Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search

Messages from 9375

Article: 9375
Subject: Re: The case for Linux and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 7 Mar 1998 23:31:05 GMT
Links: << >>  << T >>  << A >>
rk:
: >: >don't know why people think you can't do good work with non
unix/linux
: >: >systems
 
phil:
: Perhaps some people can't understand how you can get any work done at all
: on Win95 (and to a lesser extend with NT).
: 
: Win95 and NT lock you into a graphical paradigm for tools that is
: difficult to automate.  Let's say I'd like to run several simulations
over
: night and keep track of potential failures in each.  This is easy in
: Unix/Linux.  Even if a tool has a GUI, it always has a way to execute
from
: the command line - this makes it very easy to put together a perl script
: to automate the overnight simulation runs.  In the Win world, most apps
: don't have a way to execute from the command line - so you've got to stay
: around all night and push the buttons if you want to batch up several
: runs.

rk:
don't know about all of my tools (since not the biggest issue haven't
looked into it, just run them off of the home pc, keep phone line open),
but most of them do have alternate interfaces via the command line;  the
gooey is just a shell.  and running them via the internet is trivial.


rk:
: >does microsoft office run on linux?  need that to communicate these
days.
 
phil:
: Who needs it?  There are other ways to 'communicate'.  On Linux there is
: StarOffice (can read and write M$Word doc format), Applix (can also read
: and write M$Word format as well as others), WordPerfect, LyX (a GUI
: frontend for LaTeX - great for equations, MSWord can't touch it).

rk:
if it can read and write the latest formats (word, excel, and powerpoint
are the standards) then that's fine.  and it should accept objects from all
other programs too.  are these programs available for win '95?  if so, then
there's an alternative that i think many would be willing to try.

rk: 
: >does aol run on linux?  very popular in america.

phil: 
: AOL is a joke.  Who needs it, get a real ISP ;-)

rk:
the people who use it, very popular application, easy to use.  got lots of
users and the discussion was that people use win and not linux cause they
don't want to re-install os.  but, there are apps that people want to run. 
and aol is very popular with many professionals, including those that are
involved in cae.  cheap, and can dial in from anywhere.  real isp is a long
distance call.  expensive for many.  and much harder to use for the typical
pc user.  sorry if aol is such a joke, perhaps those 10 million people are
all idiots.  or perhaps that just want to use the internet easily and are
less concious of what the 'real' professionals think.

rk: 
: >visual basic run on linux?
 
phil:
: Again, who needs it?  Especially with the existence of the following on

rk:
people with a lot of exisiting vb code.  why should the switch to linux and
rewrite code?  why should they make their life harder?  why should they
learn new tools?  

phil:
: Linux:
: 
: Perl (and Perl/TK)
: TCL/Tk
: Python (and many different graphical extentions)
: Eiffle
: Scheme/Tk
: Java
: ... others too numerous to mention...
: 
: Given the availability of these languages, who in their right mind would
: us VizBasic

rk:
personally i'm not a visual basic user.  but there are a lot of people who
are.  gets the job done for them and they say it's very easy to use and do
good thinks with.  guess they're all crazy.  i'll tell them all
immediately!  ;)  p.s. one of the gals i work with is doing some java
stuff, very good programmer.  if she could, she'd switch to vb or delphi,
but they won't let her.  much easier to use (or is it the learning curve?).

rk: 
: >national instruments, do they supply ieee-488 drivers for linux?

phil: 
: I donno... but I'll bet that someone has made a driver for Linux
somewhere
: (just may not be National Inst)

rk:
if it doesn't run *ALL* of the national cards i gotta shut my lab down. 
and so do a lot of others.  or get new cards, new driver, and then right an
interface module so i don't have to revise about 8 years of code, developed
my multiple people.  *that's* more important than the os, which i don't
particularly care about.

rk: 
: >labview run on linux?
: >matlab run on linux?

phil: 
: I believe the answer is yes here (but not too sure).  If not there are
: several free equivilents that are available (Octave comes to mind).
: Mathmatica runs on Linux.

rk:
oops, good point, left off that last one.  again, are they code compatible
and humanoid interface compatible with labview and matlab?  lot of files
and training time.  and training and converstion time, for a typical
professional, runs $50 - $100 / hour.  it's expensive to change.  it's
inertia.  and that does figure into decisions, not just technical merit.

rk: 
: >orcad run on linux?

phil: 
: Now to address the EDA apps together:
: 
: No, most of them don't run under Linux and that's the point of this
: thread.  Many of us in the design community would like to see some of
: these apps on Linux. The fact that they are not here now does not point
to
: some technical deficiency in the Linux OS.  It merely points to the fact
: that a large, well financed company with monopolistic tendencies has been
: able to convice the EDA companies that its OS will ultimately rule the
: universe.  The OSs from Redmond are not in the position they are in
: because of any technical merit, but because of the marketing muscle (and
: strong arm tactics) of Micro$oft. 

rk:
well, i agree with a lot of that <rk quickly checks his pulse>.  many of
the cae tools run on unix and not windoze, so that part doesn't hold up. 
and i'd guess that it's an easier port from unix to linux.  again, a
problem with the argument.  and, as we've discussed, there are certain
designs which won't fit the pc model but need something more.  doesn't mean
that you can't do a lot of nt or linux, it's just different niche.

now w.r.t. a certain company, who's head got his *ss hauled before congress
this week (live right outside d.c.), yes, they have used marketing muscle
and strong arm tactics to get where they are, not technical merit (is this
the olympics?).  it's interesting, when win 3.1 finally was abandoned by
s/w developers, how many developed for '95 and how many for nt.  perhaps
i'm a conspiracy theorist! ;)


rk:
: >viewlogic run on linux?
: >actel designer run on linux?
: >xilinx tools run on linux (pete)?
: >altera tools run on linux?
: >orca tools run on linux (stu)?
: >synplicity run on linux?
: >examplar run on linux?  i think this is a yes, do they still support?
: 
: Answer was yes here, not too sure now.

phil: 
: >aldec run on linux?
: >does it network directly with win 95/nt machines with drag and drop?
: >delphi run on linux?
: See the languages discussion above.

rk:
yup, but here's the key point.  let's say one is developing on both windoze
now and unix.  windoze for small to medium jobs, say fpgas.  and unix for
medium sized asics.  and win '95 is about to be replaced in a years time
with the new version.  so, it's time to think, perhaps, about upgrading. 
and when you get new machines, what do you put on it?  obviously, on a
design team, you want similar configurations for all team members.

do you go to linux, for the sake of argument, technical merit?
do you stay at '95 since it does the job and wait things out?
do you go to nt 5.0, run existing apps (eda and non-eda) and bet that
that's where the action will be?
do you go to straight unix/sun configurations?

note: for windows choices, keep unix/sun for asics.

*that* is the question, as if the chips and boards and boxes don't flow
out, $ doesn't flow in.  we're all in business of one sort.

from what i've read here, from people all over the world, it doesn't look
like linux will be a good bet in the 6-12 months time.  not from
performance, technical merit, or anything like that.  from s/w
availability.


rk: 
: >sigma plot run on linux?

phil:
: gnuplot (and its X frontend).  Several plotting programs are available
: (most free).

rk:
is it compatible?  a lot of people use this package (like the wife!) so i
gotta be compatible.  again, file conversion/training time.  this case,
mostly training time.


<snip a bunch of apps - apparently there are some equivalents if you're
resourceful and can find them, eliminating some conversion hassles>

rk:
: >can i play the cd-rom interactive disks for the kids on linux?
 
phil:
: What's this got to do with getting your design done?

rk:
oh, we use the 'puter at home for lots of stuff.  business (cae) and other
stuff.  don't want to lose capability if i upgrade.

same to other stuff below.

rk <who gotta proof read better>
: >for engineering.  don't know answers, perhaps you can feel us in.
 
phil:
: Well, I'm not that kind of guy.  ;-)

rk:
good, me neither.  just sometimes types too fast.


rk:
: >like i say, "it's the s/w, stupid."  [paraphrased from the pres.]  who
: >cares about the os or cpu.  it's the apps.

phil: 
: Yeah, and its the stupid software too ;-)  But seriously, the OS does
have
: an impact in how you get your work done.  If the GUI gets in the way (as
: it does with WinXX) it can be a hinderance.

rk:
well, you have showed some good alternatives and answered some questions. 
thanks!  and you did better than the guy who doesn't want to move his hand
all the way over to the arrow keys.

yup, some programs, too much clicking.  even the 'explorer.'  i do a lot of
my navigating with the keyboard, much faster.  sometimes better with mouse.

a lot of the cae you can run from the command line and batch files.  that's
something that's easy for eda vendors to change for many apps.

but, at day job, they run many of the same apps on both unix/sun and
win/pc.  and that keeps training time/costs reasonable and shields you, to
a certain extent, from the os.

if the mainstrain cae apps aren't there, it will be hard to switch to
linux.  chicken and egg.  and will vendors translate sun --> win or sun -->
linux or none?  and on the other side will they go from pc --> linux or pc
--> unix or none?

it'll be interesting to see!
-- 
--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------
Article: 9376
Subject: Re: The case for free operating systems and EDA
From: steve <steve@sj.co.uk>
Date: Sat, 07 Mar 1998 23:37:41 +0000
Links: << >>  << T >>  << A >>
Mark Willey wrote:

> This conversation has evolved(?) beyond the scope of any of the forums that
> it's posted to.  Please move to a more appropriate venue.  I would suggest
> some advocacy groups.... which I never read.  :)

Indeed. Unfortunately, this discussion, and any discussion like it, has
no home to go to. Perhaps comp.cad.synthesis is closest, but this isn't
a comp. discussion, really. comp.arch.fpga never seems to get used for
its intended purpose, so perhaps that's the correct place for toolset
issues (for those of us hacking fpgas, at least), and, despite
appearances, this isn't really an OS war, certainly from where I'm
standing...
If I knew how to set followups in this godforsaken newsreader, I'd point
at comp.cad.synthesis. If anyone else would like to steer us over there,
that'd be great.  

  Cheers, 
    Steve
Article: 9377
Subject: Re: Correlation-continued
From: "rk" <stellare@erols.com.NOSPAM>
Date: 8 Mar 1998 00:37:37 GMT
Links: << >>  << T >>  << A >>
try signal processing technologies, there are others, too.  spt can take
you up to 1 GHz.
-- 
--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------

Lucian Fogoros <l.fogoros@popmail.csuohio.edu> wrote in article
<35019C53.1421@popmail.csuohio.edu>...
: We would like to know where we could get 100 MHz A/D's?
: 
Article: 9378
Subject: Re: The case for free operating systems and EDA
From: "Walter Daniel Gallegos" <walter@chasque.apc.org>
Date: 8 Mar 1998 03:43:34 GMT
Links: << >>  << T >>  << A >>
Hi :

Are you check Oberon S3 Native, In my opinion one of the best OS today
Take one second to check it at 
http://inf.ethz.ch

Walter.

rk <rich.katz@gsfc.nasa.gov.NOSPAM> escribió en artículo
<01bd493d$15a1cd40$1e83accf@homepc>...
> misha:
> : There is a significant advantage of Linux over Windows concerning
> portables.
> : I was very surprised when I touched the processor inside my desktop
when
> : it was running Linux. It was COLD. So was the power regulator. Under
> : Windows95 the processor is always hot, even when OSR2 System Monitor
> shows 1%
> : CPU utilisation for an hour. Same about NT.
> : 
> : Of course, when CPU-intensive application runs under Linux, the
processor
> : warms up. But for typing text on the portable Linux should give much
> longer
> : battery life.
> 
> interesting.  i haven't seen any of the rags that report on portables and
> battery life consider this.  it would make a good benchmark since a lot
of
> people do x-country traveling.  now, the portables do have a lot of power
> savings features and do a bunch of stuff to extend battery life so the
> difference on the desktop pc may not equal difference on a lapdog.  but
it
> would be interesting to see, for a portable, a comparison for running a
> simple text editor between: 
> 
> 	a. win '95
> 
> 	b. win nt
> 
> 	c. linux
> 
> 	d. unix/pc
> 
> 	e. unix/risc (pick hp, sun, ibm, etc.)
> 
> 	f. mac
> 
> may the best win!
> 
> --------------------------------------------------------------
> rk
> 
> "there's nothing like real data to screw up a great theory" 
> - me (modified from original, slightly more colorful version)
> --------------------------------------------------------------
> 
Article: 9379
Subject: Re: The case for free operating systems and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 8 Mar 1998 05:15:58 GMT
Links: << >>  << T >>  << A >>
wen:
: ;: Sorry, I can't do that.  The machine I sometimes need to use would
have
: :: so many processors, disks, and memory and powersupply that it would
take
: ;: a forklift to carry.  I also routinely use manage jobs that runs on a
: :: farm of workstations and I need to do it remotely and service them at
: ;: odd hours of the day.

rk:
: :for an application like that, at the high end, you need the high end
: ;horsepower.  that's not a good spot for NT.  but that has nothing to do
: :with suitability of nt for eda, as peter pointed out.  it only has to do
: ;with the suitability of nt for a particular class of job.  don't think
: :there's any disagreement there.  and i've bought some big machines for
big
: ;jobs (and spent big $).
 

wen:
: But that is the direction of EDA future.  Larger chips more simulation
: and analysis, and more difficult place and route.  I already have jobs
: that takes 24 hours per random seed on the fastest pc available.

rk:
a bit confused, thought you were working on
super-duper-need-a-forlift-'puter.

there's a large amount of design work going on in the < 100 k gates and
under arena.  with a lot of designs < 25 k gates.  these are being
increasingly filled with fpgas.  and a lot of people are currently
designing these with (dare i say) pc's and win and some still with dos, as
we've seen here.  as the pc hardware grows in performance, and it seems to
increase almost every week (and memory is $3/megabyte!), pc developer's are
setting their heights higher doing larger designs.  and they'll be a whole
generation of chip designers using these tools.  just like we got hit with
a whole generation of programmers who used c and unix in school and we know
the impact of that.

now, we have previously heard stories about some fpgas taking ~ 24 hours
for p&r while their competitors take an order of magnitude or more less. 
part of this is the fpga architecture.  part of this is the software.  i
think the fpga market will be shaken out some if they require 24 hour p&r. 
people will switch to devices that don't.

now, for the really big jobs, those that require a farm to run, the really
mega-jobs at the extreme end of performance.  no doubt that current win
technology is no place for that.  but, as an aside, perhaps you could
comment on technologies such as quickturns, where they use boards full of
fpgas (and now custom chips, iirc) to emulate the functionality of the asic
(pentium 'simulated' this way) and permit running at a sizeable fraction of
actual device speed, in-circuit.  

	will this be preferable to farms of computers, multi-processors, or other
simulation techniques?

	will formal equivalence checking of final netlists to original vhdl be
accepted for signoff with only a limited amount of gate level simulation?

	will the use of rapid prototyping (say chip express 1 day turn laser
programmable gate array) replace a lot of the gate level simulations?

