void
timer_end(const char *label)
{
timer_stop();
fprintf(stderr, "%20s: ", label ? label : "");
timer_print();
timer_reset();
}
int test_timersleep()
{
timer_init();
TimerData timer;
timer_start(&timer);
timer_sleep(1E4);
timer_stop(&timer);
if (timer_print(&timer) < 0.01)
goto fail;
if (timer_print(&timer) > 0.015)
goto fail;
timer_finalize();
return 1;
fail:
timer_finalize();
return 0;
}
int test_timerprint_noinit()
{
TimerData timer;
timer_reset(&timer);
double time = timer_print(&timer);
if (time != 0)
goto fail;
return 1;
fail:
return 0;
}
static void
timer_callback(int fd, enum ev_type type, void *param)
{
static int i;
if (i >= TEVSZ)
abort();
tevbuf[i++] = rdtsc();
if (i == TEVSZ) {
mevent_delete(tevp);
timer_print();
}
}
int main(int argc, char **argv) {
int opt, exp = -1, var = -1;
bool err = false;
while ((opt = getopt(argc, argv, "n:v:")) != -1) {
if (opt == 'n')
exp = atoi(optarg);
else if (opt == 'v')
var = atoi(optarg);
else
err = true;
}
if (err || exp < 0 || var < 0 || var >= 2) {
fprintf(stderr, "Usage: %s -n log2(size) -v variant\n", argv[0]);
return 1;
}
int n = 1 << exp;
size_t size = n * n * sizeof(int);
int *src = NULL, *dst = NULL;
posix_memalign((void **)&src, getpagesize(), size);
posix_memalign((void **)&dst, getpagesize(), size);
printf("Generate matrix %d x %d (%ld KiB)\n", n, n, size >> 10);
fill(src, n);
bzero(dst, size);
flush_cache();
printf("Performing matrix transposition.\n");
_timer_t timer;
timer_reset(&timer);
timer_start(&timer);
if (var == 0)
transpose1(dst, src, n);
else
transpose2(dst, src, n);
timer_stop(&timer);
timer_print(&timer);
free(src);
free(dst);
return 0;
}
int test_timerprint()
{
TimerData timer;
timer_reset(&timer);
timer_init();
double time = timer_print(&timer);
if (time != 0)
goto fail;
uint64_t cycles = timer_printCycles(&timer);
if (cycles != 0)
goto fail;
timer_finalize();
return 1;
fail:
timer_finalize();
return 0;
}
static int lua_likwid_getClock(lua_State* L)
{
TimerData timer;
double runtime, start, stop;
if (timer_isInitialized == 0)
{
timer_init();
timer_isInitialized = 1;
}
start = lua_tonumber(L,1);
stop = lua_tonumber(L,2);
timer.start.int64 = (uint64_t)start;
timer.stop.int64 = (uint64_t)stop;
runtime = timer_print(&timer);
lua_pushnumber(L, runtime);
return 1;
}
void
daemon_start(struct timespec interval)
{
daemon_run = 1;
perfmon_startCounters();
timer_start(&timeData);
while (1)
{
if (daemon_run)
{
timer_stop(&timeData);
perfmon_readCounters();
perfmon_logCounterResults( timer_print(&timeData) );
timer_start(&timeData);
}
nanosleep( &interval, NULL);
}
}
int test_timerprint_stop()
{
TimerData timer;
timer_init();
timer_reset(&timer);
timer_start(&timer);
timer_stop(&timer);
double time = timer_print(&timer);
if (time > 1)
goto fail;
if (time == 0)
goto fail;
uint64_t cycles = timer_printCycles(&timer);
if (cycles == 0)
goto fail;
if (cycles > timer_getCpuClock())
goto fail;
timer_finalize();
return 1;
fail:
timer_finalize();
return 0;
}
void
malloc_test (void)
{
struct timespec start_time;
struct timespec end_time;
void **buffer = (void*) calloc(REPETITIONS, sizeof(void*));
timer_start(&start_time);
for(size_t i = 0; i < REPETITIONS; i++)
{
buffer[i] = malloc(BUFFER_SIZE);
}
timer_stop(&end_time);
timer_print("malloc", &start_time, &end_time);
for(size_t i = 0; i < REPETITIONS; i++)
{
free(buffer[i]);
}
free(buffer);
}
int main(int argc, char *argv[])
{
exit_if_false(argc == 2,"exactly one input argument required");
timer_start();
//-------------------------------------------------------------------//
// counters
int i, n;
// file paths
char *input_file_path, *geometry_file_path, *case_file_path, *data_file_path, *data_numbered_file_path, *display_file_path, *display_numbered_file_path;
