=== shinmen [n=shinmen@nat1.inalambrica.net] has left #ubuntu-ports ["Ex-Chat"] | ||
=== braddr_ drools over a gcc build with gmake -j 32.. and actually using all 16 cpus for a lot of the build. | ||
fabbione | braddr_: it's much more fun if you just do make -j | 07:58 |
---|---|---|
braddr_ | heh | 07:58 |
braddr_ | I figured 2x cpu's is probably good enough.. but yes, I'll give that a whirl. | 07:59 |
braddr_ | this is the first build I've done on the box, and it's having to do a bunch of configure scripts run, really limiting the parallelism. | 07:59 |
fabbione | i can imagine | 08:00 |
braddr_ | after it finishes, a make clean; make ought to get much more opportunity | 08:00 |
braddr_ | any word from davem? I assume he's still wrestling gdb and mono to the ground. | 08:01 |
fabbione | he is fighting with a bad cold | 08:01 |
fabbione | so he is not hacking much around | 08:02 |
=== braddr_ nods.. "I know the feeling.. knock on wood, it's been a while since my last illness." | ||
braddr_ | 370 processes: 213 sleeping, 139 running, 1 zombie, 17 on cpu | 08:37 |
braddr_ | yes, unrestricted make -j is kinda fun | 08:37 |
fabbione | braddr_: when we started testing Niagara, davem was soft with -j32 | 08:42 |
fabbione | after i started with -j 512 we found all kind of smp bugs :P | 08:43 |
braddr_ | heh | 08:43 |
braddr_ | not suprisingly, solaris is holding up quite well under the load | 08:43 |
fabbione | we do too now | 08:43 |
braddr_ | I'm willing to be convinced. :) | 08:44 |
fabbione | i have been sort of bashed by my boss, because nagios woke up the sysadmins due to "a few machines complaining load over 1000" | 08:44 |
fabbione | (i was preparing a kernel security release a few months back) | 08:45 |
braddr_ | whee | 08:45 |
fabbione | well i was make -j on all arches at the same time | 08:45 |
fabbione | :) | 08:45 |
braddr_ | with that kinda load, you're wasting an awful lot of time in context swapping | 08:45 |
braddr_ | ok, maybe not a lot, but.. | 08:46 |
braddr_ | 11m build for gcc 4.0.3 for c,c++ | 08:46 |
braddr_ | pretty decent | 08:47 |
fabbione | point is that i was more waiting for I/O | 08:47 |
fabbione | since the kernel was already in ccache | 08:47 |
=== braddr_ nods. | ||
=== daq4th [n=darkness@netstation-005.cafe.zSeries.org] has joined #ubuntu-ports | ||
braddr_ | fabbione: re: that 16 cpu t2000.. I'll be keeping it rather than getting the 32 cpu version. Not worth the hassle to get the right model. | 11:06 |
braddr_ | so no worries about loosing the test case | 11:07 |
=== jb-home tries to imagine the noise from a 32 CPU sparc box... | ||
braddr_ | exactly the nose of a 16 cpu box.. in both of the cases for the t2000 series, its 1 physical core. | 11:09 |
braddr_ | er.. 1 physical unit, just a varying number of cores. | 11:09 |
braddr_ | the t2000 hardware is _nice_, from a physical layout standpoint. It's also nice from a performance standpoint. | 11:10 |
braddr_ | it's unfortunatly pricy from a suck you dry standpoint. | 11:10 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!