PMEM_MEMMOVE_PERSIST(3) | PMDK Programmer's Manual | PMEM_MEMMOVE_PERSIST(3) |
pmem_memmove(), pmem_memcpy(), pmem_memset(), pmem_memmove_persist(), pmem_memcpy_persist(), pmem_memset_persist(), pmem_memmove_nodrain(), pmem_memcpy_nodrain(), pmem_memset_nodrain() - functions that provide optimized copying to persistent memory
#include <libpmem.h> void *pmem_memmove(void *pmemdest, const void *src, size_t len, unsigned flags); void *pmem_memcpy(void *pmemdest, const void *src, size_t len, unsigned flags); void *pmem_memset(void *pmemdest, int c, size_t len, unsigned flags); void *pmem_memmove_persist(void *pmemdest, const void *src, size_t len); void *pmem_memcpy_persist(void *pmemdest, const void *src, size_t len); void *pmem_memset_persist(void *pmemdest, int c, size_t len); void *pmem_memmove_nodrain(void *pmemdest, const void *src, size_t len); void *pmem_memcpy_nodrain(void *pmemdest, const void *src, size_t len); void *pmem_memset_nodrain(void *pmemdest, int c, size_t len);
pmem_memmove(), pmem_memcpy() and pmem_memset() functions provide the same memory copying as their namesakes memmove(3), memcpy(3) and memset(3), and ensure that the result has been flushed to persistence before returning (unless PMEM_F_MEM_NOFLUSH flag was used).
For example, the following code is functionally equivalent to pmem_memmove() (with flags equal to 0):
memmove(dest, src, len); pmem_persist(dest, len);
Calling pmem_memmove() may out-perform the above code, because libpmem(7) implementation may take advantage of the fact that pmemdest is persistent memory and use instructions such as non-temporal stores to avoid the need to flush processor caches.
WARNING: Using these functions where pmem_is_pmem(3) returns false may not do anything useful. Use libc functions in that case.
Unlike libc implementation, libpmem functions guarantee that if destination buffer address and length are 8 byte aligned then all stores will be performed using at least 8 byte store instructions. This means that a series of 8 byte stores followed by pmem_persist(3) can be safely replaced by a single call to one of the above functions.
The flags argument of all of the above functions has the same meaning. It can be 0 or a bitwise OR of one or more of the following flags:
/* ... write several ranges to pmem ... */ pmem_memcpy(pmemdest1, src1, len1, PMEM_F_MEM_NODRAIN); pmem_memcpy(pmemdest2, src2, len2, PMEM_F_MEM_NODRAIN); /* ... */ /* wait for any pmem stores to drain from HW buffers */ pmem_drain();
The remaining flags say how the operation should be done, and are merely hints.
Using an invalid combination of flags has undefined behavior.
Without any of the above flags libpmem will try to guess the best strategy based on size. See PMEM_MOVNT_THRESHOLD description in libpmem(7) for details.
pmem_memmove_persist() is an alias for pmem_memmove() with flags equal to 0.
pmem_memcpy_persist() is an alias for pmem_memcpy() with flags equal to 0.
pmem_memset_persist() is an alias for pmem_memset() with flags equal to 0.
pmem_memmove_nodrain() is an alias for pmem_memmove() with flags equal to PMEM_F_MEM_NODRAIN.
pmem_memcpy_nodrain() is an alias for pmem_memcpy() with flags equal to PMEM_F_MEM_NODRAIN.
pmem_memset_nodrain() is an alias for pmem_memset() with flags equal to PMEM_F_MEM_NODRAIN.
All of the above functions return address of the destination buffer.
After calling any of the functions with PMEM_F_MEM_NODRAIN flag you should not expect memory to be visible to other threads before calling pmem_drain(3) or any of the _persist functions. This is because on x86_64 those functions may use non-temporal store instructions, which are weakly ordered. See “Intel 64 and IA-32 Architectures Software Developer's Manual”, Volume 1, “Caching of Temporal vs. Non-Temporal Data” section for details.
memcpy(3), memmove(3), memset(3), libpmem(7) and <https://pmem.io>
2022-08-25 | PMDK - pmem API version 1.1 |