Alec Ten Harmsel wrote: > On Sat, Jun 15, 2019, at 14:19, Walter Dnes wrote: >> On Thu, Jun 13, 2019 at 05:15:41PM +0300, Alexey Eschenko wrote >>> Thank you. Didn't think about that. Don't know why though. My >>> MAKEOPTS was "-j32". Looks like that was too many for package like >>> qtwebengine. Solved the problem with creating specific environment >>> for qtwebengine and setting it up in /etc/portage/package.env/ It was >>> veeeery long build process but this time it finished successfully. I >>> think I'll try to find more appropriate value for qtwebengine which >>> could be used with 32GB of RAM. >> Please use plain text, not HTML. >> >> According to >> https://blogs.gentoo.org/ago/2013/01/14/makeopts-jcore-1-is-not-the-best-optimization/ >> the fastest compiles come with setting MAKEOPTS to the number of cores >> in your machine. E.g. for a dual core cpu, use "-j2", for a 4 core cpu >> use "-j4", etc. To check the number of cpus in your machine, execute... >> >> grep -c ^flags /proc/cpuinfo > That's a good rule but not necessarily always true. My old machine was an > i7-3930K (6 cores, 12 threads) w/ 32G RAM. I had /var/tmp on tmpfs. I > benchmarked firefox, chromium, and some other big projects once and -j13 was > consistently the fastest on that box. > > As that blog post says: > >> I’m just saying, ${core} + 1 is not the best optimization for me >> and the test confirms the part:“but this guideline isn’t always perfect” > Depends on available RAM, how fast your disk is, etc. > > Alec > >
As a AMD CPU user, I always set mine to number of cores plus one. That seems to always be the most efficient for me. I have portages work directory on tmpfs for all but a couple large packages. I used to make exceptions for Firefox, Seamonkey, Libreoffice and qtweb something or other. Since I upgraded my memory to 32GBs I removed all but Libreoffice. My biggest problem was when more than one of those wanted to compile at the same time. The best way to know what to set it to, test it. Set it to cores plus one and test. Then set it to half the number of cores, twice the number of cores etc until you find that sweet spot. One could even do that during normal updates although it may not be as accurate. I suspect if ten people with ten different systems tested this, we'd get half a dozen different results, maybe more. The best thing in my opinion, start emerge, go to bed and then hope it is done when you wake up. lol Dale :-) :-)
