News for Linux/Mingo:
Mingo has just done the greatest scheduler patch I could have wished
for. I applied his stuff to 2.1.132, and here is what I noticed
immediately:
interactive response time is *instant* even with 4 compute bound threads
running. If I press the return key, I get prompts at the same speed as
the keyboard typamatic rate, which is something like 30/sec. Before
it was *very* sluggish.
mouse curser movement is *perfect*. IE it behaves under heavy load just
like it does with all 4 processors idle.
Processor affinity seems perfect also. IE I ran a single thread for
several minutes... early on it bounced once or twice as it was doing
some console I/O, but after either one or two bounces, it "stuck" on a
specific cpu and *stayed* there.
I say we give Mingo a raise. :) And that these patches get merged into
2.1.133 since they aren't huge (a couple of printk's could be removed as
they might cause questions, such as "why is cpu1 waiting for commence?, I
told it to 'commence' when I hit the power on switch..." :)
In any case, I hereby announce my plans to guard this scheduler code with
my life, and will personally fit it to newer systems as they come out if
it is deemed to be a 2.3 issue. It is _very_ good. And it meets my goal
of good processor performance, plus the goal (as stated by Linus but
seconded by myself and others) of also not neglecting good interactive
performance.
Now of course, once I hit the send butting, the damned thing will crash
and I'll look like an idiot... but hey, I 'm used to looking like an idiot
anyway. :)
*good work* mingo. this was a lot easier than trying to find out where
the boot was hanging by stuffing 0xff into video ram to make little white
blocks.. :)
Now another question for anyone using an intel 450NX based machine: I
have three 9.1 gig LVDS ultra-2 wide disks (10k) from IBM. If I boot
the machine normally and let the bios go out and do its normal scsi bus
scan, it reports 80mb/sec but the boot hangs with lots of "timeout,
resetting, etc." If I disable the bios scsi bus scan, the system boots,
but the ncr53c8xx driver then reports that the disks are fast-20
(40mb/sec) drives. I am using the stock NCR driver. Do I need something
newer? or is this normal/ok to see fast-20 from the driver rather than
fast-40? I note that comments in the driver mentionthe 895 version of the
scsi controller, but not the 896 version that is in this machine...
but I'll gladly accept this glitch since my machine now performs
interactively like a quad-xeon should...
Bob
Robert Hyatt Computer and Information Sciences
[EMAIL PROTECTED] University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
