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 19800

Article: 19800
Subject: Re: Design security
From: Peter Alfke <peter@xilinx.com>
Date: Wed, 12 Jan 2000 09:03:32 -0800
Links: << >>  << T >>  << A >>


Steve Rencontre wrote:

>
> I've just received an info sheet on the Cypress Delta39k family. Guess
> what? An SRAM CPLD and a flash memory in a 2-dice-per-chip package!
>
> This still isn't going to stop the bad guy with lots of money to spend on
> disassembling the chip, but it's got to be a good thing for intermediate
> security.
>
> --
> Steve Rencontre, Design Consultant
> http://www.rsn-tech.demon.co.uk

What does it take to dissolve some plastic? Twenty bucks?  Fifteen pound
sterling?
Looks like a token security to me. Somewhat like the door-lock on my house...

Peter Alfke, Xilinx Applications


Article: 19801
Subject: Re: Xilinx Spartan2
From: Greg Neff <gregneff@my-deja.com>
Date: Wed, 12 Jan 2000 18:06:14 GMT
Links: << >>  << T >>  << A >>
In article <387A38D6.9C2C0EE9@xilinx.com>,
  peter.alfke@xilinx.com wrote:
(snip)
>
> Spartan2 was officially announced today, Jan 10.
> If you are interested in details, click on
>
> http://www.xilinx.com/products/spartan2/index.htm
>
(snip)

I downloaded the data sheet, but all parts are shown as commercial
temperature only.  Any plans to offer industrial temperature grade?

--
Greg Neff
VP Engineering
*Microsym* Computers Inc.
greg@guesswhichwordgoeshere.com


Sent via Deja.com http://www.deja.com/
Before you buy.
Article: 19802
Subject: Lattice
From: elynum@my-deja.com
Date: Wed, 12 Jan 2000 19:17:05 GMT
Links: << >>  << T >>  << A >>
We have an ISPLSI2128VE Lattice cpld.  It keeps latching up.  The chip
gets really hot, intermittently. At first, I thought it was the fact
that we have 2 different power supplies on the board 3.3V and 5V and
that they were ramping up at different speeds.  Now, I think maybe it is
something with the chip.  Any suggestions?


Sent via Deja.com http://www.deja.com/
Before you buy.
Article: 19803
Subject: Re: Xilinx Spartan2
From: Ray Andraka <randraka@ids.net>
Date: Wed, 12 Jan 2000 20:04:28 GMT
Links: << >>  << T >>  << A >>
What are the differences between an XC2S50 (spartan 2) and a Virtex
XCV50 other than packaging options?

Mike Butts wrote:

> Peter Alfke wrote:
> > Spartan2 was officially announced today, Jan 10.
>
> Getting the Virtex features, especially the 4K
> block RAMs, in a small, lower cost form is great.
>
> You can put a whole CPU and its memory inside
> a Spartan2.  This will be very desirable for
> students and teachers of CPU design.  Also for
> anyone who wants an easily programmed controller
> inside their FPGA design.
>
> So, Peter, when will Spartan2 be supported by the
> Student Edition software?  The whole family,
> as with Spartan, right?
>
> Thanks!
>
>    --Mike Butts

--
-Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930     Fax 401/884-7950
email randraka@ids.net
http://users.ids.net/~randraka


Article: 19804
Subject: Re: 100 MHz counters
From: Ray Andraka <randraka@ids.net>
Date: Wed, 12 Jan 2000 20:09:50 GMT
Links: << >>  << T >>  << A >>
The tough part about doing it this way is capturing the count reliably, especially if
the counted clock period is shorter than the ripple propagation time.  In order to
reliably measure the count, you need to gate the input clock and allow the counter to
settle before taking a reading.

