UNIX for Advanced & Expert Users

How can you make a program as Memory resident in AIX.
If I make a program as a memory resident program whether all the parts of the program like code and data (stack) segements of the program will be loaded in to the Memory.

For Ex:
I have a C code which is creating array of 10000 long ints and will be initialized with some data. Whether this array will be there in the memory along with the code or not.

Thanks a lot for your time and information.

Thanks and Regads,
Pola Balaji

Short answer: yes the data will be there in memory. The entire array.

By memory-resident do you mean a daemon?

In this context, I think memory-resident might mean a process which cannot be swapped out to disk.

I checked with Mr. Google, and found this link which seems to confirm my thoughts.

Yeah, I think he is talking about memory locking. If physical memory is in short supply, some of that array may get paged out...but not if it is locked. I don't use AIX, but I can make some comments about unix in general. Processes can be locked into physical memory, programs cannot. If the process exits or execs most locks are released. The system calls are:

plock()
This was the original system call. It is not in the Posix Standard and it is showing its age. It knows about two memory segments: text and data. In the old days, the stack was in the data segment. The docs tiptoe around what DATLOCK really means today. PROCLOCK is supposed to be the entire process with one possible exception. Today a process can grow. Will new pages be locked? More tapdancing. And see below for System V shared memory concerns. Do not use this call for new code.

mlockall(), munlockall()
Part of the Posix REALTIME stuff. A flag controls whether or not future pages will lock.

mlock(), munlock()
A optional extention to Posix REALTIME. Individual memory pages are locked and unlocked.

Shared memory mapped into the process via mmap() is subject to the above Posix calls. An interesting question is the status of memory created by shm_open() prior to a mmap() call. Another interesting question is what if two processes are disagreeing over the status of a shared memory segment. Posix does not seem to spell these out.

Posix reluctantly mentions System V shared memory obtained via shmget() and shmat() and this is the nasty exception to the various rules. Usually these are not subject to the Posix locking calls. Instead, shmctl() controls the locking. Also a System V shared memory segment can continue to exist after all processes that had it attached have exited. And it can continue to be locked into core in this state. None of this is acknowledged by Posix.

Yes, now that I think about it, it sound reasonable that an atomic IPC like semaphores could be used.