static int lua_likwid_getCpuInfo(lua_State* L)
{
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
lua_newtable(L);
lua_pushstring(L,"family");
lua_pushunsigned(L,cpuinfo->family);
lua_settable(L,-3);
lua_pushstring(L,"model");
lua_pushunsigned(L,cpuinfo->model);
lua_settable(L,-3);
lua_pushstring(L,"stepping");
lua_pushunsigned(L,cpuinfo->stepping);
lua_settable(L,-3);
lua_pushstring(L,"clock");
lua_pushunsigned(L,cpuinfo->clock);
lua_settable(L,-3);
lua_pushstring(L,"turbo");
lua_pushinteger(L,cpuinfo->turbo);
lua_settable(L,-3);
lua_pushstring(L,"name");
lua_pushstring(L,cpuinfo->name);
lua_settable(L,-3);
lua_pushstring(L,"osname");
lua_pushstring(L,cpuinfo->osname);
lua_settable(L,-3);
lua_pushstring(L,"short_name");
lua_pushstring(L,cpuinfo->short_name);
lua_settable(L,-3);
lua_pushstring(L,"features");
lua_pushstring(L,cpuinfo->features);
lua_settable(L,-3);
lua_pushstring(L,"isIntel");
lua_pushinteger(L,cpuinfo->isIntel);
lua_settable(L,-3);
lua_pushstring(L,"featureFlags");
lua_pushunsigned(L,cpuinfo->featureFlags);
lua_settable(L,-3);
lua_pushstring(L,"perf_version");
lua_pushunsigned(L, cpuinfo->perf_version);
lua_settable(L,-3);
lua_pushstring(L,"perf_num_ctr");
lua_pushunsigned(L,cpuinfo->perf_num_ctr);
lua_settable(L,-3);
lua_pushstring(L,"perf_width_ctr");
lua_pushunsigned(L,cpuinfo->perf_width_ctr);
lua_settable(L,-3);
lua_pushstring(L,"perf_num_fixed_ctr");
lua_pushunsigned(L,cpuinfo->perf_num_fixed_ctr);
lua_settable(L,-3);
return 1;
}
static int lua_likwid_init(lua_State* L)
{
int ret;
int nrThreads = luaL_checknumber(L,1);
luaL_argcheck(L, nrThreads > 0, 1, "CPU count must be greater than 0");
int cpus[nrThreads];
if (!lua_istable(L, -1)) {
lua_pushstring(L,"No table given as second argument");
lua_error(L);
}
for (ret = 1; ret<=nrThreads; ret++)
{
lua_rawgeti(L,-1,ret);
cpus[ret-1] = lua_tounsigned(L,-1);
lua_pop(L,1);
}
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
cputopo = get_cpuTopology();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cputopo == NULL))
{
cputopo = get_cpuTopology();
}
if (numa_isInitialized == 0)
{
numa_init();
numa_isInitialized = 1;
numainfo = get_numaTopology();
}
if ((numa_isInitialized) && (numainfo == NULL))
{
numainfo = get_numaTopology();
}
if (perfmon_isInitialized == 0)
{
ret = perfmon_init(nrThreads, &(cpus[0]));
if (ret != 0)
{
lua_pushstring(L,"Cannot initialize likwid perfmon");
lua_error(L);
return 1;
}
perfmon_isInitialized = 1;
timer_isInitialized = 1;
lua_pushinteger(L,ret);
}
return 1;
}
int test_perfmonstop_noadd()
{
CpuInfo_t cpuinfo;
int cpu = 0;
int group;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0)
goto fail;
ret = perfmon_stopCounters();
if (ret == 0)
goto fail;
perfmon_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
topology_finalize();
return 0;
}
int test_perfmonstart()
{
CpuInfo_t cpuinfo;
int group1, group2;
int cpu = 0;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0)
goto fail;
ret = perfmon_addEventSet(eventset_ok);
if (ret != 0)
goto fail;
group1 = ret;
ret = perfmon_setupCounters(group1);
if (ret != 0)
goto fail;
if (perfmon_getIdOfActiveGroup() != group1)
goto fail;
ret = perfmon_startCounters();
if (ret != 0)
goto fail;
perfmon_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
topology_finalize();
return 0;
}
int test_perfmonstop_nostart()
{
CpuInfo_t cpuinfo;
int cpu = 0;
int group;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0)
goto fail;
ret = perfmon_addEventSet(eventset_ok);
if (ret != 0)
goto fail;
group = ret;
ret = perfmon_setupCounters(group);
if (ret != 0)
goto fail;
ret = perfmon_stopCounters();
if (ret == 0)
goto fail;
perfmon_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
topology_finalize();
return 0;
}
int test_perfmonresult_noadd()
{
CpuInfo_t cpuinfo;
int cpu = 0;
int group;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0)
goto fail;
double result = perfmon_getResult(0,0,0);
if (result != 0)
goto fail;
perfmon_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
topology_finalize();
return 0;
}
int main(int argc, char* argv[])
{
int i;
int* cpus;
int gid;
double result = 0.0;
// Load the topology module and print some values.
