void print_current_time(FILE *out, char *prefix, char *suffix)
{
#ifdef CHORD_PRINT_LONG_TIME
fprintf(out, "%s%"PRIu64"%s", prefix, wall_time(), suffix);
#else
fprintf(out, "%s%"PRIu64"%s", prefix, (wall_time() << 32) >> 32, suffix);
#endif
}
int main (int argc, char** argv)
{ /* main */
Particle* particle_array = (Particle*)NULL;
Particle* particle_array2 = (Particle*)NULL;
int timestep;
int i;
FILE *input_data = fopen(argv[1], "r");
Particle_input_arguments(input_data);
particle_array = Particle_array_construct(number_of_particles);
particle_array2 = Particle_array_construct(number_of_particles);
Particle_array_initialize(particle_array, number_of_particles);
// for (i = 0; i < number_of_particles; i++) {
// particle_array2[i].mass = particle_array[i].mass;
// }
FILE * fileptr = fopen("nbody_out.xyz", "w");
Particle_array_output_xyz(fileptr, particle_array, number_of_particles);
if (number_of_particles <= 1) return 0;
double start = wall_time();
for (timestep = 1; timestep <= number_of_timesteps; timestep++) {
if ((timestep % timesteps_between_outputs) == 0 ) fprintf(stderr, "Starting timestep #%d.\n", timestep);
Particle_array_calculate_forces_cuda(particle_array, particle_array2, number_of_particles, time_interval );
/* swap arrays */
Particle * tmp = particle_array;
particle_array = particle_array2;
particle_array2 = tmp;
} /* for timestep */
#pragma omp taskwait
double end = wall_time ();
printf("Time in seconds: %g s.\n", end - start);
printf("Particles per second: %g \n", (number_of_particles*number_of_timesteps)/(end-start));
if ((number_of_timesteps % timesteps_between_outputs) != 0) {
Particle_array_output_xyz(fileptr, particle_array, number_of_particles);
}
particle_array = Particle_array_destruct(particle_array, number_of_particles);
if (fclose(fileptr) != 0) {
fprintf(stderr, "ERROR: can't close the output file.\n");
exit(program_failure_code);
}
return program_success_code;
} /* main */
bool LoadingScreen::next_timer(wall_time& time) {
switch (_state) {
case TYPING:
case DONE: time = _next_update; return true;
case LOADING: time = wall_time(); return true;
}
return false;
}
void DebriefingScreen::become_front() {
_typed_chars = 0;
if (_state == TYPING) {
_next_update = now() + kTypingDelay;
} else {
_next_update = wall_time();
}
}
void fields::finished_working() {
double now = wall_time();
if (last_wall_time >= 0)
times_spent[working_on] += now - last_wall_time;
last_wall_time = now;
working_on = was_working_on[0];
for (int i = 0; i+1 < MEEP_TIMING_STACK_SZ; ++i)
was_working_on[i] = was_working_on[i+1];
was_working_on[MEEP_TIMING_STACK_SZ-1] = Other;
}
void fields::am_now_working_on(time_sink s) {
double now = wall_time();
if (last_wall_time >= 0)
times_spent[working_on] += now - last_wall_time;
last_wall_time = now;
for (int i = 0; i+1 < MEEP_TIMING_STACK_SZ; ++i)
was_working_on[i+1] = was_working_on[i];
was_working_on[0] = working_on;
working_on = s;
}
double
wall_dtime(double &t)
{
const double tnew = wall_time();
const double dt = tnew - t;
t = tnew ;
return dt ;
}
EXPORT void startClock(const char* name)
{
struct clock *cp = clocks;
while (cp != NULL) {
if (strcmp(cp->name,name) == 0) {
cp->startTime = wall_time();
return;
}
cp = cp->next;
}
cp = (struct clock*)malloc(sizeof(struct clock));
cp->name = (char*) malloc(strlen(name)+1);
strcpy(cp->name,name);
cp->totalTime = 0;
cp->startTime = wall_time();
cp->next = clocks;
clocks = cp;
return;
}
EXPORT void stopClock(const char* name)
{
struct clock *cp = clocks;
while (cp && strcmp(cp->name,name)) {
cp = cp->next;
}
if (cp && cp->startTime) {
cp->totalTime = (wall_time() - cp->startTime);
printf("%-20s %ld micros\n",cp->name,cp->totalTime);
cp->startTime = 0;
}
}
int main(int argc, char *argv[])
{
int i, j, n, r, step;
double s, time, ttime, *a, *b;
n = atoi(argv[1]);
r = atoi(argv[2]);
step = atoi(argv[3]);
a = (double *) malloc(step * n * sizeof(double));
b = (double *) malloc(step * n * sizeof(double));
time = wall_time();
time = wall_time();
for (i = 0; i < r; i++)
empty();
ttime = wall_time() - time;
for (i = 0; i < n; i++) {
a[i] = i;
b[i] = 1.0 / (i + 1);
}
time = wall_time();
for (j = 0; j < r; j++) {
s = skalar(a, b, n, step);
}
time = wall_time() - time;
printf("Skalarprodukt : %f\n", s);
printf("Laufzeit : %f s\n", time);
printf("Overhead : %g s\n", ttime);
printf("Zeit pro Wdhlg : %g s\n", time / r);
printf("Overhead : %g s\n", ttime / r);
// printf("Rechenleistung : %6.1f MFlop/s\n", 2.0 * n * r * 1e-6 / (time - ttime));
printf("Rechenleistung : %6.1f MFlop/s\n", 2.0 * (n/step+1) * r * 1e-6 / (time - ttime));
free(a);
free(b);
return 0;
}
/***********************************************************************
* cl_ctx_set_timer - set timer
*
* input:
* ctx - context
* tv - timeout after which the timer expires
* fun - function to be invoked when the timer expires
* data - application data passed back to the application when the
* callback is invoked
*
************************************************************************/
cl_timer *cl_ctx_set_timer(cl_context *ctx, struct timeval *tv,
void (*fun)(), void *data)
{
uint64_t when;
Event *ev;
if (ctx == NULL)
return NULL;
when = wall_time();
when = when + UMILLION*tv->tv_sec + tv->tv_usec;
ev = newEvent(fun, data, when);
insertEvent(&ctx->timer_heap, ev);
return (cl_timer *)ev;
}
void DebriefingScreen::fire_timer() {
if (_state != TYPING) {
throw Exception(format("DebriefingScreen::fire_timer() called but _state is {0}", _state));
}
sys.sound.teletype();
wall_time now = antares::now();
while (_next_update <= now) {
if (_typed_chars < _score->size()) {
_next_update += kTypingDelay;
++_typed_chars;
} else {
_next_update = wall_time();
_state = DONE;
break;
}
}
}
void ObjectDataScreen::fire_timer() {
wall_time now = antares::now();
if (_next_sound <= now) {
sys.sound.teletype();
_next_sound += 3 * kTypingDelay;
while (_next_sound <= now) {
_next_sound += kTypingDelay;
}
}
while (_next_update <= now) {
if (_typed_chars < _text->size()) {
_next_update += kTypingDelay;
++_typed_chars;
} else {
_next_update = wall_time();
_state = DONE;
break;
}
}
}
unsigned int __stdcall ping_thread_entry(void *data)
#endif
{
PingThreadData *pdata = (PingThreadData *)data;
int maxfd, ret;
fd_set all_rset, rset;
struct timeval to;
I3ServerList *list = pdata->list;
char *url = pdata->url;
uint64_t *ping_start_time = pdata->ping_start_time;
int num_pings;
I3ServerListNode *next_to_ping;
uint64_t last_ping_time, curr_time;
uint64_t last_add_new_i3servers, last_update_serverlist;
FD_ZERO(&all_rset);
FD_ZERO(&rset);
/* socket init */
#ifdef ICMP_PING
if (init_icmp_socket(&ping_sock) == -1)
abort();
#else
if (init_udp_socket(&ping_sock) == -1)
abort();
#endif
FD_SET(ping_sock, &all_rset);
maxfd = ping_sock + 1;
/* initial populate the list of i3 servers */
update_i3_server_list(url, list, &next_to_ping);
/* determine coordinates */
init_coordinates(list);
/* add some close-by servers from the list based on coordinates */
change_ping_list(list, &next_to_ping, 1);
/* eternal loop */
last_ping_time = last_add_new_i3servers = last_update_serverlist = wall_time();
set_status(ping_start_time, last_ping_time);
for (;;) {
rset = all_rset;
to.tv_sec = 0; to.tv_usec = 10000;
if ((ret = select(maxfd, &rset, NULL, NULL, &to)) < 0) {
if (errno == EINTR)
continue;
else {
perror("select");
abort();
}
}
/* message received on icmp socket */
if (FD_ISSET(ping_sock, &rset)) {
uint32_t addr; uint16_t port, seq; uint64_t rtt;
#ifdef ICMP_PING
if (recv_echo_reply(ping_sock, &addr, &seq, &rtt)) {
#else
if (recv_i3_echo_reply(ping_sock, &addr, &port, &seq, &rtt)) {
#endif
update_ping_information(list, addr, seq, rtt);
}
}
/* need to ping */
curr_time = wall_time();
if (list->num_ping_list > 0) {
char status = get_status(ping_start_time, curr_time);
num_pings = (curr_time - last_ping_time)/
(period_ping[status]/list->num_ping_list);
if (num_pings > 0) {
if (NULL == next_to_ping) {
I3_PRINT_DEBUG0(I3_DEBUG_LEVEL_MINIMAL,
"No servers to ping. Aborting\n");
}
send_npings(ping_sock, list, &next_to_ping, num_pings);
last_ping_time = curr_time;
}
}
/* change the list of i3 servers */
if (curr_time - last_add_new_i3servers >
period_pick_new_server[get_status(ping_start_time, curr_time)]) {
/* testing just the best server */
uint32_t best_addr; uint16_t best_port; uint64_t best_rtt;
struct in_addr ia;
int required_k = 1;
int ret = get_top_k(list, required_k, &best_addr, &best_port, &best_rtt);
if (ret != required_k) {
// We couldn't find the request k top nodes.
