A memory pool implemented by C.
Sharing a same pool in multi-thread is not recommanded.
At first you should create a pool handler for your data (or structure).
xmem_pool_handler xmem_create_pool(unsigned int block_size);
xmem_pool_handeris only a pointer whichtypedefed at xmempool.h.typedef void* xmem_pool_handler;
For an example:
typedef struct stct {
int id;
char str[16];
} stct;
xmem_pool_handler pool1 = xmem_create_pool(sizeof(stct));
xmem_pool_handler pool2 = xmem_create_pool(sizeof(int));
if(!pool1 || !pool2) printf("Can't alloc more space\n");You can get a space of a certain size via function:
void* xmem_alloc(xmem_pool_handle handle);For an example:
stct* my_stct = xmem_alloc(pool1);
if(!my_stct) printf("Can't alloc more space.\n");And then you will get a whole empty space that fits stct.
After your scope you want to recover your memory, don't use free(). There's a function below:
int xmem_free(xmem_pool_handle handle, void* pointer);The return value:
0: invalid pool handler or the pointer not belongs to this pool, can't be recovered.1: succeeded.
For an example:
int result = xmem_free(pool1, my_stct);
if(!result) printf("Cannot be recovered.\n");void xmem_destroy_pool(xmem_pool_handle pool);Caution! This function will destroy the whole pool,
freeing all your memory allocated by that pool, even if you haven'txmem_freeyet.
xmem_destroy_pool(pool1);
xmem_destroy_pool(pool2);You can refer to xmempool.h for any more functions.
$ npm installInstall dependencies via command above and then type follow command:
$ node perf.jsPerformance of "./perf_xmem"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 88934466.62 CPU cycles
Average of free in one case: 34828792.76 CPU cycles
Avarage of total in one case: 139718323.94 CPU cycles
Performance of "./perf_alloc"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 144649249.46 CPU cycles
Average of free in one case: 269216562.16 CPU cycles
Avarage of total in one case: 413865891.36 CPU cycles
Effect rate: (malloc - xmempool) / malloc
Alloc: 38.52%
Free: 87.06%
Total: 66.24%
This report was ran on my MacBook Air (11-inch, Mid 2013).
- CPU: 1.3GHz Intel Core i5
- Memo: 4GB 1600 MHz DDR3
- System: OS X Yosemite 10.10.2
Performance of "./perf_xmem"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 88497438.3 CPU cycles
Average of free in one case: 38895793.32 CPU cycles
Avarage of total in one case: 145739467.06 CPU cycles
Performance of "./perf_alloc"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 112048273.8 CPU cycles
Average of free in one case: 40331425.12 CPU cycles
Avarage of total in one case: 152379754.92 CPU cycles
Effect rate: (malloc - xmempool) / malloc
Alloc: 21.02%
Free: 3.56%
Total: 4.36%
This report was ran on my work computer which is under Deepin.
- CPU: Dual Intel(R) Pentium(R) CPU G620 @ 2.60GHz
- Memo: 4GB 1600 MHz DDR3 x 2
- System: Deepin 2014 (3.13.0-30-generic)
Performance of "./perf_xmem"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 192807375.4 CPU cycles
Average of free in one case: 68994110.58 CPU cycles
Avarage of total in one case: 318962417.64 CPU cycles
Performance of "./perf_alloc"
Test cases: 50
Times of each case: 1000000
Average of alloc in one case: 197417483.18 CPU cycles
Average of free in one case: 59714845.4 CPU cycles
Avarage of total in one case: 257132389.86 CPU cycles
Effect rate: (malloc - xmempool) / malloc
Alloc: 2.34%
Free: -15.54%
Total: -24.05%
This report was ran on my own VPS which is under Ubuntu.
- CPU: Dual Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz
- Memo: 1G
- System: Ubuntu 12.04.2 LTS (3.5.0-23-generic)
You're welcome to pull requests!
「雖然我覺得不怎麼可能有人會關注我」