// initialize our arenas (this is required to be able to set the chunk hooks)
static void initialize_arenas(void) {
size_t s_narenas;
unsigned narenas;
unsigned arena;
// "thread.arena" takes an unsigned, but num_arenas is a size_t.
s_narenas = get_num_arenas();
if (s_narenas > (size_t) UINT_MAX) {
chpl_internal_error("narenas too large to fit into unsigned");
}
narenas = (unsigned) s_narenas;
// for each non-zero arena, set the current thread to use it (this
// initializes each arena). arena 0 is automatically initialized.
//
// jemalloc 4.0.4 man: "If the specified arena was not initialized
// beforehand, it will be automatically initialized as a side effect of
// calling this interface."
for (arena=1; arena<narenas; arena++) {
if (je_mallctl("thread.arena", NULL, NULL, &arena, sizeof(arena)) != 0) {
chpl_internal_error("could not change current thread's arena");
}
}
// then set the current thread back to using arena 0
arena = 0;
if (je_mallctl("thread.arena", NULL, NULL, &arena, sizeof(arena)) != 0) {
chpl_internal_error("could not change current thread's arena back to 0");
}
}
size_t
mallctl_int64(const char* name, size_t* newval) {
size_t v = 0;
size_t len = sizeof(v);
if(newval) {
je_mallctl(name, &v, &len, newval, sizeof(size_t));
} else {
je_mallctl(name, &v, &len, NULL, 0);
}
// printf("name: %s, value: %zd\n", name, v);
return v;
}
void set_jemalloc()
{
//je_malloc_conf = "narenas:4";
int narenas = 0;
size_t sz = sizeof(narenas);
je_mallctl("opt.narenas", &narenas, &sz, nullptr, 0);
::wprintf(L"number of arenas: %d\n", narenas);
//je_malloc_stats_print(nullptr, nullptr, nullptr);
sz = sizeof(jemalloc_pre_allocated);
je_mallctl("stats.active", &jemalloc_pre_allocated, &sz, nullptr, 0);
}
// replace the chunk hooks for each arena with the hooks we provided above
static void replaceChunkHooks(void) {
size_t narenas;
size_t arena;
// set the pointers for the new_hooks to our above functions
chunk_hooks_t new_hooks = {
chunk_alloc,
null_dalloc,
null_commit,
null_decommit,
null_purge,
null_split,
null_merge
};
// for each arena, change the chunk hooks
narenas = get_num_arenas();
for (arena=0; arena<narenas; arena++) {
char path[128];
snprintf(path, sizeof(path), "arena.%zu.chunk_hooks", arena);
if (je_mallctl(path, NULL, NULL, &new_hooks, sizeof(chunk_hooks_t)) != 0) {
chpl_internal_error("could not update the chunk hooks");
}
}
}
int
mallctl_opt(const char* name, int* newval) {
int v = 0;
size_t len = sizeof(v);
if(newval) {
int ret = je_mallctl(name, &v, &len, newval, sizeof(int));
if(ret == 0) {
printf("set new value(%d) for (%s) succeed\n", *newval, name);
} else {
printf("set new value(%d) for (%s) failed: error -> %d\n", *newval, name, ret);
}
} else {
je_mallctl(name, &v, &len, NULL, 0);
}
return v;
}
void je_malloc_enable() {
jemalloc_postfork_parent();
if (malloc_disabled_tcache) {
je_mallctl("thread.tcache.enabled", NULL, NULL,
&malloc_disabled_tcache, sizeof(malloc_disabled_tcache));
}
pthread_mutex_unlock(&malloc_disabled_lock);
}
// get the number of arenas
static size_t get_num_arenas(void) {
size_t narenas;
size_t sz;
sz = sizeof(narenas);
if (je_mallctl("opt.narenas", &narenas, &sz, NULL, 0) != 0) {
chpl_internal_error("could not get number of arenas from jemalloc");
}
return narenas;
}
int zmalloc_get_allocator_info(size_t *allocated,
size_t *active,
size_t *resident) {
uint64_t epoch = 1;
size_t sz;
*allocated = *resident = *active = 0;
/* Update the statistics cached by mallctl. */
sz = sizeof(epoch);
je_mallctl("epoch", &epoch, &sz, &epoch, sz);
sz = sizeof(size_t);
/* Unlike RSS, this does not include RSS from shared libraries and other non
* heap mappings. */
je_mallctl("stats.resident", resident, &sz, NULL, 0);
/* Unlike resident, this doesn't not include the pages jemalloc reserves
* for re-use (purge will clean that). */
je_mallctl("stats.active", active, &sz, NULL, 0);
/* Unlike zmalloc_used_memory, this matches the stats.resident by taking
* into account all allocations done by this process (not only zmalloc). */
je_mallctl("stats.allocated", allocated, &sz, NULL, 0);
return 1;
}
void je_malloc_disable() {
static pthread_once_t once_control = PTHREAD_ONCE_INIT;
pthread_once(&once_control, je_malloc_disable_init);
pthread_mutex_lock(&malloc_disabled_lock);
bool new_tcache = false;
size_t old_len = sizeof(malloc_disabled_tcache);
je_mallctl("thread.tcache.enabled",
&malloc_disabled_tcache, &old_len,
&new_tcache, sizeof(new_tcache));
jemalloc_prefork();
}
void as_join(int id) {
if (as_flags[id] != 0) {
log_gas("address space %d already joined\n", id);
return;
}
const chunk_hooks_t *hooks = _hooks[id];
// If there aren't any custom hooks set for this space, then the basic local
// allocator is fine, which means that we don't need any special flags for
// this address space.