I3_PRINT_INFO0 (
I3_INFO_LEVEL_WARNING,
"I3 Ping Thread: Unable to obtain top k nodes.\n"
);
// Dilip: Feb 20, 2006. I don't think the following works.
// TODO: Start
// We set the last_add_new_servers to fool the thread
// to wait for some time before trying again to get
// the top k nodes.
//last_add_new_i3servers = curr_time;
// TODO: End
// Sleep for some time before trying again.
# if defined (_WIN32)
Sleep ( 25 ); // 25 milliseconds
# else
usleep(25 * 1000); // 25 milliseconds
# endif
continue;
}
ia.s_addr = htonl(best_addr);
I3_PRINT_DEBUG3(I3_INFO_LEVEL_MINIMAL,
"Best node: %s:%d with RTT %Ld\n",
inet_ntoa(ia), best_port, best_rtt
);
I3_PRINT_DEBUG0(I3_DEBUG_LEVEL_VERBOSE, "Adding new servers to list\n");
change_ping_list(list, &next_to_ping, 0);
last_add_new_i3servers = curr_time;
}
/* update (wget) i3 server list */
if (curr_time - last_update_serverlist > PERIOD_SERVERLIST_WGET) {
I3_PRINT_DEBUG0( I3_DEBUG_LEVEL_VERBOSE,
"Updating server list from server\n");
update_i3_server_list(url, list, &next_to_ping);
last_update_serverlist = curr_time;
}
}
#ifndef _WIN32
pthread_exit(0);
#endif
return 0;
}
/* BiCGSTAB(L) algorithm for the n-by-n problem Ax = b */
ptrdiff_t bicgstabL(const int L, const size_t n, realnum *x, bicgstab_op A, void *Adata,
const realnum *b, const double tol, int *iters, realnum *work,
const bool quiet) {
if (!work) return (2 * L + 3) * n; // required workspace
prealnum *r = new prealnum[L + 1];
prealnum *u = new prealnum[L + 1];
for (int i = 0; i <= L; ++i) {
r[i] = work + i * n;
u[i] = work + (L + 1 + i) * n;
}
double bnrm = norm2(n, b);
if (bnrm == 0.0) bnrm = 1.0;
int iter = 0;
double last_output_wall_time = wall_time();
double *gamma = new double[L + 1];
double *gamma_p = new double[L + 1];
double *gamma_pp = new double[L + 1];
double *tau = new double[L * L];
double *sigma = new double[L + 1];
int ierr = 0; // error code to return, if any
const double breaktol = 1e-30;
/**** FIXME: check for breakdown conditions(?) during iteration ****/
// rtilde = r[0] = b - Ax
realnum *rtilde = work + (2 * L + 2) * n;
A(x, r[0], Adata);
for (size_t m = 0; m < n; ++m)
rtilde[m] = r[0][m] = b[m] - r[0][m];
{ /* Sleipjen normalizes rtilde in his code; it seems to help slightly */
double s = 1.0 / norm2(n, rtilde);
for (size_t m = 0; m < n; ++m)
rtilde[m] *= s;
}
memset(u[0], 0, sizeof(realnum) * n); // u[0] = 0
double rho = 1.0, alpha = 0, omega = 1;
double resid;
while ((resid = norm2(n, r[0])) > tol * bnrm) {
++iter;
if (!quiet && wall_time() > last_output_wall_time + MEEP_MIN_OUTPUT_TIME) {
master_printf("residual[%d] = %g\n", iter, resid / bnrm);
last_output_wall_time = wall_time();
}
rho = -omega * rho;
for (int j = 0; j < L; ++j) {
if (fabs(rho) < breaktol) {
ierr = -1;
goto finish;
}
double rho1 = dot(n, r[j], rtilde);
double beta = alpha * rho1 / rho;
rho = rho1;
for (int i = 0; i <= j; ++i)
for (size_t m = 0; m < n; ++m)
u[i][m] = r[i][m] - beta * u[i][m];
A(u[j], u[j + 1], Adata);
alpha = rho / dot(n, u[j + 1], rtilde);
for (int i = 0; i <= j; ++i)
xpay(n, r[i], -alpha, u[i + 1]);
A(r[j], r[j + 1], Adata);
xpay(n, x, alpha, u[0]);
}
for (int j = 1; j <= L; ++j) {
for (int i = 1; i < j; ++i) {
int ij = (j - 1) * L + (i - 1);
tau[ij] = dot(n, r[j], r[i]) / sigma[i];
xpay(n, r[j], -tau[ij], r[i]);
}
sigma[j] = dot(n, r[j], r[j]);
gamma_p[j] = dot(n, r[0], r[j]) / sigma[j];
}
omega = gamma[L] = gamma_p[L];
for (int j = L - 1; j >= 1; --j) {
gamma[j] = gamma_p[j];
for (int i = j + 1; i <= L; ++i)
gamma[j] -= tau[(i - 1) * L + (j - 1)] * gamma[i];
}
for (int j = 1; j < L; ++j) {
gamma_pp[j] = gamma[j + 1];
for (int i = j + 1; i < L; ++i)
gamma_pp[j] += tau[(i - 1) * L + (j - 1)] * gamma[i + 1];
}
xpay(n, x, gamma[1], r[0]);
xpay(n, r[0], -gamma_p[L], r[L]);
xpay(n, u[0], -gamma[L], u[L]);
for (int j = 1; j < L; ++j) { /* TODO: use blas DGEMV (for L > 2) */
xpay(n, x, gamma_pp[j], r[j]);
xpay(n, r[0], -gamma_p[j], r[j]);
xpay(n, u[0], -gamma[j], u[j]);
}
if (iter == *iters) {
ierr = 1;
break;
}
}
if (!quiet) master_printf("final residual = %g\n", norm2(n, r[0]) / bnrm);
finish:
delete[] sigma;
delete[] tau;
delete[] gamma_pp;
delete[] gamma_p;
delete[] gamma;
delete[] u;
delete[] r;
*iters = iter;
return ierr;
}
/*---------------------------------------------------------------------------
* (function: do_high_level_synthesis)
*-------------------------------------------------------------------------*/
void do_high_level_synthesis()
{
double elaboration_time = wall_time();
printf("--------------------------------------------------------------------\n");
printf("High-level synthesis Begin\n");
/* Perform any initialization routines here */
#ifdef VPR6
find_hard_multipliers();
find_hard_adders();
//find_hard_adders_for_sub();
register_hard_blocks();
#endif
global_param_table_sc = sc_new_string_cache();
/* parse to abstract syntax tree */
printf("Parser starting - we'll create an abstract syntax tree. "
"Note this tree can be viewed using GraphViz (see documentation)\n");
parse_to_ast();
/* Note that the entry point for ast optimzations is done per module with the
* function void next_parsed_verilog_file(ast_node_t *file_items_list) */
/* after the ast is made potentiatlly do tagging for downstream links to verilog */
if (global_args.high_level_block != NULL)
{
add_tag_data();
}
/* Now that we have a parse tree (abstract syntax tree [ast]) of
* the Verilog we want to make into a netlist. */
printf("Converting AST into a Netlist. "
"Note this netlist can be viewed using GraphViz (see documentation)\n");
create_netlist();
// Can't levelize yet since the large muxes can look like combinational loops when they're not
check_netlist(verilog_netlist);
/* point for all netlist optimizations. */
printf("Performing Optimizations of the Netlist\n");
netlist_optimizations_top(verilog_netlist);
if (configuration.output_netlist_graphs )
{
/* Path is where we are */
graphVizOutputNetlist(configuration.debug_output_path, "optimized", 1, verilog_netlist);
}
/* point where we convert netlist to FPGA or other hardware target compatible format */
printf("Performing Partial Map to target device\n");
partial_map_top(verilog_netlist);
#ifdef VPR5
/* check for problems in the partial mapped netlist */
printf("Check for liveness and combinational loops\n");
levelize_and_check_for_combinational_loop_and_liveness(TRUE, verilog_netlist);
#endif
/* point for outputs. This includes soft and hard mapping all structures to the
* target format. Some of these could be considred optimizations */
printf("Outputting the netlist to the specified output format\n");
output_top(verilog_netlist);
elaboration_time = wall_time() - elaboration_time;
printf("Successful High-level synthesis by Odin in ");
print_time(elaboration_time);
printf("\n");
printf("--------------------------------------------------------------------\n");
// FIXME: free contents?
sc_free_string_cache(global_param_table_sc);
}
int main(int argc, char **argv)
{
void
(*hook_print_SE)(const qnu*, const Lagr*, const nmpc&) = NULL;
void
(*hook_print_LG)(const qnu*, const Lagr*, const nmpc&, const float*) = NULL;
void
(*hook_print_SD)(const unsigned int*, const double*) = NULL;
void
(*hook_print_TR)(const unsigned int*, const double*) = NULL;
double
(*hook_exec_control_horiz)(qnu*, const nmpc&, robot*) = NULL;
char errnote[256];
unsigned int sd_loop = 0, k = 0, current_tgt_no = 0;
float tgtdist, grad_dot_grad = 0.;
robot vme;
double sd_loop_time, time_last_cmd_sent = 0, now;
nmpc C;
cl_opts clopts =
{ false };
parse_command_line(argc, argv, &vme, &clopts);
parse_input_file(C, vme.conffile());
print_greeting(C);
qnu* qu = (qnu*) calloc(C.N, sizeof(qnu));
Lagr* p = (Lagr*) calloc(C.N, sizeof(Lagr));
//cmd* cmd_horiz = (cmd*) calloc(C.C, sizeof(cmd));
float* grad = (float*) calloc(C.N + 1, sizeof(float));
float* last_grad = (float*) calloc(C.N + 1, sizeof(float));
double* time_to_tgt = (double*) calloc(C.ntgt, sizeof(float));
C.cur_tgt = C.tgt;
C.control_step = 0;
init_qu_and_p(qu, p, C);
tgtdist = C.tgttol + 1; // Just to get us into the waypoint loop.