topology_init();
// CpuInfo_t contains global information like name, CPU family, ...
CpuInfo_t info = get_cpuInfo();
// CpuTopology_t contains information about the topology of the CPUs.
CpuTopology_t topo = get_cpuTopology();
printf("Likwid example on a %s with %d CPUs\n", info->name, topo->numHWThreads);
cpus = malloc(topo->numHWThreads * sizeof(int));
if (!cpus)
return 1;
for (i=0;i<topo->numHWThreads;i++)
{
cpus[i] = topo->threadPool[i].apicId;
}
// Must be called before perfmon_init() but only if you want to use another
// access mode as the pre-configured one. For direct access (0) you have to
// be root.
//accessClient_setaccessmode(0);
// Initialize the perfmon module.
perfmon_init(topo->numHWThreads, cpus);
// Add eventset string to the perfmon module.
gid = perfmon_addEventSet(EVENTSET);
// Setup the eventset identified by group ID (gid).
perfmon_setupCounters(gid);
// Start all counters in the previously set up event set.
perfmon_startCounters();
// Perform something
sleep(2);
// Stop all counters in the previously started event set.
perfmon_stopCounters();
// Print the result of every thread/CPU.
for (i = 0;i < topo->numHWThreads; i++)
{
result = perfmon_getResult(gid, 0, i);
printf("Measurement result for event set %s at CPU %d: %f\n", EVENTSET, cpus[i], result);
}
// Uninitialize the perfmon module.
perfmon_finalize();
// Uninitialize the topology module.
topology_finalize();
return 0;
}
int test_perfmonresult()
{
CpuInfo_t cpuinfo;
int cpu = 0;
int group;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0)
goto fail;
ret = perfmon_addEventSet(eventset_ok);
if (ret != 0)
goto fail;
group = ret;
ret = perfmon_setupCounters(group);
if (ret != 0)
goto fail;
ret = perfmon_startCounters();
if (ret != 0)
goto fail;
sleep(1);
ret = perfmon_stopCounters();
if (ret != 0)
goto fail;
if ((perfmon_getResult(group,0,0) == 0)||(perfmon_getResult(group,1,0) == 0))
goto fail;
if (perfmon_getTimeOfGroup(group) == 0)
goto fail;
perfmon_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
topology_finalize();
return 0;
}
int test_perfmonperfgroup()
{
CpuInfo_t cpuinfo;
int i;
int cpu = 0;
topology_init();
cpuinfo = get_cpuInfo();
int ret = perfmon_init(1, &cpu);
if (ret != 0) {
printf("Perfmon init failed\n");
goto fail;
}
char** glist = NULL;
char** slist = NULL;
char** llist = NULL;
ret = perfmon_getGroups(&glist, &slist, &llist);
if (ret <= 0)
{
goto fail;
}
ret = perfmon_addEventSet(glist[0]);
if (ret != 0) {
printf("Perfmon addEventSet(%s) failed\n", glist[0]);
goto fail;
}
if (perfmon_getNumberOfEvents(ret) == 0) {
printf("Perfmon number of events == 0\n");
goto fail;
}
if (perfmon_getNumberOfMetrics(ret) == 0) {
printf("Perfmon number of metrics == 0\n");
goto fail;
}
for (i=0; i<perfmon_getNumberOfEvents(ret); i++) {
if (strcmp(perfmon_getEventName(ret, i), "") == 0)
goto fail;
if (strcmp(perfmon_getCounterName(ret, i), "") == 0)
goto fail;
}
if (strcmp(perfmon_getGroupName(ret), "Custom") == 0)
{
goto fail;
}
