Re: [Maxima-discuss] Large memory allocations on Linux & WSL (Windows Subsystem for Linux)

[Raymond T. <toy...@gm...>]

On 5/21/23 14:07, Henry Baker wrote:
> Hi all:
>
> I want to be able to test some extremely large Maxima calculations,
> so I've been playing with my computer to see how large a memory
> space a program could utilize before it started paging itself to death.
>
> I just ran some tests on my current laptop with 16GB of RAM running
> WSL2 (Windows Subsystem for Linux) on top of Windows 11.
>
> My tests involved little more than a "Hello World" C program which also
> did a large malloc and then initialized the data.
>
> Actually, I used 'aligned_alloc' instead of 'malloc', but I expect that they
> both do roughly the same things.
>
> I was able to allocate a single object up to 2^32 *bytes* long (i.e., 4GB),
> and this object initialized properly and quite quickly (i.e., no paging).
>
> However, when I tried 2^33 bytes long (i.e., 8GB), the program works, but it
> pages like crazy.  But I have 10GB of RAM free, so there should be plenty
> of room for an 8GB object completely in RAM in a trivial program like this.
>
> I checked 'prlimit -l', and both the hard & soft limits seem to allow 131072
> pages of 'memlock'ed (non-pageable) memory.
>
> However 'ulimit -l' reports a limit of only 65536 pages of memlock'ed
> memory. Even after fiddling with /etc/security/limits.conf, I still couldn't
> increase the amount that 'ulimit -l' returns.
>
> I'm currently not running a pure Linux on this machine, so I can't tell if
> the problem is Linux (Ubuntu) itself, or whether the problem is WSL.
>
> Any thoughts ?
I don't really know, but if you send me the code, I can run it on a 
Linux system with 16GB or 32 GB of memory if you want to see if there's 
a difference.