if (!hooks) {
log_gas("no custom allocator for %d, using local\n", id);
return;
}
// Create an arena that uses the right hooks.
unsigned arena;
size_t sz = sizeof(arena);
dbg_check( je_mallctl("arenas.extend", &arena, &sz, NULL, 0) );
char path[128];
snprintf(path, 128, "arena.%u.chunk_hooks", arena);
dbg_check( je_mallctl(path, NULL, NULL, (void*)hooks, sizeof(*hooks)) );
// // Disable dirty page purging for this arena
// snprintf(path, 124, "arena.%u.lg_dirty_mult", arena);
// ssize_t i = -1;
// dbg_check( je_mallctl(path, NULL, NULL, (void*)&i, sizeof(i)) );
// Create a cache.
unsigned cache;
sz = sizeof(cache);
dbg_check( je_mallctl("tcache.create", &cache, &sz, NULL, 0) );
// And set the flags.
as_flags[id] = MALLOCX_ARENA(arena) | MALLOCX_TCACHE(cache);
}
int main(void) {
/* Check how many areans we have */
unsigned narenas;
size_t len = sizeof(narenas);
if (je_mallctl("arenas.narenas", &narenas, &len, NULL, 0) != 0) {
fprintf(stderr, "Failed to get narenas");
return 1;
}
if (narenas != 1) {
fprintf(stderr, "Incorrect number of arenas: %d != 1\n", narenas);
return 2;
}
return 0;
}
void
stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
const char *opts)
{
int err;
uint64_t epoch;
size_t u64sz;
bool general = true;
bool merged = true;
bool unmerged = true;
bool bins = true;
bool large = true;
/*
* Refresh stats, in case mallctl() was called by the application.
*
* Check for OOM here, since refreshing the ctl cache can trigger
* allocation. In practice, none of the subsequent mallctl()-related
* calls in this function will cause OOM if this one succeeds.
* */
epoch = 1;
u64sz = sizeof(uint64_t);
err = je_mallctl("epoch", &epoch, &u64sz, &epoch, sizeof(uint64_t));
if (err != 0) {
if (err == EAGAIN) {
malloc_write("<jemalloc>: Memory allocation failure in "
"mallctl(\"epoch\", ...)\n");
return;
}
malloc_write("<jemalloc>: Failure in mallctl(\"epoch\", "
"...)\n");
abort();
}
if (opts != NULL) {
unsigned i;
for (i = 0; opts[i] != '\0'; i++) {
switch (opts[i]) {
case 'g':
general = false;
break;
case 'm':
merged = false;
break;
case 'a':
unmerged = false;
break;
case 'b':
bins = false;
break;
case 'l':
large = false;
break;
default:;
}
}
}
malloc_cprintf(write_cb, cbopaque,
"___ Begin jemalloc statistics ___\n");
if (general) {
int err;
const char *cpv;
bool bv;
unsigned uv;
ssize_t ssv;
size_t sv, bsz, ssz, sssz, cpsz;
bsz = sizeof(bool);
ssz = sizeof(size_t);
sssz = sizeof(ssize_t);
cpsz = sizeof(const char *);
CTL_GET("version", &cpv, const char *);
malloc_cprintf(write_cb, cbopaque, "Version: %s\n", cpv);
CTL_GET("config.debug", &bv, bool);
malloc_cprintf(write_cb, cbopaque, "Assertions %s\n",
bv ? "enabled" : "disabled");
#define OPT_WRITE_BOOL(n) \
if ((err = je_mallctl("opt."#n, &bv, &bsz, NULL, 0)) \
== 0) { \
malloc_cprintf(write_cb, cbopaque, \
" opt."#n": %s\n", bv ? "true" : "false"); \
}
#define OPT_WRITE_SIZE_T(n) \
if ((err = je_mallctl("opt."#n, &sv, &ssz, NULL, 0)) \
== 0) { \
malloc_cprintf(write_cb, cbopaque, \
" opt."#n": %zu\n", sv); \
}
#define OPT_WRITE_SSIZE_T(n) \
