[Gllug] VM talk (was: Linux 2.4.10 is out with better VM)

[Matt Amos]

> If anyone wants to argue Matt's point above, compare FreeBSD (based on
> BSD/Mach VM) with Linux (based on optimised for x86 VM) under heavy load.
> I'd take FreeBSD every time. 

i was not evangelising the linux VM. it is, as you say, a mess. the UVM
model is superior the linux VM model, and performs better. optimisation
starts in the algorithm and ends in the code, not the other way around.

> Remember, heavy load is where you actually
> really NEED the optimisations. Theres no point saving the odd CPU cycle
> here and there if a CPU is 99% idle. 

unless, of course, you care about latency. but i dont think anyone running
linux cares about latency that much, or theyd be using something else.

> And under high memory load, the
> biggest bottleneck is the disk anyway, so choosing the right pages for
> swapping is vastly more important than how fast they were chosen. The
> Linux problems are mainly the wrong pages being chosen.

it was mot my intention to say that everything ought to be written in
assembly. this is not the case, demonstrably. however, it is always
desirable to optimise for specific architectures, especially if they are
in widespread use. by "optimise" i dont necessarily mean "write in
assembly", but using special hardware or features available in the
platform ISA can increase performance radically. 

for a case in point look at L3/L4 (http://l4ka.org/) and liedtke's papers.
these are not about the VM per se, but the ideas can be extended to cover
every aspect of systems software.

cya,

matt


-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug