best way to allocate memory in C for parallel programming openmp and mpi

Question

If one were to write C code, not C++, and were to use openmp directives to make a parallel program then what is the best way to allocate memory? I am used to the traditional malloc() and calloc() but i have also come across

#include <sys/sem.h>

for doing semaphores. Is one better or worse than the other?

The program i am considering writing would be using large amounts of RAM, on the order of 256GB or more. And it would also be reading/writing scratch files greater than 4GB. Therefore i would not be using the 32-bit version of the lseek() command but would use lseek64() . With all that in mind, can someone offer advice for the best way to write such a program? I had learned openmp a long time ago, and that generally will suffice for running on a single system having enough memory, but I mentioned MPI because i would be ok learning and using that instead if it makes everything more robust and portable. Or some other method of parallel programming in C.

Answer 1

考虑使用mmap和mremap来更快地重新映射内存，因为mremap（）使用Linux页表方案。