/* =============================================================================
* main
* =============================================================================
*/
MAIN(argc, argv)
{
GOTO_REAL();
SETUP_NUMBER_TASKS(6);
/*
* Initialization
*/
parseArgs(argc, (char** const)argv);
SIM_GET_NUM_CPU(global_numThread);
TM_STARTUP(global_numThread, 0);
P_MEMORY_STARTUP(global_numThread);
SETUP_NUMBER_THREADS(global_numThread);
thread_startup(global_numThread);
int repeat = global_repeats;
double time_total = 0.0;
double energy_total = 0.0;
for (; repeat > 0; --repeat) {
global_meshPtr = mesh_alloc();
assert(global_meshPtr);
long initNumElement = mesh_read(global_meshPtr, global_inputPrefix);
global_workHeapPtr = heap_alloc(1, &element_heapCompare);
assert(global_workHeapPtr);
long initNumBadElement = initializeWork(global_workHeapPtr, global_meshPtr);
TIMER_T start;
TIMER_READ(start);
GOTO_SIM();
thread_start(process, NULL);
GOTO_REAL();
TIMER_T stop;
TIMER_READ(stop);
double time_tmp = TIMER_DIFF_SECONDS(start, stop);
PRINT_STATS();
time_total += time_tmp;
}
printf("Time = %0.3lf\n",
time_total);
fflush(stdout);
TM_SHUTDOWN();
P_MEMORY_SHUTDOWN();
GOTO_SIM();
thread_shutdown();
MAIN_RETURN(0);
}
MAIN(argc, argv)
{
GOTO_REAL();
SETUP_NUMBER_TASKS(10);
/*
* Tuple for Scalable Data Generation
* stores startVertex, endVertex, long weight and other info
*/
graphSDG* SDGdata;
/*
* The graph data structure for this benchmark - see defs.h
*/
graph* G;
#ifdef ENABLE_KERNEL2
/*
* Kernel 2
*/
edge* maxIntWtList;
edge* soughtStrWtList;
long maxIntWtListSize;
long soughtStrWtListSize;
#endif /* ENABLE_KERNEL2 */
#ifdef ENABLE_KERNEL3
# ifndef ENABLE_KERNEL2
# error KERNEL3 requires KERNEL2
# endif
/*
* Kernel 3
*/
V* intWtVList = NULL;
V* strWtVList = NULL;
Vl** intWtVLList = NULL;
Vl** strWtVLList = NULL;
Vd* intWtVDList = NULL;
Vd* strWtVDList = NULL;
#endif /* ENABLE_KERNEL3 */
double totalTime = 0.0;
/* -------------------------------------------------------------------------
* Preamble
* -------------------------------------------------------------------------
*/
/*
* User Interface: Configurable parameters, and global program control
*/
getUserParameters(argc, (char** const) argv);
SIM_GET_NUM_CPU(THREADS);
TM_STARTUP(THREADS, 0);
P_MEMORY_STARTUP(THREADS);
SETUP_NUMBER_THREADS(THREADS);
thread_startup(THREADS);
double time_total = 0.0;
int repeat = REPEATS;
for (; repeat > 0; --repeat) {
SDGdata = (graphSDG*)malloc(sizeof(graphSDG));
assert(SDGdata);
genScalData_seq(SDGdata);
G = (graph*)malloc(sizeof(graph));
assert(G);
computeGraph_arg_t computeGraphArgs;
computeGraphArgs.GPtr = G;
computeGraphArgs.SDGdataPtr = SDGdata;
TIMER_T start;
TIMER_READ(start);
GOTO_SIM();
thread_start(computeGraph, (void*)&computeGraphArgs);
GOTO_REAL();
TIMER_T stop;
TIMER_READ(stop);
double time_tmp = TIMER_DIFF_SECONDS(start, stop);
PRINT_STATS();
time_total += time_tmp;
}
totalTime += time_total;
#ifdef ENABLE_KERNEL2
/* -------------------------------------------------------------------------
* Kernel 2 - Find Max weight and sought string
* -------------------------------------------------------------------------
*/
printf("\nKernel 2 - getStartLists() beginning execution...\n");
maxIntWtListSize = 0;
soughtStrWtListSize = 0;
maxIntWtList = (edge*)malloc(sizeof(edge));
assert(maxIntWtList);
soughtStrWtList = (edge*)malloc(sizeof(edge));
assert(soughtStrWtList);
getStartLists_arg_t getStartListsArg;
getStartListsArg.GPtr = G;
getStartListsArg.maxIntWtListPtr = &maxIntWtList;
getStartListsArg.maxIntWtListSize = &maxIntWtListSize;
getStartListsArg.soughtStrWtListPtr = &soughtStrWtList;
getStartListsArg.soughtStrWtListSize = &soughtStrWtListSize;
TIMER_READ(start);
GOTO_SIM();
#ifdef OTM
#pragma omp parallel
{
getStartLists((void*)&getStartListsArg);
}
#else
thread_start(getStartLists, (void*)&getStartListsArg);
#endif
GOTO_REAL();
TIMER_T stop;
TIMER_READ(stop);
time = TIMER_DIFF_SECONDS(start, stop);
totalTime += time;
printf("\n\tgetStartLists() completed execution.\n");
printf("\nTime taken for kernel 2 is %9.6f sec.\n\n", time);
#endif /* ENABLE_KERNEL2 */
#ifdef ENABLE_KERNEL3
/* -------------------------------------------------------------------------
* Kernel 3 - Graph Extraction
* -------------------------------------------------------------------------
*/
printf("\nKernel 3 - findSubGraphs() beginning execution...\n");
if (K3_DS == 0) {
intWtVList = (V*)malloc(G->numVertices * maxIntWtListSize * sizeof(V));
assert(intWtVList);
strWtVList = (V*)malloc(G->numVertices * soughtStrWtListSize * sizeof(V));
assert(strWtVList);
findSubGraphs0_arg_t findSubGraphs0Arg;
findSubGraphs0Arg.GPtr = G;
findSubGraphs0Arg.intWtVList = intWtVList;
findSubGraphs0Arg.strWtVList = strWtVList;
findSubGraphs0Arg.maxIntWtList = maxIntWtList;
findSubGraphs0Arg.maxIntWtListSize = maxIntWtListSize;
findSubGraphs0Arg.soughtStrWtList = soughtStrWtList;
findSubGraphs0Arg.soughtStrWtListSize = soughtStrWtListSize;
TIMER_READ(start);
GOTO_SIM();
#ifdef OTM
#pragma omp parallel
{
findSubGraphs0((void*)&findSubGraphs0Arg);
}
#else
thread_start(findSubGraphs0, (void*)&findSubGraphs0Arg);
#endif
GOTO_REAL();
TIMER_READ(stop);
} else if (K3_DS == 1) {
intWtVLList = (Vl**)malloc(maxIntWtListSize * sizeof(Vl*));
assert(intWtVLList);
strWtVLList = (Vl**)malloc(soughtStrWtListSize * sizeof(Vl*));
assert(strWtVLList);
findSubGraphs1_arg_t findSubGraphs1Arg;
findSubGraphs1Arg.GPtr = G;
findSubGraphs1Arg.intWtVLList = intWtVLList;
findSubGraphs1Arg.strWtVLList = strWtVLList;
findSubGraphs1Arg.maxIntWtList = maxIntWtList;
findSubGraphs1Arg.maxIntWtListSize = maxIntWtListSize;
findSubGraphs1Arg.soughtStrWtList = soughtStrWtList;
findSubGraphs1Arg.soughtStrWtListSize = soughtStrWtListSize;
TIMER_READ(start);
GOTO_SIM();
#ifdef OTM
#pragma omp parallel
{
findSubGraphs1((void*)&findSubGraphs1Arg);
}
#else
thread_start(findSubGraphs1, (void*)&findSubGraphs1Arg);
#endif
GOTO_REAL();
TIMER_READ(stop);
/* Verification
on_one_thread {
for (i=0; i<maxIntWtListSize; i++) {
printf("%ld -- ", i);
currV = intWtVLList[i];
while (currV != NULL) {
printf("[%ld %ld] ", currV->num, currV->depth);
currV = currV->next;
}
printf("\n");
}
for (i=0; i<soughtStrWtListSize; i++) {
printf("%ld -- ", i);
currV = strWtVLList[i];
while (currV != NULL) {
printf("[%ld %ld] ", currV->num, currV->depth);
currV = currV->next;
}
printf("\n");
}
}
*/
} else if (K3_DS == 2) {
intWtVDList = (Vd *) malloc(maxIntWtListSize * sizeof(Vd));
assert(intWtVDList);
strWtVDList = (Vd *) malloc(soughtStrWtListSize * sizeof(Vd));
assert(strWtVDList);
findSubGraphs2_arg_t findSubGraphs2Arg;
findSubGraphs2Arg.GPtr = G;
findSubGraphs2Arg.intWtVDList = intWtVDList;
findSubGraphs2Arg.strWtVDList = strWtVDList;
findSubGraphs2Arg.maxIntWtList = maxIntWtList;
findSubGraphs2Arg.maxIntWtListSize = maxIntWtListSize;
findSubGraphs2Arg.soughtStrWtList = soughtStrWtList;
findSubGraphs2Arg.soughtStrWtListSize = soughtStrWtListSize;
TIMER_READ(start);
GOTO_SIM();
#ifdef OTM
#pragma omp parallel
{
findSubGraphs2((void*)&findSubGraphs2Arg);
}
#else
thread_start(findSubGraphs2, (void*)&findSubGraphs2Arg);
#endif
GOTO_REAL();
TIMER_READ(stop);
/* Verification */
/*
on_one_thread {
printf("\nInt weight sub-graphs \n");
for (i=0; i<maxIntWtListSize; i++) {
printf("%ld -- ", i);
for (j=0; j<intWtVDList[i].numArrays; j++) {
printf("\n [Array %ld] - \n", j);
for (k=0; k<intWtVDList[i].arraySize[j]; k++) {
printf("[%ld %ld] ", intWtVDList[i].vList[j][k].num, intWtVDList[i].vList[j][k].depth);
}
}
printf("\n");
}
printf("\nStr weight sub-graphs \n");
for (i=0; i<soughtStrWtListSize; i++) {
printf("%ld -- ", i);
for (j=0; j<strWtVDList[i].numArrays; j++) {
printf("\n [Array %ld] - \n", j);
for (k=0; k<strWtVDList[i].arraySize[j]; k++) {
printf("[%ld %ld] ", strWtVDList[i].vList[j][k].num, strWtVDList[i].vList[j][k].depth);
}
}
printf("\n");
}
}
*/
} else {
assert(0);
}
time = TIMER_DIFF_SECONDS(start, stop);
totalTime += time;
printf("\n\tfindSubGraphs() completed execution.\n");
printf("\nTime taken for kernel 3 is %9.6f sec.\n\n", time);
#endif /* ENABLE_KERNEL3 */
#ifdef ENABLE_KERNEL4
/* -------------------------------------------------------------------------
* Kernel 4 - Graph Clustering
* -------------------------------------------------------------------------
*/
printf("\nKernel 4 - cutClusters() beginning execution...\n");
TIMER_READ(start);
GOTO_SIM();
#ifdef OTM
#pragma omp parallel
{
cutClusters((void*)G);
}
#else
thread_start(cutClusters, (void*)G);
#endif
GOTO_REAL();
TIMER_READ(stop);
time = TIMER_DIFF_SECONDS(start, stop);
totalTime += time;
printf("\n\tcutClusters() completed execution.\n");
printf("\nTime taken for Kernel 4 is %9.6f sec.\n\n", time);
#endif /* ENABLE_KERNEL4 */
printf("Time = %9.6f \n", totalTime);
/* -------------------------------------------------------------------------
* Cleanup
* -------------------------------------------------------------------------
*/
P_FREE(G->outDegree);
P_FREE(G->outVertexIndex);
P_FREE(G->outVertexList);
P_FREE(G->paralEdgeIndex);
P_FREE(G->inDegree);
P_FREE(G->inVertexIndex);
P_FREE(G->inVertexList);
P_FREE(G->intWeight);
P_FREE(G->strWeight);
#ifdef ENABLE_KERNEL3
LONGINT_T i;
LONGINT_T j;
Vl* currV;
Vl* tempV;
if (K3_DS == 0) {
P_FREE(strWtVList);
P_FREE(intWtVList);
}
if (K3_DS == 1) {
for (i = 0; i < maxIntWtListSize; i++) {
currV = intWtVLList[i];
while (currV != NULL) {
tempV = currV->next;
P_FREE(currV);
currV = tempV;
}
}
for (i = 0; i < soughtStrWtListSize; i++) {
currV = strWtVLList[i];
while (currV != NULL) {
tempV = currV->next;
P_FREE(currV);
currV = tempV;
}
}
P_FREE(strWtVLList);
P_FREE(intWtVLList);
}
if (K3_DS == 2) {
for (i = 0; i < maxIntWtListSize; i++) {
for (j = 0; j < intWtVDList[i].numArrays; j++) {
P_FREE(intWtVDList[i].vList[j]);
}
P_FREE(intWtVDList[i].vList);
P_FREE(intWtVDList[i].arraySize);
}
for (i = 0; i < soughtStrWtListSize; i++) {
for (j = 0; j < strWtVDList[i].numArrays; j++) {
P_FREE(strWtVDList[i].vList[j]);
}
P_FREE(strWtVDList[i].vList);
P_FREE(strWtVDList[i].arraySize);
}
P_FREE(strWtVDList);
P_FREE(intWtVDList);
}
P_FREE(soughtStrWtList);
P_FREE(maxIntWtList);
#endif /* ENABLE_KERNEL2 */
P_FREE(SOUGHT_STRING);
P_FREE(G);
P_FREE(SDGdata);
TM_SHUTDOWN();
P_MEMORY_SHUTDOWN();
GOTO_SIM();
thread_shutdown();
MAIN_RETURN(0);
}
static void
print_performance_data (void)
{
GList *iter;
gint i = 0;
GstClockTime last_start = 0;
CaptureTimingStats max;
CaptureTimingStats min;
CaptureTimingStats avg;
CaptureTimingStats avg_wo_first;
GstClockTime shot_to_shot;
if (!performance_measure)
return;
parse_target_values ();
/* Initialize stats */
min.shot_to_shot = -1;
min.shot_to_save = -1;
min.shot_to_snapshot = -1;
min.preview_to_precapture = -1;
min.shot_to_buffer = -1;
memset (&avg, 0, sizeof (CaptureTimingStats));
memset (&avg_wo_first, 0, sizeof (CaptureTimingStats));
memset (&max, 0, sizeof (CaptureTimingStats));
g_print ("-- Performance results --\n");
g_print ("Startup time: %" TIME_FORMAT "; Target: %" TIME_FORMAT "\n",
TIME_ARGS (startup_time - initial_time), TIME_ARGS (target_startup));
g_print ("Change mode time: %" TIME_FORMAT "; Target: %" TIME_FORMAT "\n",
TIME_ARGS (change_mode_after - change_mode_before),
TIME_ARGS (target_change_mode));
g_print
("\n | Shot to save |Shot to snapshot| Shot to shot |"
"Preview to precap| Shot to buffer\n");
capture_times = g_list_reverse (capture_times);
for (iter = capture_times; iter; iter = g_list_next (iter)) {
CaptureTiming *t = (CaptureTiming *) iter->data;
CaptureTimingStats stats;
stats.shot_to_save = SHOT_TO_SAVE (t);
stats.shot_to_snapshot = SHOT_TO_SNAPSHOT (t);
stats.shot_to_shot = i == 0 ? 0 : t->start_capture - last_start;
stats.preview_to_precapture = PREVIEW_TO_PRECAPTURE (t);
stats.shot_to_buffer = SHOT_TO_BUFFER (t);
PRINT_STATS (i, &stats);
if (i != 0) {
capture_timing_stats_add (&avg_wo_first, &stats);
}
capture_timing_stats_add (&avg, &stats);
if (stats.shot_to_save < min.shot_to_save) {
min.shot_to_save = stats.shot_to_save;
}
if (stats.shot_to_snapshot < min.shot_to_snapshot) {
min.shot_to_snapshot = stats.shot_to_snapshot;
}
if (stats.shot_to_shot < min.shot_to_shot && stats.shot_to_shot > 0) {
min.shot_to_shot = stats.shot_to_shot;
