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 24400

Article: 24400
Subject: Re: FPGA selection
From: korthner@hotmail.nospam.com (K. Orthner)
Date: 7 Aug 2000 01:38:06 GMT
Links: << >>  << T >>  << A >>
jimmy75@my-deja.com wrote in <8mjom3$6vc$1@nnrp1.deja.com>:

>Spartan-II is a cheap version of Virtex devices with some features
>removed (e.g. assynchronous BlockRAMs). But you probably do not need
>these features anyway. I believe that the prices are < $10 a unit.
>Check it out with Xilinx.

ack!
I thought that the functionality of the BlockRAMs in the Spartan-II were 
exactly the same as the Virtex.  

Can someone please explain to me the differences?

Thanks!

-Kent
Article: 24401
Subject: Crossing Clock Domains.
From: korthner@hotmail.nospam.com (K. Orthner)
Date: 7 Aug 2000 03:11:28 GMT
Links: << >>  << T >>  << A >>
I'm working on a design where I have two clock domains.

I've decided to implement everything crossing the two domains using FIFOs, 
both BlockRAM-based, and distributed RAM based.

I've looked at two application notes that deal with this, One aimed at the 
Spatan-II, and one for the 4000 series.

In the appnote for the 4000 series (XAPP051, 1996, Peter Afke), Grey 
counters are used to improve speed and to simplify the generation of the 
FULL and EMPTY signals to prevent "weird" states.  The new appnote for 
Spartan-II (XAPP175, 11/99) discusses pretty much the same thing, using 
grey counters and what-not, as well as keeping copies of "next_pointer"s 
make full and empty synchronous signals.

The new appnote doesn't seem to consider how to prevent problems arising 
from using the two clock domains, however.  I've copied what it said below, 
for anyone who wants to read it.

In the old appnote, it suggested using latches, to latch the empty/fuill 
signal that was generated in the other clock domain, and then re-clocking 
the output of the latch with the desired clock.

So, my question is, is it really suggested to use a latch at the output of 
simple combinatorial logic to "stretch" the full/empty signal and then re-
clock it?

I'm somewhat new to working in two clock domains, and I would rather figure 
it all out now, as I'm doing my design, then later when I'm dabugging it.

Thanks a million.

-Kent


From XAPP175:

"To solve this problem, and to maximize the speed of the control logic, 
additional logic complexity is accepted for increased performance. There 
are primary 9-bit Read and Write binary address counters, which drive the 
address inputs to the Block RAM. The binary addresses are converted to 
Gray-code, and pipelined for a few stages to create several address 
pointers (read_addrgray, read_nextgray, read_lastgray, write_addrgray,
write_nextgray) which are used to generate the Full and Empty flags as 
quickly as possible. Gray-code addresses are used so that the registered 
Full and Empty flags are always clean, and never in an unknown state due to 
the asynchronous relationship of the Read and Write clocks. In the worst 
case scenario, Full and Empty would simply stay active one cycle longer,
but this would not generate an error."

Article: 24402
Subject: Re: FPGA selection
From: Peter Alfke <palfke@earthlink.net>
Date: Mon, 07 Aug 2000 03:15:17 GMT
Links: << >>  << T >>  << A >>


"K. Orthner" wrote:

> jimmy75@my-deja.com wrote in <8mjom3$6vc$1@nnrp1.deja.com>:
>
> >Spartan-II is a cheap version of Virtex devices with some features
> >removed (e.g. assynchronous BlockRAMs). But you probably do not need
> >these features anyway. I believe that the prices are < $10 a unit.
> >Check it out with Xilinx.
>
> ack!
> I thought that the functionality of the BlockRAMs in the Spartan-II were
> exactly the same as the Virtex.
>

You are right, there is no functional difference at all, but the speed
parameters are slightly different.
And the temperature-measuring diode has been traded for a power-down input
in Spartan-II.
And the available packages are quite different.

Peter Alfke, Xilinx Applications

Article: 24403
Subject: Re: FPGA selection
From: Peter Alfke <palfke@earthlink.net>
Date: Mon, 07 Aug 2000 03:18:17 GMT
Links: << >>  << T >>  << A >>


"K. Orthner" wrote:

> jimmy75@my-deja.com wrote in <8mjom3$6vc$1@nnrp1.deja.com>:
>
> >Spartan-II is a cheap version of Virtex devices with some features
> >removed (e.g. assynchronous BlockRAMs). But you probably do not need
> >these features anyway. I believe that the prices are < $10 a unit.
> >Check it out with Xilinx.
>
> ack!
> I thought that the functionality of the BlockRAMs in the Spartan-II were
> exactly the same as the Virtex.
>

You are right, there is no functional difference at all, but the speed
parameters are slightly different.
The LUT-RAM writing became synchronous in the olden days of XC4000-E. Much
nicer, so you don't have to time the write pulse so tightly.
The Virtex temperature-measuring diode has been traded for a power-down
input in Spartan-II.
And the available packages are quite different.


Peter Alfke, Xilinx Applications

Article: 24404
Subject: Re: FPGA selection
From: korthner@hotmail.nospam.com (K. Orthner)
Date: 7 Aug 2000 03:43:05 GMT
Links: << >>  << T >>  << A >>
If there's no functional difference at all, does that mean that jimmy75 was 
mistaken?

-Kent


>You are right, there is no functional difference at all, but the speed
>parameters are slightly different.
>
>Peter Alfke, Xilinx Applications
>
>

Article: 24405
Subject: Re: Crossing Clock Domains. (Hey Peter Alfke!)
From: Phil Hays <spampostmaster@sprynet.com>
Date: Sun, 06 Aug 2000 21:39:25 -0700
Links: << >>  << T >>  << A >>
"K. Orthner" wrote:

> I'm working on a design where I have two clock domains.

Always remember that flipflops clocked by one domain and receiving data from the
other domain can be metastable -- for more detail see application note 97, which
gives data for 4000 series.

(Peter Alfke, when can we expect that some more modern devices will get
measured?)


A gotcha that needs checking for is that Xilinx PAR can and will duplicate FF's
for it's own purposes, such as to improve routing or timing.  The answer below
discusses this for the xc4000, but I would assume that PAR is still guilty until
proven innocent.  This is fine for most of the designs, most of the time, but
not for signals that cross clock domains.

http://support.xilinx.com/techdocs/3813.htm


-- 
Phil Hays
Article: 24406
Subject: Can i see Gate-delay and Interconnection-delay of circuit on FPGA and where??
From: "Phunjapa Ruangsinsup" <mpr@siampage.com>
Date: Mon, 7 Aug 2000 12:24:10 +0700
Links: << >>  << T >>  << A >>
I must design asynchronous circuit on FPGA and I must know Gate-delay and
interconnection delay for create Complete signal check (the signal which
have delay longest than another in circuit). I design by RTL-VHDL code and
synthesis by Xilinx Foundation 2.1i, my synthesis output is VHDL and SDF
code. So can i use delay report from SDF code to be a Gate-delay?? and Where
i can see interconection-delay of each signal??


Article: 24407
Subject: Re: Memory specification
From: gstse@home.com (Gary Tse)
Date: Mon, 07 Aug 2000 06:36:08 GMT
Links: << >>  << T >>  << A >>
In article <398B3965.8BA09BD3@ic.unicamp.br>, Rodolfo Jardim de Azevedo wrote:
>I need SRAM modules (I think some
>PCs use them as cache memory, that would be the best choice). Where can
>I get some specification of such modules? How to make an interface?
>(pinouts, time diagrams, ...)

http://www.idt.com/products/pages/SRAM-PL101_Sub244_Dev215.html , for 
example.  I am sure other SRAM vendors sell modules also.

-- 
Gary Tse, gstse@home.com			"Will write Verilog for food."