Peter Alfke wrote:

> Different people, different views...
>
> I would use 4 individual clocks, and even use ripple counters, since they offer
> highest frequency resolution ( if needed ) and lowest power consuumption.
> Obviously, you have to accomodate the ripple delay when the counter stops, but
> that seems to be easily done in this design that looks like an event or frequency
> counter.
>
> I am a fan of global clocks and synchronous design, but that's not the best method
>
> for everything.
>
> Once a year I may disagree with Ray. Would become boring otherwise.
>
> Peter Alfke, Xilinx Applications
> ================================
>
> Ray Andraka wrote:
>
> > The way to do this is to synchronize your independent inputs to the global
> > clock then use the synchronized inputs to enable counters working on the global
> > clock.  You'll need a synchronizer circuit that generates a one clock wide
> > pulse for each 0 to 1 transition of the input clocks.  Your local clock will
> > have to be at least as fast as the fastest incoming clock, and if you can do 2x
> > it makes the synchronizing a whole lot easier.  The whole local counter doesn't
> > have to work at 2x, only the first few bits.  For that matter, you could
> > prescale the input clocks using small counters and then synchronize the
> > prescaled clocks to the global clock.  The prescalers should not be more than 2
> > bits (in the same slice or CLB) so that you can run them with a clock that is
> > routed on the general routing rather than on a global clock net.  If you are
> > careful about it, you can run the prescaler at a considerably higher frequency
> > than the global clock tree is capable of.  I think Peter Alfke did an app note
> > to this effect for a 400MHz freq counter a couple years ago (before virtex).
> >
> > "Kresten Nørgaard" wrote:
> >
> > > Andy Peters skrev i meddelelsen <85d9lj$18d7$1@noao.edu>...
> > > >Kresten Nørgaard wrote in message <857h4h$96j$1@news.inet.tele.dk>...
> > > >>Hi group!
> > > >>I'm looking into a new design, consisting of 4 pcs. of  32-bit 100 MHz
> > > >>asynchronous counters. When stopped, the counters are emptied into a FIFO
> > > >>(common to all counters - 32 kbyte size total). The FIFO's will be read
> > > >>through an ordinary 8 MHz CPU interface.
> > > >
> > > >Question: why an async counter?  Especially at 100 MHz?  you'd better off
> > > >with a synchronous counter and some logic that generates count enables.
> > >
> > > Quite right, but I figured, that I would need 4 "global" clocks to make 4
> > > counters, and not all FPGA families features so many distributed clocks.
> > >
> > > Another issue is power consumption. I reckoned, to lower the power
> > > dissipation, if a chose a ripple counter, but I might be wrong on that?
> > >
> > > Kresten
> >
> > --
> > -Ray Andraka, P.E.
> > President, the Andraka Consulting Group, Inc.
> > 401/884-7930     Fax 401/884-7950
> > email randraka@ids.net
> > http://users.ids.net/~randraka

--
-Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930     Fax 401/884-7950
email randraka@ids.net
http://users.ids.net/~randraka


Article: 19805
Subject: Re: HW resources increased
From: "Marinos J. Yannikos" <mjy@pobox.com>
Date: Wed, 12 Jan 2000 21:01:17 GMT
Links: << >>  << T >>  << A >>
Peter da Silva wrote:
> I just checked around my net and while most systems were rebooted the weekend
> of 1/1/00 for political reasons, there's a few who escaped. I haven't found
> a UNIX box with an uptime below 10 days, and most are 50-150 days including
> some that are 10 year old 386 boxes.

I have to ask out of curiosity: what exactly do you do with those
computers? They must consume an awful lot of power... I'm all for
retro-computing and putting old hardware to good use, but I draw
the border where the cost of operating the old systems is so high
that replacing several of them with a single new system makes sense.

