Skip to content

adding json stats support #1238

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
feldergast merged 2 commits into from
Mar 13, 2025
Merged

adding json stats support #1238

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
96fa3a8
adding json statistics support
jleidel Mar 7, 2025
790c39d
adding stats to messagemesh
jleidel Mar 7, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
82 changes: 81 additions & 1 deletion src/sst/core/cfgoutput/jsonConfigOutput.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,19 @@ struct StatPair
SST::ConfigComponent const* comp;
};

struct StatGroupPair
{
std::pair<const std::string, SST::ConfigStatGroup> const& group;
std::vector<std::string> vec;
SST::ConfigGraph const* graph;
};

struct StatGroupParamPair
{
const std::string name;
const SST::Params& stat;
};

void
to_json(json::ordered_json& j, StatPair const& sp)
{
Expand Down Expand Up @@ -136,6 +149,49 @@ to_json(json::ordered_json& j, LinkConfPair const& pair)
j["right"]["latency"] = link->latency_str[1];
}

void
to_json(json::ordered_json& j, StatGroupParamPair const& pair)
{
auto const& outParams = pair.stat;

j["name"] = pair.name;

for ( auto const& param : outParams.getKeys() ) {
j["params"][param] = outParams.find<std::string>(param);
}
}

void
to_json(json::ordered_json& j, StatGroupPair const& pair)
{
auto const& grp = pair.group.second;
auto const* graph = pair.graph;
auto vec = pair.vec;

j["name"] = grp.name;

if ( grp.outputFrequency.getValue() != 0 ) { j["frequency"] = grp.outputFrequency.toStringBestSI(); }

if ( grp.outputID != 0 ) {
const SST::ConfigStatOutput& out = graph->getStatOutput(grp.outputID);
j["output"]["type"] = out.type;
if ( !out.params.empty() ) {
const SST::Params& outParams = out.params;
for ( auto const& param : vec ) {
j["output"]["params"][param] = outParams.find<std::string>(param);
}
}
}

for ( auto& i : grp.statMap ) {
if ( !i.second.empty() ) { j["statistics"].emplace_back(StatGroupParamPair { i.first, i.second }); }
}

for ( SST::ComponentId_t id : grp.components ) {
const SST::ConfigComponent* comp = graph->findComponent(id);
j["components"].emplace_back(comp->name);
}
}
} // namespace

void
Expand Down Expand Up @@ -164,14 +220,38 @@ JSONConfigGraphOutput::generate(const Config* cfg, ConfigGraph* graph)
outputJson["program_options"]["checkpoint-sim-period"] = cfg->checkpoint_sim_period();
outputJson["program_options"]["checkpoint-wall-period"] = std::to_string(cfg->checkpoint_wall_period());


// Put in the global param sets
for ( const auto& set : getGlobalParamSetNames() ) {
for ( const auto& kvp : getGlobalParamSet(set) ) {
if ( kvp.first != "<set_name>" ) outputJson["global_params"][set][kvp.first] = kvp.second;
}
}

// Global statistics
if ( 0 != graph->getStatLoadLevel() ) {
outputJson["statistics_options"]["statisticLoadLevel"] = (uint64_t)graph->getStatLoadLevel();
}

if ( !graph->getStatOutput().type.empty() ) {
outputJson["statistics_options"]["statisticOutput"] = graph->getStatOutput().type.c_str();
const Params& outParams = graph->getStatOutput().params;
if ( !outParams.empty() ) {
// generate the parameters
for ( auto const& paramsItr : getParamsLocalKeys(outParams) ) {
outputJson["statistics_options"]["params"][paramsItr] = outParams.find<std::string>(paramsItr);
}
}
}

// Generate the stat groups
if ( !graph->getStatGroups().empty() ) {
outputJson["statistics_group"];
for ( auto& grp : graph->getStatGroups() ) {
auto vec = getParamsLocalKeys(graph->getStatOutput(grp.second.outputID).params);
outputJson["statistics_group"].emplace_back(StatGroupPair { grp, vec, graph });
}
}

// no components exist in this rank
if ( const_cast<ConfigComponentMap_t&>(compMap).size() == 0 ) { outputJson["components"]; }

Expand Down
178 changes: 177 additions & 1 deletion src/sst/core/model/json/jsonmodel.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,6 +71,7 @@ SSTJSONModelDefinition::recursiveSubcomponent(ConfigComponent* Parent, const nlo
{
std::string Name;
std::string Type;
std::string StatName;
ConfigComponent* Comp = nullptr;
int Slot = 0;

Expand Down Expand Up @@ -119,6 +120,32 @@ SSTJSONModelDefinition::recursiveSubcomponent(ConfigComponent* Parent, const nlo
}
}