exit_if_false(geometry_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating geometry file path");
exit_if_false(case_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating case file path");
exit_if_false(data_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating data file path");
exit_if_false(data_numbered_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating data numbered file path");
exit_if_false(display_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating display file path");
exit_if_false(display_numbered_file_path = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating display numbered file path");
// print string
char *print;
exit_if_false(print = (char *)malloc(MAX_STRING_LENGTH * sizeof(char)),"allocating the print string");
// mesh structures
int n_nodes, n_faces, n_elements, n_boundaries_old = 0, n_boundaries, n_terms;
struct NODE *node;
struct FACE *face;
struct ELEMENT *element;
struct BOUNDARY *boundary_old = NULL, *boundary;
struct TERM *term;
EXPRESSION *initial = NULL;
// solution vectors
int n_u_old = 0, n_u;
double *u_old = NULL, *u;
// files
FILE *input_file, *case_file, *data_file, *geometry_file, *display_file;
//-------------------------------------------------------------------//
// opening the input file
print_info("opening the input file %s",argv[1]);
input_file_path = argv[1];
input_file = fopen(input_file_path,"r");
exit_if_false(input_file != NULL,"opening %s",input_file_path);
// reading the case file path
print_info("reading the case file path");
exit_if_false(fetch_value(input_file,"case_file_path",'s',case_file_path) == FETCH_SUCCESS,"reading case_file_path from %s",input_file_path);
print_continue(case_file_path);
// reading the number of variables, variable names and orders
print_info("reading the variables");
int n_variables_old = 0, n_variables;
exit_if_false(fetch_value(input_file,"number_of_variables",'i',&n_variables) == FETCH_SUCCESS,"reading number_of_variables from %s",input_file_path);
int *variable_order_old = NULL, *variable_order = (int *)malloc(n_variables * sizeof(int));
exit_if_false(variable_order != NULL,"allocating orders");
exit_if_false(fetch_vector(input_file, "variable_order", 'i', n_variables, variable_order) == FETCH_SUCCESS,"reading variable_order from %s",input_file_path);
char **variable_name = allocate_character_matrix(NULL,n_variables,MAX_STRING_LENGTH);
exit_if_false(variable_name != NULL,"allocating variable names");
warn_if_false(fetch_vector(input_file,"variable_name",'s',n_variables,variable_name) == FETCH_SUCCESS,"reading variable_name from %s",input_file_path);
for(i = 0; i < n_variables; i ++) print_continue("%s order %i",variable_name[i],variable_order[i]);
// reading the number of inner and outer iterations to perform
print_info("reading the numbers of iterations");
int outer_iteration = 0, inner_iteration;
int n_outer_iterations, n_inner_iterations, data_n_outer_iterations, display_n_outer_iterations;
exit_if_false(fetch_value(input_file,"number_of_inner_iterations",'i',&n_inner_iterations) == FETCH_SUCCESS,"reading number_of_inner_iterations from %s",input_file_path);
exit_if_false(fetch_value(input_file,"number_of_outer_iterations",'i',&n_outer_iterations) == FETCH_SUCCESS,"reading number_of_outer_iterations from %s",input_file_path);
print_continue("%i outer of %i inner iterations to be done",n_outer_iterations,n_inner_iterations);
// read existing case and data
case_file = fopen(case_file_path,"r");
if(case_file != NULL)
{
print_info("reading existing case file %s",case_file_path);
read_case(case_file, &n_variables_old, &variable_order_old, &n_nodes, &node, &n_faces, &face, &n_elements, &element, &n_boundaries_old, &boundary_old);
fclose(case_file);