if ((err = je_mallctl("opt."#n, &ssv, &sssz, NULL, 0)) \
== 0) { \
malloc_cprintf(write_cb, cbopaque, \
" opt."#n": %zd\n", ssv); \
}
#define OPT_WRITE_CHAR_P(n) \
if ((err = je_mallctl("opt."#n, &cpv, &cpsz, NULL, 0)) \
== 0) { \
malloc_cprintf(write_cb, cbopaque, \
" opt."#n": \"%s\"\n", cpv); \
}
malloc_cprintf(write_cb, cbopaque,
"Run-time option settings:\n");
OPT_WRITE_BOOL(abort)
OPT_WRITE_SIZE_T(lg_chunk)
OPT_WRITE_CHAR_P(dss)
OPT_WRITE_SIZE_T(narenas)
OPT_WRITE_SSIZE_T(lg_dirty_mult)
OPT_WRITE_BOOL(stats_print)
OPT_WRITE_BOOL(junk)
OPT_WRITE_SIZE_T(quarantine)
OPT_WRITE_BOOL(redzone)
OPT_WRITE_BOOL(zero)
OPT_WRITE_BOOL(utrace)
OPT_WRITE_BOOL(valgrind)
OPT_WRITE_BOOL(xmalloc)
OPT_WRITE_BOOL(tcache)
OPT_WRITE_SSIZE_T(lg_tcache_max)
OPT_WRITE_BOOL(prof)
OPT_WRITE_CHAR_P(prof_prefix)
OPT_WRITE_BOOL(prof_active)
OPT_WRITE_SSIZE_T(lg_prof_sample)
OPT_WRITE_BOOL(prof_accum)
OPT_WRITE_SSIZE_T(lg_prof_interval)
OPT_WRITE_BOOL(prof_gdump)
OPT_WRITE_BOOL(prof_final)
OPT_WRITE_BOOL(prof_leak)
#undef OPT_WRITE_BOOL
#undef OPT_WRITE_SIZE_T
#undef OPT_WRITE_SSIZE_T
#undef OPT_WRITE_CHAR_P
malloc_cprintf(write_cb, cbopaque, "CPUs: %u\n", ncpus);
CTL_GET("arenas.narenas", &uv, unsigned);
malloc_cprintf(write_cb, cbopaque, "Arenas: %u\n", uv);
malloc_cprintf(write_cb, cbopaque, "Pointer size: %zu\n",
sizeof(void *));
CTL_GET("arenas.quantum", &sv, size_t);
malloc_cprintf(write_cb, cbopaque, "Quantum size: %zu\n", sv);
CTL_GET("arenas.page", &sv, size_t);
malloc_cprintf(write_cb, cbopaque, "Page size: %zu\n", sv);
CTL_GET("opt.lg_dirty_mult", &ssv, ssize_t);
if (ssv >= 0) {
malloc_cprintf(write_cb, cbopaque,
"Min active:dirty page ratio per arena: %u:1\n",
(1U << ssv));
} else {
malloc_cprintf(write_cb, cbopaque,
"Min active:dirty page ratio per arena: N/A\n");
}
if ((err = je_mallctl("arenas.tcache_max", &sv, &ssz, NULL, 0))
== 0) {
malloc_cprintf(write_cb, cbopaque,
"Maximum thread-cached size class: %zu\n", sv);
}
if ((err = je_mallctl("opt.prof", &bv, &bsz, NULL, 0)) == 0 &&
bv) {
CTL_GET("opt.lg_prof_sample", &sv, size_t);
malloc_cprintf(write_cb, cbopaque,
"Average profile sample interval: %"PRIu64
" (2^%zu)\n", (((uint64_t)1U) << sv), sv);
CTL_GET("opt.lg_prof_interval", &ssv, ssize_t);
if (ssv >= 0) {
malloc_cprintf(write_cb, cbopaque,
"Average profile dump interval: %"PRIu64
" (2^%zd)\n",
(((uint64_t)1U) << ssv), ssv);
} else {
malloc_cprintf(write_cb, cbopaque,
"Average profile dump interval: N/A\n");
}
}
CTL_GET("opt.lg_chunk", &sv, size_t);
malloc_cprintf(write_cb, cbopaque, "Chunk size: %zu (2^%zu)\n",
(ZU(1) << sv), sv);
}
if (config_stats) {
size_t *cactive;
size_t allocated, active, mapped;
size_t chunks_current, chunks_high;
uint64_t chunks_total;
size_t huge_allocated;
uint64_t huge_nmalloc, huge_ndalloc;
CTL_GET("stats.cactive", &cactive, size_t *);
CTL_GET("stats.allocated", &allocated, size_t);
CTL_GET("stats.active", &active, size_t);
CTL_GET("stats.mapped", &mapped, size_t);
malloc_cprintf(write_cb, cbopaque,
"Allocated: %zu, active: %zu, mapped: %zu\n",
allocated, active, mapped);