// read the statistics
if ( subArray.contains("statistics") ) {
for ( auto& stats : compArray.at("statistics") ) {
// -- stat name
if ( stats.contains("name") ) {
auto sn = stats.find("name");
if ( sn != stats.end() ) { StatName = sn.value(); }
else {
output->fatal(
CALL_INFO, 1, "Error discovering component stat name from script: %s\n",
scriptName.c_str());
}
}

// -- stat params
Params StatParams;
if ( stats.contains("params") ) {
for ( auto& paramArray : stats.at("params").items() ) {
StatParams.insert(paramArray.key(), paramArray.value());
}
}

Comp->enableStatistic(StatName, StatParams);
}
}

// recursively build up the subcomponents
if ( subArray.contains("subcomponents") ) {
auto& subsubArray = subArray["subcomponents"];
Expand All @@ -132,6 +159,7 @@ SSTJSONModelDefinition::discoverComponents(const json& jFile)
{
std::string Name;
std::string Type;
std::string StatName;
ComponentId_t Id;
ConfigComponent* Comp = nullptr;
uint32_t rank = 0;
Expand Down Expand Up @@ -185,6 +213,32 @@ SSTJSONModelDefinition::discoverComponents(const json& jFile)
}
}

// read the statistics
if ( compArray.contains("statistics") ) {
for ( auto& stats : compArray.at("statistics") ) {
// -- stat name
if ( stats.contains("name") ) {
auto sn = stats.find("name");
if ( sn != stats.end() ) { StatName = sn.value(); }
else {
output->fatal(
CALL_INFO, 1, "Error discovering component stat name from script: %s\n",
scriptName.c_str());
}
}

// -- stat params
Params StatParams;
if ( stats.contains("params") ) {
for ( auto& paramArray : stats.at("params").items() ) {
StatParams.insert(paramArray.key(), paramArray.value());
}
}

Comp->enableStatistic(StatName, StatParams);
}
}

// set the rank information
RankInfo Rank(rank, thread);
Comp->setRank(Rank);
Expand Down Expand Up @@ -294,6 +348,127 @@ SSTJSONModelDefinition::discoverGlobalParams(const json& jFile)
}
}

void
SSTJSONModelDefinition::setStatGroupOptions(const json& jFile)
{
std::string Name;
std::string Frequency;
std::string Type;
std::string StatName;

for ( auto& statArray : jFile.at("statistics_group") ) {
// -- name
auto x = statArray.find("name");
if ( x != statArray.end() ) { Name = x.value(); }
else {
output->fatal(
CALL_INFO, 1, "Error discovering statistics group name from script: %s\n", scriptName.c_str());
}

auto* csg = graph->getStatGroup(Name);
if ( csg == nullptr ) {
output->fatal(
CALL_INFO, 1, "Error creating statistics group from script %s; name=%s\n", scriptName.c_str(),
Name.c_str());
}

// -- frequency
auto f = statArray.find("frequency");
if ( f != statArray.end() ) {
Frequency = f.value();
if ( !csg->setFrequency(Frequency) ) {
output->fatal(CALL_INFO, 1, "Error setting frequency for statistics group: %s\n", Name.c_str());
}
}

// -- output
if ( statArray.contains("output") ) {
auto& statOuts = graph->getStatOutputs();
if ( statArray.at("output").contains("type") ) { statArray.at("output").at("type").get_to(Type); }
else {
output->fatal(
CALL_INFO, 1, "Error discovering statistics group output type for group: %s\n", Name.c_str());
}

statOuts.emplace_back(ConfigStatOutput(Type));
csg->setOutput(statOuts.size() - 1);

if ( statArray.at("output").contains("params") ) {
for ( auto& paramArray : statArray.at("output").at("params").items() ) {
statOuts.back().addParameter(paramArray.key(), paramArray.value());
}
}
}

// -- statistics
if ( statArray.contains("statistics") ) {
for ( auto& stats : statArray.at("statistics") ) {
// -- stat name
if ( stats.contains("name") ) {
auto sn = stats.find("name");
if ( sn != stats.end() ) { StatName = sn.value(); }
else {
output->fatal(
CALL_INFO, 1, "Error discovering statistics group stat name from script: %s\n",
scriptName.c_str());
}
}

// -- stat params
Params StatParams;
if ( stats.contains("params") ) {
for ( auto& paramArray : stats.at("params").items() ) {
StatParams.insert(paramArray.key(), paramArray.value());
}
}

csg->addStatistic(StatName, StatParams);

bool verified;
std::string reason;
std::tie(verified, reason) = csg->verifyStatsAndComponents(graph);
if ( !verified ) {
output->fatal(CALL_INFO, 1, "Error verifying statistics and components: %s\n", reason.c_str());
}
}
}