if (strcmp(perfmon_getGroupInfoShort(ret), "Custom") == 0)
{
goto fail;
}
if (strcmp(perfmon_getGroupInfoLong(ret), "Custom") == 0)
{
goto fail;
}
if (perfmon_getLastTimeOfGroup(ret) != 0)
{
goto fail;
}
if (perfmon_getTimeOfGroup(ret) != 0)
{
goto fail;
}
for (i=0; i<perfmon_getNumberOfMetrics(ret); i++) {
if (strcmp(perfmon_getMetricName(ret, i), "") == 0)
goto fail;
if (perfmon_getMetric(ret, i, 0) != 0.0)
goto fail;
}
free(glist);
free(slist);
free(llist);
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 1;
fail:
if (glist)
free(glist);
if (slist)
free(slist);
if (llist)
free(llist);
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 0;
}
int test_perfmoncustomgroup()
{
CpuInfo_t cpuinfo;
int cpu = 0;
topology_init();
cpuinfo = get_cpuInfo();
int ret = perfmon_init(1, &cpu);
if (ret != 0) {
printf("Perfmon init failed\n");
goto fail;
}
ret = perfmon_addEventSet(eventset_ok);
if (ret != 0) {
printf("Perfmon addEventSet(ok) failed\n");
goto fail;
}
if (perfmon_getNumberOfEvents(ret) != 3) {
printf("Perfmon number of events != 3\n");
goto fail;
}
if (perfmon_getNumberOfMetrics(ret) != 0) {
printf("Perfmon number of metrics != 0\n");
goto fail;
}
if (strcmp(perfmon_getEventName(ret, 0), event1_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getEventName(ret, 1), event2_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getEventName(ret, 2), event3_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getCounterName(ret, 0), ctr1_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getCounterName(ret, 1), ctr2_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getCounterName(ret, 2), ctr3_ok) != 0)
{
goto fail;
}
if (strcmp(perfmon_getGroupName(ret), "Custom") != 0)
{
goto fail;
}
if (strcmp(perfmon_getGroupInfoShort(ret), "Custom") != 0)
{
goto fail;
}
if (strcmp(perfmon_getGroupInfoLong(ret), "Custom") != 0)
{
goto fail;
}
if (perfmon_getLastTimeOfGroup(ret) != 0)
{
goto fail;
}
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 0;
}
int test_perfmonaddeventset()
{
char eventset_fail1[] = "INSTR_RETIRED.ANY:FIXC0";
char eventset_fail2[] = "INSTR_RETIRED-ANY:FIXC0";
CpuInfo_t cpuinfo;
int cpu = 0;
topology_init();
cpuinfo = get_cpuInfo();
if (cpuinfo->isIntel == 0)
{
topology_finalize();
return 1;
}
int ret = perfmon_init(1, &cpu);
if (ret != 0) {
printf("Perfmon init failed\n");
goto fail;
}
if (perfmon_getNumberOfGroups() != 0) {
printf("Perfmon number of groups != 0\n");
goto fail;
}
if (perfmon_getNumberOfThreads() != 1) {
printf("Perfmon number of threads != 1\n");
goto fail;
}
if (perfmon_getIdOfActiveGroup() != -1) {
printf("Perfmon id of active group != -1\n");
goto fail;
}
ret = perfmon_addEventSet(eventset_ok);
if (ret != 0) {
printf("Perfmon addEventSet(ok) failed\n");
goto fail;
}
if (perfmon_getNumberOfGroups() != 1) {
printf("Perfmon number of groups != 1\n");
goto fail;
}
if (perfmon_getNumberOfEvents(ret) != 3) {
printf("Perfmon number of events != 3\n");
goto fail;
}
if (perfmon_getIdOfActiveGroup() != -1) {
printf("Perfmon id of active group != -1\n");
goto fail;
}
ret = perfmon_addEventSet(eventset_option);
if (ret != 1) {
printf("Perfmon addEventSet(options) failed\n");
goto fail;
}
if (perfmon_getNumberOfGroups() != 2) {
printf("Perfmon number of groups != 2\n");
goto fail;
}
if (perfmon_getNumberOfEvents(ret) != 3) {
printf("Perfmon number of events != 3\n");
goto fail;
}
if (perfmon_getIdOfActiveGroup() != -1) {
printf("Perfmon id of active group != -1\n");
goto fail;
}
ret = perfmon_addEventSet(eventset_fail1);
if (ret >= 0) {
printf("Perfmon addEventSet(fail1) failed\n");
goto fail;
}
if (perfmon_getNumberOfGroups() != 2) {
printf("Perfmon number of groups != 2\n");
goto fail;
}
ret = perfmon_addEventSet(eventset_fail2);
if (ret >= 0) {
printf("Perfmon addEventSet(fail2) failed\n");
goto fail;
}
if (perfmon_getNumberOfGroups() != 2) {
printf("Perfmon number of groups != 2\n");
goto fail;
}
if (perfmon_getIdOfActiveGroup() != -1) {
printf("Perfmon id of active group != -1\n");
goto fail;
}
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 1;
fail:
perfmon_finalize();
affinity_finalize();
topology_finalize();
return 0;
}
static int lua_likwid_getAffinityInfo(lua_State* L)
{
int i,j;
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
cputopo = get_cpuTopology();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cputopo == NULL))
{
cputopo = get_cpuTopology();
}
if (numa_isInitialized == 0)
{
if (numa_init() == 0)
{
numa_isInitialized = 1;
numainfo = get_numaTopology();
}
}
if ((numa_isInitialized) && (numainfo == NULL))
{
numainfo = get_numaTopology();
}
if (affinity_isInitialized == 0)
{
affinity_init();
affinity_isInitialized = 1;
affinity = get_affinityDomains();
}
if ((affinity_isInitialized) && (affinity == NULL))
{
affinity = get_affinityDomains();
}
if (!affinity)
{
lua_pushstring(L,"Cannot initialize affinity groups");
lua_error(L);
}
lua_newtable(L);
lua_pushstring(L,"numberOfAffinityDomains");
lua_pushunsigned(L,affinity->numberOfAffinityDomains);
lua_settable(L,-3);
lua_pushstring(L,"numberOfSocketDomains");
lua_pushunsigned(L,affinity->numberOfSocketDomains);
lua_settable(L,-3);
lua_pushstring(L,"numberOfNumaDomains");
lua_pushunsigned(L,affinity->numberOfNumaDomains);
lua_settable(L,-3);
lua_pushstring(L,"numberOfProcessorsPerSocket");
lua_pushunsigned(L,affinity->numberOfProcessorsPerSocket);
lua_settable(L,-3);
lua_pushstring(L,"numberOfCacheDomains");
lua_pushunsigned(L,affinity->numberOfCacheDomains);
lua_settable(L,-3);
lua_pushstring(L,"numberOfCoresPerCache");
lua_pushunsigned(L,affinity->numberOfCoresPerCache);
lua_settable(L,-3);
lua_pushstring(L,"numberOfProcessorsPerCache");
lua_pushunsigned(L,affinity->numberOfProcessorsPerCache);
lua_settable(L,-3);
lua_pushstring(L,"domains");
lua_newtable(L);
for(i=0;i<affinity->numberOfAffinityDomains;i++)
{
lua_pushunsigned(L, i+1);
lua_newtable(L);
lua_pushstring(L,"tag");
lua_pushstring(L,bdata(affinity->domains[i].tag));
lua_settable(L,-3);
lua_pushstring(L,"numberOfProcessors");
lua_pushunsigned(L,affinity->domains[i].numberOfProcessors);
lua_settable(L,-3);
lua_pushstring(L,"numberOfCores");
lua_pushunsigned(L,affinity->domains[i].numberOfCores);
lua_settable(L,-3);
lua_pushstring(L,"processorList");
lua_newtable(L);
for(j=0;j<affinity->domains[i].numberOfProcessors;j++)
{
lua_pushunsigned(L,j+1);
lua_pushunsigned(L,affinity->domains[i].processorList[j]);
lua_settable(L,-3);
}
lua_settable(L,-3);
lua_settable(L,-3);
}
lua_settable(L,-3);
return 1;
}
static int lua_likwid_getNumaInfo(lua_State* L)
{
uint32_t i,j;
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
cputopo = get_cpuTopology();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cputopo == NULL))
{
cputopo = get_cpuTopology();
}
if (numa_isInitialized == 0)
{
if (numa_init() == 0)
{
numa_isInitialized = 1;
numainfo = get_numaTopology();
}
else
{
lua_newtable(L);
lua_pushstring(L,"numberOfNodes");
lua_pushunsigned(L,0);
lua_settable(L,-3);
lua_pushstring(L,"nodes");
lua_newtable(L);
lua_settable(L,-3);
return 1;
}
}
if ((numa_isInitialized) && (numainfo == NULL))
{
numainfo = get_numaTopology();
}
if (affinity_isInitialized == 0)
{
affinity_init();
affinity_isInitialized = 1;
affinity = get_affinityDomains();
}
if ((affinity_isInitialized) && (affinity == NULL))
{
affinity = get_affinityDomains();
}
lua_newtable(L);
lua_pushstring(L,"numberOfNodes");
lua_pushunsigned(L,numainfo->numberOfNodes);
lua_settable(L,-3);
lua_pushstring(L,"nodes");
lua_newtable(L);
for(i=0;i<numainfo->numberOfNodes;i++)
{
lua_pushinteger(L, i+1);
lua_newtable(L);
lua_pushstring(L,"id");
lua_pushunsigned(L,numainfo->nodes[i].id);
lua_settable(L,-3);
lua_pushstring(L,"totalMemory");
lua_pushunsigned(L,numainfo->nodes[i].totalMemory);
lua_settable(L,-3);
lua_pushstring(L,"freeMemory");
lua_pushunsigned(L,numainfo->nodes[i].freeMemory);
lua_settable(L,-3);
lua_pushstring(L,"numberOfProcessors");
lua_pushunsigned(L,numainfo->nodes[i].numberOfProcessors);
lua_settable(L,-3);
lua_pushstring(L,"numberOfDistances");
lua_pushunsigned(L,numainfo->nodes[i].numberOfDistances);
lua_settable(L,-3);
lua_pushstring(L,"processors");
lua_newtable(L);
for(j=0;j<numainfo->nodes[i].numberOfProcessors;j++)
{
lua_pushunsigned(L,j+1);
lua_pushunsigned(L,numainfo->nodes[i].processors[j]);
lua_settable(L,-3);
}
lua_settable(L,-3);
/*lua_pushstring(L,"processorsCompact");
lua_newtable(L);
for(j=0;j<numa->nodes[i].numberOfProcessors;j++)
{
lua_pushunsigned(L,j);
lua_pushunsigned(L,numa->nodes[i].processorsCompact[j]);
lua_settable(L,-3);
}
lua_settable(L,-3);*/
lua_pushstring(L,"distances");
lua_newtable(L);
for(j=0;j<numainfo->nodes[i].numberOfDistances;j++)
{
lua_pushinteger(L,j+1);
lua_newtable(L);
lua_pushinteger(L,j);
lua_pushunsigned(L,numainfo->nodes[i].distances[j]);
lua_settable(L,-3);
lua_settable(L,-3);
}
lua_settable(L,-3);
lua_settable(L,-3);
}
lua_settable(L,-3);
return 1;
}
static int lua_likwid_getEventsAndCounters(lua_State* L)
{
int i;
char optString[1024];
int optStringIndex = 0;
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
perfmon_init_maps();
lua_newtable(L);
lua_pushstring(L,"Counters");
lua_newtable(L);
for(i=1;i<=perfmon_numCounters;i++)
{
optStringIndex = 0;
optString[0] = '\0';
lua_pushunsigned(L,i);
lua_newtable(L);
lua_pushstring(L,"Name");
lua_pushstring(L,counter_map[i-1].key);
lua_settable(L,-3);
lua_pushstring(L,"Options");
for(int j=1; j<NUM_EVENT_OPTIONS; j++)
{
if (counter_map[i-1].optionMask & REG_TYPE_MASK(j))
{
optStringIndex += sprintf(&(optString[optStringIndex]), "%s|", eventOptionTypeName[j]);
}
}
optString[optStringIndex-1] = '\0';
lua_pushstring(L,optString);
lua_settable(L,-3);
lua_pushstring(L,"Type");
lua_pushunsigned(L, counter_map[i-1].type);
lua_settable(L,-3);
lua_pushstring(L,"TypeName");
lua_pushstring(L, RegisterTypeNames[counter_map[i-1].type]);
lua_settable(L,-3);
lua_pushstring(L,"Index");
lua_pushunsigned(L,counter_map[i-1].index);
lua_settable(L,-3);
lua_settable(L,-3);
}
lua_settable(L,-3);
lua_pushstring(L,"Events");
lua_newtable(L);
for(i=1;i<=perfmon_numArchEvents;i++)
{
optStringIndex = 0;
optString[0] = '\0';
lua_pushunsigned(L,i);
lua_newtable(L);
lua_pushstring(L,"Name");
lua_pushstring(L,eventHash[i-1].name);
lua_settable(L,-3);
lua_pushstring(L,"ID");
lua_pushunsigned(L,eventHash[i-1].eventId);
lua_settable(L,-3);
lua_pushstring(L,"UMask");
lua_pushunsigned(L,eventHash[i-1].umask);
lua_settable(L,-3);
lua_pushstring(L,"Limit");
lua_pushstring(L,eventHash[i-1].limit);
lua_settable(L,-3);
lua_pushstring(L,"Options");
for(int j=1; j<NUM_EVENT_OPTIONS; j++)
{
if (eventHash[i-1].optionMask & REG_TYPE_MASK(j))
{
optStringIndex += sprintf(&(optString[optStringIndex]), "%s|", eventOptionTypeName[j]);
}
}
optString[optStringIndex-1] = '\0';
lua_pushstring(L,optString);
lua_settable(L,-3);
lua_settable(L,-3);
}
lua_settable(L,-3);
return 1;
}
static int lua_likwid_getPowerInfo(lua_State* L)
{
int i;
if (topology_isInitialized == 0)
{
topology_init();
topology_isInitialized = 1;
cpuinfo = get_cpuInfo();
cputopo = get_cpuTopology();
}
if ((topology_isInitialized) && (cpuinfo == NULL))
{
cpuinfo = get_cpuInfo();
}
if ((topology_isInitialized) && (cputopo == NULL))
{
cputopo = get_cpuTopology();
}
if (power_isInitialized == 0)
{
power_hasRAPL = power_init(0);
if (power_hasRAPL)
{
power_isInitialized = 1;
power = get_powerInfo();
}
else
{
return 0;
}
}
lua_newtable(L);
lua_pushstring(L,"hasRAPL");
lua_pushboolean(L,power_hasRAPL);
lua_settable(L,-3);
lua_pushstring(L,"baseFrequency");
lua_pushnumber(L,power->baseFrequency);
lua_settable(L,-3);
lua_pushstring(L,"minFrequency");
lua_pushnumber(L,power->minFrequency);
lua_settable(L,-3);
lua_pushstring(L,"powerUnit");
lua_pushnumber(L,power->powerUnit);
lua_settable(L,-3);
lua_pushstring(L,"timeUnit");
lua_pushnumber(L,power->timeUnit);
lua_settable(L,-3);
lua_pushstring(L,"turbo");
lua_newtable(L);
lua_pushstring(L,"numSteps");
lua_pushunsigned(L,power->turbo.numSteps);
lua_settable(L,-3);
lua_pushstring(L,"steps");
