Profile externally loaded mechanisms

CMakeLists.txt

@@ -420,9 +420,6 @@ set(ARB_MODCC_FLAGS)
 if(ARB_VECTORIZE)
     list(APPEND ARB_MODCC_FLAGS "--simd")
 endif()
-if(ARB_WITH_PROFILING)
-    list(APPEND ARB_MODCC_FLAGS "--profile")
-endif()
 
 #----------------------------------------------------------
 # Set up install paths, permissions.

arbor/benchmark_cell_group.cpp

@@ -61,7 +61,7 @@ void benchmark_cell_group::advance(epoch ep,
     using std::chrono::high_resolution_clock;
     using duration_type = std::chrono::duration<double, std::micro>;
 
-    PE(advance_bench_cell);
+    PE(advance:bench:cell);
     // Micro-seconds to advance in this epoch.
     auto us = 1e3*(ep.duration());
     for (auto i: util::make_span(0, gids_.size())) {

arbor/communication/communicator.cpp

@@ -139,12 +139,12 @@ time_type communicator::min_delay() {
 }
 
 gathered_vector<spike> communicator::exchange(std::vector<spike> local_spikes) {
-    PE(communication_exchange_sort);
+    PE(communication:exchange:sort);
     // sort the spikes in ascending order of source gid
     util::sort_by(local_spikes, [](spike s){return s.source;});
     PL();
 
-    PE(communication_exchange_gather);
+    PE(communication:exchange:gather);
     // global all-to-all to gather a local copy of the global spike list on each node.
     auto global_spikes = distributed_->gather_spikes(local_spikes);
     num_spikes_ += global_spikes.size();

arbor/fvm_lowered_cell_impl.hpp

@@ -205,7 +205,7 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
     set_gpu();
 
     // Integration setup
-    PE(advance_integrate_setup);
+    PE(advance:integrate:setup);
     threshold_watcher_.clear_crossings();
 
     auto n_samples = staged_samples.size();
@@ -234,11 +234,11 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
 
         // Deliver events and accumulate mechanism current contributions.
 
-        PE(advance_integrate_events);
+        PE(advance:integrate:events);
         state_->deliverable_events.mark_until_after(state_->time);
         PL();
 
-        PE(advance_integrate_current_zero);
+        PE(advance:integrate:current:zero);
         state_->zero_currents();
         PL();
         for (auto& m: mechanisms_) {
@@ -255,7 +255,7 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
         // Add current contribution from gap_junctions
         state_->add_gj_current();
 
-        PE(advance_integrate_events);
+        PE(advance:integrate:events);
         state_->deliverable_events.drop_marked_events();
 
         // Update event list and integration step times.
@@ -270,24 +270,24 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
         // want to use mean current contributions as opposed to point
         // sample.)
 
-        PE(advance_integrate_stimuli)
+        PE(advance:integrate:stimuli)
         state_->add_stimulus_current();
         PL();
 
         // Take samples at cell time if sample time in this step interval.
 
-        PE(advance_integrate_samples);
+        PE(advance:integrate:samples);
         sample_events_.mark_until(state_->time_to);
         state_->take_samples(sample_events_.marked_events(), sample_time_, sample_value_);
         sample_events_.drop_marked_events();
         PL();
 
         // Integrate voltage by matrix solve.
 
-        PE(advance_integrate_matrix_build);
+        PE(advance:integrate:matrix:build);
         matrix_.assemble(state_->dt_intdom, state_->voltage, state_->current_density, state_->conductivity);
         PL();
-        PE(advance_integrate_matrix_solve);
+        PE(advance:integrate:matrix:solve);
         matrix_.solve(state_->voltage);
         PL();
 
@@ -299,17 +299,17 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
 
         // Update ion concentrations.
 
-        PE(advance_integrate_ionupdate);
+        PE(advance:integrate:ionupdate);
         update_ion_state();
         PL();
 
         // Update time and test for spike threshold crossings.
 
-        PE(advance_integrate_threshold);
+        PE(advance:integrate:threshold);
         threshold_watcher_.test(&state_->time_since_spike);
         PL();
 
-        PE(advance_integrate_post)
+        PE(advance:integrate:post)
         if (post_events_) {
             for (auto& m: mechanisms_) {
                 m->post_event();
@@ -323,14 +323,14 @@ fvm_integration_result fvm_lowered_cell_impl<Backend>::integrate(
         // Check for non-physical solutions:
 
         if (check_voltage_mV_>0) {
-            PE(advance_integrate_physicalcheck);
+            PE(advance:integrate:physicalcheck);
             assert_voltage_bounded(check_voltage_mV_);
             PL();
         }
 
         // Check for end of integration.
 
-        PE(advance_integrate_stepsupdate);
+        PE(advance:integrate:stepsupdate);
         if (!--remaining_steps) {
             tmin_ = state_->time_bounds().first;
             remaining_steps = dt_steps(tmin_, tfinal, dt_max);

arbor/include/arbor/mechanism.hpp

@@ -9,6 +9,8 @@
 #include <arbor/fvm_types.hpp>
 #include <arbor/mechanism_abi.h>
 #include <arbor/mechinfo.hpp>
+#include <arbor/profile/profiler.hpp>
+#include <arbor/version.hpp>
 
 namespace arb {
 
@@ -32,6 +34,8 @@ class mechanism {
     mechanism(const arb_mechanism_type m,
               const arb_mechanism_interface& i): mech_{m}, iface_{i}, ppack_{} {
         if (mech_.abi_version != ARB_MECH_ABI_VERSION) throw unsupported_abi_error{mech_.abi_version};
+        state_prof_id   = profile::profiler_region_id("advance:integrate:state:"+internal_name());
+        current_prof_id = profile::profiler_region_id("advance:integrate:current:"+internal_name());
     }
     mechanism() = default;
     mechanism(const mechanism&) = delete;
@@ -55,8 +59,8 @@ class mechanism {
 
     // Forward to interface methods
     void initialize()     { ppack_.vec_t = *time_ptr_ptr; iface_.init_mechanism(&ppack_); }
-    void update_current() { ppack_.vec_t = *time_ptr_ptr; iface_.compute_currents(&ppack_); }
-    void update_state()   { ppack_.vec_t = *time_ptr_ptr; iface_.advance_state(&ppack_); }
+    void update_current() { prof_enter(current_prof_id); ppack_.vec_t = *time_ptr_ptr; iface_.compute_currents(&ppack_); prof_exit(); }
+    void update_state()   { prof_enter(state_prof_id);   ppack_.vec_t = *time_ptr_ptr; iface_.advance_state(&ppack_);    prof_exit(); }
     void update_ions()    { ppack_.vec_t = *time_ptr_ptr; iface_.write_ions(&ppack_); }
     void post_event()     { ppack_.vec_t = *time_ptr_ptr; iface_.post_event(&ppack_); }
     void deliver_events(arb_deliverable_event_stream& stream) { ppack_.vec_t  = *time_ptr_ptr; iface_.apply_events(&ppack_, &stream); }
@@ -68,6 +72,21 @@ class mechanism {
     arb_mechanism_interface iface_;
     arb_mechanism_ppack ppack_;
     arb_value_type** time_ptr_ptr = nullptr;
+
+private:
+#ifdef ARB_PROFILE_ENABLED
+    void prof_enter(profile::region_id_type id) {
+        profile::profiler_enter(id);
+    }
+    void prof_exit() {
+        profile::profiler_leave();
+    }
+#else
+    void prof_enter(profile::region_id_type) {}
+    void prof_exit() {}
+#endif
+    profile::region_id_type state_prof_id;
+    profile::region_id_type current_prof_id;
 };
 
 struct mechanism_layout {

arbor/include/arbor/profile/profiler.hpp

@@ -39,7 +39,7 @@ void profiler_enter(std::size_t region_id);
 void profiler_leave();
 
 profile profiler_summary();
-std::size_t profiler_region_id(const char* name);
+std::size_t profiler_region_id(const std::string& name);
 
 std::ostream& operator<<(std::ostream&, const profile&);
 

arbor/lif_cell_group.cpp

@@ -40,7 +40,7 @@ cell_kind lif_cell_group::get_cell_kind() const {
 }
 
 void lif_cell_group::advance(epoch ep, time_type dt, const event_lane_subrange& event_lanes) {
-    PE(advance_lif);
+    PE(advance:lif);
     if (event_lanes.size() > 0) {
         for (auto lid: util::make_span(gids_.size())) {
             // Advance each cell independently.

arbor/mc_cell_group.cpp

@@ -388,7 +388,7 @@ void mc_cell_group::advance(epoch ep, time_type dt, const event_lane_subrange& e
 
     // Bin and collate deliverable events from event lanes.
 
-    PE(advance_eventsetup);
+    PE(advance:eventsetup);
     staged_events_.clear();
 
     // Skip event handling if nothing to deliver.
@@ -446,7 +446,7 @@ void mc_cell_group::advance(epoch ep, time_type dt, const event_lane_subrange& e
     // value as defined below, grouping together all the samples of the
     // same probe for this callback in this association.
 
-    PE(advance_samplesetup);
+    PE(advance:samplesetup);
     std::vector<sampler_call_info> call_info;
 
     std::vector<sample_event> sample_events;
@@ -527,7 +527,7 @@ void mc_cell_group::advance(epoch ep, time_type dt, const event_lane_subrange& e
     // vector of sample entries from the lowered cell sample times and values
     // and then call the callback.
 
-    PE(advance_sampledeliver);
+    PE(advance:sampledeliver);
     std::vector<sample_record> sample_records;
     sample_records.reserve(max_samples_per_call);
 

arbor/profile/profiler.cpp

@@ -20,24 +20,24 @@ using util::make_span;
 namespace {
     // Check whether a string describes a valid profiler region name.
     bool is_valid_region_string(const std::string& s) {
-        if (s.size()==0u || s.front()=='_' || s.back()=='_') return false;
+        if (s.size()==0u || s.front()==':' || s.back()==':') return false;
         return s.find("__") == s.npos;
     }
 
     //
-    // Return a list of the words in the string, using '_' as the delimiter
+    // Return a list of the words in the string, using ':' as the delimiter
     // string, e.g.:
     //      "communicator"             -> {"communicator"}
     //      "communicator_events"      -> {"communicator", "events"}
     //      "communicator_events_sort" -> {"communicator", "events", "sort"}
     std::vector<std::string> split(const std::string& str) {
         std::vector<std::string> cont;
         std::size_t first = 0;
-        std::size_t last = str.find('_');
+        std::size_t last = str.find(':');
         while (last != std::string::npos) {
             cont.push_back(str.substr(first, last - first));
             first = last + 1;
-            last = str.find('_', first);
+            last = str.find(':', first);
         }
         cont.push_back(str.substr(first, last - first));
         return cont;
@@ -91,7 +91,7 @@ class profiler {
     // The regions are assigned consecutive indexes in the order that they are
     // added to the profiler with calls to `region_index()`, with the first
     // region numbered zero.
-    std::unordered_map<const char*, region_id_type> name_index_;
+    std::unordered_map<std::string, region_id_type> name_index_;
 
     // The name of each region being recorded, with index stored in name_index_
     // is used to index into region_names_.
@@ -108,10 +108,10 @@ class profiler {
 
     void initialize(task_system_handle& ts);
     void enter(region_id_type index);
-    void enter(const char* name);
+    void enter(const std::string& name);
     void leave();
     const std::vector<std::string>& regions() const;
-    region_id_type region_index(const char* name);
+    region_id_type region_index(const std::string& name);
     profile results() const;
 
     static profiler& get_global_profiler() {
@@ -186,7 +186,7 @@ void profiler::enter(region_id_type index) {
     recorders_[thread_ids_.at(std::this_thread::get_id())].enter(index);
 }
 
-void profiler::enter(const char* name) {
+void profiler::enter(const std::string& name) {
     if (!init_) return;
     const auto index = region_index(name);
     recorders_[thread_ids_.at(std::this_thread::get_id())].enter(index);
@@ -197,7 +197,7 @@ void profiler::leave() {
     recorders_[thread_ids_.at(std::this_thread::get_id())].leave();
 }
 
-region_id_type profiler::region_index(const char* name) {
+region_id_type profiler::region_index(const std::string& name) {
     // The name_index_ hash table is shared by all threads, so all access
     // has to be protected by a mutex.
     std::lock_guard<std::mutex> guard(mutex_);
@@ -249,6 +249,20 @@ profile profiler::results() const {
 
     p.num_threads = recorders_.size();
 
+    // Remove elements with count == 0
+    for(unsigned i=0; i<p.counts.size();) {
+        if (p.counts[i] != 0) {
+            ++i;
+            continue;
+        }
+        std::swap(p.counts[i], p.counts.back());
+        std::swap(p.times[i],  p.times.back());
+        std::swap(p.names[i],  p.names.back());
+        p.counts.pop_back();
+        p.times.pop_back();
+        p.names.pop_back();
+    }
+
     return p;
 }
 
@@ -332,7 +346,7 @@ void profiler_leave() {
     profiler::get_global_profiler().leave();
 }
 
-region_id_type profiler_region_id(const char* name) {
+region_id_type profiler_region_id(const std::string& name) {
     if (!is_valid_region_string(name)) {
         throw std::runtime_error(std::string("'")+name+"' is not a valid profiler region name.");
     }
@@ -370,7 +384,7 @@ void profiler_enter(region_id_type) {}
 profile profiler_summary();
 void profiler_print(const profile& prof, float threshold) {};
 profile profiler_summary() {return profile();}
-region_id_type profiler_region_id(const char*) {return 0;}
+region_id_type profiler_region_id(const std::string&) {return 0;}
 std::ostream& operator<<(std::ostream& o, const profile&) {return o;}
 
 #endif // ARB_HAVE_PROFILING