-mjy
Article: 19806
Subject: Re: Assignment of pins for thousand+ pin packages
From: jonathan@oxfordbromley.u-net.com (Jonathan Bromley)
Date: Wed, 12 Jan 2000 22:15:45 GMT
Links: << >>  << T >>  << A >>
On Tue, 11 Jan 2000 21:54:04 -0800, Phil Hays
<Spampostmaster@sprynet.com> wrote:

>Are there any users of big parts that care to share a better way to assign pins?
>In the past, I handled pin assignments in the three ways listed below.
<snip explanations>

Can anyone explain to me why we still have to spend gigabucks to get
an EDA environment in which there is at least a teeny weeny little
bit of a relationship between schematic/PCB layout software and FPGA
software, so that FPGA pinning could be an integral part of the PCB
layout backannotation process? :-)

Jonathan Bromley
Article: 19807
Subject: Re: Xilinx Spartan2
From: Peter Alfke <peter@xilinx.com>
Date: Wed, 12 Jan 2000 14:22:45 -0800
Links: << >>  << T >>  << A >>


Greg Neff wrote:

>
> I downloaded the data sheet, but all parts are shown as commercial
> temperature only.  Any plans to offer industrial temperature grade?
>
>

Xilinx will be offering Industrial grade in the -5 speed grade, but we
are still
deciding  on part/package combinations. We are interested in knowing
which
parts have the highest demand for I-grade.
The low price of Spartan is partly due to manufacturing efficiencies
and inventory management. We could easilyt hurt that cost structure by
being too generous with Industrial options.

Peter Alfke, Xilinx Applications



Article: 19808
Subject: Re: HW resources increased
From: "Jeffrey S. Dutky" <dutky@bellatlantic.net>
Date: Wed, 12 Jan 2000 18:23:16 -0500
Links: << >>  << T >>  << A >>
"Marinos J. Yannikos" wrote:
> 
> Peter da Silva wrote:
> > I just checked around my net... I haven't found a UNIX
> > box with an uptime below 10 days, and most are 50-150
> > days including some that are 10 year old 386 boxes.
> 
> I have to ask out of curiosity: what exactly do you do with
> those computers? They must consume an awful lot of power...
> I'm all for retro-computing and putting old hardware to good
> use, but I draw the border where the cost of operating the
> old systems is so high that replacing several of them with
> a single new system makes sense.

The break even point on 'old' super-micros (as opposed to
normal-micros like Apple IIs or IBM PC-XTs) can be pretty
obscure. Anything that can run a unix-derivative and can be
connected to ethernet at nominal cost, can still be pretty
usefull. I run an old 486-66 as my firewall at home, and
we are using some 486-50s as secure X-Terminals (you run
X linux on them and 'tunnle' X through ssh) at my work.

In both cases, it wouldn't make a whole lot of sense to
replace thoses machines with new, low-cost computers (much
less with new, high-cost ones) since the tasks they are
doing don't demand any more processing power than they
provide, they don't consume any more electricity than a
modern machine, and they cost almost nothing to aquire.
(my firewall cost me $15 and the psuedo-X-terminals at
work were just sitting in a closet gathering dust)

It's not as if we're talking about running an old work-
station or a minicomputer (my PDP-11/44 and Sun 3/60 stay
powered off unless I need to get something off old media)
where wattage alone would be prohibitive.

- Jeff Dutky
Article: 19809
Subject: Re: Altera Flex10K bitstream compatibility ?
From: steve (Steve Rencontre)
Date: Thu, 13 Jan 2000 03:00 +0000 (GMT Standard Time)
Links: << >>  << T >>  << A >>
In article <85gbbu$n2h$1@agate.berkeley.edu>, ying@soda.CSUA.Berkeley.EDU 
(Ying C.) wrote:

> Nothing will happen. 10K20 simply can't be configured with 10K10 
> bitstream.
> Since they are pin to pin compatible, the device will try to configure 
> and then get
> stuck in the error out mode. (ie. nStatus goes high)
> 
> Ying
> 
> In article <387B0CFA.FBFC56FE@dotcom.fr>,
> Nicolas Matringe  <nicolas@dotcom.fr> wrote:
> >Hi
> >One of our engineers who doesn't know much about FPGAs just told me 
> "All
> >right(, I replaced the 10K10 with a 10K20, this should work fine". I
> >stopped him, telling that it's not because they are pin to pin
> >compatible that they can be programmed with the same bitstream.
> >Any of you knows what would happen if we tried? Just curious...

