Subject: Re: MicroVax 3400 performance with NetBSD 1.5.1_BETA again
To: Anders Magnusson <ragge@ludd.luth.se>
From: Saku Setala <setala@elisa.fi>
List: port-vax
Date: 04/27/2001 15:20:09
Hi again,

thanks for the hint. Seems that the setting is reset during booting, but I tried
it on-the-fly and it seems to work better;

--clip--
% make
Running dhrystone (Level 1 optimization, without registers)
./dhrystone
Dhrystone(1.1) time for 50000 passes = 35
This machine benchmarks at 1422 dhrystones/second
./dhrystone

~
?02 EXT HLT
        PC = 00001DD7
>>>D/I 25 F0
>>>cont

^C% make
Running dhrystone (Level 1 optimization, without registers)
./dhrystone
Dhrystone(1.1) time for 50000 passes = 20
This machine benchmarks at 2472 dhrystones/second
./dhrystone
Dhrystone(1.1) time for 50000 passes = 20
This machine benchmarks at 2469 dhrystones/second

--clip--


On Fri, 27 Apr 2001, Anders Magnusson wrote:

> Try to do:
> 
> >>>D/I 25 F0
> 
> before you load the kernel, and see if it runs any faster.
> 
> The KA640 do not have any L2 cache, but it has a L1 cache which
> I think I miss to enable.
> 
> -- Ragge
> 
> > Hi Ragge,
> > 
> > some documents say that it should have 1K on-chip cache.
> > 
> > Anyway, the same document says KA630 is 0.9 VUPS and KA640 is 2.4 VUPS.
> > 
> > Now I get almost similar speed results with dhrystone 1.1, so there must
> > be something funny?
> > 
> > On Fri, 27 Apr 2001, Anders Magnusson wrote:
> > 
> > > > Hi again,
> > > > 
> > > > could it be so that NetBSD does not enable the CPU cache on KA640?
> > > > 
> > > The KA640 doesn't have a cache, so hopefully NetBSD do not try to
> > > turn it on :-)
> > > 
> > > -- Ragge
> > > 
> > 
> > --
> > Saku Setälä
> > System Planning Manager
> > Kolumbus Oy 	http://www.kolumbus.com/
> > Elisa Communications Group
> > 
> > 
> 
> 

--
Saku Setälä
System Planning Manager
Kolumbus Oy 	http://www.kolumbus.com/
Elisa Communications Group