===============================


wen:
: ;: The problem is not with NT, the problem is with how Microsoft is
having
: :: an effect on how EDA programs were written.  There is absolutely no
reason
: :: why EDA vendors couldn't compile their program on NT but link them
with
: ;: X-window libraries so that the NT machine would be useful as a remote
: :: compute engine within a network of unix workstations.  NT still has a
: ;: long way to go to match Unix for networking and user interface.  

rk:
: :yup, microsoft monopoly, which is not good [see my earlier post on
: ;competition and open standards].  and, i believe, microsoft does care
about
: :you; well about your $, anyways.  for some networking features, the
: ;microsoft stuff is pretty good (and win 3.11 was horrible); on others
it's
: :a pain and not up to unix or another of other os'es.  their user
interface
: ;is not bad, very mac-like, and it's easy to write your own programs with
: :nice, easy to use and understand humanoid interfaces.  having programmed
: ;both, i'd go with the win '95/nt interface.
 
wen:
:Humanoid interface?  You should know by now that engineers are more
:machine than human.  :-)

rk:

no comment. :-)

--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------
 
Article: 9380
Subject: Re: The case for Linux and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 8 Mar 1998 05:15:59 GMT
Links: << >>  << T >>  << A >>
steve:
: As for basing my choice of synthesis / simulation tools on which
: platform rund Word / VB / whatever, heck, that's what the sub-$K PC /
: portable was put on this earth for. Don't want to be polluting my 'real'
: machine with cycle-hungry dancing paperclips now, do I? 


rk:
for small businesspeople, running two 'puters is not really that good of an
option.  unless you'd like to make an investment!

for the stupid paperclip, geeze, how'd that get past any review cycle at
all.

steve: 
: Of cource, I could roll out the panacea that is java, and we can all run
: the same tools. After all, cycles are cheap *8-) 

rk:
yeah, right, sure. ;)
Article: 9381
Subject: ISR Programming Cable for FLASH370i from CYPRESS
From: gerhard@zipfel.franken.de (Gerhard Zipfel)
Date: 08 Mar 1998 12:44:00 +0200
Links: << >>  << T >>  << A >>
How to build a programming cable for the FLASH370i devices
from CYPRESS (PC version)?
Is there an application note?

Thanks
Article: 9382
Subject: Re: Altera MaxPlus II version 8.1 delays
From: pmyg@ix.netcom.com
Date: Sun, 08 Mar 1998 08:59:03 -0800
Links: << >>  << T >>  << A >>
Russell May wrote:
> 
> I am looking for ideas why version 8.1 of Altera's MaxPlus II
> has a delay of about 0.5 to 1 second each time a new window is
> selected, either automatically by the program or by clicking
> on a different window. Does anyone else experience this?
> 
> I have been using version 8.1 at home on my Pentium 233 MMX
> system, which has a 64 Mb of EDO memory and a large fast disk drive,
> and version 8.0 at work on a classic Pentium 133 system, which
> has 32 Mb of FPM memory and a slower disk drive. My home system
> runs the Altera compiler nearly twice as fast as the one at work,
> but has these annoying and continual pauses. The system at work
> does not have the pauses.
> 
> Altera tech support has been no help, only insisting that I must not
> have sufficient memory or something else must be running. The pauses
> happen as soon as the program starts, and I have not compiled anything
> which takes over about 9 Mb of working memory. Nothing else (except
> clock) is visibly running. My system at work is connected to a 3COM
> network, the one at home is not connected to a network (except thru
> DUN to my Internet provider).

Many times, I've seen unexplained delays in programs due to an invalid
path on the PATH tree. For example, if the CD-ROM is on the PATH tree,
and no CD is present, there could be unusual delays in your program.
Especially if the program you're having a problem with is after the
invalid path location in the PATH tree.

Hope this helps,
pmyg
Article: 9383
Subject: Re: The case for Linux and EDA
From: ptkwt@user1.teleport.com (Phil Ptkwt Kristin)
Date: 8 Mar 1998 10:29:19 -0800
Links: << >>  << T >>  << A >>
In article <01bd49d4$198ce380$3a80accf@homepc>,
rk <stellare@erols.comNOSPAM> wrote:
>rk:
>: >does microsoft office run on linux?  need that to communicate these
>days.
> 
>thomas:
>: There is StarOffice, that understand the Microsoft "standards" and is
>free
>: for personnal use. Works pretty well. Really looks like Office.
>
>rk:
>ok, but does it understand for reading and writing *all* of the latest file
>formats, not just the standards to be 'office compatible.'  one can argue
>whether we should be using office, pagemaker, tex, latex, or whatever. 
>but, micro$oft has one, and to communicate effectively with others and do
>joint writing, it *is* the standard.  also, i can take the output of my cae

How can anything be a serious standard that changes every two years?
Office 95 can't read the latest Office 97 formats either.  On the
otherhand, things like TeX and Postscript don't change nearly so often.
But of course M$ needs to constantly 'update' these formats to 'encourage'
you to go out and buy the latest Office every two years - someday, maybe,
people will figure out this scam.

>rk: 
>: >visual basic run on linux?
>rk:
>this is important as there is a lot of investment in visual basic code,
>amongst engineers.  not me (delphi fan), but i know a lot of people who do
>it.  very important application.

There was a Vestigial Basic clone that ran on Linux - it allowed you to
run VB byte codes.  But again, there are so many superior languages
available on Linux - why would anyone in their right mind use Vestigial
Basic?

>rk: 
>: >national instruments, do they supply ieee-488 drivers for linux?
> 
>thomas:
>: I know a physicist who drives some experiments with a Linux-PC and IEEE
>: cards. I'll have to ask him about this version.
>
>rk:
>the national cards, at least in america, are most probably the most
>popular, they are fairly 'standard,' and give us code portability from dos
>them and for us users.  and yes, machines that i use for cae i also use for
>laboratory use.  

The drivers may not be supplied by Nat Inst, but someone else may have
developed some drivers on their own.

>thomas: 
>: I don't have time to comment upon the others. Some (especially Xilinx
>tools)
>: do not have a Linux version, which is explanable by market analysis but
>: not technical reasons.
>rk:
>what succeeds in the market is not determined solely by technical
>considerations, which is the point of this whole argument.  we can discuss
>os/2 vs. win 3.1.  microchannel vs. isa vs. eisa.  c vs. pascal.  etc. etc.
>etc.

Sure, but the point that the Linux folks are trying to make is that Linux
is enjoying INCREASING success in the market.  The number of Linux users
is increasing at a high rate.  Of the other technologies you mention:
OS/2: Killed primarily by IBM's incompetance.  Linux is not controlled by
any one entity, the source code is available everywhere - it can't be
killed by one company's incompetance.
Win3.1: Killed by Micro$oft.  See OS/2 above.
MicroChannel,ISA, EISA: Stages in hardware evolution.
C vs Pascal: Maybe Unix had something to do with the popularity of C?

The point here is that Linux is different from all of those other
technologies you mention.  The number of Linux users worldwide is on the
increase, and the numbers don't seem to be peaking out anytime soon.
At this point, conservative estimates place the number of Linux users
worldwide at 5Million.  Other estimates range as high as 15 Million.
It is quite possible that there are 7million.

>thomas:
>:                   The whole point is that if all vendors made Linux
>: versions, everybody would be happier: the vendors who would have an
>: extended market for cheap (good software should be easy to port) and
>: the clients, who could benefit from the power and stability of Linux.
>thomas:
>: Oh well, maybe Bill Gates would not be so happy (niark niark niark).
>
>rk:
>don't underestimate that <fill in choice word here>.  i'm sure he'd find a
>way to make more billions off of it, selling more upgrades, convertors,
>bleah, bleah, bleah.  

You're invited to get off the treadmill of upgrades, rk.  Save yourself
$thousands$.

>he'd use it to avoid the justice department and any
>attempt to break his company up for being a monopoly.  look what he's doing
>to netscape.  first he got them to give away their product for free.  now
>he has them giving away their source code for free.  what do they give away

Netscape's move to release source code is brilliant.  And it is one that
M$ cannot follow; its too much against their mind$et.  By releasing source
code, netscape has ensured that their browser will never go away, even if
Netscape dies as a company.  They've also gained thousands of new
developers for free.

>rk:
>performance and interface is secondary.  like i said, you can have the best
>os, the nicest interface <one where the truly lazy don't have to move their
>hands over to the arrow keys, a windows complaint i have recently heard>
>and have a worthless machine.
>if the applications aren't there, no work gets done.  go ask ibm about
>os/2.  ask apollo about their os?  well, you can ask some apollo users,
>still running the machines, keeping them for their applications, locked in.
>

And in the open source world, if the apps aren't there we write them
ourselves.  We're getting quite a lot of them nowadays. In the meantime, I
don't see Linux going away for lack of apps.

 phil

Article: 9384
Subject: LARGE SELECTION OF FPGA BOARDS & KITS
From: Richard Schwarz <aps@associatedpro.com>
Date: Sun, 08 Mar 1998 16:00:41 -0500
Links: << >>  << T >>  << A >>
APS has al arge selection of FPGA boards/routers and fronte end tools
including the XILINX Foundation Series with Synopsis FPGA express,
Lucent routers and test boards, and VHDL simulators, as well as
ATMEL's new 20K kits.

Check out these and other tools at :

http://www.associatedpro.com/aps

--
__/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/

Richard Schwarz, President              EDA & Engineering Tools
Associated Professional Systems (APS)   http://www.associatedpro.com
3003 Latrobe Court                      richard@associatedpro.com
Abingdon, Maryland 21009
Phone: 410.569.5897                     Fax:410.661.2760

__/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/


Article: 9385
Subject: Re: The case for Linux and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 8 Mar 1998 23:31:42 GMT
Links: << >>  << T >>  << A >>
rk:
: >: >does microsoft office run on linux?  need that to communicate these
: >days.
 
thomas:
: >: There is StarOffice, that understand the Microsoft "standards" and is
: >free
: >: for personnal use. Works pretty well. Really looks like Office.

rk:
: >ok, but does it understand for reading and writing *all* of the latest
file
: >formats, not just the standards to be 'office compatible.'  one can
argue
: >whether we should be using office, pagemaker, tex, latex, or whatever. 
: >but, micro$oft has one, and to communicate effectively with others and
do
: >joint writing, it *is* the standard.  also, i can take the output of my
cae
 
phil: How can anything be a serious standard that changes every two years?
: Office 95 can't read the latest Office 97 formats either.  On the
: otherhand, things like TeX and Postscript don't change nearly so often.
: But of course M$ needs to constantly 'update' these formats to
'encourage'
: you to go out and buy the latest Office every two years - someday, maybe,
: people will figure out this scam.

rk:
er, everyone's already have figured out this scam.  and we're stuck. 
heeeeeellllpppp!!!!!

and, of course, a microshafty can say, "well, office 97 can write office
95/6.0 file formats."  and they'd be correct.  and everytime one down
converts (and i'm not sure if the file i sent you via email to try was down
converted or not, i'll have to check) you get one giant bloat.  <insert
expletive here>.  they did do good ONCE, office 95 used same file formats
as word 6.0.  but they figured out their mistake and quickly corrected it,
since there was little reason for anyone to update.  and, i'll give you a
hint, the paperclip ain't making it!

and bill g. was downtown this week, chatting about this, and how it's good
for the world.

on the other hand, having used tex, latex, etc., ms word is easier to use. 
and when i worked in unix joint, tex was required, i fought tooth and nail,
easier to format straight ascii with my turbo pascal code editor on my
little 286 pc at home.

and it is rather standard.

and it is often required.

now, if you read back, say, like 2,000 posts or so ago, i was on the
proverbial soapbox about open standards for eda to enhance competition. 
same here.

the update-thing sucks, gotta do it too, too often.  and at day job, take
the word update and multiply by 4.  gets expensive.  personally, i was
pretty happy with word 2.0.



rk: 
: >: >visual basic run on linux?

rk:
: >this is important as there is a lot of investment in visual basic code,
: >amongst engineers.  not me (delphi fan), but i know a lot of people who
do
: >it.  very important application.
 
phil:
: There was a Vestigial Basic clone that ran on Linux - it allowed you to
: run VB byte codes.  But again, there are so many superior languages
: available on Linux - why would anyone in their right mind use Vestigial
: Basic?

rk:
personally, i don't program in it.  but a lot of people do.  and they can
do it quickly.  lot of support for it, esp. in instrumentation, can develop
good, custom, easy to use apps quickly and cheaply with a lot of
functionality.  for some example, i'd suggest you look at the national
instruments catalog.  have some rather amazing stuff there.  

for performance, code legacy, sanity, etc. reasons, we're using the delphi
(pascal) compiler.  very, very nice.  and, of course, part of microshafts
great world domination strategy is to program the world in basic, bill
gates' favorite language.  that's why they're doing away with macros for
all of their apps and sticking in vb.  bleah!  



rk: 
: >: >national instruments, do they supply ieee-488 drivers for linux?
 
: >thomas:
: >: I know a physicist who drives some experiments with a Linux-PC and
IEEE
: >: cards. I'll have to ask him about this version.

rk:
: >the national cards, at least in america, are most probably the most
: >popular, they are fairly 'standard,' and give us code portability from
dos
: >them and for us users.  and yes, machines that i use for cae i also use
for
: >laboratory use.  
 
phil:
: The drivers may not be supplied by Nat Inst, but someone else may have
: developed some drivers on their own.

rk:
for exercise purposes, and to keep an open mind, i checked out the national
instruments catalog and their www site for the latest info.  they had mac,
unix, dos, win 3.1, win nt, win '95, exec 8, no linux.

right now i gotta run dos/win 95(or nt) and unix.  linux would make a
third.  now, if the unix apps would all migrate to linux, *that* would be
good.  maintaining skills and systems for three environments is
undesirable, to say the least.


thomas: 
: >: I don't have time to comment upon the others. Some (especially Xilinx
tools)
: >: do not have a Linux version, which is explanable by market analysis
but
: >: not technical reasons.

: >rk:
: >what succeeds in the market is not determined solely by technical
: >considerations, which is the point of this whole argument.  we can
discuss
: >os/2 vs. win 3.1.  microchannel vs. isa vs. eisa.  c vs. pascal.  etc.
etc.
: >etc.
 
phil:
: Sure, but the point that the Linux folks are trying to make is that Linux
: is enjoying INCREASING success in the market.  The number of Linux users
: is increasing at a high rate.  Of the other technologies you mention:
: OS/2: Killed primarily by IBM's incompetance.  Linux is not controlled by
: any one entity, the source code is available everywhere - it can't be
: killed by one company's incompetance.
: Win3.1: Killed by Micro$oft.  See OS/2 above.
: MicroChannel,ISA, EISA: Stages in hardware evolution.
: C vs Pascal: Maybe Unix had something to do with the popularity of C?
: 
: The point here is that Linux is different from all of those other
: technologies you mention.  The number of Linux users worldwide is on the
: increase, and the numbers don't seem to be peaking out anytime soon.
: At this point, conservative estimates place the number of Linux users
: worldwide at 5Million.  Other estimates range as high as 15 Million.
: It is quite possible that there are 7million.

rk:
7 million is a large number, worldwide.  very impressive.  and if it can do
the job with little work, is as easy and cost-effective as the other
system, then it's gonna be a winner.  *IF* the conversion costs aren't too
high, which is the point of above.  there is inertia and for good reasons -
people have invested a lot of time and $ in what they have and don't change
easily.  and one guy at work keeps on bugging me to upgrade some dos apps
i'm running; real-time programs that have to work correctly everytime. 
can't justify, even to myself, the time to upgrade it to 95/nt.  rather
move forward.

c vs. pascal?  another thread, another time, that one's *ALWAYS* fun.  just
like the unix <--> dos ones from a decade ago, where the unix guys said
messy-dos is dead, after 386 becomes popular, unix will be everything. 
same arguments as we've seen here, for the most part.  big argument for
unix in the '80s was that they knew where to put the <crtl> key in unix,
dos machines couldn't even get that right, slowed down the entire machine! 
now they complain about arrow keys.  what a difference a decade makes!