n = 0; for(i = 0; i < n_variables; i ++) n += n_elements*ORDER_TO_N_BASIS(variable_order[i]);
if(fetch_value(input_file,"initial_data_file_path",'s',data_file_path) == FETCH_SUCCESS)
{
print_info("reading existing data file %s",data_file_path);
exit_if_false(data_file = fopen(data_file_path,"r"),"opening %s",data_file_path);
read_data(data_file, &n_u_old, &u_old, &outer_iteration);
fclose(data_file);
exit_if_false(n_u_old == n,"case and initial data does not match");
}
}
// construct new case
else
{
print_info("reading the geometry file path");
exit_if_false(fetch_value(input_file,"geometry_file_path",'s',geometry_file_path) == FETCH_SUCCESS,"reading geometry_file_path from %s",input_file_path);
print_continue(geometry_file_path);
print_info("reading the geometry file %s",geometry_file_path);
exit_if_false(geometry_file = fopen(geometry_file_path,"r"),"opening %s",geometry_file_path);
read_geometry(geometry_file, &n_nodes, &node, &n_faces, &face, &n_elements, &element);
fclose(geometry_file);
print_continue("%i nodes, %i faces and %i elements",n_nodes,n_faces,n_elements);
print_info("generating additional connectivity and geometry");
process_geometry(n_nodes, node, n_faces, face, n_elements, element);
}
// read the data file path and output frequency
print_info("reading the data file path and output frequency");
exit_if_false(fetch_value(input_file,"data_file_path",'s',data_file_path) == FETCH_SUCCESS,"reading data_file_path from %s",input_file_path);
if(fetch_value(input_file,"data_number_of_outer_iterations",'i',&data_n_outer_iterations) != FETCH_SUCCESS)
data_n_outer_iterations = n_outer_iterations + outer_iteration;
print_continue("data to be written to %s every %i outer iterations",data_file_path,data_n_outer_iterations);
// read boundaries
print_info("reading boundaries");
boundaries_input(input_file, n_faces, face, &n_boundaries, &boundary);
print_continue("%i boundaries",n_boundaries);
// read terms
print_info("reading PDE terms");
terms_input(input_file, &n_terms, &term);
print_continue("%i terms",n_terms);
// update unknown indices and values
print_info("updating the numbering of the degrees of freedom");
update_element_unknowns(n_variables_old, n_variables, variable_order_old, variable_order, n_elements, element, n_u_old, &n_u, u_old, &u);
print_continue("%i degrees of freedom",n_u);
// update face boundaries
print_info("updating the face boundary associations");
update_face_boundaries(n_variables, n_faces, face, n_boundaries, boundary);
// update integration
print_info("updating integration");
i = update_face_integration(n_variables_old, n_variables, variable_order_old, variable_order, n_faces, face);
if(i) print_continue("updated %i face quadratures",i);
i = update_element_integration(n_variables_old, n_variables, variable_order_old, variable_order, n_elements, element);
if(i) print_continue("updated %i element quadratures",i);
// update numerics
print_info("updating numerics");
i = update_face_numerics(n_variables_old, n_variables, variable_order_old, variable_order, n_faces, face, n_boundaries_old, boundary_old);
if(i) print_continue("updated %i face interpolations",i);
i = update_element_numerics(n_variables_old, n_variables, variable_order_old, variable_order, n_elements, element);
if(i) print_continue("updated %i element interpolations",i);
// write case file
print_info("writing case file %s",case_file_path);
exit_if_false(case_file = fopen(case_file_path,"w"),"opening %s",case_file_path);
write_case(case_file, n_variables, variable_order, n_nodes, node, n_faces, face, n_elements, element, n_boundaries, boundary);
fclose(case_file);
// read the display file path and output frequency
print_info("reading the display file path and output frequency");
if(