// -- components
if ( statArray.contains("components") ) {
for ( auto& compArray : statArray["components"].items() ) {
csg->addComponent(findComponentIdByName(compArray.value()));
}
}
}
}

void
SSTJSONModelDefinition::discoverStatistics(const json& jFile)
{
// discover the global statistics options
if ( jFile.contains("statistics_options") ) {
if ( jFile.at("statistics_options").contains("statisticLoadLevel") ) {
uint8_t loadLevel;
jFile.at("statistics_options").at("statisticLoadLevel").get_to(loadLevel);
graph->setStatisticLoadLevel(loadLevel);
}

if ( jFile.at("statistics_options").contains("statisticOutput") ) {
std::string output;
jFile.at("statistics_options").at("statisticOutput").get_to(output);
graph->setStatisticOutput(output);
}

if ( jFile.at("statistics_options").contains("params") ) {
for ( auto& paramArray : jFile.at("statistics_options").at("params").items() ) {
graph->addStatisticOutputParameter(paramArray.key(), paramArray.value());
}
}
}
// discover the statistics groups
if ( jFile.contains("statistics_group") ) { setStatGroupOptions(jFile); }
}

ConfigGraph*
SSTJSONModelDefinition::createConfigGraph()
{
Expand Down Expand Up @@ -322,7 +497,8 @@ SSTJSONModelDefinition::createConfigGraph()
// discover the links
discoverLinks(jFile);

// TODO: discover statistics
// discover statistics
discoverStatistics(jFile);

return graph;
}
2 changes: 2 additions & 0 deletions src/sst/core/model/json/jsonmodel.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,11 +65,13 @@ class SSTJSONModelDefinition : public SSTModelDescription
double start_time;

private:
void setStatGroupOptions(const json& jFile);
void recursiveSubcomponent(ConfigComponent* Parent, const nlohmann::basic_json<>& compArray);
void discoverProgramOptions(const json& jFile);
void discoverComponents(const json& jFile);
void discoverLinks(const json& jFile);
void discoverGlobalParams(const json& jFile);
void discoverStatistics(const json& jFile);
ComponentId_t findComponentIdByName(const std::string& Name);
};

Expand Down
4 changes: 3 additions & 1 deletion src/sst/core/testElements/message_mesh/enclosingComponent.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,7 +118,8 @@ RouteMessage::RouteMessage(
ports(parent_ports),
my_id(nid)
{
rng = new SST::RNG::MersenneRNG(my_id + 100);
rng = new SST::RNG::MersenneRNG(my_id + 100);
mcnt = registerStatistic<uint64_t>("msg_count");
}

void
Expand All @@ -127,4 +128,5 @@ RouteMessage::send(MessageEvent* ev, int UNUSED(incoming_port))
// Route to random port
int nextport = rng->generateNextUInt32() % ports.size();
ports[nextport]->send(ev);
mcnt->addData(1);
}
3 changes: 3 additions & 0 deletions src/sst/core/testElements/message_mesh/enclosingComponent.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -212,6 +212,7 @@ class RouteMessage : public RouteInterface
)

SST_ELI_DOCUMENT_STATISTICS(
{"msg_count", "Message counter", "count", 1},
)

SST_ELI_DOCUMENT_PORTS(
Expand All @@ -229,6 +230,8 @@ class RouteMessage : public RouteInterface
const std::vector<PortInterface*> ports;
int my_id;
SST::RNG::Random* rng;

Statistic<uint64_t>* mcnt;
};

} // namespace SST::CoreTest::MessageMesh
Expand Down
3 changes: 3 additions & 0 deletions tests/test_MessageMesh.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,9 @@
x_size = int(sys.argv[1])
y_size = int(sys.argv[2])

sst.setStatisticOutput("sst.statOutputCSV")
Copy link
Contributor

@feldergast feldergast Mar 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Using the default name for stats output file in the test could cause random failures when running testsuites in parallel. However, I think this is a problem in other testsuites as well and we don't currently run the tests in parallel so we can reevaluate this one when we look at the others.

Copy link
Contributor Author

@jleidel jleidel Mar 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I can override it with a specific name if need be

Copy link
Contributor

@feldergast feldergast Mar 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's leave it as is for now. This just got me thinking about parallel testsuites running and I found errors in other testsuites as well. We'll tackle them all together.

sst.enableAllStatisticsForAllComponents()

# Calculate number of routers and endpoints
num_routers = x_size * y_size

Expand Down
Loading