C.horizon_loop = 0;
/*
* This next block of decisions sort out the hooks used to print output.
* Rather than include the ifs in each loop, I'm using function pointers
* which are set at runtime based on CL flags for verbosity.
*/
if (clopts.selec_verbose)
{
hook_print_SE = &print_pathnerr;
hook_print_LG = &print_LG;
hook_print_SD = &print_SD;
hook_print_TR = &print_TR;
}
else if (clopts.selec_verbose)
{
hook_print_SE = &empty_output_hook;
hook_print_LG = &empty_output_hook;
hook_print_SD = &empty_output_hook;
hook_print_TR = &empty_output_hook;
}
else
{
if (clopts.selec_state_and_error_SE)
hook_print_SE = &print_pathnerr;
else
hook_print_SE = &empty_output_hook;
if (clopts.selec_lagrange_grad_LG)
hook_print_LG = &print_LG;
else
hook_print_LG = &empty_output_hook;
if (clopts.selec_SD_converged_SD)
hook_print_SD = &print_SD;
else
hook_print_SD = &empty_output_hook;
if (clopts.selec_target_reached_TR)
hook_print_TR = &print_TR;
else
hook_print_TR = &empty_output_hook;
}
if (clopts.selec_sim)
hook_exec_control_horiz = &exec_control_horiz_dummy;
else
hook_exec_control_horiz = &exec_control_horiz_vme;
if (!clopts.selec_sim)
{
vme.tcp_connect();
vme.update_poshead(qu, C);
}
/*
* Enter the loop which will take us through all waypoints.
*/
while (current_tgt_no < C.ntgt)
{
time_to_tgt[current_tgt_no] = -wall_time();
C.cur_tgt = &C.tgt[current_tgt_no * 2];
tgtdist = C.tgttol + 1;
while (tgtdist > .1)
{
C.horizon_loop += 1;
sd_loop = 0;
sd_loop_time = -wall_time();
/*
* SD loop.
*/
while (1)
{
sd_loop += 1;
grad_dot_grad = 0.;
/*
* The core of the gradient decent is in the next few lines:
*/
tgtdist = predict_horizon(qu, p, C);
get_gradient(qu, p, C, grad);
for (k = 0; k < C.N; k++)
{
grad_dot_grad += grad[k] * last_grad[k];
}
/*
* Detect direction changes in the gradient by inspecting the
* product <grad, last_grad>. If it is positive, then the
* iterations are successfully stepping to the minimum, and we
* can accelerate by increasing dg. If we overshoot (and the
* product becomes negative), then backstep and drop dg to a
* safe value.
*/
if (grad_dot_grad > 0)
{
C.dg *= 2;
for (k = 0; k < C.N; ++k)
{
qu[k].Dth -= C.dg * grad[k];
}
}
else
{
C.dg = 0.1; // TODO: Adaptive.
for (k = 0; k < C.N; ++k)
{
qu[k].Dth += C.dg * grad[k];
}
}
swap_fptr(&grad, &last_grad);
if (last_grad[C.N] < .1)
break;
if (sd_loop >= MAX_SD_ITER)
{
sprintf(errnote, "Reached %d SD iterations. Stopping.",
sd_loop);
report_error(EXCEEDED_MAX_SD_ITER, errnote);
}
}
hook_print_SE(qu, p, C);
hook_print_LG(qu, p, C, grad);
sd_loop_time += wall_time();
hook_print_SD(&sd_loop, &sd_loop_time);
C.control_step += C.C;
hook_exec_control_horiz(qu, C, &vme);
for (k = 0; k < C.N - C.C - 1; ++k)
{
// cmd_horiz[k].v = qu[k].v;
// cmd_horiz[k].Dth = qu[k].Dth;
qu[k].v = qu[k + C.C].v;
qu[k].Dth = qu[k + C.C].Dth;
}
if (C.control_step > MAX_NMPC_ITER * C.C)
{
sprintf(errnote,
"Reached %d NMPC steps without reaching tgt. Stopping.",
MAX_NMPC_ITER);
report_error(TRAPPED_IN_NMPC_LOOP, errnote);
}
/*
* The last thing we do is get the new position and heading for
* the next SD calculation.
*/
// vme.update_poshead(qu);
}
time_to_tgt[current_tgt_no] += wall_time();
hook_print_TR(¤t_tgt_no, &time_to_tgt[current_tgt_no]);
++current_tgt_no;
}
return 0;
}
void LBFGSSolver::solve(const Function& function,
SolverResults* results) const
{
double global_start_time = wall_time();
// Dimension of problem.
size_t n = function.get_number_of_scalars();
if (n == 0) {
results->exit_condition = SolverResults::FUNCTION_TOLERANCE;
return;
}
// Current point, gradient and Hessian.
double fval = std::numeric_limits<double>::quiet_NaN();
double fprev = std::numeric_limits<double>::quiet_NaN();
double normg0 = std::numeric_limits<double>::quiet_NaN();
double normg = std::numeric_limits<double>::quiet_NaN();
double normdx = std::numeric_limits<double>::quiet_NaN();
Eigen::VectorXd x, g;
// Copy the user state to the current point.
function.copy_user_to_global(&x);
Eigen::VectorXd x2(n);
// L-BFGS history.
std::vector<Eigen::VectorXd> s_data(this->lbfgs_history_size),
y_data(this->lbfgs_history_size);
std::vector<Eigen::VectorXd*> s(this->lbfgs_history_size),
y(this->lbfgs_history_size);
for (int h = 0; h < this->lbfgs_history_size; ++h) {
s_data[h].resize(function.get_number_of_scalars());
s_data[h].setZero();
y_data[h].resize(function.get_number_of_scalars());
y_data[h].setZero();
s[h] = &s_data[h];
y[h] = &y_data[h];
}
Eigen::VectorXd rho(this->lbfgs_history_size);
rho.setZero();
Eigen::VectorXd alpha(this->lbfgs_history_size);
alpha.setZero();
Eigen::VectorXd q(n);
Eigen::VectorXd r(n);
// Needed from the previous iteration.
Eigen::VectorXd x_prev(n), s_tmp(n), y_tmp(n);
CheckExitConditionsCache exit_condition_cache;
//
// START MAIN ITERATION
//
results->startup_time += wall_time() - global_start_time;
results->exit_condition = SolverResults::INTERNAL_ERROR;
int iter = 0;
bool last_iteration_successful = true;
int number_of_line_search_failures = 0;
int number_of_restarts = 0;
while (true) {
//
// Evaluate function and derivatives.
//
double start_time = wall_time();
// y[0] should contain the difference between the gradient
// in this iteration and the gradient from the previous.
// Therefore, update y before and after evaluating the
// function.
if (iter > 0) {
y_tmp = -g;
}
fval = function.evaluate(x, &g);
normg = std::max(g.maxCoeff(), -g.minCoeff());
if (iter == 0) {
normg0 = normg;
}
results->function_evaluation_time += wall_time() - start_time;
//
// Update history
//
start_time = wall_time();
if (iter > 0 && last_iteration_successful) {
s_tmp = x - x_prev;
y_tmp += g;
double sTy = s_tmp.dot(y_tmp);
if (sTy > 1e-16) {
// Shift all pointers one step back, discarding the oldest one.
Eigen::VectorXd* sh = s[this->lbfgs_history_size - 1];
Eigen::VectorXd* yh = y[this->lbfgs_history_size - 1];
for (int h = this->lbfgs_history_size - 1; h >= 1; --h) {
s[h] = s[h - 1];
y[h] = y[h - 1];
rho[h] = rho[h - 1];
}
// Reuse the storage of the discarded data for the new data.
s[0] = sh;
y[0] = yh;
*y[0] = y_tmp;
*s[0] = s_tmp;
rho[0] = 1.0 / sTy;
}
}
results->lbfgs_update_time += wall_time() - start_time;
//
// Test stopping criteriea
//
start_time = wall_time();
if (iter > 1 && this->check_exit_conditions(fval, fprev, normg,
normg0, x.norm(), normdx,
last_iteration_successful,
&exit_condition_cache, results)) {
break;
}
if (iter >= this->maximum_iterations) {
results->exit_condition = SolverResults::NO_CONVERGENCE;
break;
}
if (this->callback_function) {
CallbackInformation information;
information.objective_value = fval;
information.x = &x;
information.g = &g;
if (!callback_function(information)) {
results->exit_condition = SolverResults::USER_ABORT;
break;
}
}
results->stopping_criteria_time += wall_time() - start_time;
//
// Compute search direction via L-BGFS two-loop recursion.
//
start_time = wall_time();
bool should_restart = false;
double H0 = 1.0;
if (iter > 0) {
// If the gradient is identical two iterations in a row,
// y will be the zero vector and H0 will be NaN. In this
// case the line search will fail and L-BFGS will be restarted
// with a steepest descent step.