for linux to suceed in eda (and in an attempt to keep this thread somewhat
relevant), it needs something to make people want to use it.  better is the
enemy of good enough and if people are reasonably happy with what they got
and know that it will get their job done (i.e., unix, mac, pc/win), why
change to an unknown, NOT be able to deliver on time.  when you don't
deliver, they never say, "ah, but he used linux" (forth, or smalltalk, or
nueral nets, or fuzzy logic, or expert systems, or tagauchi method, or next
(jobs machine) or whatever the trendy technology  of the day is.  they say,
"phil <expletive> up, playing with that crap."  and that's not good.  i
know people who spend all their time messing with tools, never have time to
design the damn logic!

anyways, the thought of having to run *three* systems, is, well, not nice. 
now for linux to succeed, imho, it can do the following:

	1. be a unix on a pc.  be able to run most of the major unix apps. 
perhaps
		a bit slower, but be able to go to the synopsis store, pick up your
		copy, and run.

or

	2. be a better pc.  minimal hassles and conversion costs for people to
dump
		the guys from the northwest.  don't disrupt the world.  look at the
hassles
		in just converting from win 3.1 to 95.  people went nuts.  ok, you might
say,
		they were mostly aol, vb idiots.  and you may be right.  but we're
talking
		what, 100,000,000 machines or so?  *that's* a big market. world-wide
there
		is, iifc, 100,000 eda people?  really quite a small market.

so, it is my conclusion, for linux to succeed, they gotta hit critical mass
in applications so one of the two cases above is met.  for general use,
they gotta run *all* of the apps.  like 95 does, mostly, in the pc world. 
or for unix.  why give up a sun for something that can't run all of the
s/w?  and price of the unix boxes is falling, towards pc prices.  while pc
performance moves up, closing in on workstations.

same holds true for the eda world.  gotta have the flow.  only having some
of them on the machine will make it little more than a pain in the butt. 
let's say they get 10% of the eda users.  and let's say my 100,000 eda
users is the correct number.  and let's say that the eda community, on
pc's, is to move, and we're to have low-cost s/w (currently there is a lot
of that, check out rich, programmable logic jump statoin, xilinx, actel,
quicklogic, bleah, bleah bleah - notice viewlogic isn't on this list;).

anyways, take the 10,000 converts, and 10% is a big number, and say they
spend $1,000 for eda tools.  $10,000,000.  big deal.  @ $100,000/man-year,
that's 100 man-years of effort.  and that's not much, considering the
variety of tools that are necessary for eda, with all of the vendors,
libraries, manuals to be written, machines to run them and check them out
on, app notes to right, bleah, bleah, bleah.  and let's say the users pay
maintenance, $250.00 per year.  that's 25%, *higher* than industry average,
from what i've seen.  another $2,500,000 or 25 man-years of effort.  bfd. 
so, if these assumptions are correct, gotta raise prices.  but, for eda,
fpga users, for example, are getting back end tools real cheap or even free
(actel, quicklogic, etc.).  and getting free vhdl compilers (not the best,
but they work, and *outstanding* macro generators).  and viewlogic
maintenance, i think it's $500.00/year, time to re-up anyways, and
synopsys/viewlogic said no linux, no customer demand.  and orcad is cheap
on pc.  ==> touch switch for pc users.  on the unix side, expensive tools,
cost of machine negligible to the s/w bill and the nre for the super-big
asics.  perhaps tough sell, esp. if all the apps aren't ported, then they
have to run two machines.  not three, unix guys love tex!

anyways,

from all that i've seen here over the last week, i think it will be tough,
either in eda or general use, for linux to take over.  don't think that
it's because that's the os that comes on the machine, i think it will be
s/w availability and the desire of a lot of people to go with what they
know will do the job, even if it's a bit more $ and technically inferior. 
in electronics, my output is chips and boards and boxes (and some paper
too!) and the tools are, to a large extent, just a means to an end.  and no
way would i jeopardize output for a 'better' tool that isn't well
supported.  chicken and egg, yes, but i think that will be the issue. 
unless there's something made available that is way better and won't cause
you to screw up.

thomas:
: >:                   The whole point is that if all vendors made Linux
: >: versions, everybody would be happier: the vendors who would have an
: >: extended market for cheap (good software should be easy to port) and
: >: the clients, who could benefit from the power and stability of Linux.

thomas:
: >: Oh well, maybe Bill Gates would not be so happy (niark niark niark).

: >rk:
: >don't underestimate that <fill in choice word here>.  i'm sure he'd find
a
: >way to make more billions off of it, selling more upgrades, convertors,
: >bleah, bleah, bleah.  
 
phil:
: You're invited to get off the treadmill of upgrades, rk.  Save yourself
: $thousands$.

rk:
love to, if it can do the job.  see above.

rk: 
: >he'd use it to avoid the justice department and any
: >attempt to break his company up for being a monopoly.  look what he's
doing
: >to netscape.  first he got them to give away their product for free. 
now
: >he has them giving away their source code for free.  what do they give
away
 
phil:
: Netscape's move to release source code is brilliant.  And it is one that
: M$ cannot follow; its too much against their mind$et.  By releasing
source
: code, netscape has ensured that their browser will never go away, even if
: Netscape dies as a company.  They've also gained thousands of new
: developers for free.

rk:
yup, but a lousy way to run a business.  not a good move if you need to
convince those funding guys to fork over $ for your next startup.  

again, the key is public domain standards.  ibm opening up the pc vs. apple
closing the mac.  pc development was competitive as hell and look where
they are.  and the technically superior mac (my favorite, btw, but could
never afford one), needs handout from bill gates just to keep the doors
open.  going back a computer generation, the open apple ii got people
working on it.  it's when a single man can control what lots of people do
and behave because he controls, in secret, developments that can affect the
world, we're in trouble.  (is this iraq or seattle? ;)

so, for eda, i'd like to see more open standards, not less.  and they're
brawling about that now, right, just like in os.  and if standards are
open, idiots like us will write utilities, people will form startups,
competition and greed will kick in, and capitalism will work.


rk:
: >performance and interface is secondary.  like i said, you can have the
best
: >os, the nicest interface <one where the truly lazy don't have to move
their
: >hands over to the arrow keys, a windows complaint i have recently heard>
: >and have a worthless machine.
: >if the applications aren't there, no work gets done.  go ask ibm about
: >os/2.  ask apollo about their os?  well, you can ask some apollo users,
: >still running the machines, keeping them for their applications, locked
in.

phil: 
: And in the open source world, if the apps aren't there we write them
: ourselves.  We're getting quite a lot of them nowadays. In the meantime,
I
: don't see Linux going away for lack of apps.

rk:
sounds like the unix argument in the '80s.  unix is obviously not going
away.  and it's much improved over what it was.  but the open source world
hasn't caused it to take off the way the pc's have.  with all of the great
unix programmers, really smart guys and all, they missed the big apps that
caused pc's to take off: visicalc[iirc].  and games.  and easy to setup and
use.  and cheap.  back in '87, got an inexpensive '286 (which has slowly
morphed into the one i'm typing on now), could program anything (yes, wrote
some of my own analog and digital simulators), could develop pals, run
abel, spice, do documentation, useful machine.  never could afford it for
unix.  and unix was harder to program.  did it professionally for a while
so i'm speaking from experience.  and any idiot could design their own
add-in card for the pc.  from general use card to a laboratory instrument.

which leads to my great big b*tch about pc's: they're getting rid of the
isa
bus!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!

now *that* move will cost me
tons!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

yeah, yeah, yeah, pci is so much better technically.  but it will cost me a
fortune in time, effort, and $ to convert so i can stay at the same place i
am now! [got a lot of semi-custom and home-designed cards, as well as a
pile of other cards that need replacing.  i've started to prepare for this
but it will be a disaster financially - run about 12 'puters total].

aaaarrrrgggghhhh!!!!

have a nice weekend,


-- 
--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------
Article: 9386
Subject: Re: The case for free operating systems and EDA
From: wen-king@myri.com (Wen-King Su)
Date: 8 Mar 1998 18:26:29 -0800
Links: << >>  << T >>  << A >>
In a previous article "rk" <stellare@erols.com.NOSPAM> writes:
:
;wen:
:: ;: Sorry, I can't do that.  The machine I sometimes need to use would
;have
:: :: so many processors, disks, and memory and powersupply that it would
;take
:: ;: a forklift to carry.  I also routinely use manage jobs that runs on a
;: :: farm of workstations and I need to do it remotely and service them at
:: ;: odd hours of the day.
;
:wen:
;: But that is the direction of EDA future.  Larger chips more simulation
:: and analysis, and more difficult place and route.  I already have jobs
;: that takes 24 hours per random seed on the fastest pc available.
:
;rk:
:a bit confused, thought you were working on
;super-duper-need-a-forlift-'puter.

I said that is how long it takes.  I didn't say that is what I use to
do actual productive work.  I als said that it is for one random seed,
and I would have used several hundred in each job, and that is still
not complete to my liking.

;there's a large amount of design work going on in the < 100 k gates and
:under arena.  with a lot of designs < 25 k gates.  these are being
;increasingly filled with fpgas.  and a lot of people are currently
:designing these with (dare i say) pc's and win and some still with dos, as
;we've seen here.  as the pc hardware grows in performance, and it seems to
:increase almost every week (and memory is $3/megabyte!), pc developer's are
;setting their heights higher doing larger designs.  and they'll be a whole
:generation of chip designers using these tools.  just like we got hit with
;a whole generation of programmers who used c and unix in school and we know
:the impact of that.

When pc's get faster, so would unix running on pc's.  It is not an issue
of hardware.  The pc will never scale as fast as the demand of computing
power.  While speed grows much less than linear compared to the physical
measures of the processor (area, gates, channel length, etc), the computing
demand grows much faster.  There are a lot of things that would tend to
grow with n*n or n*(log n) if not managed carefully.  What is wrong with
EDA tools running on NT is that it is short sighted and because of poor
network services, it precludes effective scaling of computational power
by aggregating a network of computing engines. 

:now, for the really big jobs, those that require a farm to run, the really
;mega-jobs at the extreme end of performance.  no doubt that current win
:technology is no place for that.  but, as an aside, perhaps you could
;comment on technologies such as quickturns, where they use boards full of
:fpgas (and now custom chips, iirc) to emulate the functionality of the asic
;(pentium 'simulated' this way) and permit running at a sizeable fraction of
:actual device speed, in-circuit.  

Witness the long list of pentium bugs since the processor was released
as evidence that insufficient verification cycles, even with the aid of
quickturns, has been devoted to check the design. 

:	will this be preferable to farms of computers, multi-processors, or other
;simulation techniques?
:
;	will formal equivalence checking of final netlists to original vhdl be
:accepted for signoff with only a limited amount of gate level simulation?
;
:	will the use of rapid prototyping (say chip express 1 day turn laser
;programmable gate array) replace a lot of the gate level simulations?

I am afraid none of that will be good enough.  It is going to take a group
of really sharp programmers many years in an effort that is not driven
by the short sighted market forces to come up with a real solution that
scales.  And only multicomputers of some sort can provide that scaling
of computing power required.
Article: 9387
Subject: Re: The case for Linux and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 9 Mar 1998 02:29:26 GMT
Links: << >>  << T >>  << A >>
phil:
: Win95 and NT lock you into a graphical paradigm for tools that is
: difficult to automate.  Let's say I'd like to run several simulations
over
: night and keep track of potential failures in each.  This is easy in
: Unix/Linux.  Even if a tool has a GUI, it always has a way to execute
from
: the command line - this makes it very easy to put together a perl script
: to automate the overnight simulation runs.  In the Win world, most apps
: don't have a way to execute from the command line - so you've got to stay
: around all night and push the buttons if you want to batch up several
: runs.

rk:
there is a difference in the win and unix world, since in the win world you
can carry your cycles with you or perhaps have a home pc (my case).

now we have discussed this before with myself and one or two others that
some tools are command line driven too, as you mention for unix.  i know
that the actel tools for fpga, which i did initially in dos and now in win,
had command line i/f's and ascii configuration files.  this makes it
suitable for remote operation with very simple software over the internet:
i.e., ftp, drag and drop over the internet, and a small program to stay
resident and start other programs.

i just finished looking at the simulator and related tools (viewlogic) and
see that these tools, which i normally operate via the gooey, also have
command line forms, so that is not an issue here.  perhaps this is
something, as our designs get bigger and more complex, win users should
watch for and put in their $0.02 to the vendors to enable this sort of
operation.

can't speak, obviously, for all apps but did take a closer look at this
one.

now, however, i'm using a new tool, statecad, for graphical entry, vhdl
output, using the gooey.  dunno how good these graphic tools will work over
a slow phone line and perhaps the cycle here for win and x (although one
respondent said that there a program to speed up x) starts to break down. 
btw, statecad has a built in simulator.  so enough there to keep you busy
on your home pc.  or ftp in the generated vhdl and continue from there.

cheers!
-- 
--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------
Article: 9388
Subject: Re: Help with ViewLogic 4
From: Rick Collins <redsp@yahoo.com>
Date: Sun, 08 Mar 1998 22:13:01 -0500
Links: << >>  << T >>  << A >>
Gavin Melville wrote:

> On Thu, 05 Mar 1998 14:45:00 GMT, k.rozniak@XXX.ien.gda.pl (Krzysztof
> Rozniak) wrote:
>
> >Thank you for reply, Gavin and Peter.
> >
> >I managed to find the reason of VL malfunction. The problem was with
> >too long path. I had path about 180 chars long. None of old programs I
> >was using complained about it. VL was the first one. I had to shorten
> >path (now ~65 chars). Since then all works perfectly. Probably VL has
> >a workspace for path limited to 128 chars (DOS <=6.22 limit IIRC).
> >Hope it helps someone.
>
> Oh -- THAT version.   I had not read the original post very carefully.
> It is actually a little more subtle than that -- not only must the
> path be less than 127 characters long, but ALL the directories in that
> path must exist.
> --
> Gavin Melville
> gavin@cypher.co.nz

I remember dealing with this once. I was told that the limitation was that
the comamnd line which included this path had to fit 128 characters. So the
Path had to be somewhat less than that. I seem to recall a practical limit
of about 90 characters.


Rick Collins




Article: 9389
Subject: Re: Viewlogic file format for schematic symbols
From: Rick Collins <redsp@yahoo.com>
Date: Sun, 08 Mar 1998 22:36:00 -0500
Links: << >>  << T >>  << A >>


Peter wrote:

> I have no extra info beyond what you can work out by looking at the
> files. However, do please post your final results - ESPECIALLY if you
> can work out how the magic number is calculated.
>
> This magic number is used to stop things like e.g. using a Xilinx/LCA
> version of Viewdraw to open a sch or lib file which was created in the
> unrestricted version.
>
> I believe that unless you can generate the magic number, you will not
> be able to generate valid Viewlogic files.
>
> The sch file starts with something like
>
> V 50
> K 257863394400 FRED
> Y 0
> D 0 0 1653 1169
>
> Where the 2nd line, after the "K", has the magic number. This, I
> believe, is computed with an algorithm which takes in the filename
> ("FRED" above), a text string identifying the company selling the
> restricted version (e.g. "Xilinx") and some other stuff I don't know.
>
> I doubt Viewlogic will tell you, since they went to a lot of trouble
> to prevent the restricted versions (sold for much less than their own
> full version - $30k at one time) being used to design normal stuff,
> with general libraries like LSTTL.
>

But you don't need to know how it is calculated if you use the same filename
and the same key number. I have done this with success. The key number does
not depend on anything else in the file.

