diff options
Diffstat (limited to 'Objects/obmalloc.c')
-rw-r--r-- | Objects/obmalloc.c | 63 |
1 files changed, 45 insertions, 18 deletions
diff --git a/Objects/obmalloc.c b/Objects/obmalloc.c index 3916262120..9cd6a50466 100644 --- a/Objects/obmalloc.c +++ b/Objects/obmalloc.c @@ -2,6 +2,13 @@ #ifdef WITH_PYMALLOC +#ifdef HAVE_MMAP + #include <sys/mman.h> + #ifdef MAP_ANONYMOUS + #define ARENAS_USE_MMAP + #endif +#endif + #ifdef WITH_VALGRIND #include <valgrind/valgrind.h> @@ -75,7 +82,8 @@ static int running_on_valgrind = -1; * Allocation strategy abstract: * * For small requests, the allocator sub-allocates <Big> blocks of memory. - * Requests greater than 256 bytes are routed to the system's allocator. + * Requests greater than SMALL_REQUEST_THRESHOLD bytes are routed to the + * system's allocator. * * Small requests are grouped in size classes spaced 8 bytes apart, due * to the required valid alignment of the returned address. Requests of @@ -107,10 +115,11 @@ static int running_on_valgrind = -1; * 57-64 64 7 * 65-72 72 8 * ... ... ... - * 241-248 248 30 - * 249-256 256 31 + * 497-504 504 62 + * 505-512 512 63 * - * 0, 257 and up: routed to the underlying allocator. + * 0, SMALL_REQUEST_THRESHOLD + 1 and up: routed to the underlying + * allocator. */ /*==========================================================================*/ @@ -139,14 +148,17 @@ static int running_on_valgrind = -1; * small enough in order to use preallocated memory pools. You can tune * this value according to your application behaviour and memory needs. * + * Note: a size threshold of 512 guarantees that newly created dictionaries + * will be allocated from preallocated memory pools on 64-bit. + * * The following invariants must hold: - * 1) ALIGNMENT <= SMALL_REQUEST_THRESHOLD <= 256 + * 1) ALIGNMENT <= SMALL_REQUEST_THRESHOLD <= 512 * 2) SMALL_REQUEST_THRESHOLD is evenly divisible by ALIGNMENT * * Although not required, for better performance and space efficiency, * it is recommended that SMALL_REQUEST_THRESHOLD is set to a power of 2. */ -#define SMALL_REQUEST_THRESHOLD 256 +#define SMALL_REQUEST_THRESHOLD 512 #define NB_SMALL_SIZE_CLASSES (SMALL_REQUEST_THRESHOLD / ALIGNMENT) /* @@ -174,15 +186,15 @@ static int running_on_valgrind = -1; /* * The allocator sub-allocates <Big> blocks of memory (called arenas) aligned * on a page boundary. This is a reserved virtual address space for the - * current process (obtained through a malloc call). In no way this means - * that the memory arenas will be used entirely. A malloc(<Big>) is usually - * an address range reservation for <Big> bytes, unless all pages within this - * space are referenced subsequently. So malloc'ing big blocks and not using - * them does not mean "wasting memory". It's an addressable range wastage... + * current process (obtained through a malloc()/mmap() call). In no way this + * means that the memory arenas will be used entirely. A malloc(<Big>) is + * usually an address range reservation for <Big> bytes, unless all pages within + * this space are referenced subsequently. So malloc'ing big blocks and not + * using them does not mean "wasting memory". It's an addressable range + * wastage... * - * Therefore, allocating arenas with malloc is not optimal, because there is - * some address space wastage, but this is the most portable way to request - * memory from the system across various platforms. + * Arenas are allocated with mmap() on systems supporting anonymous memory + * mappings to reduce heap fragmentation. */ #define ARENA_SIZE (256 << 10) /* 256KB */ @@ -440,6 +452,9 @@ static poolp usedpools[2 * ((NB_SMALL_SIZE_CLASSES + 7) / 8) * 8] = { , PT(48), PT(49), PT(50), PT(51), PT(52), PT(53), PT(54), PT(55) #if NB_SMALL_SIZE_CLASSES > 56 , PT(56), PT(57), PT(58), PT(59), PT(60), PT(61), PT(62), PT(63) +#if NB_SMALL_SIZE_CLASSES > 64 +#error "NB_SMALL_SIZE_CLASSES should be less than 64" +#endif /* NB_SMALL_SIZE_CLASSES > 64 */ #endif /* NB_SMALL_SIZE_CLASSES > 56 */ #endif /* NB_SMALL_SIZE_CLASSES > 48 */ #endif /* NB_SMALL_SIZE_CLASSES > 40 */ @@ -525,6 +540,8 @@ new_arena(void) { struct arena_object* arenaobj; uint excess; /* number of bytes above pool alignment */ + void *address; + int err; #ifdef PYMALLOC_DEBUG if (Py_GETENV("PYTHONMALLOCSTATS")) @@ -577,8 +594,15 @@ new_arena(void) arenaobj = unused_arena_objects; unused_arena_objects = arenaobj->nextarena; assert(arenaobj->address == 0); - arenaobj->address = (uptr)malloc(ARENA_SIZE); - if (arenaobj->address == 0) { +#ifdef ARENAS_USE_MMAP + address = mmap(NULL, ARENA_SIZE, PROT_READ|PROT_WRITE, + MAP_PRIVATE|MAP_ANONYMOUS, -1, 0); + err = (address == MAP_FAILED); +#else + address = malloc(ARENA_SIZE); + err = (address == 0); +#endif + if (err) { /* The allocation failed: return NULL after putting the * arenaobj back. */ @@ -586,6 +610,7 @@ new_arena(void) unused_arena_objects = arenaobj; return NULL; } + arenaobj->address = (uptr)address; ++narenas_currently_allocated; #ifdef PYMALLOC_DEBUG @@ -1054,7 +1079,11 @@ PyObject_Free(void *p) unused_arena_objects = ao; /* Free the entire arena. */ +#ifdef ARENAS_USE_MMAP + munmap((void *)ao->address, ARENA_SIZE); +#else free((void *)ao->address); +#endif ao->address = 0; /* mark unassociated */ --narenas_currently_allocated; @@ -1775,7 +1804,6 @@ _PyObject_DebugMallocStats(void) * will be living in full pools -- would be a shame to miss them. */ for (i = 0; i < maxarenas; ++i) { - uint poolsinarena; uint j; uptr base = arenas[i].address; @@ -1784,7 +1812,6 @@ _PyObject_DebugMallocStats(void) continue; narenas += 1; - poolsinarena = arenas[i].ntotalpools; numfreepools += arenas[i].nfreepools; /* round up to pool alignment */ |