Since the '10 bitstream is shorter than the '20, I'd personally expect 
the latter just to hang expecting more bits.

--
Steve Rencontre, Design Consultant
http://www.rsn-tech.demon.co.uk
Article: 19810
Subject: Re: SDRAM controller ?
From: "Joel Kolstad" <Joel.Kolstad@USA.Net>
Date: Wed, 12 Jan 2000 20:38:10 -0800
Links: << >>  << T >>  << A >>
Simon D. Wibowo <simon.wibowo@infineon.com> wrote in message
news:85etg6$964$1@mosquito.HL.Siemens.DE...
> What, do you think, is the best FPGA for SDRAM controller ? Say, up to
PC100

Xilinx has a pretty large number of applicaion notes for interfacing their
Virtex parts to various flavors of high speed RAM, including SDRAM.  It's
pretty impressive how fast Virtex can fly, especially the Virtex-E parts.

The application notes usually describe relatively simple-minded controllers
that are more suitable for applications such as data logging rather than as
processor/RAM bridges, but they're a good starting point.

I couldn't really say about "best."  I can say that Xilinx works.

---Joel Kolstad



Article: 19811
Subject: Re: SDRAM controller ?
From: "Joel Kolstad" <Joel.Kolstad@USA.Net>
Date: Wed, 12 Jan 2000 20:38:10 -0800
Links: << >>  << T >>  << A >>
Simon D. Wibowo <simon.wibowo@infineon.com> wrote in message
news:85etg6$964$1@mosquito.HL.Siemens.DE...
> What, do you think, is the best FPGA for SDRAM controller ? Say, up to
PC100

Xilinx has a pretty large number of applicaion notes for interfacing their
Virtex parts to various flavors of high speed RAM, including SDRAM.  It's
pretty impressive how fast Virtex can fly, especially the Virtex-E parts.

The application notes usually describe relatively simple-minded controllers
that are more suitable for applications such as data logging rather than as
processor/RAM bridges, but they're a good starting point.

I couldn't really say about "best."  I can say that Xilinx works.

---Joel Kolstad



Article: 19812
Subject: Reliability of programming SRAM FPGAs
From: Richard Lamoreaux <lumpy@rtd.com>
Date: Thu, 13 Jan 2000 04:59:48 GMT
Links: << >>  << T >>  << A >>
What is the reliability of the PROM based download programming as
opposed to FLASH based devices or antifuse devices? I am looking at
using a Xilinx Virtex device, but I need to have 100% reliability in
the functionality after the download ends, passes CRC, and comes out
of tristate. I understand that if the CRC is bad, the device will shut
down, and that's OK.

Rich

Article: 19813
Subject: Re: Reliability of programming SRAM FPGAs
From: Ray Andraka <randraka@ids.net>
Date: Thu, 13 Jan 2000 06:07:14 GMT
Links: << >>  << T >>  << A >>
From what I've seen it's pretty good.  The registers that hold the
configuration are pretty robust, and hold the configuration to some
amazingly low supply voltages.  You are not likely to see problems with
the FPGA losing its marbles if that is what you are asking.  For space
applications, you will probably be concerned about upset due to radiation
etc.  In that case, you need to monitor the configuration somehow and
reconfigure if the thing breaks.  As far as the configuration itself goes,
as long as you keep the configuration clock clean and follow the
recommended sequence and timing in the databooks, you shouldn't have any
problems.