lua_newtable(L);
for(i=0;i<power->turbo.numSteps;i++)
{
lua_pushunsigned(L,i+1);
lua_pushnumber(L,power->turbo.steps[i]);
lua_settable(L,-3);
}
lua_settable(L,-3);
lua_settable(L,-3);
lua_pushstring(L,"domains");
lua_newtable(L);
for(i=0;i<NUM_POWER_DOMAINS;i++)
{
lua_pushstring(L,power_names[i]);
lua_newtable(L);
lua_pushstring(L, "ID");
lua_pushnumber(L, power->domains[i].type);
lua_settable(L,-3);
lua_pushstring(L, "energyUnit");
lua_pushnumber(L, power->domains[i].energyUnit);
lua_settable(L,-3);
lua_pushstring(L,"supportStatus");
if (power->domains[i].supportFlags & POWER_DOMAIN_SUPPORT_STATUS)
{
lua_pushboolean(L, 1);
}
else
{
lua_pushboolean(L, 0);
}
lua_settable(L,-3);
lua_pushstring(L,"supportPerf");
if (power->domains[i].supportFlags & POWER_DOMAIN_SUPPORT_PERF)
{
lua_pushboolean(L, 1);
}
else
{
lua_pushboolean(L, 0);
}
lua_settable(L,-3);
lua_pushstring(L,"supportPolicy");
if (power->domains[i].supportFlags & POWER_DOMAIN_SUPPORT_POLICY)
{
lua_pushboolean(L, 1);
}
else
{
lua_pushboolean(L, 0);
}
lua_settable(L,-3);
lua_pushstring(L,"supportLimit");
if (power->domains[i].supportFlags & POWER_DOMAIN_SUPPORT_LIMIT)
{
lua_pushboolean(L, 1);
}
else
{
lua_pushboolean(L, 0);
}
lua_settable(L,-3);
if (power->domains[i].supportFlags & POWER_DOMAIN_SUPPORT_INFO)
{
lua_pushstring(L,"supportInfo");
lua_pushboolean(L, 1);
lua_settable(L,-3);
lua_pushstring(L,"tdp");
lua_pushnumber(L, power->domains[i].tdp);
lua_settable(L,-3);
lua_pushstring(L,"minPower");
lua_pushnumber(L, power->domains[i].minPower);
lua_settable(L,-3);
lua_pushstring(L,"maxPower");
lua_pushnumber(L, power->domains[i].maxPower);
lua_settable(L,-3);
lua_pushstring(L,"maxTimeWindow");
lua_pushnumber(L, power->domains[i].maxTimeWindow);
lua_settable(L,-3);
}
else
{
lua_pushstring(L,"supportInfo");
lua_pushboolean(L, 0);
lua_settable(L,-3);
}
lua_settable(L,-3);
}
lua_settable(L,-3);
return 1;
}
int test_topologyinit()
{
int i, j;
int ret = topology_init();
if (ret != 0)
goto fail;
CpuInfo_t cpuinfo = get_cpuInfo();
if (cpuinfo == NULL)
goto fail;
if (cpuinfo->family == 0)
goto fail;
if (cpuinfo->model == 0)
goto fail;
if (cpuinfo->osname == NULL)
goto fail;
if (cpuinfo->name == NULL)
goto fail;
if (cpuinfo->features == NULL)
goto fail;
CpuTopology_t cputopo = get_cpuTopology();
if (cputopo->threadPool == NULL)
goto fail;
if (cputopo->cacheLevels == NULL)
goto fail;
if (cputopo->numHWThreads == 0)
goto fail;
if (cputopo->activeHWThreads == 0)
goto fail;
if (cputopo->numSockets == 0)
goto fail;
if (cputopo->numCoresPerSocket < 1)
goto fail;
if (cputopo->numThreadsPerCore < 1)
goto fail;
if (cputopo->numHWThreads > 0)
{
for (i = 0; i < cputopo->numHWThreads; i++)
{
for (j=0;j< cputopo->numHWThreads; j++)
{
if ((i != j) && (cputopo->threadPool[i].apicId == cputopo->threadPool[j].apicId))
goto fail;
}
if (cputopo->threadPool[i].threadId >= cputopo->numThreadsPerCore)
{
goto fail;
}
if (cputopo->threadPool[i].packageId >= cputopo->numSockets)
{
goto fail;
}
}
}