H0 = s[0]->dot(*y[0]) / y[0]->dot(*y[0]);
// If isinf(H0) || isnan(H0)
if (H0 == std::numeric_limits<double>::infinity() ||
H0 == -std::numeric_limits<double>::infinity() ||
H0 != H0) {
should_restart = true;
}
}
q = -g;
for (int h = 0; h < this->lbfgs_history_size; ++h) {
alpha[h] = rho[h] * s[h]->dot(q);
q = q - alpha[h] * (*y[h]);
}
r = H0 * q;
for (int h = this->lbfgs_history_size - 1; h >= 0; --h) {
double beta = rho[h] * y[h]->dot(r);
r = r + (*s[h]) * (alpha[h] - beta);
}
// If the function improves very little, the approximated Hessian
// might be very bad. If this is the case, it is better to discard
// the history once in a while. This allows the solver to correctly
// solve some badly scaled problems.
double restart_test = std::fabs(fval - fprev) /
(std::fabs(fval) + std::fabs(fprev));
if (iter > 0 && iter % 100 == 0 && restart_test
< this->lbfgs_restart_tolerance) {
should_restart = true;
}
if (! last_iteration_successful) {
should_restart = true;
}
if (should_restart) {
if (this->log_function) {
char str[1024];
if (number_of_restarts <= 10) {
std::sprintf(str, "Restarting: fval = %.3e, deltaf = %.3e, max|g_i| = %.3e, test = %.3e",
fval, std::fabs(fval - fprev), normg, restart_test);
this->log_function(str);
}
if (number_of_restarts == 10) {
this->log_function("NOTE: No more restarts will be reported.");
}
number_of_restarts++;
}
r = -g;
for (int h = 0; h < this->lbfgs_history_size; ++h) {
(*s[h]).setZero();
(*y[h]).setZero();
}
rho.setZero();
alpha.setZero();
// H0 is not used, but its value will be printed.
H0 = std::numeric_limits<double>::quiet_NaN();
}
results->lbfgs_update_time += wall_time() - start_time;
//
// Perform line search.
//
start_time = wall_time();
double start_alpha = 1.0;
// In the first iteration, start with a much smaller step
// length. (heuristic used by e.g. minFunc)
if (iter == 0) {
double sumabsg = 0.0;
for (size_t i = 0; i < n; ++i) {
sumabsg += std::fabs(g[i]);
}
start_alpha = std::min(1.0, 1.0 / sumabsg);
}
double alpha_step = this->perform_linesearch(function, x, fval, g,
r, &x2, start_alpha);
if (alpha_step <= 0) {
if (this->log_function) {
this->log_function("Line search failed.");
char str[1024];
std::sprintf(str, "%4d %+.3e %9.3e %.3e %.3e %.3e %.3e",
iter, fval, std::fabs(fval - fprev), normg, alpha_step, H0, rho[0]);
this->log_function(str);
}
if (! last_iteration_successful || number_of_line_search_failures++ > 10) {
// This happens quite seldom. Every time it has happened, the function
// was actually converged to a solution.
results->exit_condition = SolverResults::GRADIENT_TOLERANCE;
break;
}
last_iteration_successful = false;
}
else {
// Record length of this step.
normdx = alpha_step * r.norm();
// Compute new point.
x_prev = x;
x = x + alpha_step * r;
last_iteration_successful = true;
}
results->backtracking_time += wall_time() - start_time;
//
// Log the results of this iteration.
//
start_time = wall_time();
int log_interval = 1;
if (iter > 30) {
log_interval = 10;
}
if (iter > 200) {
log_interval = 100;
}
if (iter > 2000) {
log_interval = 1000;
}
if (this->log_function && iter % log_interval == 0) {
if (iter == 0) {
this->log_function("Itr f deltaf max|g_i| alpha H0 rho");
}
this->log_function(
to_string(
std::setw(4), iter, " ",
std::setw(10), std::setprecision(3), std::scientific, std::showpos, fval, std::noshowpos, " ",
std::setw(9), std::setprecision(3), std::scientific, std::fabs(fval - fprev), " ",
std::setw(9), std::setprecision(3), std::setprecision(3), std::scientific, normg, " ",
std::setw(9), std::setprecision(3), std::scientific, alpha_step, " ",
std::setw(9), std::setprecision(3), std::scientific, H0, " ",
std::setw(9), std::setprecision(3), std::scientific, rho[0]
)
);
}
results->log_time += wall_time() - start_time;
fprev = fval;
iter++;
}
function.copy_global_to_user(x);
results->total_time += wall_time() - global_start_time;
if (this->log_function) {
char str[1024];
std::sprintf(str, " end %+.3e %.3e", fval, normg);
this->log_function(str);
}
}
wall_time now() const { return wall_time(_ticks); }
void fields::step() {
// however many times the fields have been synched, we want to restore now
int save_synchronized_magnetic_fields = synchronized_magnetic_fields;
if (synchronized_magnetic_fields) {
synchronized_magnetic_fields = 1; // reset synchronization count
restore_magnetic_fields();
}
am_now_working_on(Stepping);
if (!t) {
last_step_output_wall_time = wall_time();
last_step_output_t = t;
}
if (!quiet && wall_time() > last_step_output_wall_time + MIN_OUTPUT_TIME) {
master_printf("on time step %d (time=%g), %g s/step\n", t, time(),
(wall_time() - last_step_output_wall_time) /
(t - last_step_output_t));
if (save_synchronized_magnetic_fields)
master_printf(" (doing expensive timestepping of synched fields)\n");
last_step_output_wall_time = wall_time();
last_step_output_t = t;
}
phase_material();
// update cached conductivity-inverse array, if needed
for (int i=0;i<num_chunks;i++) chunks[i]->s->update_condinv();
calc_sources(time()); // for B sources
step_db(B_stuff);
step_source(B_stuff);
step_boundaries(B_stuff);
calc_sources(time() + 0.5*dt); // for integrated H sources
update_eh(H_stuff);
step_boundaries(WH_stuff);
update_pols(H_stuff);
step_boundaries(PH_stuff);
step_boundaries(H_stuff);
if (fluxes) fluxes->update_half();
calc_sources(time() + 0.5*dt); // for D sources
step_db(D_stuff);
step_source(D_stuff);
step_boundaries(D_stuff);
calc_sources(time() + dt); // for integrated E sources
update_eh(E_stuff);
step_boundaries(WE_stuff);
update_pols(E_stuff);
step_boundaries(PE_stuff);
step_boundaries(E_stuff);
if (fluxes) fluxes->update();
t += 1;
update_dfts();
finished_working();
// re-synch magnetic fields if they were previously synchronized
if (save_synchronized_magnetic_fields) {
synchronize_magnetic_fields();
synchronized_magnetic_fields = save_synchronized_magnetic_fields;
}
}
void NelderMeadSolver::solve(const Function& function,
SolverResults* results) const
{
double global_start_time = wall_time();
// Dimension of problem.
size_t n = function.get_number_of_scalars();
if (n == 0) {
results->exit_condition = SolverResults::FUNCTION_TOLERANCE;
return;
}
// The Nelder-Mead simplex.
std::vector<SimplexPoint> simplex(n + 1);
// Copy the user state to the current point.
Eigen::VectorXd x;
function.copy_user_to_global(&x);
initialize_simplex(function, x, &simplex);
SimplexPoint mean_point;
SimplexPoint reflection_point;
SimplexPoint expansion_point;
mean_point.x.resize(n);
reflection_point.x.resize(n);
expansion_point.x.resize(n);
double fmin = std::numeric_limits<double>::quiet_NaN();
double fmax = std::numeric_limits<double>::quiet_NaN();
double fval = std::numeric_limits<double>::quiet_NaN();
double area = std::numeric_limits<double>::quiet_NaN();
double area0 = std::numeric_limits<double>::quiet_NaN();
double length = std::numeric_limits<double>::quiet_NaN();
double length0 = std::numeric_limits<double>::quiet_NaN();
Eigen::MatrixXd area_mat(n, n);
//
// START MAIN ITERATION
//
results->startup_time += wall_time() - global_start_time;
results->exit_condition = SolverResults::INTERNAL_ERROR;
int iter = 0;
int n_shrink_in_a_row = 0;
while (true) {
//
// In each iteration, the worst point in the simplex
// is replaced with a new one.
//
double start_time = wall_time();
mean_point.x.setZero();
fval = 0;
// Compute the mean of the best n points.
for (size_t i = 0; i < n; ++i) {
mean_point.x += simplex[i].x;
fval += simplex[i].value;
}
fval /= double(n);
mean_point.x /= double(n);
fmin = simplex[0].value;
fmax = simplex[n].value;
const char* iteration_type = "n/a";
// Compute the reflexion point and evaluate it.
reflection_point.x = 2.0 * mean_point.x - simplex[n].x;
reflection_point.value = function.evaluate(reflection_point.x);
bool is_shrink = false;
if (simplex[0].value <= reflection_point.value &&
reflection_point.value < simplex[n - 1].value) {
// Reflected point is neither better nor worst in the
// new simplex.
std::swap(reflection_point, simplex[n]);
iteration_type = "Reflect 1";
}
else if (reflection_point.value < simplex[0].value) {
// Reflected point is better than the current best; try
// to go farther along this direction.
