Re: Xcode 1.5 distributed build test results
Re: Xcode 1.5 distributed build test results
- Subject: Re: Xcode 1.5 distributed build test results
- From: Scott Tooker <email@hidden>
- Date: Wed, 25 Aug 2004 21:19:13 -0700
The reason using the local machine is a bad idea is that all
preprocessing of files is done locally.
Scott
On Aug 25, 2004, at 11:22 AM, Dave Camp wrote:
I sent this out internally in my company and thought I would share it
here as well. I figured someone else might like to see some real world
numbers.
Build a medium sized project - 125 cpp files. Project has no external
dependencies. All tests run twice to check for anomalous results.
All machines in the test were Dual 2.0 G5's with 1 Gig of RAM. All
machines were connected via a 24 port gigabit switch.
The first set of tests were building the project only on the selected
machine.
Dave(2) 55.3s
Monroe(2) 48.5s
Mick(2) 45.0s
Jim(2) NA
Interestingly, it's always faster to build on someone else's machine
than yours. I'm assuming that is because there is other Xcode stuff
going on at the same time that would be competing for CPU time with
the gcc processes.
I think the gerbils in Jim's machine are dead. It took 2 minutes to
compile the first two files. I gave up at that point. This problem was
present yesterday as well. Jim's machine has been dropped from the
remaining tests...
Next up were the combined machine tests.
Dave(2) + Monroe(2) 36s 35% faster than my machine
Dave(2) + Mick(2) 37s About the same, as expected
Monroe(2) + Mick(2) 32s 42% faster...
Dave(2) + Monroe(2) + Mick(2) 33s Not helping
As expected, adding an external box to the mix yields a nice speed
boost. Using just external boxes is even better. Adding both of your
CPU's into the mix adds negative value. Again, I think this indicates
that there are other Xcode activities competing with the two gcc
processes.
Final test using a single local processor and distributing builds.
Dave(1) + Monroe(2) 32s
Dave(1) + Mick(2) 31s
Dave(1) + Monroe(2) + Mick(2) 26s Woot! 53% faster than a
local-only compile.
This set seems to prove that combining external boxes with a single
local CPU yields the best results. I also observed that Both of my
CPUs were pegged, so the other CPU must have been busy with build
overhead.
When we figure out why the fourth G5 is not playing well with the
others I'll run the remaining tests with it to see if there is further
improvement.
Dave
---
The path of least resistance, it's not just for electricity any more.
_______________________________________________
xcode-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/xcode-users
Do not post admin requests to the list. They will be ignored.
_______________________________________________
xcode-users mailing list | email@hidden
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/xcode-users
Do not post admin requests to the list. They will be ignored.