Re: Memory fragmentation?

Jean-Marc Desperrier wrote:
> As a result, when you call VirtualFree, you can decommit any amount of 
> memory you want inside the VirtualAlloc block, but you can free only the 
> whole block. When you decommit memory with VirtualFree, you must 
> remember the area to recommit it later by giving it's address to 
> VirtualAlloc (with the flag MEM_COMMIT). VirtualAlloc with the flag 
> MEM_RESERVE will never give you back the address of that memory.

A little more :
There's also the MEM_RESET flag which can be used to say to the OS the 
data in the page is not important, so don't store it to the paging file.

It's not very clear what this does it, but my understanding is this has 
the gain of memory decommitting, without having to do an OS call before 
using the memory again.

But the memory will very certainly still "appear" as owned by the 
proccess. Some testing would be needed also to check how decommited 
memory is reported WRT freed memory.

I've found a codeproject sample that plays with all that to release 
thread stack memory:
http://www.codeproject.com/cpp/StackShrink.asp?df=100&forumid=253775&exp=0&select=1376692

BTW : To determine the size of a page and the allocation granularity on 
the host computer, use the GetSystemInfo function.
http://msdn2.microsoft.com/en-us/library/ms724381.aspx
The code project sample shows how to do the call and get the info.