depmod memory usage.

depmod memory usage.

[Dave Jones]

I pulled out an older laptop last night to test something, and built
a kernel on it.  When I did the make install part, the machine almost
ground to a halt.

It turned out that depmod used at peak 1.7GB of memory, (the laptop only had 1GB,
so it was swapping to the point of locking up interactivity for minutes
at a time).  It took an hour to complete.

The config file is pretty pathological (it's a distro build, so tons of
modules -- http://codemonkey.org.uk/junk/fedora-config-x86-64)

It's been a while since I built on a machine with this amount of memory,
but it seems that depmod has taken a big step backwards as it used to
cope with this just fine.

This is with the m-i-t that shipped in Fedora 16 (3.16)

	Dave

Re: depmod memory usage.

[Jon Masters]

On Fri, 2011-11-18 at 15:35 -0500, Dave Jones wrote:
> I pulled out an older laptop last night to test something, and built
> a kernel on it.  When I did the make install part, the machine almost
> ground to a halt.
> 
> It turned out that depmod used at peak 1.7GB of memory, (the laptop only had 1GB,
> so it was swapping to the point of locking up interactivity for minutes
> at a time).  It took an hour to complete.
> 
> The config file is pretty pathological (it's a distro build, so tons of
> modules -- http://codemonkey.org.uk/junk/fedora-config-x86-64)
> 
> It's been a while since I built on a machine with this amount of memory,
> but it seems that depmod has taken a big step backwards as it used to
> cope with this just fine.
> 
> This is with the m-i-t that shipped in Fedora 16 (3.16)

I'll do some digging at the distro specific side of things. It's odd
because I've got a couple of netbooks and embedded systems with *less*
memory than 1GB that run depmod just fine. There were some memory leaks
in older versions that should not be a problem any more.

Jon.