Hilco Wijbenga posted on Sat, 12 Nov 2011 16:05:36 -0800 as excerpted:
> By the way, is there a noticeable difference in build time (for
> relatively large builds) when logging to the console is off? Not
> important, just curious.
There can be. AFAIK it's not too bad when output is to a vgacon text-
console, but to frame-buffer (kms mode text console or X) tends to be
more CPU intensive and to slow things down if you're bottlenecking on CPU
as is often (but not always, especially on multi-core systems without
emerge --jobs set to allow multiple parallel package emerges at once, as
well as MAKEOPTS=-j, to allow multiple make jobs with a single merge job)
the case. Terminal windows tend to use what a text console framebuffer
does, often more, depending on effects and how much is hardware
Perhaps the biggest efficiency gain, however, if you have 4 gig plus of
memory (and a 64-bit system to handle it reasonably efficiently), is to
point PORTAGE_TMPDIR at a tmpfs and control its size and the number of
parallel jobs to avoid more than a few MB of swappage. Based on my
experience here, with four or even at two cores, the load average doesn't
bog down the system anywhere close to what disk I/O does, regardless of
whether it's swap or conventional disk I/O access. Putting the temp-dir
in tmpfs means a lot of scratch files are never written to disk at all,
thus saving that disk access, and theoretically it could even save if
you're into swap (forums or user list for the details), but obviously far
less then, so I now set jobs and load average to best control, if
indirectly, for memory usage including the tmpfs, as opposed to direct
CPU load control.
But there's whole threads on optimizing emerges on the forums and user
list. This really isn't the place for further discussion on that.
Duncan - List replies preferred. No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master." Richard Stallman