malloc_cprintf(write_cb, cbopaque,
"Current active ceiling: %zu\n", atomic_read_z(cactive));
/* Print chunk stats. */
CTL_GET("stats.chunks.total", &chunks_total, uint64_t);
CTL_GET("stats.chunks.high", &chunks_high, size_t);
CTL_GET("stats.chunks.current", &chunks_current, size_t);
malloc_cprintf(write_cb, cbopaque, "chunks: nchunks "
"highchunks curchunks\n");
malloc_cprintf(write_cb, cbopaque,
" %13"PRIu64" %12zu %12zu\n",
chunks_total, chunks_high, chunks_current);
/* Print huge stats. */
CTL_GET("stats.huge.nmalloc", &huge_nmalloc, uint64_t);
CTL_GET("stats.huge.ndalloc", &huge_ndalloc, uint64_t);
CTL_GET("stats.huge.allocated", &huge_allocated, size_t);
malloc_cprintf(write_cb, cbopaque,
"huge: nmalloc ndalloc allocated\n");
malloc_cprintf(write_cb, cbopaque,
" %12"PRIu64" %12"PRIu64" %12zu\n",
huge_nmalloc, huge_ndalloc, huge_allocated);
if (merged) {
unsigned narenas;
CTL_GET("arenas.narenas", &narenas, unsigned);
{
VARIABLE_ARRAY(bool, initialized, narenas);
size_t isz;
unsigned i, ninitialized;
isz = sizeof(bool) * narenas;
xmallctl("arenas.initialized", initialized,
&isz, NULL, 0);
for (i = ninitialized = 0; i < narenas; i++) {
if (initialized[i])
ninitialized++;
}
if (ninitialized > 1 || unmerged == false) {
/* Print merged arena stats. */
malloc_cprintf(write_cb, cbopaque,
"\nMerged arenas stats:\n");
stats_arena_print(write_cb, cbopaque,
narenas, bins, large);
}
}
}
if (unmerged) {
unsigned narenas;
/* Print stats for each arena. */
CTL_GET("arenas.narenas", &narenas, unsigned);
{
VARIABLE_ARRAY(bool, initialized, narenas);
size_t isz;
unsigned i;
isz = sizeof(bool) * narenas;
xmallctl("arenas.initialized", initialized,
&isz, NULL, 0);
for (i = 0; i < narenas; i++) {
if (initialized[i]) {
malloc_cprintf(write_cb,
cbopaque,
"\narenas[%u]:\n", i);
stats_arena_print(write_cb,
cbopaque, i, bins, large);
}
}
}
}
}
malloc_cprintf(write_cb, cbopaque, "--- End jemalloc statistics ---\n");
}
int main()
{
#if defined(CRT_LFH)
::wprintf(L"crt lfh\n");
set_lfh();
#elif defined(TBBMALLOC)
::wprintf(L"tbbmalloc\n");
#elif defined(TCMALLOC)
::wprintf(L"tcmalloc\n");
#elif defined(JEMALLOC)
::wprintf(L"jemalloc\n");
set_jemalloc();
#elif defined(MS_CONCURRENCY)
::wprintf(L"ms concurrency\n");
#elif defined(MY_ALLOCATOR)
::wprintf(L"my allocator\n");
if (allocator.create(true, 0) == false)
{
::wprintf(L"allocator.create() failed.\n");
return 0;
}
#endif
// for crt tbb tcmalloc
char* dummy = new char[10];
SecureZeroMemory(dummy, 10);
::wprintf(L"start memory info\n");
print_memory_info();
HANDLE thread_handle[thread_count] = { nullptr, };
ULONGLONG start = ::GetTickCount64();
for (std::size_t i = 0; i < thread_count; ++i)
{
unsigned int id = 0;
thread_handle[i] = reinterpret_cast< HANDLE >(::_beginthreadex(nullptr, 0, memory_alloc_test_thread_func, nullptr, 0, &id));
if (thread_handle[i] == nullptr)
{
::wprintf(L"%llu: _beginthreadex failed.\n", i);
}
}
::WaitForMultipleObjects(thread_count, thread_handle, TRUE, INFINITE);
ULONGLONG run_time = ::GetTickCount64() - start;
::wprintf(L"run time: %llu\n", run_time);
