Load compile limit on CPU

[Mr-Biscuit]

I have a tendency to run make on my machines until the swap is being used.
The G3 is still building with six active processes. Before the amd64 died, I would run the same number of [make install clean]s. Just wondering- and a little bored while I wait- if anyone else does the same.

[jggimi]

OpenBSD view:

I used to do manual parallel builds -- primarily when building ports where a fetch of the software was part of it.

I no longer do this, because the fighting for resource (CPU/Disk/RAM) would take longer if the jobs were working instead of waiting for downloads.

Now, I use make -j x <target>, where x = my number of CPUs. For ports, and for kernel and userland builds. I don't use -j for X builds, nor for releases.

Is it supported? No. Does it work? For these limited uses, yes.

[Mr-Biscuit]

I think I'll try that statement out

make -j

when I get home tomorrow.

[TerryP]

using login classes (man login.conf), it is possible to define limits to how much resources a user may use (cpu, memory, open files, processes, etc). If necessary you can always define one or several of these, and use su to run make under it, i.e. $ su -c loginclass -c 'make -j4 target'. That can be handy if you need to keep compiling something within a given limit, but remember that it can also cause builds to fail.


On FreeBSD at least, the -j switch to make, can significantly decrease compile times for the kernel and base, but doesn't play well with ports at times. The # of processors (or cores) is often used, but I typically use -j4 to -j6 on my sempron uniprocessor and -j8 to -j12 on my dual cored pentium d. Tends to work quite well.