On Thu, Feb 4, 2010 at 2:02 PM, Eric Sproul <[email protected]> wrote:
> On 02/ 4/10 02:54 PM, William Yang wrote:
>> We currently have an M4000 (8 cores SPARC64-VI 2.1 GHz, 2 threads per core)
>> that we do our compiles on, and that's been able to compile almost as fast as
>> the identical software on an 8 core Intel X5460.
>
> I should be so lucky. :)  I think that's a *bit* out of my price range.
>
> Certainly sounds like some sort of parallel make is the way to go, though my
> current build box has only 2 cores (the V210) so it's not likely to make a
> difference there.

Many times setting the parallelism to a bit more than the number of
cores is beneficial because you can have CPU-intensive things running
while waing for I/O.  That is, with "gmake -j2" you may get a 1.7x
speedup and with "gmake -j 3" you may get a 2.1x speedup.  Your
experience will likely vary by the software you are building and the
overall performance characteristics of the build machine.

-- 
Mike Gerdts
http://mgerdts.blogspot.com/
_______________________________________________
sysadmin-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/sysadmin-discuss

Reply via email to