Rick Collins



Article: 9390
Subject: Re: The case for free operating systems and EDA
From: "rk" <stellare@erols.com.NOSPAM>
Date: 9 Mar 1998 03:45:21 GMT
Links: << >>  << T >>  << A >>
Wen-King Su <wen-king@myri.com> wrote in article
<6dvk0l$mcp@neptune.myri.com>...
: In a previous article "rk" <stellare@erols.com.NOSPAM> writes:
: :
: ;wen:
: :: ;: Sorry, I can't do that.  The machine I sometimes need to use would
: ;have
: :: :: so many processors, disks, and memory and powersupply that it would
: ;take
: :: ;: a forklift to carry.  I also routinely use manage jobs that runs on
a
: ;: :: farm of workstations and I need to do it remotely and service them
at
: :: ;: odd hours of the day.
: ;
: :wen:
: ;: But that is the direction of EDA future.  Larger chips more simulation
: :: and analysis, and more difficult place and route.  I already have jobs
: ;: that takes 24 hours per random seed on the fastest pc available.
: :
: ;rk:
: :a bit confused, thought you were working on
: ;super-duper-need-a-forlift-'puter.
: 
: I said that is how long it takes.  I didn't say that is what I use to
: do actual productive work.  I als said that it is for one random seed,
: and I would have used several hundred in each job, and that is still
: not complete to my liking.
: 
: ;there's a large amount of design work going on in the < 100 k gates and
: :under arena.  with a lot of designs < 25 k gates.  these are being
: ;increasingly filled with fpgas.  and a lot of people are currently
: :designing these with (dare i say) pc's and win and some still with dos,
as
: ;we've seen here.  as the pc hardware grows in performance, and it seems
to
: :increase almost every week (and memory is $3/megabyte!), pc developer's
are
: ;setting their heights higher doing larger designs.  and they'll be a
whole
: :generation of chip designers using these tools.  just like we got hit
with
: ;a whole generation of programmers who used c and unix in school and we
know
: :the impact of that.
: 
: When pc's get faster, so would unix running on pc's.  It is not an issue
: of hardware.  The pc will never scale as fast as the demand of computing
: power.  While speed grows much less than linear compared to the physical
: measures of the processor (area, gates, channel length, etc), the
computing
: demand grows much faster.  There are a lot of things that would tend to
: grow with n*n or n*(log n) if not managed carefully.  What is wrong with
: EDA tools running on NT is that it is short sighted and because of poor
: network services, it precludes effective scaling of computational power
: by aggregating a network of computing engines. 

rk:
i agree, for the really large jobs.  and as i said, before, doesn't seem to
be a place for nt.  and that's why i invest in unix workstations, too. 
now, for fpga jobs, for instance, with the improvement in pc technology,
and moving from dos -> 95, and incresing the size and complexity of the
jobs, i find time spent on machine running is decreasing, over all, and i
work more efficiently.  two separate problems, two solutions.

while nt would be shortsighted since it "precludes scaling", it is also
efficient at the lower end of design size, where a large volume of designs
are currently being done.  i know that doing my "lower end" designs on unix
box would be more expensive.  and with some here advocating same price of
tools for all platforms, that would drive development costs up through the
roof, as compared to what i use for dos/win designs.

to get back on thread, a little bit, what will linux mean?  unix
scalability with unix tools at unix prices?  the synopsys guys said same
price, unix and nt.  my guess would be that the people who pay that $ and
use that capability are already using it on unix and there would not be a
great reason to switch to nt and they will have a very big infrastructure. 
some will move but i would predict not the majority.  and the amount of win
people wanting unix software with the limits of nt are not going to fork
over that kind of money.  will win apps port to linux, a unix-type
environment for the scalability and charge win-type prices?  it seems from
the discussion here (and at work) the unix die-hards wouldn't go for that,
they'd want their unix tools for many reasons, an important one not having
to retrain, write new scripts, etc.  and the win people would have to
maintain two machines for the near term.  and idiots like me, would have to
work with three environments.  aaaarrrrgggghhhh!!!!!


rk: 
: :now, for the really big jobs, those that require a farm to run, the
really
: ;mega-jobs at the extreme end of performance.  no doubt that current win
: :technology is no place for that.  but, as an aside, perhaps you could
: ;comment on technologies such as quickturns, where they use boards full
of
: :fpgas (and now custom chips, iirc) to emulate the functionality of the
asic
: ;(pentium 'simulated' this way) and permit running at a sizeable fraction
of
: :actual device speed, in-circuit.  
 
wen:
: Witness the long list of pentium bugs since the processor was released
: as evidence that insufficient verification cycles, even with the aid of
: quickturns, has been devoted to check the design. 

rk:
verification for the mega-chips is a real issue.  and historically, the
list of processor bugs, even for simple ones like the 8086, has been, well,
non-zero.  and, iirc, the 386 multiply bug was a result of the test
consisting of a bunch of random vectors thrown at the chip, not the most
thorough.  can't remember the details of the pentium fdiv bug, other than
they tried to cover it up!  scary problem, really, and these chips are
getting into more and more 'critical' applications.

rk: 
: :	will this be preferable to farms of computers, multi-processors, or
other
: ;simulation techniques?
: :
: ;	will formal equivalence checking of final netlists to original vhdl be
: :accepted for signoff with only a limited amount of gate level
simulation?
: ;
: :	will the use of rapid prototyping (say chip express 1 day turn laser
: ;programmable gate array) replace a lot of the gate level simulations?
 
wen:
: I am afraid none of that will be good enough.  It is going to take a
group
: of really sharp programmers many years in an effort that is not driven
: by the short sighted market forces to come up with a real solution that
: scales.

rk:
and please don't forget the test/verification engineers.  tough (and
thankless) job.


wen:
:                    And only multicomputers of some sort can provide that
scaling
: of computing power required.

rk:
not sure i understand/agree here.  if one can run logic simulations (not
timing, different story) on a multicomputer (pick your favorite type) to
scale simulation performance, how would this logically differ from running
on a quick-turn or rapid asic prototype?  

thanks for the info and discussion,
-- 
--------------------------------------------------------------
rk

"there's nothing like real data to screw up a great theory" 
- me (modified from original, slightly more colorful version)
--------------------------------------------------------------


 
Article: 9391
Subject: Problems with Atmel IDS 5.0 installation
From: ivan@caseware.com (Ivan)
Date: Mon, 09 Mar 1998 07:35:28 GMT
Links: << >>  << T >>  << A >>

	I was wondering if anyone had successfully installed the Atmel
IDS 5.0 (the free version - http://www.atmel.com/fpga_software.html).
I get the following message after trying to run it: "A primitive has
failed". 
	Also, there is WorkView Office on the same CD, and for its
installation a floppy disk is required. I didn't get any disk in the
package though. Anyone had luck with that?

	Thanks,
	Ivan.
Article: 9392
Subject: Re: Viewlogic file format for schematic symbols
From: z80@ds2.com (Peter)
Date: Mon, 09 Mar 1998 08:53:27 GMT
Links: << >>  << T >>  << A >>

>But you don't need to know how it is calculated if you use the same filename
>and the same key number. I have done this with success. The key number does
>not depend on anything else in the file.

OK, others have suggested this too. Therefore, can anyone post a few
magic numbers generated by the unrestricted version of VL? That would
cover all options.

What about libraries? One of the objectives, I assume, of this scheme
was to prevent owners of the restricted versions from using them for
any normal work, e.g. using 74HC libraries from VL.


Peter.

Return address is invalid to help stop junk mail.
E-mail replies to zX80@digiYserve.com but
remove the X and the Y.
Article: 9393
Subject: Floating point representation on FPGA
From: satish_me@hotmail.com
Date: Mon, 09 Mar 1998 05:39:54 -0600
Links: << >>  << T >>  << A >>
I am new to this field (FPGA).

I want to know how to represent the floating point on FPGA?
How floating point operations are carried out in FPGA?

Please answer to these questions
my e-mailaddress is satish_me@hotmail.com

(M.E.Satish Kumar) India.

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/   Now offering spam-free web-based newsreading
Article: 9394
Subject: Re: Whats wrong with this method
From: Jan Zegers <janz@easics.be>
Date: Mon, 09 Mar 1998 14:32:02 +0100
Links: << >>  << T >>  << A >>
Allan Redenbaugh wrote:
> 
> Given the following code for a control register where a single bit has to
> have
> a seperate async reset :
> (target device = Xilinx 4ke)
> 
> process ( strobe, reset, clear_bit0)
> 
> begin
> if reset = '0' then
>    cntl_reg <= DEFAULT_REG;
> elsif clear_bit0 = '1' then
>    cntl_reg(0) <= DEFAULT_REG(0);
> elsif strobe'event and strobe = '1' then
>    cntl_reg <= data;
> end if;
> end process;

There is nothing wrong with your code. I tried it with Synopsys
and it didn't complain.

> 
> I assign reset to GSR so I expect clear_bit0 to drive the reset line of a
> dff
> on bit 0 and the reset of the register bit to only have the GSR reset.
> 

I don't understand what you say, but what you have coded will
work as follows:
  - if reset is asserted '0', then all FFs get DEFAULT_REG value
  - if reset is asserted '1' and clear_bit0 is '1', then only
    bit 0 will get DEFAULT_REG(0).
  - else if there is a positive edge on strobe, sample data.

> My synthesis tool (Leonardo) says that since I have nothing defined for the
> upper bits under the clear_bit0 condition it defaults to a preset which is
> not what I inteded.
> 
> I have pulled out bit 0 into its own process and everythings happy, I just
> don't
> understand why this method doesn't work.
> 
> Any ideas?
> 
> Thanks,
> Allan Redenbaugh
> Raytheon Systems

Kind regards,
Jan
-- 
===================================================================
Jan Zegers                ===              Easics               ===
General Manager           ===  VHDL-based ASIC design services  ===
NEW Tel: +32-16-395 601      ===================================
NEW Fax: +32-16-395 619  Interleuvenlaan 86, B-3001 Leuven, BELGIUM
mailto:janz@easics.be              http://www.easics.com
Article: 9395
Subject: Re: Problems with Atmel IDS 5.0 installation
From: "Martin Mason" <mtmason@ix.netcom.com>
Date: Mon, 9 Mar 1998 07:56:43 -0800
Links: << >>  << T >>  << A >>


Ivan wrote in article <35039a7a.1493447@news.sprint.ca>...

>
> I was wondering if anyone had successfully installed the Atmel
>IDS 5.0 (the free version - http://www.atmel.com/fpga_software.html).
>I get the following message after trying to run it: "A primitive has
>failed".

Please make sure you have the latest patch from :

http://www.atmel.com/atmel/products/prod180.htm


Martin.


Article: 9396
Subject: Re: fghk
From: clown <clown11@geocities.com>
Date: Mon, 09 Mar 1998 12:02:28 -0400
Links: << >>  << T >>  << A >>
Bigwheel...you suck
comp.arch.fpga
too easy
ya gotta change em both.
Article: 9397
Subject: Re: The case for free operating systems and EDA
From: wen-king@myri.com (Wen-King Su)
Date: 9 Mar 1998 09:02:21 -0800
Links: << >>  << T >>  << A >>
In a previous article "rk" <stellare@erols.com.NOSPAM> writes:

:wen:
;:                    And only multicomputers of some sort can provide that
:scaling
;: of computing power required.
:
;rk:
:not sure i understand/agree here.  if one can run logic simulations (not
;timing, different story) on a multicomputer (pick your favorite type) to
:scale simulation performance, how would this logically differ from running
;on a quick-turn or rapid asic prototype?  
:
;thanks for the info and discussion,

With those, you are limited by what the physical test environment can
check.  If you verify your PCI design by plugging a protptype into the
PCI bus of a machine you happen to have, for example, you will get a very
inadequate test coverage.
Article: 9398
Subject: Re: Floating point representation on FPGA
From: Marc Daumas <Marc.Daumas@ENS-Lyon.Fr>
Date: Mon, 09 Mar 1998 18:26:49 +0100
Links: << >>  << T >>  << A >>
satish_me@hotmail.com wrote:
> I want to know how to represent the floating point on FPGA?
> How floating point operations are carried out in FPGA?

I would be interested in any real life (64 bits) documented
implementation of fp arithmetic (possibly IEEE standard) on FPGAs.

-- 
Marc Daumas - Charge de recherches au CNRS (LIP - ENS de Lyon)
mailto:Marc.Daumas@ENS-Lyon.Fr - http://www.ens-lyon.fr/~daumas
ENS de Lyon - 46, allee d'Italie - 69364 Lyon Cedex 07 - FRANCE
Phone: (+33) 4 72 72 83 52 - Fax: (+33) 4 72 72 80 80
Article: 9399
Subject: Re: Whats wrong with this method
From: Mike Treseler <tres@tc.fluke.com>
Date: Mon, 09 Mar 1998 10:46:30 -0800
Links: << >>  << T >>  << A >>
This is a multi-part message in MIME format.
--------------8AD7F76A08015AD1ABADAF78
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Allan Redenbaugh wrote:
> 
> Given the following code for a control register where a single bit has to
> have
> a seperate async reset :
> (target device = Xilinx 4ke)
> 
> process ( strobe, reset, clear_bit0)
> 
> begin
> if reset = '0' then
>    cntl_reg <= DEFAULT_REG;
> elsif clear_bit0 = '1' then
>    cntl_reg(0) <= DEFAULT_REG(0);
> elsif strobe'event and strobe = '1' then
>    cntl_reg <= data;
> end if;
> end process;
> 
> I assign reset to GSR so I expect clear_bit0 to drive the reset line of a
> dff
> on bit 0 and the reset of the register bit to only have the GSR reset.
> 
> My synthesis tool (Leonardo) says that since I have nothing defined for the
> upper bits under the clear_bit0 condition it defaults to a preset which is
> not what I inteded.
> 
> I have pulled out bit 0 into its own process and everythings happy, I just
> don't
> understand why this method doesn't work.
> 
> Any ideas?
> 
> Thanks,
> Allan Redenbaugh
> Raytheon Systems

Try this:

process ( strobe, reset, clear_bit0)
 
begin
if reset = '0' then                  -- reset
    cntl_reg <= DEFAULT_REG;         -- default
    if (clear_bit0 = '1') then
       cntl_reg <= DEFAULT_REG_ALT;  -- alternate default
    end if;                              
elsif strobe'event and strobe = '1' then
    cntl_reg <= data;
end if;
end process;

-- overwriting should be ok since evaluation is at the end
of the process
--------------8AD7F76A08015AD1ABADAF78
Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Mike Treseler
Content-Disposition: attachment; filename="vcard.vcf"

begin:          vcard
fn:             Mike Treseler
n:              Treseler;Mike
org:            Fluke Networks Division
adr;dom:        6920 Seaway Blvd;;;Everett WA;;98203;
email;internet: tres@tc.fluke.com
title:          Staff Engineer
tel;work:       425.356.5409
tel;fax:        425.356.5043
x-mozilla-cpt:  tc.fluke.com;2
x-mozilla-html: TRUE
version:        2.1
end:            vcard


--------------8AD7F76A08015AD1ABADAF78--



Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

Authors:A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Custom Search