Skip to content

GitLab

Commit 3bbbc738 authored by K-H-Ismail's avatar K-H-Ismail Committed by Gauthier Quesnel
Browse files

benchmark: add fully connected network benchmark

parent 17155b25
Pipeline #15158 passed with stage
in 1 minute and 32 seconds
Hide whitespace changes
Inline Side-by-side
lib/benchmark/benchmark.cpp
View file @ 3bbbc738
......@@ -91,7 +91,7 @@ dot_graph_save(const irt::simulation& sim, std::FILE* os)
}
}
}
enum neuron_type { gener,leaky_int_fire,izhikevich };
struct neuron_lif {
irt::dynamics_id sum;
irt::dynamics_id prod;
......@@ -142,7 +142,7 @@ struct synapse {
irt::dynamics_id accumulator_syn;
};
struct neuron_izhikevich
make_neuron_izhikevich(irt::simulation* sim, long unsigned int i, double a, double b, double c, double d, double I, double vini ) noexcept
make_neuron_izhikevich(irt::simulation* sim, long unsigned int i, double quantum, double a, double b, double c, double d, double I, double vini ) noexcept
{
using namespace boost::ut;
......@@ -175,7 +175,7 @@ make_neuron_izhikevich(irt::simulation* sim, long unsigned int i, double a, doub
quantifier_a.default_adapt_state =
irt::quantifier::adapt_state::possible;
quantifier_a.default_zero_init_offset = true;
quantifier_a.default_step_size = 0.1;
quantifier_a.default_step_size = quantum;
quantifier_a.default_past_length = 3;
integrator_b.default_current_value = 0.0;
......@@ -183,7 +183,7 @@ make_neuron_izhikevich(irt::simulation* sim, long unsigned int i, double a, doub
quantifier_b.default_adapt_state =
irt::quantifier::adapt_state::possible;
quantifier_b.default_zero_init_offset = true;
quantifier_b.default_step_size = 0.01;
quantifier_b.default_step_size = quantum;
quantifier_b.default_past_length = 3;
product.default_input_coeffs[0] = 1.0;
......@@ -303,10 +303,10 @@ make_neuron_gen(irt::simulation* sim, long unsigned int i, double offset, double
return neuron_model;
}
struct neuron_lif
make_neuron_lif(irt::simulation* sim, long unsigned int i) noexcept
make_neuron_lif(irt::simulation* sim, long unsigned int i, double quantum, double tau) noexcept
{
using namespace boost::ut;
double tau_lif = 5.0 + static_cast <float> (rand()) /( static_cast <float> (RAND_MAX/(10.0-5.0)));
double tau_lif = tau;
double Vr_lif = 0.0;
double Vt_lif = 1.0;
......@@ -334,7 +334,7 @@ make_neuron_lif(irt::simulation* sim, long unsigned int i) noexcept
quantifier_lif.default_adapt_state =
irt::quantifier::adapt_state::possible;
quantifier_lif.default_zero_init_offset = true;
quantifier_lif.default_step_size = 0.1;
quantifier_lif.default_step_size = quantum;
quantifier_lif.default_past_length = 3;
cross_lif.default_threshold = Vt_lif;
......@@ -562,18 +562,18 @@ struct synapse make_synapse(irt::simulation* sim, long unsigned int source, long
return synapse_model;
}
void network(long unsigned int N, double simulation_duration, double quantum)
void network(neuron_type T, long unsigned int N, double simulation_duration, double quantum_synapse, double quantum_neuron, double spike_rate)
{
using namespace boost::ut;
irt::simulation sim;
constexpr size_t base{ 10000000 };
constexpr size_t base{ 100000000 };
constexpr size_t ten{ 10 };
constexpr size_t two{ 2 };
expect(irt::is_success(sim.model_list_allocator.init(base + (two * N * N + N) * ten)));
expect(irt::is_success(sim.message_list_allocator.init(base + (two * N * N + N) * ten)));
expect(irt::is_success(sim.message_list_allocator.init(base*two + (two * N * N + N) * ten)));
expect(irt::is_success(sim.input_port_list_allocator.init(base + (two * N * N + N) * ten * ten )));
expect(irt::is_success(sim.output_port_list_allocator.init(base + (two * N * N + N) * ten * ten)));
expect(irt::is_success(sim.emitting_output_port_allocator.init(base + (two * N * N + N) * ten)));
......@@ -581,7 +581,7 @@ void network(long unsigned int N, double simulation_duration, double quantum)
expect(irt::is_success(sim.sched.init(base + (two * N * N + N))) * ten);
expect(irt::is_success(sim.models.init(base + (two * N * N + N))));
expect(irt::is_success(sim.init_messages.init(base + (two * N * N + N))));
expect(irt::is_success(sim.init_messages.init(base*two + (two * N * N + N))));
expect(irt::is_success(sim.messages.init(base + (two * N * N + N))));
expect(irt::is_success(sim.input_ports.init(base + (two * N * N + N) * 16)));
expect(irt::is_success(sim.output_ports.init(base + (two * N * N + N) * 7)));
......@@ -605,39 +605,79 @@ void network(long unsigned int N, double simulation_duration, double quantum)
printf(">> Allocating neurones ... ");
auto start = std::chrono::steady_clock::now();
/*std::vector<struct neuron_gen> neurons;
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_gen neuron_model = make_neuron_gen(&sim,i,i+1.0,N+1.0);
neurons.emplace_back(neuron_model);
}*/
std::vector<struct neuron_lif> neurons;
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_lif neuron_model = make_neuron_lif(&sim,i);
neurons.emplace_back(neuron_model);
std::vector<struct neuron_gen> neurons_gen;
std::vector<struct neuron_izhikevich> neurons_izhikevich;
std::vector<struct neuron_lif> neurons_lif;
switch (T)
{
case gener:
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_gen neuron_model = make_neuron_gen(&sim,i,0.0 + static_cast <float> (rand()) /( static_cast <float> (RAND_MAX/(1.0-0.0))),spike_rate);
neurons_gen.emplace_back(neuron_model);
}
break;
case izhikevich:
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_izhikevich neuron_model = make_neuron_izhikevich(&sim,i,quantum_neuron,(spike_rate/2.0) + static_cast <float> (rand()) /( static_cast <float> (RAND_MAX/(spike_rate/2.0))),0.2,-65.0,8.0,10.0,0.0);
neurons_izhikevich.emplace_back(neuron_model);
}
break;
case leaky_int_fire:
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_lif neuron_model = make_neuron_lif(&sim,i,quantum_neuron,spike_rate + static_cast <float> (rand()) /( static_cast <float> (RAND_MAX/(spike_rate))));
neurons_lif.emplace_back(neuron_model);
}
break;
}
/*std::vector<struct neuron_izhikevich> neurons;
for (long unsigned int i = 0 ; i < N; i++) {
struct neuron_izhikevich neuron_model = make_neuron_izhikevich(&sim,i,0.02,0.2,-65.0,8.0,10.0,0.0);
neurons.emplace_back(neuron_model);
}*/
auto end = std::chrono::steady_clock::now();
printf(" [%f] ms.\n" ,static_cast<double>(std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()));
printf(">> Allocating synapses ... ");
printf(">> Allocating synapses ... \n ");
start = std::chrono::steady_clock::now();
std::vector<struct synapse> synapses;
synapses.reserve(256*N);
for (long unsigned int i = 0 ; i < N; i++) {
for (long unsigned int j = 0 ; j < N; j++) {
struct synapse synapse_model = make_synapse(&sim,i,j,
neurons[i].out_port,
neurons[i].out_port,quantum);
synapses.emplace_back(synapse_model);
}
switch(T)
{
case gener:
printf(" -Neurons type gen. \n");
for (long unsigned int i = 0 ; i < N; i++) {
for (long unsigned int j = 0 ; j < N; j++) {
struct synapse synapse_model = make_synapse(&sim,i,j,
neurons_gen[i].out_port,
neurons_gen[i].out_port,quantum_synapse);
synapses.emplace_back(synapse_model);
}
}
break;
case izhikevich:
printf(" -Neurons type Izhikevich. \n");
for (long unsigned int i = 0 ; i < N; i++) {
for (long unsigned int j = 0 ; j < N; j++) {
struct synapse synapse_model = make_synapse(&sim,i,j,
neurons_izhikevich[i].out_port,
neurons_izhikevich[i].out_port,quantum_synapse);
synapses.emplace_back(synapse_model);
}
}
break;
case leaky_int_fire:
printf(" -Neurons type LIF. \n");
for (long unsigned int i = 0 ; i < N; i++) {
for (long unsigned int j = 0 ; j < N; j++) {
struct synapse synapse_model = make_synapse(&sim,i,j,
neurons_lif[i].out_port,
neurons_lif[i].out_port,quantum_synapse);
synapses.emplace_back(synapse_model);
}
}
break;
}
end = std::chrono::steady_clock::now();
printf(" [%f] s.\n" ,static_cast<double>(std::chrono::duration_cast<std::chrono::seconds>(end - start).count()));
printf(">> Synapses allocated in [%f] s.\n" ,static_cast<double>(std::chrono::duration_cast<std::chrono::seconds>(end - start).count()));
printf(">> synapses size %ld \n",static_cast<long int>(synapses.capacity()));
irt::time t = 0.0;
......@@ -663,13 +703,36 @@ void network(long unsigned int N, double simulation_duration, double quantum)
}
BENCHMARK_P(Network, N, 1, 1,(long unsigned int N, double simulation_duration, double quantum))
BENCHMARK_P(Network, N, 1, 1,(neuron_type T,long unsigned int N, double simulation_duration, double quantum_synapse, double quantum_neuron, double spike_rate))
{
network(N,simulation_duration,quantum);
network(T,N,simulation_duration,quantum_synapse,quantum_neuron,spike_rate);
}
BENCHMARK_P_INSTANCE(Network, N, (10,70,1e-5));
BENCHMARK_P_INSTANCE(Network, N, (100,30,1e-5));
//BENCHMARK_P_INSTANCE(Network, N, (500,30,1e-5));
BENCHMARK_P_INSTANCE(Network, N, (gener,10,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,50,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,100,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,250,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,500,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,750,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (gener,1000,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,10,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,50,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,100,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,250,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,500,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,750,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (leaky_int_fire,1000,500,1e-5,0.1,250));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,10,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,50,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,100,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,250,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,500,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,750,500,1e-5,0.1,0.002));
BENCHMARK_P_INSTANCE(Network, N, (izhikevich,1000,500,1e-5,0.1,0.002));
int
main()
{
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment