Add serialization for TimeVortex and several action types

Add CheckpointAction to trigger checkpoints

src/sst/core/Makefile.am

@@ -21,6 +21,7 @@ nobase_dist_sst_HEADERS = \
 	activity.h \
 	clock.h \
 	baseComponent.h \
+	checkpointAction.h \
 	component.h \
 	componentExtension.h \
 	componentInfo.h \
@@ -92,11 +93,13 @@ nobase_dist_sst_HEADERS = \
 	serialization/serialize.h \
 	serialization/serialize_impl_fwd.h \
 	serialization/serialize_array.h \
+	serialization/serialize_atomic.h \
 	serialization/serialize_buffer_accessor.h \
 	serialization/serialize_deque.h \
 	serialization/serialize_list.h \
 	serialization/serialize_map.h \
 	serialization/serialize_packer.h \
+	serialization/serialize_priority_queue.h \
 	serialization/serialize_serializable.h \
 	serialization/serialize_set.h \
 	serialization/serialize_sizer.h \
@@ -160,6 +163,7 @@ sst_core_sources = \
 	action.cc \
 	clock.cc \
 	baseComponent.cc \
+	checkpointAction.cc \
 	component.cc \
 	componentExtension.cc \
 	componentInfo.cc \

src/sst/core/action.cc

@@ -29,4 +29,10 @@ Action::endSimulation(SimTime_t end)
     Simulation_impl::getSimulation()->endSimulation(end);
 }
 
+void
+Action::serialize_order(SST::Core::Serialization::serializer& ser)
+{
+    SST::Activity::serialize_order(ser);
+}
+
 } // namespace SST

src/sst/core/action.h

@@ -32,12 +32,14 @@ class Action : public Activity
 protected:
     /** Called to signal to the Simulation object to end the simulation */
     void endSimulation();
+
     /** Called to signal to the Simulation object to end the simulation
      * @param end Simulation cycle when the simulation finishes
      */
     void endSimulation(SimTime_t end);
 
-    NotSerializable(SST::Action)
+    void serialize_order(SST::Core::Serialization::serializer& ser) override;
+    ImplementVirtualSerializable(SST::Action)
 };
 
 } // namespace SST

src/sst/core/cfgoutput/jsonConfigOutput.cc

@@ -154,12 +154,14 @@ JSONConfigGraphOutput::generate(const Config* cfg, ConfigGraph* graph)
     outputJson["program_options"]["print-timing-info"]  = cfg->print_timing() ? "true" : "false";
     // Ignore stopAfter for now
     // outputJson["program_options"]["stopAfter"] = cfg->stopAfterSec();
-    outputJson["program_options"]["heartbeat-period"]   = cfg->heartbeatPeriod();
+    outputJson["program_options"]["heartbeat-period"]   = cfg->heartbeat_period();
     outputJson["program_options"]["timebase"]           = cfg->timeBase();
     outputJson["program_options"]["partitioner"]        = cfg->partitioner();
     outputJson["program_options"]["timeVortex"]         = cfg->timeVortex();
     outputJson["program_options"]["interthread-links"]  = cfg->interthread_links() ? "true" : "false";
     outputJson["program_options"]["output-prefix-core"] = cfg->output_core_prefix();
+    outputJson["program_options"]["checkpoint-period"]  = cfg->checkpoint_period();
+
 
     // Put in the global param sets
     for ( const auto& set : getGlobalParamSetNames() ) {

src/sst/core/cfgoutput/pythonConfigOutput.cc

@@ -229,7 +229,7 @@ PythonConfigGraphOutput::generate(const Config* cfg, ConfigGraph* graph)
         outputFile, "sst.setProgramOption(\"print-timing-info\", \"%s\")\n", cfg->print_timing() ? "true" : "false");
     // Ignore stopAfter for now
     // fprintf(outputFile, "sst.setProgramOption(\"stopAfter\", \"%" PRIu32 "\")\n", cfg->stopAfterSec);
-    fprintf(outputFile, "sst.setProgramOption(\"heartbeat-period\", \"%s\")\n", cfg->heartbeatPeriod().c_str());
+    fprintf(outputFile, "sst.setProgramOption(\"heartbeat-period\", \"%s\")\n", cfg->heartbeat_period().c_str());
     fprintf(outputFile, "sst.setProgramOption(\"timebase\", \"%s\")\n", cfg->timeBase().c_str());
     fprintf(outputFile, "sst.setProgramOption(\"partitioner\", \"%s\")\n", cfg->partitioner().c_str());
     fprintf(outputFile, "sst.setProgramOption(\"timeVortex\", \"%s\")\n", cfg->timeVortex().c_str());
@@ -238,6 +238,8 @@ PythonConfigGraphOutput::generate(const Config* cfg, ConfigGraph* graph)
         cfg->interthread_links() ? "true" : "false");
     fprintf(outputFile, "sst.setProgramOption(\"output-prefix-core\", \"%s\")\n", cfg->output_core_prefix().c_str());
 
