[TriLUG] kernel question

Matt Hoover Matt.Hoover at haht.com
Tue Jul 15 10:50:25 EDT 2003


Darn required work email client :-(, sorry about the 
HTML mail.

Also, one other thing that I forgot to say, a lot of
work goes into tuning spinlocks, so there are as
efficient as possible (read hand crafted assembly),
so the overhead is fairly low.
--matt

-----Original Message-----
From: Matt Hoover [mailto:Matt.Hoover at haht.com] 
Sent: Tuesday, July 15, 2003 10:44 AM
To: 'trilug at trilug.org'
Subject: RE: [TriLUG] kernel question


The good news is that a SMP kernel will work fine on a 
Uniprocessor system.  The bad news is that there is a 
performance hit.  
Now for those interested in the details, the reason for 
the performance hit is that the kernel has to do additional 
locking on SMP.  On a Uniprocessor system, the kernel can 
disable interrupts, and be assured that nothing else is going 
to modify memory.  But on a SMP system, another processor 
can modify memory.  So the solution is to add locks (often 
called spinlocks) for various parts of memory.  What happens 
is when a processor wants to access memory protected by a 
lock, it first gets the lock.  In order to get the lock, 
the processor goes into a tight loop until the lock is available.  
Once it has acquired the lock, it accesses the memory, and 
then releases the lock so other processors can get it.  Acquiring 
and releasing locks has a certain amount of overhead, even if 
there are no other processors making the current processor wait 
for the lock.  So in the Uniprocessor kernel, where spinlocks 
aren't needed, the kernel replaces the get and release spinlock 
code with no-ops, which get optimized away. 
One other thing to note is that this applies to the 2.4.x and 2.2.x 
series kernels.  In 2.5.x, and now 2.6.x, pre-emption support has 
been added to the kernel.  Therefore, spinlocks are needed even 
on a Uniprocessor system (if pre-emption has been enabled in the 
config) 
> -----Original Message----- 
> From: Greg Brown [mailto:gregbrown at mindspring.com] 
> Sent: Tuesday, July 15, 2003 8:31 AM 
> To: trilug at trilug.org 
> Subject: [TriLUG] kernel question 
> 
> 
> If I accidently included symmetric multi-processing support into my 
> kernel on a machine which has only one processor.  By including this 
> option do I face and kind of a performance hit on my single processor 
> machine?  Even though I did include this option built-into my 
> kernel I 
> did not compile the kernel with options telling it to compile for N 
> number of processors.  I'm running the 2.4.21 kernel.  If the 
> kernel is 
> fine as-is I'll just leave well enough alone until I upgrade 
> again - if 
> not, I'll recompile.  Thoughts?  Anyone?  Bueller?  Fry? 
> 
> Greg 
> 
> -- 
> TriLUG mailing list        : 
> http://www.trilug.org/mailman/listinfo/trilug 
> TriLUG 
> Organizational FAQ  : http://trilug.org/faq/ 
> TriLUG Member Services FAQ : http://members.trilug.org/services_faq/ 
> TriLUG PGP Keyring         : http://trilug.org/~chrish/trilug.asc 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.trilug.org/pipermail/trilug/attachments/20030715/ca3ab13d/attachment.html>


More information about the TriLUG mailing list