for (std::size_t i = 0; i < thread_count; ++i)
{
if (thread_handle[i] != nullptr)
{
::CloseHandle(thread_handle[i]);
thread_handle[i] = nullptr;
}
}
delete[] dummy;
dummy = nullptr;
::wprintf(L"end memory info\n");
print_memory_info();
#if defined(MY_ALLOCATOR)
print_allocator_stat(allocator.get_statistics());
allocator.destroy();
#elif defined(JEMALLOC)
{
//je_malloc_stats_print(nullptr, nullptr, nullptr);
size_t sz = sizeof(jemalloc_post_allocated);
je_mallctl("stats.active", &jemalloc_post_allocated, &sz, nullptr, 0);
size_t leaked = jemalloc_post_allocated - jemalloc_pre_allocated;
::wprintf(L"\nDone. Leaked: %zd bytes\n", leaked);
bool failed = leaked > 65536; // in case C++ runtime allocated something (e.g. iostream locale or facet)
::wprintf(L"\nTest %s!\n", (failed ? L"FAILED" : L"successful"));
}
#endif
return 0;
}
/// Launch the Trinity server
extern int main(int argc, char** argv)
{
///- Command line parsing to get the configuration file name
char const* cfg_file = _TRINITY_CORE_CONFIG;
int c = 1;
while (c < argc)
{
if (!strcmp(argv[c], "-c"))
{
if (++c >= argc)
{
printf("Runtime-Error: -c option requires an input argument");
usage(argv[0]);
return 1;
}
else
cfg_file = argv[c];
}
#ifdef _WIN32
if (strcmp(argv[c], "-s") == 0) // Services
{
if (++c >= argc)
{
printf("Runtime-Error: -s option requires an input argument");
usage(argv[0]);
return 1;
}
if (strcmp(argv[c], "install") == 0)
{
if (WinServiceInstall())
printf("Installing service\n");
return 1;
}
else if (strcmp(argv[c], "uninstall") == 0)
{
if (WinServiceUninstall())
printf("Uninstalling service\n");
return 1;
}
else
{
printf("Runtime-Error: unsupported option %s", argv[c]);
usage(argv[0]);
return 1;
}
}
if (strcmp(argv[c], "--service") == 0)
WinServiceRun();
#endif
++c;
}
if (!sConfigMgr->LoadInitial(cfg_file))
{
printf("Invalid or missing configuration file : %s\n", cfg_file);
printf("Verify that the file exists and has \'[worldserver]' written in the top of the file!\n");
return 1;
}
TC_LOG_INFO("server.worldserver", "Using configuration file %s.", cfg_file);
TC_LOG_INFO("server.worldserver", "Using SSL version: %s (library: %s)", OPENSSL_VERSION_TEXT, SSLeay_version(SSLEAY_VERSION));
TC_LOG_INFO("server.worldserver", "Using ACE version: %s", ACE_VERSION);
#ifdef HAVE_JEMALLOC
{
char const *jmVersion;
size_t jmVersionSize = sizeof(jmVersion);
je_mallctl("version", &jmVersion, &jmVersionSize, NULL, 0);
TC_LOG_INFO("server.worldserver", "Using Jemalloc: %s", jmVersion);
}
#endif
///- and run the 'Master'
/// @todo Why do we need this 'Master'? Can't all of this be in the Main as for Realmd?
int ret = sMaster->Run();
// at sMaster return function exist with codes
// 0 - normal shutdown
// 1 - shutdown at error
// 2 - restart command used, this code can be used by restarter for restart Trinityd
return ret;
}
size_t as_bytes_per_chunk(void) {
size_t log2_bytes_per_chunk = 0;
size_t sz = sizeof(log2_bytes_per_chunk);
je_mallctl("opt.lg_chunk", &log2_bytes_per_chunk, &sz, NULL, 0);
return (1lu << log2_bytes_per_chunk);
}