}
if (stats.preview_to_precapture < min.preview_to_precapture) {
min.preview_to_precapture = stats.preview_to_precapture;
}
if (stats.shot_to_buffer < min.shot_to_buffer) {
min.shot_to_buffer = stats.shot_to_buffer;
}
if (stats.shot_to_save > max.shot_to_save) {
max.shot_to_save = stats.shot_to_save;
}
if (stats.shot_to_snapshot > max.shot_to_snapshot) {
max.shot_to_snapshot = stats.shot_to_snapshot;
}
if (stats.shot_to_shot > max.shot_to_shot) {
max.shot_to_shot = stats.shot_to_shot;
}
if (stats.preview_to_precapture > max.preview_to_precapture) {
max.preview_to_precapture = stats.preview_to_precapture;
}
if (stats.shot_to_buffer > max.shot_to_buffer) {
max.shot_to_buffer = stats.shot_to_buffer;
}
last_start = t->start_capture;
i++;
}
if (i > 1)
shot_to_shot = avg.shot_to_shot / (i - 1);
else
shot_to_shot = GST_CLOCK_TIME_NONE;
capture_timing_stats_div (&avg, i);
avg.shot_to_shot = shot_to_shot;
if (i > 1)
capture_timing_stats_div (&avg_wo_first, i - 1);
else {
memset (&avg_wo_first, 0, sizeof (CaptureTimingStats));
}
g_print ("\n Stats | MIN | MAX |"
" AVG | AVG wo First | Target | Diff \n");
g_print ("Shot to shot | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIME_FORMAT " | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIMEDIFF_FORMAT "\n",
TIME_ARGS (min.shot_to_shot), TIME_ARGS (max.shot_to_shot),
TIME_ARGS (avg.shot_to_shot),
TIME_ARGS (avg_wo_first.shot_to_shot),
TIME_ARGS (target_shot_to_shot),
TIMEDIFF_ARGS (TIME_DIFF (avg.shot_to_shot, target_shot_to_shot)));
g_print ("Shot to save | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIME_FORMAT " | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIMEDIFF_FORMAT "\n",
TIME_ARGS (min.shot_to_save), TIME_ARGS (max.shot_to_save),
TIME_ARGS (avg.shot_to_save),
TIME_ARGS (avg_wo_first.shot_to_save),
TIME_ARGS (target_shot_to_save),
TIMEDIFF_ARGS (TIME_DIFF (avg.shot_to_save, target_shot_to_save)));
g_print ("Shot to snapshot | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIME_FORMAT " | %" TIME_FORMAT " | %" TIME_FORMAT
" | %" TIMEDIFF_FORMAT "\n",
TIME_ARGS (min.shot_to_snapshot),
TIME_ARGS (max.shot_to_snapshot),
TIME_ARGS (avg.shot_to_snapshot),
TIME_ARGS (avg_wo_first.shot_to_snapshot),
TIME_ARGS (target_shot_to_snapshot),
TIMEDIFF_ARGS (TIME_DIFF (avg.shot_to_snapshot,
target_shot_to_snapshot)));
g_print ("Preview to precapture | %" TIME_FORMAT " | %" TIME_FORMAT " | %"
TIME_FORMAT " | %" TIME_FORMAT " | %" TIME_FORMAT " | %" TIMEDIFF_FORMAT
"\n", TIME_ARGS (min.preview_to_precapture),
TIME_ARGS (max.preview_to_precapture),
TIME_ARGS (avg.preview_to_precapture),
TIME_ARGS (avg_wo_first.preview_to_precapture),
TIME_ARGS (target_preview_to_precapture),
TIMEDIFF_ARGS (TIME_DIFF (avg.preview_to_precapture,
target_preview_to_precapture)));
g_print ("Shot to buffer | %" TIME_FORMAT " | %" TIME_FORMAT " | %"
TIME_FORMAT " | %" TIME_FORMAT " | %" TIME_FORMAT " | %" TIMEDIFF_FORMAT
"\n", TIME_ARGS (min.shot_to_buffer), TIME_ARGS (max.shot_to_buffer),
TIME_ARGS (avg.shot_to_buffer), TIME_ARGS (avg_wo_first.shot_to_buffer),
