Re: [gentoo-dev] limiting memory and mathematica

[Denis Shcherbakov]

Matt,

That's the thing.  It uses like 2% CPU and shows like 90% MEM in that
table.  So that is a real problem for me.

In /etc/limits, which option sets the max memory use?  Is it max stack
size, or max address space?  I have no idea which option I should choose -
I am just a layman :)

Thanks
Denis


On Tue, 9 Mar 2004, Matthew Kennedy wrote:
>
>It sounds like MathKernel uses an over-commit strategy (like what
>CMUCL and SBCL Common Lisp implementations on GNU/Linux use).  They
>are not really using that much memory.  They over-commit in order to
>have a linear address space in which to implement their own memory
>management schemes.
>
>For example, CMUCL on Gentoo, looks like this in the top(1) output:
>
>$ top
>
>top - 15:43:40 up 21 days,  5:47,  1 user,  load average: 1.03, 0.74, 0.36
>Tasks: 111 total,   2 running, 108 sleeping,   0 stopped,   1 zombie
>Cpu(s): 97.4% us,  2.3% sy,  0.0% ni,  0.0% id,  0.0% wa,  0.3% hi,  0.0% si
>Mem:    514180k total,   497184k used,    16996k free,    50088k buffers
>Swap:  1004052k total,   122512k used,   881540k free,   225592k cached
>
>  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
>22549 mkennedy  25   0 1284m 8244  22m R  1.0  1.6   5:24.32 lisp
>
>
>You should look in the %MEM column for the "real" measure.
>
>However, you wrote "graphic applications came to a screeching halt" --
>over-committing would not explain that though (presumably very high
>CPU usage on Mathematica's behalf??).  You might want to try Linux
>2.6.x kernels which provide much improved desktop responsiveness.
>
>Matt
>
>--
>Matthew Kennedy
>Gentoo Linux Developer
>

--
gentoo-dev@g.o mailing list