// Compute expansion point.
expansion_point.x = 3.0 * mean_point.x - 2.0 * simplex[n].x;
expansion_point.value = function.evaluate(expansion_point.x);
if (expansion_point.value < reflection_point.value) {
std::swap(expansion_point, simplex[n]);
iteration_type = "Expansion";
}
else {
std::swap(reflection_point, simplex[n]);
iteration_type = "Reflect 2";
}
}
else {
// Reflected point is still worse than x[n]; contract.
bool success = false;
if (simplex[n - 1].value <= reflection_point.value &&
reflection_point.value < simplex[n].value) {
// Try to perform "outside" contraction.
expansion_point.x = 1.5 * mean_point.x - 0.5 * simplex[n].x;
expansion_point.value = function.evaluate(expansion_point.x);
if (expansion_point.value <= reflection_point.value) {
std::swap(expansion_point, simplex[n]);
success = true;
iteration_type = "Outside contraction";
}
}
else {
// Try to perform "inside" contraction.
expansion_point.x = 0.5 * mean_point.x + 0.5 * simplex[n].x;
expansion_point.value = function.evaluate(expansion_point.x);
if (expansion_point.value < simplex[n].value) {
std::swap(expansion_point, simplex[n]);
success = true;
iteration_type = "Inside contraction";
}
}
if (! success) {
// Neither outside nor inside contraction was acceptable;
// shrink the simplex toward the best point.
for (size_t i = 1; i < n + 1; ++i) {
simplex[i].x = 0.5 * (simplex[0].x + simplex[i].x);
simplex[i].value = function.evaluate(simplex[i].x);
iteration_type = "Shrink";
is_shrink = true;
}
}
}
std::sort(simplex.begin(), simplex.end());
results->function_evaluation_time += wall_time() - start_time;
//
// Test stopping criteriea
//
start_time = wall_time();
// Compute the area of the simplex.
length = 0;
for (size_t i = 0; i < n; ++i) {
area_mat.col(i) = simplex[i].x - simplex[n].x;
length = std::max(length, area_mat.col(i).norm());
}
area = std::abs(area_mat.determinant());
if (iter == 0) {
area0 = area;
length0 = length;
}
if (area / area0 < this->area_tolerance) {
results->exit_condition = SolverResults::GRADIENT_TOLERANCE;
break;
}
if (area == 0) {
results->exit_condition = SolverResults::GRADIENT_TOLERANCE;
break;
}
if (length / length0 < this->length_tolerance) {
results->exit_condition = SolverResults::GRADIENT_TOLERANCE;
break;
}
if (is_shrink) {
n_shrink_in_a_row++;
}
else {
n_shrink_in_a_row = 0;
}
if (n_shrink_in_a_row > 50) {
results->exit_condition = SolverResults::GRADIENT_TOLERANCE;
break;
}
if (iter >= this->maximum_iterations) {
results->exit_condition = SolverResults::NO_CONVERGENCE;
break;
}
if (this->callback_function) {
CallbackInformation information;
information.objective_value = simplex[0].value;
information.x = &simplex[0].x;
if (!callback_function(information)) {
results->exit_condition = SolverResults::USER_ABORT;
break;
}
}
results->stopping_criteria_time += wall_time() - start_time;
//
// Restarting
//
//if (area / area1 < 1e-10) {
// x = simplex[0].x;
// initialize_simplex(function, x, &simplex);
// area1 = area;
// if (this->log_function) {
// this->log_function("Restarted.");
// }
//}
//
// Log the results of this iteration.
//
start_time = wall_time();
int log_interval = 1;
if (iter > 30) {
log_interval = 10;
}
if (iter > 200) {
log_interval = 100;
}
if (iter > 2000) {
log_interval = 1000;
}
if (this->log_function && iter % log_interval == 0) {
char str[1024];
if (iter == 0) {
this->log_function("Itr min(f) avg(f) max(f) area length type");
}
std::sprintf(str, "%6d %+.3e %+.3e %+.3e %.3e %.3e %s",
iter, fmin, fval, fmax, area, length, iteration_type);
this->log_function(str);
}
results->log_time += wall_time() - start_time;
iter++;
}
// Return the best point as solution.
function.copy_global_to_user(simplex[0].x);
results->total_time += wall_time() - global_start_time;
if (this->log_function) {
char str[1024];
std::sprintf(str, " end %+.3e %.3e %.3e", fval, area, length);
this->log_function(str);
}
}
void main(void) {
// Malloc spaces for four matrix
double *A = malloc(sizeof(double) * SIZE * SIZE);
fill_matrix(A, SIZE);
double *B = malloc(sizeof(double) * SIZE * SIZE);
fill_matrix(B, SIZE);
double *C = malloc(sizeof(double) * SIZE * SIZE);
memset(C, 0, sizeof(double) * SIZE * SIZE);
double *D = malloc(sizeof(double) * SIZE * SIZE);
memset(D, 0, sizeof(double) * SIZE * SIZE);
// struct to timing
struct timeval begin, end;
// test function
gettimeofday(&begin, NULL);
square_dgemm(SIZE, A, B, C);
gettimeofday(&end, NULL);
// niave multipily
naive_multiply(A, B, D, SIZE);
// validate result, if wrong, print four matrix
for(int i=0; i<SIZE*SIZE; i++) {
if(C[i] != D[i]) {
printf("WRONG.\n");
for(int x=0; x<SIZE; x++) {
for(int y=0; y<SIZE; y++) {
printf("%f ", A[x*SIZE+y]);
}
printf("\n");
}
printf("-----------\n");
for(int x=0; x<SIZE; x++) {
for(int y=0; y<SIZE; y++) {
printf("%f ", B[x*SIZE+y]);
}
printf("\n");
}
printf("-----------\n");
for(int x=0; x<SIZE; x++) {
for(int y=0; y<SIZE; y++) {
printf("%f ", C[x*SIZE+y]);
}
printf("\n");
}
printf("-----------\n");
for(int x=0; x<SIZE; x++) {
for(int y=0; y<SIZE; y++) {
printf("%f ", D[x*SIZE+y]);
}
printf("\n");
}
return;
}
}
printf("CORRECT.^_^\n");
printf("Single Round Time use: %ld usec.\n", (end.tv_sec-begin.tv_sec)*1000000 + (end.tv_usec - begin.tv_usec));
/* Time a "sufficiently long" sequence of calls to reduce noise */
double Gflops_s, seconds = -1.0;
double timeout = 0.1; // "sufficiently long" := at least 1/10 second.
for (int n_iterations = 1; seconds < timeout; n_iterations *= 2)
{
/* Warm-up */
square_dgemm (SIZE, A, B, C);
/* Benchmark n_iterations runs of square_dgemm */
seconds = -wall_time();
for (int it = 0; it < n_iterations; ++it)
square_dgemm (SIZE, A, B, C);
seconds += wall_time();
/* compute Mflop/s rate */
Gflops_s = 2.e-9 * n_iterations * SIZE * SIZE * SIZE / seconds;
}
printf ("Size: %d\tGflop/s: %.3g\n", SIZE, Gflops_s);
}
int main(int argc, char* argv[])
{
// Print help if necessary
bool help = read_bool(argc, argv, "--help", false);
if ((argc < 2) || (help)) {
usage(argv);
return 0;
}
// Use parameters struct for passing parameters to kernels efficiently
parameters prm;
// Parse inputs
prm.matDims[0] = read_int(argc, argv, "--m", 2);
prm.matDims[1] = read_int(argc, argv, "--k", 2);
prm.matDims[2] = read_int(argc, argv, "--n", 2);
prm.rank = read_int(argc, argv, "--rank", 7);
prm.method = read_string(argc, argv, "--method", (char *)"als");
int maxIters = read_int(argc, argv, "--maxiters", 1000);
int maxSecs = read_int(argc, argv, "--maxsecs", 1000);
double tol = read_double(argc, argv, "--tol", 1e-8);
int printItn = read_int(argc, argv, "--printitn", 0);
double printTol = read_double(argc, argv, "--printtol", 1.0);
int seed = read_int(argc, argv, "--seed", 0);
int numSeeds = read_int(argc, argv, "--numseeds", 1);
bool verbose = read_bool(argc, argv, "--verbose", false);
prm.rnd_maxVal = read_double(argc,argv,"--maxval",1.0);
prm.rnd_pwrOfTwo = read_int(argc,argv,"--pwrof2",0);
bool roundFinal = read_bool(argc, argv, "--rndfin",false);
prm.alpha = read_double(argc,argv, "--alpha", 0.1);
int M = read_int(argc,argv, "--M", 0);
if (M)
{
prm.M[0] = M;
prm.M[1] = M;
prm.M[2] = M;
} else {
prm.M[0] = read_int(argc, argv, "--M0", -1);
prm.M[1] = read_int(argc, argv, "--M1", -1);
prm.M[2] = read_int(argc, argv, "--M2", -1);
}
char * infile = read_string(argc, argv, "--input", NULL);
char * outfile = read_string(argc, argv, "--output", NULL);
if (verbose) {
setbuf(stdout, NULL);
printf("\n\n---------------------------------------------------------\n");
printf("PARAMETERS\n");
printf("dimensions = %d %d %d\n",prm.matDims[0],prm.matDims[1],prm.matDims[2]);
printf("rank = %d\n",prm.rank);
printf("method = %s\n",prm.method);
if (infile)
printf("input = %s\n",infile);
else
{
if (numSeeds == 1)
printf("input = seed %d\n",seed);
else
printf("inputs = seeds %d-%d\n",seed,seed+numSeeds-1);
}
if (outfile)
printf("output = %s\n",outfile);
else
printf("output = none\n");
if (!strcmp(prm.method,"als"))
{
printf("tol = %1.2e\n",tol);
printf("alpha = %1.2e\n",prm.alpha);
printf("maval = %1.2e\n",prm.rnd_maxVal);
printf("M's = (%d,%d,%d)\n",prm.M[0],prm.M[1],prm.M[2]);
printf("maxiters = %d\n",maxIters);
printf("maxsecs = %d\n",maxSecs);
printf("printitn = %d\n",printItn);
printf("printtol = %1.2e\n",printTol);
}
printf("---------------------------------------------------------\n");
}
// Initialize other variables
int i, j, k, numIters, mkn, tidx[3];
double err, errOld, errChange = 0.0, start_als, start_search, elapsed, threshold;
// Compute tensor dimensions
prm.dims[0] = prm.matDims[0]*prm.matDims[1];
