Text 6046, 91 rader
Skriven 2006-05-10 19:15:54 av Maurice Kinal (1:140/13.1)
   Kommentar till text 6044 av Kenneth Parrish (8:8/2)
Ärende: Re: sensors, pnpbios
============================
Hey Kenneth!

May 09 23:12 06, Kenneth Parrish wrote to Maurice Kinal:

 KP> absolutely. btw there's been a pretty good thread about flash memory 
 KP> on the linux-kernel@vger.. list. they say the wear-leveling 
 KP> algorithms aren't all the same and some are a bit better, and the 
 KP> larger the memory the longer it may tend to remain usable with 
 KP> repeated writes.

The only writes are the intitial install of ramdisk and kernel.  I don't need
to upgrade them that often and the logfiles are written to ramspace. 
Everything else, such as downloads, are to external drives.  A good flashdisk
will last longer then I do.

 KP> another message, more recent and in a different 
 KP> thread about caching data for faster application startup and booting,

The entire OS is in ramspace.  No need for caching.  There is more then enough
ram and if I use the concept for multimedia apps the external sata drive can
easily handle it ... and then some.
 
 KP> mentions that microsoft's new windows version might possibly use 
 KP> flash memory, optionally, to cache often-used data as an energy 
 KP> saver.

No need to.  ram is cheap and faster then any disk, flash or otherwise.  None
of this has been an issue here, not on the 486 and definetly not on the smp guy
... and now a p4 guy.  Only the intitial install requires writing to the
flashdisk.  It should last at least 10 years as per specs for data retention
and then it should be a simple matter of upgrading.  I figure if things go well
a good ide flash disk will outlast any computer and more then likely my
lifetime.  Writes are definetly not an issue with me and if they are I'll use
external drives.

 KP> i was thinking you've got the benefit of two processors level one 
 KP> cache, but linux may be helping with all the improvements to smp 
 KP> scheduling and perhaps other things.

Looks good from this angle.  Some source has problems but not the good stuff. 
I get a kick out of some of the multimedia apps thinking that the two p3's are
one +4GHz Coppermine processor.  Ever hear of one of those?  :-)

It works great overall, other then minor problems.  I probably have less
problems then most people do with single processors but then again I am
extremely brutal when it comes to apps and am not afraid to give them the
heave-ho when they can't play nice.

 KP> i did a 'simple profile' described at ../kernel/Documentation/.. and 
 KP> found acpi has some constant overhead, but i gather it's not bad. 

When it works.  Not an option on the smp guy as it is definetly flakey. 
Luckily it can be safely ignored and turned off in the bios.

 KP> there's a small latency to go from processor acpi sleep states to 
 KP> active, but it seems negligible: how much is saved, i wonder. :-)

I've heard a few horror stories about acpi lowering the life expectancy of HDs.
 I have no idea if that is true.  Never bothered with power management with
drives before and mine seem to outlast others with the exact same drives. 
Seems to me there could be something to the horror stories.

 KP> on.  i haven't seen much info about apm settings.

Other then what I know to be true by not using power management for drives I
can't really say for sure.

 KP> a gcc with the new fortran95 support wasn't compatible with f77 
 KP> .configure scripts and/or makefiles one time when i gave that a go, 

gfortran isn't f95.  I tried compiling a couple of math libs with it and it
seems to work fine with f77 code.  Got a couple of warnings about 'goto'
statements that I thought were amusing.  I recall when I was first learning
fortran that goto's are evil.  Seems to me they've gotten less anal about that
over time.  Tsk, tsk.

 KP> but i haven't looked at it again. is there some switch for it to work 
 KP> with f77 code or documentation about how to adapt? didn't see 
 KP> anything casually browsing.

Pretty well much compatible already other then some syntax changes.  Also
supposedly compatible with f90 code.  I don't know and it has been quite some
time for me fortran-wise.  I need to refresh my memory on this stuff and
perhaps learn f90/95.  We'll see.  Upgraded to gmp-4.2 and that took care of
the mpfr memory allocation issue.

Life is good,
Maurice

--- Msged/LNX 6.2.0
 * Origin: The Pointy Stick Society XVIII - Don't know what I want (1:140/13.1)