Richard Lamoreaux wrote:

> What is the reliability of the PROM based download programming as
> opposed to FLASH based devices or antifuse devices? I am looking at
> using a Xilinx Virtex device, but I need to have 100% reliability in
> the functionality after the download ends, passes CRC, and comes out
> of tristate. I understand that if the CRC is bad, the device will shut
> down, and that's OK.
>
> Rich

--
-Ray Andraka, P.E.
President, the Andraka Consulting Group, Inc.
401/884-7930     Fax 401/884-7950
email randraka@ids.net
http://users.ids.net/~randraka


Article: 19814
Subject: Re: HW resources increased
From: Ketil Z Malde <ketil@ii.uib.no>
Date: Thu, 13 Jan 2000 06:58:59 GMT
Links: << >>  << T >>  << A >>
"Larry Elmore" <ljelmore@montana.campuscw.net> writes:

> "Ketil Z Malde" <ketil@ii.uib.no> wrote in message

> I think you just answered your own question. If Redhat et al "are hurting
> pretty bad from their rushing out unfinished products," then they've got a
> real incentive to be more conservative and do it right next time, don't
> they? Well, maybe not. Microsoft and a lot of others do this all the time.

Interesting point, but irrelevant.  Red Hat isn't rushing development
of open source software, only packaging and sales.  And Red Hat is in
many ways the enfant terrible here, Debian and S.u.S.E seems to be a
lot more careful.

Basically, I find that it is commercial software that gets features
at the expense of bug fixes, and that gets rushed out the door
unfinished or at least poorly tested.

>> If you want to live on the cutting edge, use
>> "unstable", and deal with it.

> Like the first two versions of almost any MS product...

Sorta.  Of course, if you run unstable, you're expected to help stamp
out the bugs.  Debian has a pretty good system with package
maintainers and bug-squash parties on IRC and so on.

Are we still within the scope of any of these groups? :-)

-kzm
-- 
If I haven't seen further, it is by standing in the footprints of giants
Article: 19815
Subject: Re: Reliability of programming SRAM FPGAs
From: Nicolas Matringe <nicolas@dotcom.fr>
Date: Thu, 13 Jan 2000 09:03:11 +0100
Links: << >>  << T >>  << A >>
Ray Andraka wrote:
> 
> From what I've seen it's pretty good.  The registers that hold the
> configuration are pretty robust, and hold the configuration to some
> amazingly low supply voltages.

I sopmetimes encounter a problem with an Altera Flex10K which is on an
ISA extension card. It sometimes loses its configuration after an
electrostatic discharge on the computer case. I must reckon the
decoupling of the Vcc pins is quite poor, too :o)

Nicolas MATRINGE           DotCom S.A.
Conception electronique    16 rue du Moulin des Bruyeres
Tel 00 33 1 46 67 51 11    92400 COURBEVOIE
Fax 00 33 1 46 67 51 01    FRANCE
Article: 19816
Subject: Re: Help for EDIF format !
From: fliptron@netcom.com (Philip Freidin)
Date: 13 Jan 2000 08:15:20 GMT
Links: << >>  << T >>  << A >>

Not that I like being the bearer of bad news, but you may be far more
hosed than you realize.

If your intent is to take this old design and put into a current fpga
then your only challenges are the ones you are currently facing. (which
arent trivial. see notes on libraries below)

But...

If you are going to be updating/changing the design and then loading
it into existing XC3000 chips on boards (Not XC3000A or XC3000L or
3100A or 3100L) then you are in much more trouble, because NONE of
the Mx.y software knows how to create valid bitstreams for XC3000
devices (or XC2000 or XC4000 or XC4000H or XC4000A).

The last version of Xilinx software that knows how to do this is
XACT V5.2.1/V6.0.1

See more notes below.