TIME_ARGS (target_shot_to_buffer),
TIMEDIFF_ARGS (TIME_DIFF (avg.shot_to_buffer, target_shot_to_buffer)));
}
void dump_cache_stats(void) {
PRINT_STATS(ParsedPMinisock);
PRINT_STATS(Continuation);
}
/* =============================================================================
* main
* =============================================================================
*/
MAIN(argc, argv)
{
GOTO_REAL();
/*
* Initialization
*/
SETUP_NUMBER_TASKS(3);
parseArgs(argc, (char** const)argv);
long numThread = global_params[PARAM_THREAD];
SETUP_NUMBER_THREADS(numThread);
SIM_GET_NUM_CPU(numThread);
TM_STARTUP(numThread, 0);
P_MEMORY_STARTUP(numThread);
thread_startup(numThread);
long percentAttack = global_params[PARAM_ATTACK];
long maxDataLength = global_params[PARAM_LENGTH];
long numFlow = global_params[PARAM_NUM];
long randomSeed = global_params[PARAM_SEED];
/* printf("Percent attack = %li\n", percentAttack);
printf("Max data length = %li\n", maxDataLength);
printf("Num flow = %li\n", numFlow);
printf("Random seed = %li\n", randomSeed); */
double time_total = 0.0;
int repeats = global_params[PARAM_REPEAT];
for (; repeats > 0; --repeats) {
dictionary_t* dictionaryPtr = dictionary_alloc();
assert(dictionaryPtr);
stream_t* streamPtr = stream_alloc(percentAttack);
assert(streamPtr);
long numAttack = stream_generate(streamPtr,
dictionaryPtr,
numFlow,
randomSeed,
maxDataLength);
// printf("Num attack = %li\n", numAttack);
decoder_t* decoderPtr = decoder_alloc();
assert(decoderPtr);
vector_t** errorVectors = (vector_t**)malloc(numThread * sizeof(vector_t*));
assert(errorVectors);
long i;
for (i = 0; i < numThread; i++) {
vector_t* errorVectorPtr = vector_alloc(numFlow);
assert(errorVectorPtr);
errorVectors[i] = errorVectorPtr;
}
arg_t arg;
arg.streamPtr = streamPtr;
arg.decoderPtr = decoderPtr;
arg.errorVectors = errorVectors;
/*
* Run transactions
*/
TIMER_T startTime;
TIMER_READ(startTime);
tm_time_t start_clock=TM_TIMER_READ();
GOTO_SIM();
thread_start(processPackets, (void*)&arg);
GOTO_REAL();
TIMER_T stopTime;
tm_time_t end_clock=TM_TIMER_READ();
TIMER_READ(stopTime);
double time_tmp = TIMER_DIFF_SECONDS(startTime, stopTime);
time_total += time_tmp;
PRINT_STATS();
PRINT_CLOCK_THROUGHPUT(end_clock-start_clock);
/*
* Check solution
*/
long numFound = 0;
for (i = 0; i < numThread; i++) {
vector_t* errorVectorPtr = errorVectors[i];
long e;
long numError = vector_getSize(errorVectorPtr);
numFound += numError;
for (e = 0; e < numError; e++) {
long flowId = (long)vector_at(errorVectorPtr, e);
bool_t status = stream_isAttack(streamPtr, flowId);
assert(status);
}
}
// printf("Num found = %li\n", numFound);
assert(numFound == numAttack);
/*
* Clean up
*/
for (i = 0; i < numThread; i++) {
vector_free(errorVectors[i]);
}
free(errorVectors);
decoder_free(decoderPtr);
stream_free(streamPtr);
dictionary_free(dictionaryPtr);
}
printf("Time = %f\n", time_total);
TM_SHUTDOWN();
P_MEMORY_SHUTDOWN();
GOTO_SIM();
thread_shutdown();
MAIN_RETURN(0);
}
XPT_DumpStats(XPTArena *arena)
{
PRINT_STATS(arena);
}