prm.dims[1] = prm.matDims[1]*prm.matDims[2];
prm.dims[2] = prm.matDims[0]*prm.matDims[2];
// Compute tensor's nnz, total number of entries, and Frobenius norm
mkn = prm.matDims[0]*prm.matDims[1]*prm.matDims[2];
prm.mkn2 = mkn*mkn;
prm.xNorm = sqrt(mkn);
// Compute number of columns in matricized tensors
for (i = 0; i < 3; i++)
prm.mtCols[i] = prm.mkn2 / prm.dims[i];
// Construct three matricizations of matmul tensor
prm.X = (double**) malloc( 3 * sizeof(double*) );
for (i = 0; i < 3; i++)
prm.X[i] = (double*) calloc( prm.mkn2, sizeof(double) );
for (int mm = 0; mm < prm.matDims[0]; mm++)
for (int kk = 0; kk < prm.matDims[1]; kk++)
for (int nn = 0; nn < prm.matDims[2]; nn++)
{
tidx[0] = mm + kk*prm.matDims[0];
tidx[1] = kk + nn*prm.matDims[1];
tidx[2] = mm + nn*prm.matDims[0];
prm.X[0][tidx[0]+prm.dims[0]*(tidx[1]+prm.dims[1]*tidx[2])] = 1;
prm.X[1][tidx[1]+prm.dims[1]*(tidx[0]+prm.dims[0]*tidx[2])] = 1;
prm.X[2][tidx[2]+prm.dims[2]*(tidx[0]+prm.dims[0]*tidx[1])] = 1;
}
// Allocate factor weights and matrices: working, initial, and model
prm.lambda = (double*) malloc( prm.rank * sizeof(double) );
prm.U = (double**) malloc( 3 * sizeof(double*) );
double** U0 = (double**) malloc( 3 * sizeof(double*) );
prm.model = (double**) malloc( 3 * sizeof(double*) );
for (i = 0; i < 3; i++)
{
prm.U[i] = (double*) calloc( prm.mkn2, sizeof(double) );
U0[i] = (double*) calloc( prm.dims[i]*prm.rank, sizeof(double) );
prm.model[i] = (double*) calloc( prm.dims[i]*prm.rank, sizeof(double) );
}
// Allocate coefficient matrix within ALS (Khatri-Rao product)
int maxMatDim = prm.matDims[0];
if (maxMatDim < prm.matDims[1]) maxMatDim = prm.matDims[1];
if (maxMatDim < prm.matDims[2]) maxMatDim = prm.matDims[2];
prm.A = (double*) malloc( maxMatDim*mkn*prm.rank * sizeof(double) );
// Allocate workspaces
prm.tau = (double*) malloc( mkn * sizeof(double) );
prm.lwork = maxMatDim*mkn*prm.rank;
prm.work = (double*) malloc( prm.lwork * sizeof(double) );
prm.iwork = (int*) malloc( prm.mkn2 * sizeof(int) );
// Allocate matrices for normal equations
int maxDim = prm.dims[0];
if (maxDim < prm.dims[1]) maxDim = prm.dims[1];
if (maxDim < prm.dims[2]) maxDim = prm.dims[2];
prm.NE_coeff = (double*) malloc( prm.rank*prm.rank * sizeof(double) );
prm.NE_rhs = (double*) malloc( maxDim*prm.rank * sizeof(double) );
prm.residual = (double*) malloc( prm.mkn2 * sizeof(double) );
//--------------------------------------------------
// Search Loop
//--------------------------------------------------
int mySeed = seed, numGoodSeeds = 0, statusCnt = 0, status = 1;
start_search = wall_time();
for (int seed_cnt = 0; seed_cnt < numSeeds; ++seed_cnt)
{
// Set starting point from random seed (match Matlab Tensor Toolbox)
RandomMT cRMT(mySeed);
for (i = 0; i < 3; i++)
for (j = 0; j < prm.dims[i]; j++)
for (k = 0; k < prm.rank; k++)
U0[i][j+k*prm.dims[i]] = cRMT.genMatlabMT();
for (i = 0; i < prm.rank; i++)
prm.lambda[i] = 1.0;
// Copy starting point
for (i = 0; i < 3; i++)
cblas_dcopy(prm.dims[i]*prm.rank,U0[i],1,prm.U[i],1);
// read from file if input is given
if( infile )
read_input( infile, prm );
if (verbose)
{
printf("\nSTARTING POINT...\n");
for (i = 0; i < 3; i++)
{
printf("Factor matrix %d:\n",i);
print_matrix(prm.U[i],prm.dims[i],prm.rank,prm.dims[i]);
}
printf("\n");
}
//--------------------------------------------------
// Main ALS Loop
//--------------------------------------------------
start_als = wall_time();
err = 1.0;
threshold = 1e-4;
for (numIters = 0; numIters < maxIters && (wall_time()-start_als) < maxSecs; numIters++)
{
errOld = err;
if (!strcmp(prm.method,"als"))
{
// Perform an iteration of ALS using NE with Smirnov's penalty term
err = als( prm );
}
else if (!strcmp(prm.method,"sparsify"))
{
// print stats before sparsifying
printf("Old residual: %1.2e\n",compute_residual(prm,2,true));
printf("Old nnz (larger than %1.1e): %d %d %d\n", threshold, nnz(prm.U[0],prm.dims[0]*prm.rank,threshold), nnz(prm.U[1],prm.dims[1]*prm.rank,threshold), nnz(prm.U[2],prm.dims[2]*prm.rank,threshold) );
// sparsify and return
printf("\nSparsifying...\n\n");
sparsify( prm );
numIters = maxIters;
// print stats after sparsifying
printf("New residual: %1.2e\n",compute_residual(prm,2,true));
printf("New nnz (larger than %1.1e): %d %d %d\n", threshold, nnz(prm.U[0],prm.dims[0]*prm.rank,threshold), nnz(prm.U[1],prm.dims[1]*prm.rank,threshold), nnz(prm.U[2],prm.dims[2]*prm.rank,threshold) );
}
else if (!strcmp(prm.method,"round"))
{
// print stats before rounding
printf("Old residual: %1.2e\n",compute_residual(prm,2,true));
printf("Old nnz (larger than %1.1e): %d %d %d\n", threshold, nnz(prm.U[0],prm.dims[0]*prm.rank,threshold), nnz(prm.U[1],prm.dims[1]*prm.rank,threshold), nnz(prm.U[2],prm.dims[2]*prm.rank,threshold) );
// round and return
for (i = 0; i < 3; i++)
{
capping(prm.U[i],prm.dims[i]*prm.rank,prm.rnd_maxVal);
rounding(prm.U[i],prm.dims[i]*prm.rank,prm.rnd_pwrOfTwo);
}
numIters = maxIters;
// print stats after rounding
printf("New residual: %1.2e\n",compute_residual(prm,2,true));
printf("New nnz (larger than %1.1e): %d %d %d\n", threshold, nnz(prm.U[0],prm.dims[0]*prm.rank,threshold), nnz(prm.U[1],prm.dims[1]*prm.rank,threshold), nnz(prm.U[2],prm.dims[2]*prm.rank,threshold) );
}
else
die("Invalid method\n");
// Compute change in relative residual norm
errChange = fabs(err - errOld);
// Print info at current iteration
if ((printItn > 0) && (((numIters + 1) % printItn) == 0))
{
// print info
printf ("Iter %d: residual = %1.5e change = %1.5e\n", numIters + 1, err, errChange);
}
// Check for convergence
if ( numIters > 0 && errChange < tol )
break;
}
// If rounding, round final solution and re-compute residual
if(roundFinal)
{
// normalize columns in A and B factors, put arbitrary weights into C
normalize_model( prm, 2 );
// cap large values and round to nearest power of 2
for (i = 0; i < 3; i++)
{
capping(prm.U[i],prm.dims[i]*prm.rank,prm.rnd_maxVal);
rounding(prm.U[i],prm.dims[i]*prm.rank,prm.rnd_pwrOfTwo);
}
err = compute_residual(prm,0,true);
}
// Print status if searching over many seeds
statusCnt++;
if (numSeeds > 1000 && statusCnt == numSeeds/10)
{
printf("...%d%% complete...\n",10*status);
status++;
statusCnt = 0;
}
// Print final info
elapsed = wall_time() - start_als;
if ((printItn > 0 || verbose) && !strcmp(prm.method,"als"))
{
if (infile)
printf("\nInput %s ",infile);
else
printf("\nInitial seed %d ",mySeed);
printf("achieved residual %1.3e in %d iterations and %1.3e seconds\n \t final residual change: %1.3e\n \t average time per iteration: %1.3e s\n", err, numIters, elapsed, errChange, elapsed/numIters);
}
if (verbose)
{
printf("\nSOLUTION...\n");
for (i = 0; i < 3; i++)
{
printf("Factor matrix %d:\n",i);
if (roundFinal || !strcmp(prm.method,"round"))
print_int_matrix(prm.U[i], prm.dims[i], prm.rank, prm.dims[i], prm.rnd_pwrOfTwo);
else
print_matrix(prm.U[i],prm.dims[i],prm.rank,prm.dims[i]);
}
if (err < printTol)
numGoodSeeds++;
}
else if (err < printTol)
{
numGoodSeeds++;
printf("\n\n***************************************\n");
if (infile)
printf("Input %s: ",infile);
else
printf("Initial seed %d: ",mySeed);
printf("after %d iterations, achieved residual %1.3e with final residual change of %1.3e\n", numIters, err, errChange);
if (roundFinal)
{
for (i = 0; i < 3; i++)
{
printf("Factor matrix %d:\n",i);
print_int_matrix(prm.U[i], prm.dims[i], prm.rank, prm.dims[i], prm.rnd_pwrOfTwo);
}
int count = 0;
for (i = 0; i < 3; i++)
count += nnz(prm.U[i],prm.dims[i]*prm.rank);
printf("\ttotal nnz in solution: %d\n",count);
printf("\tnaive adds/subs: %d\n",count - prm.dims[2] - 2*prm.rank);
}
printf("***************************************\n\n\n");
}
// write to output
if( outfile )
write_output( outfile, prm );
mySeed++;
}
// Final report of processor statistics
elapsed = wall_time()-start_search;
// Print stats
if (!strcmp(prm.method,"als"))
{
printf("\n\n------------------------------------------------------------\n");
printf("Time elapsed: \t%1.1e\tseconds\n",elapsed);
printf("Total number of seeds tried: \t%d\n",numSeeds);
printf("Total number of good seeds: \t%d",numGoodSeeds);
printf("\t(residual < %2.1e)\n",printTol);
printf("------------------------------------------------------------\n");
}
// free allocated memory
for (i = 0; i < 3; i++)
{
free( prm.X[i] );
free( prm.U[i] );
free( U0[i] );
free( prm.model[i] );
}
free( prm.X );
free( prm.U );
free( U0 );
free( prm.model );
free( prm.lambda );
free( prm.A );
free( prm.NE_coeff );
free( prm.NE_rhs );
free( prm.residual );
free( prm.tau );
free( prm.work );
free( prm.iwork );
return 0;
}
void
chord_main(char *conf_file, int parent_sock)
{
fd_set interesting, readable;
int nfound, nfds;
struct in_addr ia;
char id[4*ID_LEN];
FILE *fp;
int64_t stabilize_wait;
struct timeval timeout;
setprogname("chord");
srandom(getpid() ^ time(0));
memset(&srv, 0, sizeof(Server));
srv.to_fix_finger = NFINGERS-1;
fp = fopen(conf_file, "r");
if (fp == NULL)
eprintf("fopen(%s,\"r\") failed:", conf_file);
if (fscanf(fp, "%hd", (short*)&srv.node.port) != 1)
eprintf("Didn't find port in \"%s\"", conf_file);
if (fscanf(fp, " %s\n", id) != 1)
eprintf("Didn't find id in \"%s\"", conf_file);
srv.node.id = atoid(id);
/* Figure out one's own address somehow */
srv.node.addr = ntohl(get_addr());
ia.s_addr = htonl(srv.node.addr);
fprintf(stderr, "Chord started.\n");
fprintf(stderr, "id="); print_id(stderr, &srv.node.id);
fprintf(stderr, "\n");
fprintf(stderr, "ip=%s\n", inet_ntoa(ia));
fprintf(stderr, "port=%d\n", srv.node.port);
initialize(&srv);
srv_ref = &srv;
join(&srv, fp);
fclose(fp);
FD_ZERO(&interesting);
FD_SET(srv.in_sock, &interesting);
FD_SET(parent_sock, &interesting);
nfds = MAX(srv.in_sock, parent_sock) + 1;
NumKeys = read_keys(ACCLIST_FILE, KeyArray, MAX_KEY_NUM);
if (NumKeys == -1) {
printf("Error opening file: %s\n", ACCLIST_FILE);
}
if (NumKeys == 0) {
printf("No key found in %s\n", ACCLIST_FILE);
}
/* Loop on input */
for (;;) {
readable = interesting;
stabilize_wait = (int64_t)(srv.next_stabilize_us - wall_time());
stabilize_wait = MAX(stabilize_wait,0);
timeout.tv_sec = stabilize_wait / 1000000UL;
timeout.tv_usec = stabilize_wait % 1000000UL;
nfound = select(nfds, &readable, NULL, NULL, &timeout);
if (nfound < 0 && errno == EINTR) {
continue;
}
if (nfound == 0) {
stabilize_wait = (int64_t)(srv.next_stabilize_us - wall_time());
if( stabilize_wait <= 0 ) {
stabilize( &srv );
}
continue;
}
if (FD_ISSET(srv.in_sock, &readable)) {
handle_packet(srv.in_sock);
}
else if (FD_ISSET(parent_sock, &readable)) {
handle_packet(parent_sock);
}
else {
assert(0);
}
}
}
int main(int argc, char *argv[])
{
const int length[] = {1, 10, 100, 1000, 10000, 100000, 1000000};
const int arrlength = sizeof(length)/sizeof(length[0]);
const int r_max = 1000; // wiederholungen
char *msg;
int rank, r, i, name_length, size;
MPI_Status status;
double time, ttime;
char name[MPI_MAX_PROCESSOR_NAME];
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
int destination[size]; // maps: rank => destination of messages
MPI_Get_processor_name(name, &name_length);
if(size == 2) {
destination[0] = 1;
destination[1] = 0;
} else if(size == 4) {
destination[0] = 3;
destination[2] = 1;
destination[3] = 0;
destination[1] = 2;
} else {
MPI_Abort(MPI_COMM_WORLD,0);
}
if(rank == 0) printf("size is %d\n", size);
printf("%s: got rank %d\n",name, rank);
sleep(1);
for (i = 0; i < arrlength; i++){
msg = (char *)malloc(length[i]);
if(rank%2 == 0) {
time = wall_time();
time = wall_time();
}
for (r = 0; r < r_max; r++){
// printf("rank %d, %d, %d\n", rank, i, r);
if(rank%2 == 0) {
MPI_Send(msg, length[i], MPI_CHAR, destination[rank], 0, MPI_COMM_WORLD);
MPI_Recv(msg, length[i], MPI_CHAR, destination[rank], 0, MPI_COMM_WORLD, &status);
} else {
MPI_Recv(msg, length[i], MPI_CHAR, destination[rank], 0, MPI_COMM_WORLD, &status);
MPI_Send(msg, length[i], MPI_CHAR, destination[rank], 0, MPI_COMM_WORLD);
}
}
if(rank%2 == 0) {
time = wall_time() - time;
printf("%s: Zeit um %7d Bytes 2mal zu übertragen: %g s\n",name, length[i], time / r_max);
}
free(msg);
}
MPI_Finalize();
}
/* To determine the coordinates of the local node initially
* Ping a subset of nodes and determine coordinates */
void init_coordinates(I3ServerList *list)
{
int n = MIN(NUM_LANDMARKS_COORDINATE, list->num_newservers + list->num_ping_list);
I3ServerListNode *node = list->list, *temp_node;
uint64_t start_time = wall_time();
Coordinates_RTT coord_rtt[NUM_LANDMARKS_COORDINATE];
int num_landmarks = 0; int started_full_list = 0;
struct in_addr ia;
nw_skt_t tmp_ping_sock;
#ifdef ICMP_PING
if (init_icmp_socket(&tmp_ping_sock) == -1)
abort();
#else
if (init_udp_socket(&tmp_ping_sock) == -1)
abort();
#endif
// wait for responses and accumulate
// cut and pasted from below
while ((wall_time() - start_time < COORD_INIT_PING_WAIT_TIME) &&
(num_landmarks < n)) {
fd_set rset;
struct timeval to;
int ret;
FD_ZERO(&rset);
if (!node && !started_full_list) {
node = list-> full_list;
started_full_list = 1;
}
if (node) {
ia.s_addr = htonl(node->addr);
I3_PRINT_DEBUG1(I3_DEBUG_LEVEL_VERBOSE,
"Sending ICMP echo request to %s\n", inet_ntoa(ia));
#ifdef ICMP_PING
send_echo_request(tmp_ping_sock, node->addr, 0);
#else
i3_echo_request(tmp_ping_sock, node->addr, node->port, 0);
#endif
node = node->next_list;
}
FD_SET(tmp_ping_sock, &rset);
to.tv_sec = 0; to.tv_usec = 200000ULL;
if ((ret = select(tmp_ping_sock+1, &rset, NULL, NULL, &to)) < 0) {
int err = nw_error();
if (err == EINTR)
continue;
else {
perror("select");
abort();
}
}
// message received on icmp socket
if (FD_ISSET(tmp_ping_sock, &rset)) {
uint32_t addr; uint16_t port, seq; uint64_t rtt;
#ifdef ICMP_PING
if (recv_echo_reply(tmp_ping_sock, &addr, &seq, &rtt)) {
#else
if (recv_i3_echo_reply(tmp_ping_sock, &addr, &port, &seq, &rtt)) {
#endif
temp_node = lookup_i3server(list, addr);
assert(NULL != temp_node);
coord_rtt[num_landmarks].coord = temp_node->coord;
coord_rtt[num_landmarks].rtt = rtt;
num_landmarks++;
ia.s_addr = htonl(addr);
I3_PRINT_DEBUG4(I3_DEBUG_LEVEL_VERBOSE,
"Node: %s Coordinate: %.1f:%.1f RTT: %Ld\n",
inet_ntoa(ia), temp_node->coord.latitude,
temp_node->coord.longitude, rtt);
}
}
}
nw_close(tmp_ping_sock);
// compute own coordinate
compute_coordinates(num_landmarks, coord_rtt);
}
/* Update the coordinates of a node using ping information */
void update_coordinate(I3ServerList *list, I3ServerListNode *next_to_ping)
{
Coordinates_RTT coord_rtt[NUM_LANDMARKS_COORDINATE];
int count, num_landmarks = 0;
I3ServerListNode *node;
// n1 and n2: number of landmarks from ping_list and rest in
// proportion to the number of nodes in those lists
int i, n = MIN(NUM_LANDMARKS_COORDINATE,
list->num_newservers + list->num_ping_list);
int n1 = ((float)list->num_ping_list/
(list->num_newservers + list->num_ping_list)) * n;
int n2 = n-n1;
// add from ping list
count = 0;
for (i = 0, node = list->list;
i < list->num_ping_list, count < n1;
node = node->next_list, ++i) {
if (node->n > 0) {
coord_rtt[count].rtt = get_rtt_node(node);
coord_rtt[count].coord = node->coord;
count++;
}
}
num_landmarks = count;
// add from rest
count = 0;
for (i = 0, node = list->full_list;
i < list->num_newservers, count < n2;
node = node->next_list, ++i) {
if (node->n > 0) {
coord_rtt[num_landmarks + count].rtt = get_rtt_node(node);
coord_rtt[num_landmarks + count].coord = node->coord;
count++;
}
}
num_landmarks += count;
// recompute coordinates
compute_coordinates(num_landmarks, coord_rtt);
// repopulate ping list afresh
change_ping_list(list, &next_to_ping, 1);
}
void main(int argc, char** argv) {
args::Parser parser(argv[0], "Plays a replay into a set of images and a log of sounds");
String replay_path(utf8::decode(argv[0]));
parser.add_argument("replay", store(replay_path)).help("an Antares replay script").required();
Optional<String> output_dir;
parser.add_argument("-o", "--output", store(output_dir))
.help("place output in this directory");
int interval = 60;
int width = 640;
int height = 480;
bool text = false;
bool smoke = false;
parser.add_argument("-i", "--interval", store(interval))
.help("take one screenshot per this many ticks (default: 60)");
parser.add_argument("-w", "--width", store(width)).help("screen width (default: 640)");
parser.add_argument("-h", "--height", store(height)).help("screen height (default: 480)");
parser.add_argument("-t", "--text", store_const(text, true)).help("produce text output");
parser.add_argument("-s", "--smoke", store_const(smoke, true)).help("run as smoke text");
parser.add_argument("--help", help(parser, 0)).help("display this help screen");
String error;
if (!parser.parse_args(argc - 1, argv + 1, error)) {
print(io::err, format("{0}: {1}\n", parser.name(), error));
exit(1);
}
if (output_dir.has()) {
makedirs(*output_dir, 0755);
}
Preferences preferences;
preferences.play_music_in_game = true;
NullPrefsDriver prefs(preferences);
EventScheduler scheduler;
scheduler.schedule_event(unique_ptr<Event>(new MouseMoveEvent(wall_time(), Point(320, 240))));
// TODO(sfiera): add recurring snapshots to OffscreenVideoDriver.
