From: Ole-Hjalmar Kristensen <ole-hjalmar.kristensen@substitute_employer_here.com>
Subject: Re: "everybody in the control room is using Java"
Date: 19 Jan 2004 09:21:11 +0100
Date: 2004-01-19T08:21:11+00:00 [thread overview]
Message-ID: <wvbrn08k9wnc.fsf@sun.com> (raw)
In-Reply-To: 49cbf610.0401170658.55862afb@posting.google.com
>>>>> "DL" == Dmytry Lavrov <dmytrylavrov@fsmail.net> writes:
DL> "Martin Dowie" <martin.dowie@btopenworld.com> wrote in message news:<bu9o60$h6v$1@sparta.btinternet.com>...
>> "Hyman Rosen" <hyrosen@mail.com> wrote in message
>> news:1074275354.853176@master.nyc.kbcfp.com...
>> > <http://news.com.com/2100-1007_3-5142220.html?tag=nefd_top>
>>
>> Yeah, not the real thing though!..
>>
>> http://www.ghs.com/news/230811j.html
DL> http://www.cobalt.com/aboutsun/media/features/mars.html
DL> http://www.linuxdevices.com/news/NS3674252711.html
DL> Interesting thing, that so called "java technology" ."It's powers the
DL> rover from earth" :)
DL> Major problem in mars missions is probably programs portability and
DL> execution with as much as possible limited access to hardware(no
DL> access to things like ports,legs,cameras,trusters,fans,etc) ,like java
DL> applets on my PC where i don't want it to have any access to my
DL> hardware except framebuffer memory ?
DL> Question:
DL> I just can't believe _at_all_ that realtime java will _always_ meet
DL> deadline,expecially that someone has proved that.
DL> Should be assumed that worsest possible GC(garbage collection)
DL> operations could happen between deadlines.That's mean,i think,normally
DL> load of processor should be not more than several precents,because
DL> processor should be fast enough to do GC.If there's JIT like
DL> compiler,situation are even worse,it's should be assumed that all
DL> commands are interpreted,not compiled! Or someone should do very
DL> pricely analysis.... and it's possible that there will be something
DL> wrong with it itself,so it's should be checked....
Don't know about the compiler, but real time garbage collection is
essentially a solved problem, provided you do not allocate such large
objects that the time to move a single object breaks your schedule.
Look in the book "Garbage Collection" by Lin and Jones, or do a Google
serach for Henry G. Baker (who now is a venture capitalist it seems)
to read his papers.
DL> Program should use not more than,say, 1/100th of time(assuming that GC
DL> lag 100 times more pricely than normal allocation,and that lag could
DL> happen constantly ),and run at <1/20 speed(interpretation)....damn
DL> inefficient ,because it's better to have slower and more reliable
DL> processors than ,hehe,p4 . Is there any tools for SPECIFIC java
DL> platform that will calculate maximal execution time _accurately_? It's
DL> should be very hard to make that tool that it will not assume GC lags
DL> always like i'm doing.
DL> Also,there's no doubling for software if there's same runtime in both
DL> rovers.
--
C++: The power, elegance and simplicity of a hand grenade.
next prev parent reply other threads:[~2004-01-19 8:21 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-01-16 17:49 "everybody in the control room is using Java" Hyman Rosen
2004-01-16 22:21 ` Martin Dowie
2004-01-17 12:06 ` Preben Randhol
2004-01-17 13:52 ` Martin Dowie
2004-01-17 14:58 ` Dmytry Lavrov
2004-01-18 0:41 ` Vinzent 'Gadget' Hoefler
2004-01-19 8:21 ` Ole-Hjalmar Kristensen [this message]
2004-01-19 9:38 ` Erlo Haugen
2004-01-19 10:32 ` Preben Randhol
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox