Memory/Swapping

[Michael Mattias]

When I went to memory-mapped files with absolute arrays to replace 'real' arrays, I got a HUGE reduction in the "Page Faults" number you see in the report..."Page Faults" is the counter of how many times the process requested memory which was currently swapped out and had to be swapped in.

I think there is a thread here with a discusssion of that... let me look....yes, here is the OLD forum page, which should be readable..

http://www.powerbasic.com/support/forums/Forum6/HTML/003994.html

You can see the page faults ("swap-in required") for the same test file data went from 1.1 million to just under 37,000... which is not too shabby if you ask me.

MCM

[Gert Voland]

Michael,

OK, works fine, I will test the call at different levels of memory allocation in
the project and then make a decision where to warn/inform the user.

Yes, I wondered why it all came in lower case ... Normally that is not your
style! I don't think anyone can expect you (or others) to re-post your (their)
code. Of course that effect de-values those contributions quite a bit because
not everyone is able to readily respond as you did just now.

Eventually I will also consider your suggestion to store the array data in a
disk file and memory-map it - after checking how often I get close to the
limits of installed memory.

Thanks again.

Regards, Gert

[Michael Mattias]

That's what happened when the forum was converted: a lot of posts ended up getting converted to all lower-case letters.

The token "getprocessmemoryinfo" should be "GetProcessMemoryInfo"

"Too bad we can't edit the posts if more than 15 days old" (he said, having only yesterday contacted the PBPTB about same).

Oh, well, Steve told me how to fix it, I have to send complete new text of the message to them and a 'forum adminstrator' can replace the post.

[Gert Voland]

Michael,

thanks for that piece of code. It compiles and runs on PBCC V4.04.
I get the following output:

process memory statistics for process id 976 on 11-14-2007 at 16:36:32

process memory statistics with array allocated
can't get address for getprocessmemoryinfo

process memory statistics with array de-allocated
can't get address for getprocessmemoryinfo

end of report

The relevant variables are:

szlibname = psapi.dll
szprocname = getprocessmemoryinfo
hlib = 1760755712
dwaddr = 0

Is this checking a none-existing process, or what am I doing wrong?

Regards, Gert.

[Michael Mattias]

BTW, you might do better in the long run that instead of using REDIM PRESERVE, storing your array data in a disk file and memory-mapping it. Then just re-map your view of the file and 're' REDIM AT instead of REDIM PRESERVE.

That way you are not doing constant re-allocation/de-allocation.

Memory Mapped Files instead of RANDOM disk file access 5-8-04

[Michael Mattias]

This is your lucky day:
Add Process Memory Usage Report to any program 1-12-04

(It don't get no simpler than TWO lines of code, one of them an #INCLUDE).

MCM

[Gert Voland]

Hi,

in a project I dynamically adjust the memory usage by REDIM PRESERVE of
some multi-dimensional arrays. The complexity increases up to 3^n, n being
the number of user variables. Since the algorithms predict only an upper limit
of needed memory, which is only reached sometimes, the adaptive REDIM has
proven to be a good method.

Now I wonder if there is a way to find/measure the momentary memory
allocation compared to the maximum of available non-swapped memory?
My goal would be to inform the user when swapping to disk starts
(or has just started), then causing a considerable slow-down in execution.

Does someone know a simple method for PBCC (+ Win32Api)?

Regards, Gert.