for (int64_t i = 1; i < 72000; i += interval) {
scheduler.schedule_snapshot(i);
}
unique_ptr<SoundDriver> sound;
if (!smoke && output_dir.has()) {
String out(format("{0}/sound.log", *output_dir));
sound.reset(new LogSoundDriver(out));
} else {
sound.reset(new NullSoundDriver);
}
NullLedger ledger;
MappedFile replay_file(replay_path);
if (smoke) {
TextVideoDriver video({width, height}, Optional<String>());
video.loop(new ReplayMaster(replay_file.data(), output_dir), scheduler);
} else if (text) {
TextVideoDriver video({width, height}, output_dir);
video.loop(new ReplayMaster(replay_file.data(), output_dir), scheduler);
} else {
OffscreenVideoDriver video({width, height}, output_dir);
video.loop(new ReplayMaster(replay_file.data(), output_dir), scheduler);
}
}
/* The benchmarking program */
int main (int argc, char **argv)
{
printf ("Description:\t%s\n\n", dgemm_desc);
/* Test sizes should highlight performance dips at multiples of certain powers-of-two */
int test_sizes[] =
/* Multiples-of-32, +/- 1. Currently commented. */
{31,32,33,63,64,65,95,96,97,127,128,129,159,160,161,191,192,193,223,224,225,255,256,257,287,288,289,319,320,321,351,352,353,383,384,385,415,416,417,447,448,449,479,480,481,511,512,513,543,544,545,575,576,577,607,608,609,639,640,641,671,672,673,703,704,705,735,736,737,767,768,769,799,800,801,831,832,833,863,864,865,895,896,897,927,928,929,959,960,961,991,992,993,1023,1024,1025};
/* A representative subset of the first list. Currently uncommented. */
//{ 31, 32, 96, 97, 127, 128, 129, 191, 192, 229, 255, 256, 257,
// 319, 320, 321, 417, 479, 480, 511, 512, 639, 640, 767, 768, 769 };
int nsizes = sizeof(test_sizes)/sizeof(test_sizes[0]);
/* assume last size is also the largest size */
int nmax = test_sizes[nsizes-1];
/* allocate memory for all problems */
double* buf = NULL;
buf = (double*) malloc (3 * nmax * nmax * sizeof(double));
if (buf == NULL) die ("failed to allocate largest problem size");
double Mflops_s[nsizes],per[nsizes],aveper;
/* For each test size */
for (int isize = 0; isize < sizeof(test_sizes)/sizeof(test_sizes[0]); ++isize) {
for( int block_size = 3;block_size<200;block_size++) {
/* Create and fill 3 random matrices A,B,C*/
int n = test_sizes[isize];
double* A = buf + 0;
double* B = A + nmax*nmax;
double* C = B + nmax*nmax;
fill (A, n*n);
fill (B, n*n);
fill (C, n*n);
/* Measure performance (in Gflops/s). */
/* Time a "sufficiently long" sequence of calls to reduce noise */
double Gflops_s, seconds = -1.0;
double timeout = 0.1; // "sufficiently long" := at least 1/10 second.
for (int n_iterations = 1; seconds < timeout; n_iterations *= 2) {
/* Warm-up */
square_dgemm (block_size,n, A, B, C);
/* Benchmark n_iterations runs of square_dgemm */
seconds = -wall_time();
for (int it = 0; it < n_iterations; ++it)
square_dgemm (block_size,n, A, B, C);
seconds += wall_time();
/* compute Gflop/s rate */
Gflops_s = 2.e-9 * n_iterations * n * n * n / seconds;
}
/* Storing Mflop rate and calculating percentage of peak */
Mflops_s[isize] = Gflops_s*1000;
per[isize] = Gflops_s*100/MAX_SPEED;
printf ("Size: %d\t Block Size: %d\t Mflop/s: %8g\tPercentage:%6.2lf\n", n, block_size,Mflops_s[isize],per[isize]);
/* Ensure that error does not exceed the theoretical error bound. */
/* C := A * B, computed with square_dgemm */
memset (C, 0, n * n * sizeof(double));
square_dgemm (block_size,n, A, B, C);
/* Do not explicitly check that A and B were unmodified on square_dgemm exit
* - if they were, the following will most likely detect it:
* C := C - A * B, computed with reference_dgemm */
reference_dgemm(n, -1., A, B, C);
/* A := |A|, B := |B|, C := |C| */
absolute_value (A, n * n);
absolute_value (B, n * n);
absolute_value (C, n * n);
/* C := |C| - 3 * e_mach * n * |A| * |B|, computed with reference_dgemm */
reference_dgemm (n, -3.*DBL_EPSILON*n, A, B, C);
/* If any element in C is positive, then something went wrong in square_dgemm */
for (int i = 0; i < n * n; ++i)
if (C[i] > 0)
die("*** FAILURE *** Error in matrix multiply exceeds componentwise error bounds.\n" );
}
}
free (buf);
return 0;
}
/*---------------------------------------------------------------------------------------------
* (function: simulateNextWave)
*-------------------------------------------------------------------------------------------*/
int OdinInterface::simulateNextWave()
{
if(!num_vectors){
fprintf(stderr, "No vectors to simulate.\n");
} else {
double total_time = 0;
double simulation_time = 0;
num_cycles = num_vectors*2;
num_waves = 1;
tvector = 0;
double wave_start_time = wall_time();
//create a new wave
wave++;
int cycle_offset = SIM_WAVE_LENGTH * wave;
int wave_length = SIM_WAVE_LENGTH;
// Assign vectors to lines, either by reading or generating them.
// Every second cycle gets a new vector.
for (cycle = cycle_offset; cycle < cycle_offset + wave_length; cycle++)
{
if (is_even_cycle(cycle))
{
if (input_vector_file)
{
char buffer[BUFFER_MAX_SIZE];
if (!get_next_vector(in, buffer))
error_message(SIMULATION_ERROR, 0, -1, (char*)"Could not read next vector.");
tvector = parse_test_vector(buffer);
}
else
{
tvector = generate_random_test_vector(input_lines, cycle, hold_high_index, hold_low_index);
}
}
add_test_vector_to_lines(tvector, input_lines, cycle);
if (!is_even_cycle(cycle))
free_test_vector(tvector);
}
// Record the input vectors we are using.
write_wave_to_file(input_lines, in_out, cycle_offset, wave_length, 1);
// Write ModelSim script.
write_wave_to_modelsim_file(netlist, input_lines, modelsim_out, cycle_offset, wave_length);
double simulation_start_time = wall_time();
// Perform simulation
for (cycle = cycle_offset; cycle < cycle_offset + wave_length; cycle++)
{
if (cycle)
{
simulate_cycle(cycle, stgs);
}
else
{
// The first cycle produces the stages, and adds additional
// lines as specified by the -p option.
pin_names *p = parse_pin_name_list(global_args.sim_additional_pins);
stgs = simulate_first_cycle(netlist, cycle, p, output_lines);
free_pin_name_list(p);
// Make sure the output lines are still OK after adding custom lines.
if (!verify_lines(output_lines))
error_message(SIMULATION_ERROR, 0, -1,
(char*)"Problem detected with the output lines after the first cycle.");
}
}
simulation_time += wall_time() - simulation_start_time;
// Write the result of this wave to the output vector file.
write_wave_to_file(output_lines, out, cycle_offset, wave_length, output_edge);
total_time += wall_time() - wave_start_time;
// Print netlist-specific statistics.
if (!cycle_offset)
{
print_netlist_stats(stgs, num_vectors);
fflush(stdout);
}
// Print statistics.
print_simulation_stats(stgs, num_vectors, total_time, simulation_time);
myCycle = cycle;
}
return myCycle;
}
ulong get_current_time()
{
return (ulong)wall_time();
}