fetch_value(input_file,"display_file_path",'s',display_file_path) == FETCH_SUCCESS &&
fetch_value(input_file,"display_number_of_outer_iterations",'i',&display_n_outer_iterations) == FETCH_SUCCESS
)
print_continue("display to be written to %s every %i outer iterations",display_file_path,display_n_outer_iterations);
else
{
display_n_outer_iterations = 0;
warn_if_false(0,"display files will not be written");
}
// initialise
if(initial_input(input_file, n_variables, &initial))
{
print_info("initialising the degrees of freedom");
initialise_values(n_variables, variable_order, n_elements, element, initial, u);
}
//-------------------------------------------------------------------//
// allocate and initialise the system
print_info("allocating and initialising the system");
SPARSE system = NULL;
initialise_system(n_variables, variable_order, n_elements, element, n_u, &system);
double *residual, *max_residual, *du, *max_du;
exit_if_false(residual = (double *)malloc(n_u * sizeof(double)),"allocating the residuals");
exit_if_false(max_residual = (double *)malloc(n_variables * sizeof(double)),"allocating the maximum residuals");
exit_if_false(du = (double *)malloc(n_u * sizeof(double)),"allocating du");
exit_if_false(max_du = (double *)malloc(n_variables * sizeof(double)),"allocating the maximum changes");
exit_if_false(u_old = (double *)realloc(u_old, n_u * sizeof(double)),"re-allocating u_old");
timer_reset();
// iterate
print_info("iterating");
n_outer_iterations += outer_iteration;
for(; outer_iteration < n_outer_iterations; outer_iteration ++)
{
print_output("iteration %i", outer_iteration);
for(i = 0; i < n_u; i ++) u_old[i] = u[i];
for(inner_iteration = 0; inner_iteration < n_inner_iterations; inner_iteration ++)
{
calculate_system(n_variables, variable_order, n_faces, face, n_elements, element, n_terms, term, n_u, u_old, u, system, residual);
exit_if_false(sparse_solve(system, du, residual) == SPARSE_SUCCESS,"solving system");
for(i = 0; i < n_u; i ++) u[i] -= du[i];
calculate_maximum_changes_and_residuals(n_variables, variable_order, n_elements, element, du, max_du, residual, max_residual);
for(i = 0; i < n_variables; i ++) sprintf(&print[26*i],"%.6e|%.6e ",max_du[i],max_residual[i]);
print_continue("%s",print);
}
slope_limit(n_variables, variable_order, n_nodes, node, n_elements, element, n_boundaries, boundary, u);
if(data_n_outer_iterations && outer_iteration % data_n_outer_iterations == 0)
{
generate_numbered_file_path(data_numbered_file_path, data_file_path, outer_iteration);
print_info("writing data to %s",data_numbered_file_path);
exit_if_false(data_file = fopen(data_numbered_file_path,"w"),"opening %s",data_numbered_file_path);
write_data(data_file, n_u, u, outer_iteration);
fclose(data_file);
}
if(display_n_outer_iterations && outer_iteration % display_n_outer_iterations == 0)
{
generate_numbered_file_path(display_numbered_file_path, display_file_path, outer_iteration);
print_info("writing display to %s",data_numbered_file_path);
exit_if_false(display_file = fopen(display_numbered_file_path,"w"),"opening %s",display_numbered_file_path);
write_display(display_file, n_variables, variable_name, variable_order, n_elements, element, n_u, u);
fclose(display_file);
}
timer_print();
}
//-------------------------------------------------------------------//
print_info("freeing all memory");
fclose(input_file);
free(geometry_file_path);
free(case_file_path);
free(data_file_path);
free(data_numbered_file_path);
free(display_file_path);
free(display_numbered_file_path);
free(print);
destroy_nodes(n_nodes,node);
destroy_faces(n_faces,face,n_variables);
destroy_elements(n_elements,element,n_variables);
destroy_boundaries(n_boundaries_old,boundary_old);
