James Cornell wrote:

>SMP is a must on UltraSparc platforms.  Not only are they 64-bit, but they generally have 2mb or more of level 2 data cache.  Sparc64's are a beefy Risc processor and deserve a little SMP lovin'.  I look forward to seeing SMP on NetBSD for Sparc64 platforms.  In the future I hope to see a formal release of SMP for Sparc64 in the next release of NetBSD.    Take a look around... Redhat 6.2's 2.2.x kernel has had SMP for several years.  The Gentoo distribution with a newer kernel built for Ultras also does SMP.  What my point in referencing SMP support in linux kernels is that all in all it is very possible.  
>
There is this slight problem, though, of two things:  SMP with sparc on 
Linux is actually in most cases slower than the same application running 
on a single processor and (2) SMP on Linux, in general, is quite poor.  
So I'd love to have SMP, but of NetBSD quality, certainly not of Linux 
quality.

>SMP has been a great feature for many platforms on NetBSD for some time now, but due to some amazingly annoying and trivial problems with various NetBSD ports in the past, no one has had time to even attempt to implement SMP support natively in the NetBSD kernel for Sparc64 processors.  Once my machine is complete, I will personally donate hardware time to NetBSD kernel developers if they can promise some progress towards multiple cpu support in NetBSD for Sparc64.  It wouldn't be illegal to reference the methods in which Linux kernels use SMP on the UltraSparc platform.  Don't use the code, just figure out what's missing and implement it based on existing NetBSD methods to avoid license restraints between the GPL and BSD licenses.  Just my two cents.  Thank you for your interest.
>
>James Cornell
>Unix Devil - Free Unix Shells
>http://www.unixdevil.org
>
>!DSPAM:421ba435207706058112152!
>
>  
>