocprofiler
|
||||
{
|
||||
namespace tool
|
||||
{
|
||||
struct tool_buffer_tracing_kernel_dispatch_with_stream_record_t
|
||||
struct tool_buffer_tracing_kernel_dispatch_ext_record_t
|
||||
: rocprofiler_buffer_tracing_kernel_dispatch_record_t
|
||||
{
|
||||
using base_type = rocprofiler_buffer_tracing_kernel_dispatch_record_t;
|
||||
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t(
|
||||
const base_type& _base,
|
||||
const rocprofiler_stream_id_t& _stream_id,
|
||||
const uint64_t& _kernel_rename_val)
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t(const base_type& _base,
|
||||
const rocprofiler_stream_id_t& _stream_id,
|
||||
const uint64_t& _kernel_rename_val)
|
||||
: base_type{_base}
|
||||
, stream_id{_stream_id}
|
||||
, kernel_rename_val{_kernel_rename_val}
|
||||
{}
|
||||
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t();
|
||||
~tool_buffer_tracing_kernel_dispatch_with_stream_record_t() = default;
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t(
|
||||
const tool_buffer_tracing_kernel_dispatch_with_stream_record_t&) = default;
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t(
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t& operator =(
|
||||
const tool_buffer_tracing_kernel_dispatch_with_stream_record_t&) = default;
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t& operator =(
|
||||
tool_buffer_tracing_kernel_dispatch_with_stream_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t();
|
||||
~tool_buffer_tracing_kernel_dispatch_ext_record_t() = default;
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t(
|
||||
const tool_buffer_tracing_kernel_dispatch_ext_record_t&) = default;
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t(
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t& operator =(
|
||||
const tool_buffer_tracing_kernel_dispatch_ext_record_t&) = default;
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t& operator =(
|
||||
tool_buffer_tracing_kernel_dispatch_ext_record_t&&) noexcept = default;
|
||||
|
||||
rocprofiler_stream_id_t stream_id = {};
|
||||
uint64_t kernel_rename_val = {};
|
||||
};
|
||||
|
||||
struct tool_buffer_tracing_memory_copy_with_stream_record_t
|
||||
struct tool_buffer_tracing_memory_copy_ext_record_t
|
||||
: rocprofiler_buffer_tracing_memory_copy_record_t
|
||||
{
|
||||
using base_type = rocprofiler_buffer_tracing_memory_copy_record_t;
|
||||
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t(const base_type& _base,
|
||||
const rocprofiler_stream_id_t& _stream_id)
|
||||
tool_buffer_tracing_memory_copy_ext_record_t(const base_type& _base,
|
||||
const rocprofiler_stream_id_t& _stream_id)
|
||||
: base_type{_base}
|
||||
, stream_id{_stream_id}
|
||||
{}
|
||||
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t();
|
||||
~tool_buffer_tracing_memory_copy_with_stream_record_t() = default;
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t(
|
||||
const tool_buffer_tracing_memory_copy_with_stream_record_t&) = default;
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t(
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t& operator =(
|
||||
const tool_buffer_tracing_memory_copy_with_stream_record_t&) = default;
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t& operator =(
|
||||
tool_buffer_tracing_memory_copy_with_stream_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_memory_copy_ext_record_t();
|
||||
~tool_buffer_tracing_memory_copy_ext_record_t() = default;
|
||||
tool_buffer_tracing_memory_copy_ext_record_t(
|
||||
const tool_buffer_tracing_memory_copy_ext_record_t&) = default;
|
||||
tool_buffer_tracing_memory_copy_ext_record_t(
|
||||
tool_buffer_tracing_memory_copy_ext_record_t&&) noexcept = default;
|
||||
tool_buffer_tracing_memory_copy_ext_record_t& operator =(
|
||||
const tool_buffer_tracing_memory_copy_ext_record_t&) = default;
|
||||
tool_buffer_tracing_memory_copy_ext_record_t& operator =(
|
||||
tool_buffer_tracing_memory_copy_ext_record_t&&) noexcept = default;
|
||||
|
||||
rocprofiler_stream_id_t stream_id = {};
|
||||
};
|
||||
@@ -97,8 +96,8 @@ namespace cereal
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar,
|
||||
const ::rocprofiler::tool::tool_buffer_tracing_kernel_dispatch_with_stream_record_t& data)
|
||||
save(ArchiveT& ar,
|
||||
const ::rocprofiler::tool::tool_buffer_tracing_kernel_dispatch_ext_record_t& data)
|
||||
{
|
||||
cereal::save(ar, static_cast<const rocprofiler_buffer_tracing_kernel_dispatch_record_t&>(data));
|
||||
SAVE_DATA_FIELD(stream_id);
|
||||
@@ -107,8 +106,7 @@ save(ArchiveT&
|
||||
|
||||
template <typename ArchiveT>
|
||||
void
|
||||
save(ArchiveT& ar,
|
||||
const ::rocprofiler::tool::tool_buffer_tracing_memory_copy_with_stream_record_t& data)
|
||||
save(ArchiveT& ar, const ::rocprofiler::tool::tool_buffer_tracing_memory_copy_ext_record_t& data)
|
||||
{
|
||||
cereal::save(ar, static_cast<const rocprofiler_buffer_tracing_memory_copy_record_t&>(data));
|
||||
SAVE_DATA_FIELD(stream_id);
|
||||
|
||||
Reference in new issue
Block a user