destroy_boundaries(n_boundaries,boundary);
destroy_terms(n_terms,term);
destroy_initial(n_variables,initial);
free(variable_order_old);
free(variable_order);
destroy_matrix((void *)variable_name);
free(u_old);
free(u);
sparse_destroy(system);
free(residual);
free(max_residual);
free(du);
free(max_du);
return 0;
}
int main (int argc, char** argv)
{
int socket_fd = -1;
int optInfo = 0;
int optClock = 0;
int optStethoscope = 0;
int optSockets = 0;
double runtime;
int hasDRAM = 0;
int c;
bstring argString;
bstring eventString = bfromcstr("CLOCK");
int numSockets=1;
int numThreads=0;
int threadsSockets[MAX_NUM_NODES*2];
int threads[MAX_NUM_THREADS];
threadsSockets[0] = 0;
if (argc == 1)
{
HELP_MSG;
exit (EXIT_SUCCESS);
}
while ((c = getopt (argc, argv, "+c:hiM:ps:v")) != -1)
{
switch (c)
{
case 'c':
CHECK_OPTION_STRING;
numSockets = bstr_to_cpuset_physical((uint32_t*) threadsSockets, argString);
bdestroy(argString);
optSockets = 1;
break;
case 'h':
HELP_MSG;
exit (EXIT_SUCCESS);
case 'i':
optInfo = 1;
break;
case 'M': /* Set MSR Access mode */
CHECK_OPTION_STRING;
accessClient_setaccessmode(str2int((char*) argString->data));
bdestroy(argString);
break;
case 'p':
optClock = 1;
break;
case 's':
CHECK_OPTION_STRING;
optStethoscope = str2int((char*) argString->data);
bdestroy(argString);
break;
case 'v':
VERSION_MSG;
exit (EXIT_SUCCESS);
case '?':
if (optopt == 's' || optopt == 'M' || optopt == 'c')
{
HELP_MSG;
}
else if (isprint (optopt))
{
fprintf (stderr, "Unknown option `-%c'.\n", optopt);
}
else
{
fprintf (stderr,
"Unknown option character `\\x%x'.\n",
optopt);
}
exit( EXIT_FAILURE);
default:
HELP_MSG;
exit (EXIT_SUCCESS);
}
}
if (!lock_check())
{
fprintf(stderr,"Access to performance counters is locked.\n");
exit(EXIT_FAILURE);
}
if (optClock && optind == argc)
{
fprintf(stderr,"Commandline option -p requires an executable.\n");
exit(EXIT_FAILURE);
}
if (optSockets && !optStethoscope && optind == argc)
{
fprintf(stderr,"Commandline option -c requires an executable if not used in combination with -s.\n");
exit(EXIT_FAILURE);
}
if (cpuid_init() == EXIT_FAILURE)
{
fprintf(stderr, "CPU not supported\n");
exit(EXIT_FAILURE);
}
if (numSockets > cpuid_topology.numSockets)
{
fprintf(stderr, "System has only %d sockets but %d are given on commandline\n",
cpuid_topology.numSockets, numSockets);
exit(EXIT_FAILURE);
}
numa_init(); /* consider NUMA node as power unit for the moment */
accessClient_init(&socket_fd);
msr_init(socket_fd);
timer_init();
/* check for supported processors */
if ((cpuid_info.model == SANDYBRIDGE_EP) ||
(cpuid_info.model == SANDYBRIDGE) ||
(cpuid_info.model == IVYBRIDGE) ||
(cpuid_info.model == IVYBRIDGE_EP) ||
(cpuid_info.model == HASWELL) ||
(cpuid_info.model == NEHALEM_BLOOMFIELD) ||
(cpuid_info.model == NEHALEM_LYNNFIELD) ||
(cpuid_info.model == NEHALEM_WESTMERE))
{
power_init(numa_info.nodes[0].processors[0]);
}
else
{
fprintf (stderr, "Query Turbo Mode only supported on Intel Nehalem/Westmere/SandyBridge/IvyBridge/Haswell processors!\n");
exit(EXIT_FAILURE);
}
double clock = (double) timer_getCpuClock();
printf(HLINE);
printf("CPU name:\t%s \n",cpuid_info.name);
printf("CPU clock:\t%3.2f GHz \n", (float) clock * 1.E-09);
printf(HLINE);
if (optInfo)
{
if (power_info.turbo.numSteps != 0)
{
printf("Base clock:\t%.2f MHz \n", power_info.baseFrequency );
printf("Minimal clock:\t%.2f MHz \n", power_info.minFrequency );
printf("Turbo Boost Steps:\n");
for (int i=0; i < power_info.turbo.numSteps; i++ )
{
printf("C%d %.2f MHz \n",i+1, power_info.turbo.steps[i] );
}
}
printf(HLINE);
}
if (cpuid_info.model == SANDYBRIDGE_EP)
{
hasDRAM = 1;
}
else if ((cpuid_info.model != SANDYBRIDGE) &&
(cpuid_info.model != SANDYBRIDGE_EP) &&
(cpuid_info.model != IVYBRIDGE) &&