+    fprintf(outputFile, "sst.setProgramOption(\"checkpoint-period\", \"%s\")\n", cfg->checkpoint_period().c_str());
+
     // Output the global params
     fprintf(outputFile, "# Define the global parameter sets:\n");
     std::vector<std::string> global_param_sets = getGlobalParamSetNames();

src/sst/core/checkpointAction.cc

@@ -0,0 +1,69 @@
+// Copyright 2009-2023 NTESS. Under the terms
+// of Contract DE-NA0003525 with NTESS, the U.S.
+// Government retains certain rights in this software.
+//
+// Copyright (c) 2009-2023, NTESS
+// All rights reserved.
+//
+// This file is part of the SST software package. For license
+// information, see the LICENSE file in the top level directory of the
+// distribution.
+
+#include "sst_config.h"
+
+#include "sst/core/checkpointAction.h"
+
+#include "sst/core/component.h"
+#include "sst/core/mempoolAccessor.h"
+#include "sst/core/simulation_impl.h"
+#include "sst/core/stringize.h"
+#include "sst/core/timeConverter.h"
+#include "sst/core/warnmacros.h"
+
+#ifdef SST_CONFIG_HAVE_MPI
+DISABLE_WARN_MISSING_OVERRIDE
+#include <mpi.h>
+REENABLE_WARNING
+#endif
+
+namespace SST {
+
+CheckpointAction::CheckpointAction(Config* UNUSED(cfg), int this_rank, Simulation_impl* sim, TimeConverter* period) :
+    Action(),
+    rank(this_rank),
+    m_period(period)
+{
+    sim->insertActivity(period->getFactor(), this);
+    if ( (0 == this_rank) ) { lastTime = sst_get_cpu_time(); }
+    // if( (0 == this_rank) ) {
+    //     sim->insertActivity( period->getFactor(), this );
+    //     lastTime = sst_get_cpu_time();
+    // }
+}
+
+CheckpointAction::~CheckpointAction() {}
+
+void
+CheckpointAction::execute(void)
+{
+    Simulation_impl* sim = Simulation_impl::getSimulation();
+    const double     now = sst_get_cpu_time();
+
+    Output& sim_output = sim->getSimulationOutput();
+    if ( 0 == rank ) {
+        sim->getSimulationOutput().output(
+            "# Simulation Checkpoint: Simulated Time %s (Real CPU time since last checkpoint %.5f seconds)\n",
+            sim->getElapsedSimTime().toStringBestSI().c_str(), (now - lastTime));
+
+        lastTime = now;
+    }
+
+    sim->checkpoint();
+
+    SimTime_t next = sim->getCurrentSimCycle() + m_period->getFactor();
+    sim->insertActivity(next, this);
+
+    // Print some resource usage
+}
+
+} // namespace SST

src/sst/core/checkpointAction.h

@@ -0,0 +1,55 @@
+// Copyright 2009-2023 NTESS. Under the terms
+// of Contract DE-NA0003525 with NTESS, the U.S.
+// Government retains certain rights in this software.
+//
+// Copyright (c) 2009-2023, NTESS
+// All rights reserved.
+//
+// This file is part of the SST software package. For license
+// information, see the LICENSE file in the top level directory of the
+// distribution.
+
+#ifndef SST_CORE_CHECKPOINT_ACTION_H
+#define SST_CORE_CHECKPOINT_ACTION_H
+
+#include "sst/core/action.h"
+#include "sst/core/config.h"
+#include "sst/core/cputimer.h"
+#include "sst/core/output.h"
+#include "sst/core/sst_types.h"
+
+#include <set>
+
+namespace SST {
+
+class Simulation_impl;
+class TimeConverter;
+
+/**
+  \class CheckpointAction
+    A recurring event to trigger checkpoint generation
+*/
+class CheckpointAction : public Action
+{
+public:
+    /**
+    Create a new checkpoint object for the simulation core to initiate checkpoints
+    */
+    CheckpointAction(Config* cfg, int this_rank, Simulation_impl* sim, TimeConverter* period);
+    ~CheckpointAction();
+
+    NotSerializable(SST::CheckpointAction) // Going to have to fix this
+
+        private : CheckpointAction() {};
+    CheckpointAction(const CheckpointAction&);
+
+    void           operator=(CheckpointAction const&);
+    void           execute(void) override;
+    int            rank;
+    TimeConverter* m_period;
+    double         lastTime;
+};
+
+} // namespace SST
+
+#endif // SST_CORE_CHECKPOINT_ACTION_H

src/sst/core/componentInfo.h

@@ -168,7 +168,7 @@ class ComponentInfo
     ComponentInfo();
 
