• Open Menu Close Menu
  • Apple
  • Shopping Bag
  • Apple
  • Mac
  • iPad
  • iPhone
  • Watch
  • TV
  • Music
  • Support
  • Search apple.com
  • Shopping Bag

Lists

Open Menu Close Menu
  • Terms and Conditions
  • Lists hosted on this site
  • Email the Postmaster
  • Tips for posting to public mailing lists
Re: Distributed build Issues
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Distributed build Issues


  • Subject: Re: Distributed build Issues
  • From: Israel Brewster <email@hidden>
  • Date: Tue, 11 Apr 2006 09:38:33 -0800

Watching the paging activity, I'm hardly seeing any on the iMac- during a compile, it was mostly sitting at zero in/out, with an occasional spike up to about 75 max on the page ins,nothing on the page outs. Overall, my iMac currently has an uptime of 13 days, and a total of only 287846/89690 page ins/outs. If there was any significant memory pressure I would think those numbers should be much higher, especially after 13 days runtime. The iMac does have a Gig of RAM, so only the Powermac is RAM hungry. Responsiveness doesn't seem noticeably affected either- I tried doing several normal tasks such as checking e-mail, playing music, browsing the web, etc during the compile, and all seemed as responsive as normal.

One thing I noticed: in reading through the distcc man page, it specifies that when making the host list, localhost should generally be listed first, followed by any other machines you want to distribute to, i.e setenv DISTCC_HOSTS localhost PowerMac-G4.local". I noticed, however, that when Xcode sets the DISTCC_HOSTS environment variable, at least on my system, it is always "setenv DISTCC_HOSTS PowerMac-G4.local./4" Now depending on how the build system worked, I could see it only setting a single host for each file, and then just varying the host on a per-file basis, but on my machine it never varies. Course, it does have that /4 on the end, which may accomplish the same thing (networking isn't my strongest area)

One final thing is that over on the apple discussion boards where I initially posted this question, they are now saying that this is expected behavior, unless the iMac is also set to be shared. Specifically, that as the primary machine, the iMac would ONLY be used for linking, precompiled headers, etc, and not for the actual compilation. This doesn't sound right to me, or if it is right I don't think it should be, as it would be vasty under-utilizing the available computing power (at least in my situation), but I thought I'd mention it just in case.

Thanks again for your help and attention :)
On Apr 11, 2006, at 8:03 AM, Bill Bumgarner wrote:

On Apr 10, 2006, at 7:32 PM, Israel Brewster wrote:
Yeah, top typically shows 60-70% idle on the iMac while compiling, aside from occasional spikes associated with linking or the like.

OK -- now what about I/O and memory pressure? An idle CPU does not necessarily indicate an underutilized machine (though I suspect that is the case).


top -u should give an idea of pageins/pageouts. For I/O, I usually find it useful to more judge the responsiveness of the machine when I try to use it for normal tasks.

Clearly, we need to do a better job of using the local CPU when doing distributed compilation. Yet, our testing has shown that too much use of local resources can very quickly kill the usability of the local machine. Worse, it can make distributed compilation performance suffer greatly as the local machine is no longer able to fill the remote queue.

b.bum


_______________________________________________ Do not post admin requests to the list. They will be ignored. Xcode-users mailing list (email@hidden) Help/Unsubscribe/Update your Subscription: This email sent to email@hidden
  • Follow-Ups:
    • Re: Distributed build Issues
      • From: j o a r <email@hidden>
References: 
 >Distributed build Issues (From: Israel Brewster <email@hidden>)
 >Re: Distributed build Issues (From: William Bumgarner <email@hidden>)
 >Re: Distributed build Issues (From: Israel Brewster <email@hidden>)

  • Prev by Date: Re: TechNote for building Universal Binary...
  • Next by Date: Debugger stepping into incorrect files on intel
  • Previous by thread: Re: Distributed build Issues
  • Next by thread: Re: Distributed build Issues
  • Index(es):
    • Date
    • Thread