(cpuid_info.model != IVYBRIDGE_EP) &&
(cpuid_info.model != HASWELL))
{
fprintf (stderr, "RAPL not supported on this processor!\n");
exit(EXIT_FAILURE);
}
if (optInfo)
{
printf("Thermal Spec Power: %g Watts \n", power_info.tdp );
printf("Minimum Power: %g Watts \n", power_info.minPower);
printf("Maximum Power: %g Watts \n", power_info.maxPower);
printf("Maximum Time Window: %g micro sec \n", power_info.maxTimeWindow);
printf(HLINE);
exit(EXIT_SUCCESS);
}
if (optClock)
{
affinity_init();
argString = bformat("S%u:0-%u", threadsSockets[0], cpuid_topology.numCoresPerSocket-1);
for (int i=1; i<numSockets; i++)
{
bstring tExpr = bformat("@S%u:0-%u", threadsSockets[i], cpuid_topology.numCoresPerSocket-1);
bconcat(argString, tExpr);
}
numThreads = bstr_to_cpuset(threads, argString);
bdestroy(argString);
perfmon_init(numThreads, threads, stdout);
perfmon_setupEventSet(eventString, NULL);
}
{
PowerData pDataPkg[MAX_NUM_NODES*2];
PowerData pDataDram[MAX_NUM_NODES*2];
printf("Measure on sockets: %d", threadsSockets[0]);
for (int i=1; i<numSockets; i++)
{
printf(", %d", threadsSockets[i]);
}
printf("\n");
if (optStethoscope)
{
if (optClock)
{
perfmon_startCounters();
}
else
{
for (int i=0; i<numSockets; i++)
{
int cpuId = numa_info.nodes[threadsSockets[i]].processors[0];
if (hasDRAM) power_start(pDataDram+i, cpuId, DRAM);
power_start(pDataPkg+i, cpuId, PKG);
}
}
sleep(optStethoscope);
if (optClock)
{
perfmon_stopCounters();
perfmon_printCounterResults();
perfmon_finalize();
}
else
{
for (int i=0; i<numSockets; i++)
{
int cpuId = numa_info.nodes[threadsSockets[i]].processors[0];
power_stop(pDataPkg+i, cpuId, PKG);
if (hasDRAM) power_stop(pDataDram+i, cpuId, DRAM);
}
}
runtime = (double) optStethoscope;
}
else
{
TimerData time;
argv += optind;
bstring exeString = bfromcstr(argv[0]);
for (int i=1; i<(argc-optind); i++)
{
bconchar(exeString, ' ');
bcatcstr(exeString, argv[i]);
}
printf("%s\n",bdata(exeString));
if (optClock)
{
perfmon_startCounters();
}
else
{
for (int i=0; i<numSockets; i++)
{
int cpuId = numa_info.nodes[threadsSockets[i]].processors[0];
if (hasDRAM) power_start(pDataDram+i, cpuId, DRAM);
power_start(pDataPkg+i, cpuId, PKG);
}
timer_start(&time);
}
if (system(bdata(exeString)) == EOF)
{
fprintf(stderr, "Failed to execute %s!\n", bdata(exeString));
exit(EXIT_FAILURE);
}
if (optClock)
{
perfmon_stopCounters();
perfmon_printCounterResults();
perfmon_finalize();
}
else
{
timer_stop(&time);
for (int i=0; i<numSockets; i++)
{
int cpuId = numa_info.nodes[threadsSockets[i]].processors[0];
power_stop(pDataPkg+i, cpuId, PKG);
if (hasDRAM) power_stop(pDataDram+i, cpuId, DRAM);
}
runtime = timer_print(&time);
}
}
if (!optClock)
{
printf("Runtime: %g second \n",runtime);
printf(HLINE);
for (int i=0; i<numSockets; i++)
{
printf("Socket %d\n",threadsSockets[i]);
printf("Domain: PKG \n");
printf("Energy consumed: %g Joules \n", power_printEnergy(pDataPkg+i));
printf("Power consumed: %g Watts \n", power_printEnergy(pDataPkg+i) / runtime );
if (hasDRAM)
{
printf("Domain: DRAM \n");
printf("Energy consumed: %g Joules \n", power_printEnergy(pDataDram+i));
printf("Power consumed: %g Watts \n", power_printEnergy(pDataDram+i) / runtime );
}
printf("\n");
}
}
}
#if 0
if ( cpuid_hasFeature(TM2) )
{
thermal_init(0);
printf("Current core temperatures:\n");
for (uint32_t i = 0; i < cpuid_topology.numCoresPerSocket; i++ )
{
printf("Core %d: %u C\n",
numa_info.nodes[socketId].processors[i],
thermal_read(numa_info.nodes[socketId].processors[i]));
}
}
#endif
msr_finalize();
return EXIT_SUCCESS;
}
/* Main program. */
int main(void)
{
int i, j, n_samples, max_n, step_n;
int array_size;
int radix;
test_item_t a[N_ITEMS], test_array[N_ITEMS];
test_item_t *timing_array, *copy_array;
timing_t *t;
