On Mon, Apr 17, 2017 at 08:04:51PM +0000, Christian Weisgerber wrote: > On 2017-04-13, Andreas Kusalananda Kähäri <andreas.kah...@icm.uu.se> wrote: > > > I have 2 CPUs, and dpb usually builds two ports at a time (or one port > > in parallel). However, while rebuilding llvm and ghc I observed that > > ghc was being built as one job while llvm clearly used a parallel build > > with two jobs at the same time. > > Yes, that can happen. > > > If dpb is allowed to do a parallel build of a package, does that mean > > that it'll do that regardless of whether another build is running? > > Yes. The assumption is that the other build will finish soon, and > that usually happens when building the full ports tree and on > machines with four or more cores.
Ah. > > > How should I interpret the '1*'? > > The number of build slots occupied. > Your machine has 2 build slots, occupied by ghc and 1*llvm. If ghc > finished ahead of llvm, you will see 2*llvm: llvm occupies both slots > and no new jobs are started until it has finished. > > Here's a snapshot from a setup of three machines times four cores: > > 17 Apr 14:02:29 [93925] running for 00:26:31 > 2*lang/gcc/4.9(build) [9292] on amd64-1 17% > 2*devel/boost(build) [71601] 33% > 2*devel/cmake(build) [96529] 27% > 2*graphics/cairo(build) [27962] on amd64-1 66% > lang/vala(build) [90156] on amd64-2 41% > www/libcroco(build) [26311] on amd64-2 39% > misc/gpsd(build) [14547] on amd64-2 21% > ~productivity/tryton/timesheet_cost(package) [5212] on amd64-2 27% > Hosts: amd64-1 [63886] amd64-2 [16323] localhost > I=636 B=194 Q=3986 T=5425 F=0 !=93 I was thinking dpb kept track of number of cores used in the build, but it uses job slots, and some of the jobs may be parallel. Ok. > > > This might have been the result of unlucky scheduling on dpb's part as > > both ports take a really long time to build... > > Exactly. > > -- > Christian "naddy" Weisgerber na...@mips.inka.de > > Thanks for the explanation! Regards, Kusalananda