     /**
-       Funciton used to serialize the class
+       Function used to serialize the class
      */
     void serialize_order(SST::Core::Serialization::serializer& ser);
 

src/sst/core/config.cc

@@ -162,7 +162,7 @@ class ConfigHelper
     static int setHeartbeat(Config* cfg, const std::string& arg)
     {
         /* TODO: Error checking */
-        cfg->heartbeatPeriod_ = arg;
+        cfg->heartbeat_period_ = arg;
         return 0;
     }
 
@@ -451,6 +451,22 @@ class ConfigHelper
         return 0;
     }
 
+    // Advanced options - checkpointing
+
+    // Set frequency of checkpoint generation
+    static int setCheckpointPeriod(Config* cfg, const std::string& arg)
+    {
+        /* TODO: Error checking */
+        cfg->checkpoint_period_ = arg;
+        return 0;
+    }
+
+    // Set whether to load from checkpoint
+    static int setLoadFromCheckpoint(Config* cfg, const std::string& UNUSED(arg))
+    {
+        if ( arg == "" ) { cfg->load_from_checkpoint_ = true; }
+        return 0;
+    }
 
     // Advanced options - environment
 
@@ -477,7 +493,7 @@ Config::print()
     std::cout << "stop_at = " << stop_at_ << std::endl;
     std::cout << "exit_after = " << exit_after_ << std::endl;
     std::cout << "partitioner = " << partitioner_ << std::endl;
-    std::cout << "heartbeatPeriod = " << heartbeatPeriod_ << std::endl;
+    std::cout << "heartbeat_period = " << heartbeat_period_ << std::endl;
     std::cout << "output_directory = " << output_directory_ << std::endl;
     std::cout << "output_core_prefix = " << output_core_prefix_ << std::endl;
     std::cout << "output_config_graph = " << output_config_graph_ << std::endl;
@@ -489,6 +505,7 @@ Config::print()
     std::cout << "output_partition = " << output_partition_ << std::endl;
     std::cout << "timeBase = " << timeBase_ << std::endl;
     std::cout << "parallel_load = " << parallel_load_ << std::endl;
+    std::cout << "load_checkpoint = " << load_from_checkpoint_ << std::endl;
     std::cout << "timeVortex = " << timeVortex_ << std::endl;
     std::cout << "interthread_links = " << interthread_links_ << std::endl;
 #ifdef USE_MEMPOOL
@@ -535,15 +552,15 @@ Config::Config(uint32_t num_ranks, bool first_rank) : ConfigShared(!first_rank,
     // Basic Options
     first_rank_ = first_rank;
 
-    num_ranks_       = num_ranks;
-    num_threads_     = 1;
-    configFile_      = "NONE";
-    model_options_   = "";
-    print_timing_    = false;
-    stop_at_         = "0 ns";
-    exit_after_      = 0;
-    partitioner_     = "sst.linear";
-    heartbeatPeriod_ = "";
+    num_ranks_        = num_ranks;
+    num_threads_      = 1;
+    configFile_       = "NONE";
+    model_options_    = "";
+    print_timing_     = false;
+    stop_at_          = "0 ns";
+    exit_after_       = 0;
+    partitioner_      = "sst.linear";
+    heartbeat_period_ = "";
 
     char* wd_buf = (char*)malloc(sizeof(char) * PATH_MAX);
     getcwd(wd_buf, PATH_MAX);
@@ -590,6 +607,10 @@ Config::Config(uint32_t num_ranks, bool first_rank) : ConfigShared(!first_rank,
 #endif
     rank_seq_startup_ = false;
 
+    // Advanced Options - Checkpointing
+    checkpoint_period_    = "";
+    load_from_checkpoint_ = false;
+
     // Advanced Options - environment
     enable_sig_handling_ = true;
 
@@ -780,6 +801,17 @@ Config::insertOptions()
         "disable-signal-handlers", 0, "Disable signal handlers",
         std::bind(&ConfigHelper::disableSigHandlers, this, _1));
 
+    /* Advanced Features - Checkpoint */
+    DEF_SECTION_HEADING("Advanced Options - Checkpointing (EXPERIMENTAL)");
+    DEF_ARG(
+        "checkpoint-period", 0, "PERIOD",
+        "Set frequency for checkpoints to be generated (this is an approximate timing and specified in simulated "
+        "time.\n ",
+        std::bind(&ConfigHelper::setCheckpointPeriod, this, _1), true);
+    DEF_ARG(
+        "load-checkpoint", 0, "FILE", "Set file to load checkpoint from",
+        std::bind(&ConfigHelper::setLoadFromCheckpoint, this, _1), true);
+
     enableDashDashSupport(std::bind(&ConfigHelper::setModelOptions, this, _1));
     addPositionalCallback(std::bind(&Config::positionalCallback, this, _1, _2));
 };