/* Assign random values to the key of each array element. */
rand_array(a, N_ITEMS, 100);
/* Now test quicksort(). */
memcpy(test_array, a, sizeof(test_array));
printf("array before quicksort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
quicksort(test_array, N_ITEMS, sizeof(test_item_t), item_cmp);
printf("array after quicksort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
/* Now test mergesort0(). */
memcpy(test_array, a, sizeof(test_array));
printf("array before mergesort0\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
mergesort0(test_array, N_ITEMS, sizeof(test_item_t), item_cmp);
printf("array after mergesort0\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
/* Now test mergesort(). */
memcpy(test_array, a, sizeof(test_array));
printf("array before mergesort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
mergesort(test_array, N_ITEMS, sizeof(test_item_t), item_cmp);
printf("array after mergesort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
/* Now test radix sort. */
memcpy(test_array, a, sizeof(test_array));
printf("array before radixsort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
radixsort(test_array, N_ITEMS, sizeof(test_item_t), get_value, 10);
printf("array after radixsort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
/* Now test heapsort. */
memcpy(test_array, a, sizeof(test_array));
printf("array before heapsort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
heapsort(test_array, N_ITEMS, sizeof(test_item_t), item_cmp);
printf("array after heapsort\n = ");
print_array(test_array, N_ITEMS);
printf("\n\n");
/* Time the quicksort and mergesort sorting functions. */
printf("Enter the number of samples to use: ");
scanf("%d", &n_samples);
printf("Enter the maximum array length to sort: ");
scanf("%d", &max_n);
printf("Enter the step size for array lengths: ");
scanf("%d", &step_n);
t = timing_alloc(5); /* Five different sorting algorithms. */
printf("\nResults (n, qsort, quicksort, mergesort, mergesort0, heapsort) (msec)\n"
);
for(i = step_n; i <= max_n; i += step_n) {
array_size = i * sizeof(test_item_t);
timing_array = malloc(array_size);
copy_array = malloc(array_size);
rand_array(copy_array, i, MAX_VALUE);
timing_reset(t);
for(j = 0; j < n_samples; j++) {
memcpy(timing_array, copy_array, array_size);
timing_start();
qsort(timing_array, i, sizeof(test_item_t), item_cmp);
timing_stop(t,0);
memcpy(timing_array, copy_array, array_size);
timing_start();
quicksort(timing_array, i, sizeof(test_item_t), item_cmp);
timing_stop(t,1);
memcpy(timing_array, copy_array, array_size);
timing_start();
mergesort(timing_array, i, sizeof(test_item_t), item_cmp);
timing_stop(t,2);
memcpy(timing_array, copy_array, array_size);
timing_start();
mergesort0(timing_array, i, sizeof(test_item_t), item_cmp);
timing_stop(t,3);
memcpy(timing_array, copy_array, array_size);
timing_start();
heapsort(timing_array, i, sizeof(test_item_t), item_cmp);
timing_stop(t,4);
}
printf("%d", i);
timing_print(t,"\t%.2f",n_samples);
putchar('\n');
free(timing_array);
free(copy_array);
}
timing_free(t);
/* Time radix sort on the largest array, using different radix sizes. */
printf("\nRadix Sort Results. Using n = %d\n", max_n);
printf("(radix, time)\n");
array_size = max_n * sizeof(test_item_t);
timing_array = malloc(array_size);
copy_array = malloc(array_size);
rand_array(copy_array, max_n, MAX_VALUE);
for(radix = 2; radix <= max_n; radix <<= 1) {
timer_reset();
for(j = 0; j < n_samples; j++) {
memcpy(timing_array, copy_array, array_size);
timer_start();
radixsort(timing_array, max_n, sizeof(test_item_t), get_value,
radix);
timer_stop();
}
printf("%d", radix);
timer_print("\t%.2f", n_samples);
putchar('\n');
}
free(timing_array);
free(copy_array);
return 0;
}