Re: SunOS 4.1.3 or OS/MP?

" From: Craig Dewick <[email protected]>
" 
" []
" 
" > Was it possible to run OS/MP on Sun machines?
" 
" No because at the time Sun did not have any SMP capability (OS/MP had it 
" well before Sun did). In fact Sun didn't get proper SMP until well into 
" the SunOS 5.x OS period. SunOS 5.4 I think might have been when it first 
" appeared (aka Solaris 2[.4]).

don't forget ross and their hypersparc modules!  they had dual-proc
mbus modules early on, and had a sunos4 patch early on, which was iirc
pretty crude - a massive spinlock about the kernel.  sun incorporated
it into 4.1.4, when they were selling hypersparc-optioned ss20s.

the patch was intended for hypersparc support, but i can tell you dual
supersparc ss20s ran just fine under 4.1.4 too.  mid '90s again, and i
was working at a place that did heavy fpga-based design.  the sim
tools would completely eat one cpu during runs, which lasted several
hrs - but they were single-threaded.  an uptime of 90+ days under this
kind of stress is pretty clear...  the 2nd cpu didn't accelerate a sim
but meant a fairly responsive system for other work during a sim, and
because 4.1.4 already incorporated the patch, all we had to do was
plug it in.

iirc os/mp's kernel locks were much more fine grained than ross's, so
in theory it should've performed better on the same system, if there
ever was one that could run both os/mp and 4.1.4.  otoh 4.3bsd
[sunos4] was not afaik originally designed for smp as sVr4 [solaris2]
was, so solbourne's work involved some fairly heavy lifting and my own
experience is that they never did get it running quite right.

btw os/mp 4.1c = sunos 4.1.3.  was there ever a 4.1d, or did grumman
happen instead?
________________________________________________________________________
Andrew Hay                                  the genius nature
internet rambler                            is to see what all have seen
[email protected]                       and think what none thought