In article <85hu8h$tk2$1@nnrp1.deja.com>,  <himalayas@my-deja.com> wrote:
>Thanks Philip,
>    in fact the true problem I am facing is that ,
>my design project(XC3000) is based on a design someone developed
>years ago using workview office 4.3(under dos),now my box runs on
>win98,and I can only open the design project using WVO7.53(just as
>you have said,xilinx foundation 2.1i cannot import the project),and to
>implementation the design I have to use Foundation 2.1i. During the
>implemetation(yes,the target netlist file is exported from WVO753 with
>the "level" property set to "Xilinx"),the software(flow engine) window
>shows "Writing the design to ....ngo",then "Reading ....ngo",then
>"Reading component libraries for design expansion...",then it
>stops,after some minutes,it bursts out lots of error messages such as
>"ERROR:NgdHelpers:312 - logical block "$1I1\$1I1\$1I1\$1I1" of type
>"DFF" is unexpanded." ,

The problem here is pretty convoluted. The library you must be using
is a non-unified library. Since your design obviously has DFF symbols
in it, this is the right library, but Oh dear .....  This stuff must be 
about 1991 vintage or earlier. It was around this time that the unified 
libraries were created, and among the casualties was the DFF symbol,
which I believe was replaced by the FDCE symbol. This lives on to the
current libraries that work with the M2.1i/F2.1i software.
There never was an easy way to migrate these old designs to work with
unified libraries.

Last time I did this, I ran Viewdraw with the pre-unified libraries,
and printed out everything.

Then I started a new design, using current libraries, and re-drew
everything. (there were a few black-belt secrets I used to make
the job not too hard, but "easy" it wasn't).


>"ERROR:NgdHelpers:312 - logical block "$1I1\$1I1\$1I2\$1I17\$1I1" of
>type "DFF" is unexpanded.",etc. of course the software exits.
>
>It seems that Foundation 2.1i doesn't recognize the component libraries
>(builtin,mx3000,x3000,xblox,xpal,xttl..)used in the design with WVO 4.3
>under DOS?And how can I port the project onto the current EDA
>environment?

See above.

>
>Would anyone help me?:-)
>

Not a fun project  :-(

>Thanks again,
>Jimmy
>
>
>
>Sent via Deja.com http://www.deja.com/
>Before you buy.


Philip

Article: 19817
Subject: Re: Design security
From: satish_me@hotmail.com
Date: Thu, 13 Jan 2000 09:27:14 GMT
Links: << >>  << T >>  << A >>


Hai,
Unless you use one time coding FPGA, there will not be any possibility
of getting high security. The SRAM based FPGA are good for performing
exercise, such that whether the FPGA meets our necessary requiremnts.
Once u go for One time coding FPPGAs(Altera.....), ur design have High
security, unless the FPGA is stolen or Mal functioning.
This is my view for Design security.......
In SRAM based, FPGAs the PROM contents can be occupied erased, by any
power full magnetic filed. This is one of the major disadvantage of
SRAM based FPGA's.

Satish...
In article <3878359A.69D0@DesignTools.co.nz>,
  jim.granville@DesignTools.co.nz wrote:
> Hal Murray wrote:
> >
> > In article <3870B679.D84288FD@ids.net>,
> >  Ray Andraka <randraka@ids.net> writes:
> > > The problem with putting an OTP key in the device, is that the
non-volatile
> > > cells can't be fabricated without additional process steps.  The
FPGA
> > > process is essentially the same as DRAM, which lets it be done
with bleeding
> > > edge process.  Put PROM cells on there, and you lose the speed.
> >
> > I'm not a silicon guru.  This case is different from a normal
> > PROM in that we don't need a lot of bits.
>
> But you do - the whole config memory needs to go on chip, not just the
> security bit(s), and  you'll want the config memory re-writable as
well,
> otherwise the anitfuse devices are equivalent.
>
>  A few bits of write-once is not a big problem - you just arrange a
> kill-able element.
>
>  Gatefield/Actel have FLASH FPGA, but the very leading edge will
always
> be
> RAM based.
>
>  A design option for the security paranoid is to split the design
> between
> the FPGA, and a sister CPLD - the FPGA can be cloned, but the system
> cannot.
> In a QFP44 you can get 64/72 Macrocells, which can be quite a 'lock'
>
> - jg
>
>


Sent via Deja.com http://www.deja.com/
Before you buy.
Article: 19818
Subject: Re: PCI Bus Problems with Burst Transfers
From: "Peter Lang" <Peter.Lang@rmvmachinevision.de>
Date: Thu, 13 Jan 2000 12:41:08 +0100
Links: << >>  << T >>  << A >>
Here some additional information concerning our Retry problem:
We are using the Philips SAA 7164 Multimedia Chip as PCI Master.

The main question I have is: Why is the PCI Target not abel to accept
my Burst for such a long time (answering my request with retries).
Is there a deadlock situation?
I suppose there is a problem with the Target (Intel 82443LX Chip)
on our motherboard in comination with the used DisplayController (S3 ViRGE
GX2)

thanks for any hints
... peter






Article: 19819
Subject: call for comments
From: Jamil Khaib <Khatib@ieee.org>
Date: Thu, 13 Jan 2000 14:32:53 +0200
Links: << >>  << T >>  << A >>
Hi,
OpenIPCore Project tries to define set of openhardware design
methodology and concepts

we currntly have some designs and ideas we would like to ask from you to
give us some comments on our ideas and designs

Thanks
Jamil Khatib
OpenIP Organization   http://www.openip.org
OpenIPCore Porject   http://www.openip.org/oc
OpenCores Project     http://www.opencores.org

Article: 19820
Subject: Re: fastest 32 bit RISC
From: "Ulf Samuelsson" <ulf.samuelsson@atmel.spamme.com.not>
Date: Thu, 13 Jan 2000 14:01:27 +0100
Links: << >>  << T >>  << A >>
I do not know the max speed available in current technologies,
but I simulated a 32 bit ALU and a 32 bit barrel shifter on an Atmel 40K
FPGA
architecture which is similar to the Xilinx architecture.

The critical path was around 250-300 ns for both.
Since in a real 32 bit processor, this is just part of single stage in a
processor,
you will find that high MHz processors are hard to find.

I doubt that you can get 50 Mhz with an FPGA implementation.
If you are lucky You might get  10-15 Mhz in very advanced technology.

Also they will be quite expensive.
Just the ALU and the barrel shifter is 30 kgates in itself.
I have seen data on a fullblown 32 bitter indicating about 120k gates
for a RISC processor.
An off the shelf ARM7 will be soo much cheaper.

Pretty soon, you can get silicon with a processor integrated
with the FPGA and then you will have low cost 40-50 MHz.

The 8 bit AVR RISC + FPGA (FPSLIC) will soon do 40 MHz.


--
This is a personal view which may or may not be shared
by my employer         Atmel Sweden
Ulf Samuelsson         ulf 'a't atmel 'd'o't com

Damjan Lampret skrev i meddelandet <85clgf$o03$1@planja.arnes.si>...
>Hi everybody,
>
>does anyone know how fast (in terms of clock speed or/and in terms of
>performance) and how big is the fastest 32 bit RISC in FPGA (not restricted
>to any particular FPGA vendor). Processor can be commercial or open source.
>It would be best if it is with caches and MMU. Thanks.
>
>regards, Damjan
>
>
>


Article: 19821
Subject: Re: PCI Bus Problems with Burst Transfers
From: jhallen@world.std.com (Joseph H Allen)
Date: Thu, 13 Jan 2000 13:04:34 GMT
Links: << >>  << T >>  << A >>
In article <85evl4$a9f$1@news01.btx.dtag.de>,
Peter Lang <Peter.Lang@rmvmachinevision.de> wrote:
>Hi to all,

>I am not shure this is the right newsgroup for my problem, but maybe
>someone can give me a hint or the name of a better fitting newsgroup.

>We are designing PCI-Framegrabber cards using standard PC-platforms and
>Windows98 The problem is that the Video Data DMA Burst via PCI-Bus
>sometimes stopps for about 30 microsec. While that time alle PCI
>Data-Transfers are aborted with a RETRY termination. As a consequence our
>Video Fifos get a Data Overflow!

