Date: Mon, 24 Nov 97 00:17:38 +0100 From: Ben Stuyts <benst@terminus.stuyts.nl> To: freebsd-smp@freebsd.org Subject: Re: make -j8 of kernel fails Message-ID: <199711232317.AAA06591@daneel.stuyts.nl> In-Reply-To: <3478A761.6B553EF7@orbis.net> References: <199711232023.VAA06477@daneel.stuyts.nl> <3478A761.6B553EF7@orbis.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 23 Nov 1997, "Keith A. Fredericks" wrote: > Ben Stuyts wrote: > > > A few days ago I reported having trouble getting the kernel to compile with > > make -j4. Thank you to all of you who pointed out to me that I had to do a > > make depend first. > > > > It works now with make -j2, make -j4, make -j6 but not with make -j8: > > I get the same thing if I do not do the make depend before the make -j 8. > It works if I do make depend right before the make -j 8. I just tried that, and you are right: that fixes it. I was only doing a make clean in between the various make -j # tests. I guess that wasn't sufficient. > I am using a dual Ppro on a PR440FX: > > time make -j 8 > > 193.583u 93.758s 3:08.57 152.3% 431+671k 90+9401io 0pf+0w Here it takes: 296.867u 103.535s 3:33.06 187.9% 533+753k 349+1033io 39pf+0w Unloaded system otherwise. (Well, rc5 is running in the background...) There's not that much difference in wall time. I guess it must be i/o limited. I'll have to go experiment with -async mounts. Which is similar to my -j4 and -j6 times. I guess there really is nothing to gain from anything higher than -j4 for kernel compiles. If I want to use -j4 on a make world, do I have to do a make depend first? Anything else? I remember a discussion some time ago here that the make world system wasn't set up for a parallel make. Is this still the case? Best regards, Ben
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199711232317.AAA06591>