Re: Performance Related Question
Frederico Costa writes: > And thanks for the suggestions. > I have now tested with -j option and i can confirm that my > expectations are correct the Dual CPU dual core AMD completes the > buildworld with -j4 in one hour only, while the intal core 2 does it > in 1h30m The ideal values for that parameter vary considerably with the balance between number of cores, number of virtual cores, memory (amount and speed) and disk throughput. If you want to optimize it, you'd need to experiment. But it's not worth it; you can easily spend hours shaving a couple of minutes off of your system build time. > should i stick the -j option in the make.conf? No. It sometimes causes problems with the install targets, and besides, it makes the build output very confusing (so when you have a problem, you always want to run without it). ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
Hi... And thanks for the suggestions. I have now tested with -j option and i can confirm that my expectations are correct the Dual CPU dual core AMD completes the buildworld with -j4 in one hour only, while the intal core 2 does it in 1h30m should i stick the -j option in the make.conf? Fred On 2013-02-27 22:58, Warren Block wrote: On Wed, 27 Feb 2013, Frederico Costa wrote: On 2013-02-27 22:27, Michael Ross wrote: If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Yes, i just made "make buildworld". So i should use make -j2 on the S1(dual core) and -j4 on S2 (2xdualcore)? And it also makes sense what you say about the I/O. It really depends on the system. On my dual-core systems, I use devel/ccache and found that -j8 gave the best performance. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
On 2/27/2013 4:53 PM, Michael Ross wrote: On a multi-CPU machine using an SMP configured kernel, try values between 6 and 10 and see how they speed things up. But you also do need to consider memory usage. On the areas of buildworld that are CPU intensive, they can also be memory intensive. If the active processes need to get swapped out to disk, you can wipe out any performance gain. I've noticed this some with clang. I don't care that clang uses more memory, "compile once, run many", but be aware of it if you're benchmarking. The ram amount can also influence cache sizes, and you have a major difference in memory amounts. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
On Wed, 27 Feb 2013, Frederico Costa wrote: On 2013-02-27 22:27, Michael Ross wrote: If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Yes, i just made "make buildworld". So i should use make -j2 on the S1(dual core) and -j4 on S2 (2xdualcore)? And it also makes sense what you say about the I/O. It really depends on the system. On my dual-core systems, I use devel/ccache and found that -j8 gave the best performance. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
On Wed, 27 Feb 2013 23:38:34 +0100, Frederico Costa wrote: On 2013-02-27 22:27, Michael Ross wrote: If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Yes, i just made "make buildworld". So i should use make -j2 on the S1(dual core) and -j4 on S2 (2xdualcore)? And it also makes sense what you say about the I/O. i will start another to see the results. Maybe try higher settings. Handbook ( http://www.freebsd.org/doc/handbook/makeworld.html ) says: However, since much of the compiling process is I/O bound rather than CPU bound, it is also useful on single CPU machines. On a typical single-CPU machine, run: # make -j4 buildworld make(1) will then have up to 4 processes running at any one time. Empirical evidence posted to the mailing lists shows this generally gives the best performance benefit. On a multi-CPU machine using an SMP configured kernel, try values between 6 and 10 and see how they speed things up. Thanks fred On Wed, 27 Feb 2013 23:05:44 +0100, Frederico Costa wrote: Hi everyone... I have a kind of interesting question when comes to performance of FreeBSD in different HW. i am not trying to come up with a scientific reason for measuring performance. :-) It is just a curiosity, and of course to see if i understand it and improve performance of my systems. i am running 2 systems at the moment, lets call them S1 and S2, running FreeBSD 9.1-RELEASE-p1 amd64: S1: Intel Core2 Duo E6550 @ 2.33GHz 2GB RAM 500GB disk (not important probably just for reference) S2: 2x Dual-Core AMD Opteron 2216 2.4GHz 14GB Ram 320GB disk (not important probably just for reference) Both the systems are running more or less the same sw, apache, imap server, postfix, and the needed perl/php/python and running very light load. Also both are using a GENERIC kernel and not running X, they are just text based :-) From cpubenchmark.net the cpu performance index are for s1: 1501 and s2: 1518, so very similar. As i felt the AMD system seemed slower when comes to compiling, i just done a "performance test" which was "make buildworld" on both of systems from scratch and the times are: S1: 2h 12m S2: 2h 59m If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Regards, Michael ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
On 2013-02-27 22:27, Michael Ross wrote: If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Yes, i just made "make buildworld". So i should use make -j2 on the S1(dual core) and -j4 on S2 (2xdualcore)? And it also makes sense what you say about the I/O. i will start another to see the results. Thanks fred On Wed, 27 Feb 2013 23:05:44 +0100, Frederico Costa wrote: Hi everyone... I have a kind of interesting question when comes to performance of FreeBSD in different HW. i am not trying to come up with a scientific reason for measuring performance. :-) It is just a curiosity, and of course to see if i understand it and improve performance of my systems. i am running 2 systems at the moment, lets call them S1 and S2, running FreeBSD 9.1-RELEASE-p1 amd64: S1: Intel Core2 Duo E6550 @ 2.33GHz 2GB RAM 500GB disk (not important probably just for reference) S2: 2x Dual-Core AMD Opteron 2216 2.4GHz 14GB Ram 320GB disk (not important probably just for reference) Both the systems are running more or less the same sw, apache, imap server, postfix, and the needed perl/php/python and running very light load. Also both are using a GENERIC kernel and not running X, they are just text based :-) From cpubenchmark.net the cpu performance index are for s1: 1501 and s2: 1518, so very similar. As i felt the AMD system seemed slower when comes to compiling, i just done a "performance test" which was "make buildworld" on both of systems from scratch and the times are: S1: 2h 12m S2: 2h 59m If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Regards, Michael ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Re: Performance Related Question
On Wed, 27 Feb 2013 23:05:44 +0100, Frederico Costa wrote: Hi everyone... I have a kind of interesting question when comes to performance of FreeBSD in different HW. i am not trying to come up with a scientific reason for measuring performance. :-) It is just a curiosity, and of course to see if i understand it and improve performance of my systems. i am running 2 systems at the moment, lets call them S1 and S2, running FreeBSD 9.1-RELEASE-p1 amd64: S1: Intel Core2 Duo E6550 @ 2.33GHz 2GB RAM 500GB disk (not important probably just for reference) S2: 2x Dual-Core AMD Opteron 2216 2.4GHz 14GB Ram 320GB disk (not important probably just for reference) Both the systems are running more or less the same sw, apache, imap server, postfix, and the needed perl/php/python and running very light load. Also both are using a GENERIC kernel and not running X, they are just text based :-) From cpubenchmark.net the cpu performance index are for s1: 1501 and s2: 1518, so very similar. As i felt the AMD system seemed slower when comes to compiling, i just done a "performance test" which was "make buildworld" on both of systems from scratch and the times are: S1: 2h 12m S2: 2h 59m If I read you right, you didn't ``make -jX buildworld'', with X being the number of processes to spawn, so you used just one core on either machine. Buildworld does a lot of I/O, so disk speed is relevant. Regards, Michael ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"
Performance Related Question
Hi everyone... I have a kind of interesting question when comes to performance of FreeBSD in different HW. i am not trying to come up with a scientific reason for measuring performance. :-) It is just a curiosity, and of course to see if i understand it and improve performance of my systems. i am running 2 systems at the moment, lets call them S1 and S2, running FreeBSD 9.1-RELEASE-p1 amd64: S1: Intel Core2 Duo E6550 @ 2.33GHz 2GB RAM 500GB disk (not important probably just for reference) S2: 2x Dual-Core AMD Opteron 2216 2.4GHz 14GB Ram 320GB disk (not important probably just for reference) Both the systems are running more or less the same sw, apache, imap server, postfix, and the needed perl/php/python and running very light load. Also both are using a GENERIC kernel and not running X, they are just text based :-) From cpubenchmark.net the cpu performance index are for s1: 1501 and s2: 1518, so very similar. As i felt the AMD system seemed slower when comes to compiling, i just done a "performance test" which was "make buildworld" on both of systems from scratch and the times are: S1: 2h 12m S2: 2h 59m My mind tells me that the S2 system should be faster, just because there are 2 CPU instead 1, or 4 cores and of course more RAM. But the "smaller" Intel seems to beat the crap of the AMD. Is this expected? Should i expect the S2 should be way faster? And just trying to get some advice, is there any tweaks i can do on the S2 system to make it go faster? Like i said i was expecting the S2 to be faster, but of course maybe i am wrong? Any advice/feedback will be appreciated, as i am just trying to understanding and if possible improve performance. Thanks in advance Fred ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "freebsd-questions-unsubscr...@freebsd.org"