I think you need a frame buffer on your card.  There are no bandwidth
guarantees with PCI.

-- 
/*  jhallen@world.std.com (192.74.137.5) */               /* Joseph H. Allen */
int a[1817];main(z,p,q,r){for(p=80;q+p-80;p-=2*a[p])for(z=9;z--;)q=3&(r=time(0)
+r*57)/7,q=q?q-1?q-2?1-p%79?-1:0:p%79-77?1:0:p<1659?79:0:p>158?-79:0,q?!a[p+q*2
]?a[p+=a[p+=q]=q]=q:0:0;for(;q++-1817;)printf(q%79?"%c":"%c\n"," #"[!a[q-1]]);}
Article: 19822
Subject: Re: PCI Bus Problems with Burst Transfers
From: Thomas Rathgen <trathgen@gmx_cut_.de>
Date: Thu, 13 Jan 2000 15:12:06 +0100
Links: << >>  << T >>  << A >>
Hi,

you may want to ask PCI folks,
there is a mailing list, to subscribe send a SUBSCRIBE(subject) message
to pci-all-request@znyx.com ....

As pointed out by Joseph, you probably need a larger FIFO (as large as 1
MByte ??).
BTW, I don't think the 7164 is the rigth bridge in such design.

rgds

  Tom



Peter Lang schrieb:
> 
> Here some additional information concerning our Retry problem:
> We are using the Philips SAA 7164 Multimedia Chip as PCI Master.
> 
> The main question I have is: Why is the PCI Target not abel to accept
> my Burst for such a long time (answering my request with retries).
> Is there a deadlock situation?
> I suppose there is a problem with the Target (Intel 82443LX Chip)
> on our motherboard in comination with the used DisplayController (S3 ViRGE
> GX2)
> 
> thanks for any hints
> ... peter
Article: 19823
Subject: Re: HW resources increased
From: Sander Vesik <sander@haldjas.folklore.ee>
Date: 13 Jan 2000 14:36:39 GMT
Links: << >>  << T >>  << A >>
In comp.arch Marinos J. Yannikos <mjy@pobox.com> wrote:
> Peter da Silva wrote:
>> I just checked around my net and while most systems were rebooted the weekend
>> of 1/1/00 for political reasons, there's a few who escaped. I haven't found
>> a UNIX box with an uptime below 10 days, and most are 50-150 days including
>> some that are 10 year old 386 boxes.

> I have to ask out of curiosity: what exactly do you do with those
> computers? They must consume an awful lot of power... I'm all for

They consume no more than "modern" computers and may well consume less
power than the desktop computers presently sold.

> retro-computing and putting old hardware to good use, but I draw

It's not retro-computing. There are lots of embedded 386 based computers,
intel still sells 386s, and I don't think it is "not suggested for new 
designs".

> the border where the cost of operating the old systems is so high
> that replacing several of them with a single new system makes sense.

Operating a 386 based computer in which say one of the *BSD runs is no
more costly than operating say a pentium on which one of the *BSD runs.

The same applies to say Sun IPX 8-)

> -mjy

-- 
	Sander

	There is no love, no good, no happiness and no future -
	these are all just illusions.
Article: 19824
Subject: Re: HW resources increased
From: "Marinos J. Yannikos" <mjy@pobox.com>
Date: Thu, 13 Jan 2000 14:50:22 GMT
Links: << >>  << T >>  << A >>
Sander Vesik wrote:
> > computers? They must consume an awful lot of power...
> 
> They consume no more than "modern" computers and may well consume less
> power than the desktop computers presently sold.

But if the CPU cycles are put to any use, dozens of old 386 boxes can
be replaced by a modern system, eliminating much power consumption.

I won't even switch my old P133 on, since CPU-intensive stuff can run
better on my newer systems and the P133 would make too little of 
a difference to justify the power consumption and noise.

-mjy


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