if (cputopo->numCacheLevels > 0)
{
for (i=0;i<cputopo->numCacheLevels;i++)
{
if (cputopo->cacheLevels[i].level > cputopo->numCacheLevels)
{
goto fail;
}
}
}
isIntel = cpuinfo->isIntel;
topology_finalize();
return 1;
fail:
topology_finalize();
return 0;
}
int main(int argc, char* argv[])
{
int i, j;
int err;
int* cpus;
int gid;
double result = 0.0;
char estr[] = "INSTR_RETIRED_ANY:FIXC0,CPU_CLK_UNHALTED_CORE:FIXC1,CPU_CLK_UNHALTED_REF:FIXC2,TEMP_CORE:TMP0";
// Load the topology module and print some values.
err = topology_init();
if (err < 0)
{
printf("Failed to initialize LIKWID's topology module\n");
return 1;
}
// CpuInfo_t contains global information like name, CPU family, ...
CpuInfo_t info = get_cpuInfo();
// CpuTopology_t contains information about the topology of the CPUs.
CpuTopology_t topo = get_cpuTopology();
// Create affinity domains. Commonly only needed when reading Uncore counters
//affinity_init();
printf("Likwid example on a %s with %d CPUs\n", info->name, topo->numHWThreads);
cpus = (int*)malloc(topo->numHWThreads * sizeof(int));
if (!cpus)
return 1;
for (i=0;i<topo->numHWThreads;i++)
{
cpus[i] = topo->threadPool[i].apicId;
}
// Must be called before perfmon_init() but only if you want to use another
// access mode as the pre-configured one. For direct access (0) you have to
// be root.
//accessClient_setaccessmode(0);
// Initialize the perfmon module.
err = perfmon_init(topo->numHWThreads, cpus);
if (err < 0)
{
printf("Failed to initialize LIKWID's performance monitoring module\n");
topology_finalize();
return 1;
}
// Add eventset string to the perfmon module.
gid = perfmon_addEventSet(estr);
if (gid < 0)
{
printf("Failed to add event string %s to LIKWID's performance monitoring module\n", estr);
perfmon_finalize();
topology_finalize();
return 1;
}
// Setup the eventset identified by group ID (gid).
err = perfmon_setupCounters(gid);
if (err < 0)
{
printf("Failed to setup group %d in LIKWID's performance monitoring module\n", gid);
perfmon_finalize();
topology_finalize();
return 1;
}
// Start all counters in the previously set up event set.
err = perfmon_startCounters();
if (err < 0)
{
printf("Failed to start counters for group %d for thread %d\n",gid, (-1*err)-1);
perfmon_finalize();
topology_finalize();
return 1;
}
// Perform something
sleep(2);
// Stop all counters in the previously started event set.
err = perfmon_stopCounters();
if (err < 0)
{
printf("Failed to stop counters for group %d for thread %d\n",gid, (-1*err)-1);
perfmon_finalize();
topology_finalize();
return 1;
}
// Print the result of every thread/CPU for all events in estr.
char* ptr = strtok(estr,",");
j = 0;
while (ptr != NULL)
{
for (i = 0;i < topo->numHWThreads; i++)
{
result = perfmon_getResult(gid, j, cpus[i]);
printf("Measurement result for event set %s at CPU %d: %f\n", ptr, cpus[i], result);
}
ptr = strtok(NULL,",");
j++;
}
// Uninitialize the perfmon module.
perfmon_finalize();
// Uninitialize the topology module.
topology_finalize();
return 0;
}
