[mythtv-users] compiling performance question
mwedel at sonic.net
Tue Feb 12 06:12:06 UTC 2019
On 02/11/2019 09:04 AM, Jon Heizer wrote:
> On Mon, Feb 11, 2019 at 10:12 AM Tim Draper <veehexx at zoho.com
> <mailto:veehexx at zoho.com>> wrote:
> Recently upgraded from 29 to 30 but as of yet, rpm based distro's don't have
> a v30 in the repos. Cue git and self compiling. All ran with a simple
> ./configure, make, make install.
> 1) Compiled on 3 machines, with everything running SSD's with at least 4GB
> ram. CPU's & compile times as follows:
> Laptop - i5-5200u - 3minutes (used as a test machine)
> Server - Pentium G3240 - 20minutes (multipurpose, running Myth BE)
> intel Nuc - Celeron N2830 - 2h40mins (Primary FE)
> That does seem like a long time. In comparison my J4205 (low powered quad core
> no HT) chip has similar clock speeds but twice the cores. With 8GB of ram and a
> no name SATA SSD a refreshing git pull and clean through make install takes
> about 3min 30s. Are you passing a -j 3 or similar to the make?
With it being that long, my suspicion would be that the amount of jobs it is
running (plus other work it is doing) and the amount of ram the system has, is
that it is running out of memory and paging to disk.
Even with an SSD, paging is much slower than RAM access speeds. If you really
want to experiment, passing -j 1 to make may actually result in it being faster
- its only dong 1 compile at a time, but if it saves it from paging out to disk,
that might be an overall gain.
More information about the mythtv-users