Article: 24408
Subject: Re: Help! Troubles using async FIFO cores in Virtex
From: erika_uk@my-deja.com
Date: Mon, 07 Aug 2000 08:12:38 GMT
Links: << >>  << T >>  << A >>
hey,

what are the advantages from using Johnston counters in FIFO design

--Erika

Thanks

In article <01bffffb$fa063c10$8c0af7a5@drt1>,
  "Austin Franklin" <austin@d44arkroom.com> wrote:
> > Clocks are not very fast BTW, one is 40MHz and another
> > just 10MHz...
>
> You're welcome.
>
> clocks, eh?  Does that mean the FIFO has one clock for its input
address
> counter/flags and one for its output address counter/flags?  Are you
using
> gray coded counters, and have a single point of synchronization
between
> time domains?  Your flags must be referenced in the time domain you
are
> using them in...
>
> What is your input tool, Verilog/Synplicity or
> Schematics/Viewdraw....perhaps?
>
>


Sent via Deja.com http://www.deja.com/
Before you buy.
Article: 24409
Subject: Drivers for Fujitso MB87Y237 USB core
From: "Shlomo Kut" <skut@vyyo.co.il>
Date: Mon, 7 Aug 2000 14:09:58 +0200
Links: << >>  << T >>  << A >>
Hi,

Does anybody know of any third party drivers for the Fujitsu MB87Y237 USB
Core ?

--
Shlomo K
--------------------------------------------------
Shlomo Kut
Vyyo Ltd
PO Box 45017 Jerusalem 91450
Tel 972-2-5889715
Home 02-5618098
email: skut@vyyo.co.il


Article: 24410
Subject: Xilinx Foundation 3.1i
From: Rick Filipkiewicz <rick@algor.co.uk>
Date: Mon, 07 Aug 2000 13:36:30 +0100
Links: << >>  << T >>  << A >>

Is it worth upgrading from 2.1iSP6 to the new s/w ? Any experiences good
or bad ?

Considering that I now have to pay the ``time based license'' fee aka
`the shareholders are complaining and chip prices are competitive so
lets make more money out of the s/w' I want to know what the benefits
are, if any.

I don't actually mind paying a fair bit for Xilinx s/w - our original
total outlay up to XCV1000 was about UKP3500 - since in general its
pretty good. But I do strongly object to it timing out after a year. I
want to buy the s/w, not rent it.

Note for the lawyers: Normally if you are renting something and it goes
wrong its the responsibility of the owner to get it fixed or replaced.
Does this mean I can demand an instant fix for any serious bug  I trip
over ?


Article: 24411
Subject: Re: Xilinx Foundation 3.1i
From: Nicolas Matringe <nicolas@dotcom.fr>
Date: Mon, 07 Aug 2000 14:55:50 +0200
Links: << >>  << T >>  << A >>
Rick Filipkiewicz a écrit :

> I don't actually mind paying a fair bit for Xilinx s/w - our original
> total outlay up to XCV1000 was about UKP3500 - since in general its
> pretty good. But I do strongly object to it timing out after a year.
> I want to buy the s/w, not rent it.

Hi Rick
From what I've heard, you don't rent it and it won't timeout, you just
won't be able to use new chips if you don't buy the new version (in
fact, you buy maintenance for one year)
I hope I got it right...

-- 
Nicolas MATRINGE           DotCom S.A.
Conception electronique    16 rue du Moulin des Bruyeres
Tel +33 1 46 67 51 11      F-92400 COURBEVOIE - FRANCE
Fax +33 1 46 67 51 01      http://www.dotcom.fr/
Article: 24412
Subject: Re: Abel from dataIO?
From: Paul Smart <pablo@maine.rr.com>
Date: Mon, 07 Aug 2000 09:14:12 -0400
Links: << >>  << T >>  << A >>
On Sat, 05 Aug 2000 21:04:08 -0500, Jack Lai <jwlai@mmmpcc.org> wrote:

>Does anyone know who pickup the ABEL product line Dataio?
>
>Thanks, in advance for your help.
>
>Jck

I'm pretty sure Xilinx bought it, but I don't know the details...

Article: 24413
Subject: Re: Help! Troubles using async FIFO cores in Virtex
From: David Kessner <davidk@free-ip.com>
Date: Mon, 07 Aug 2000 09:10:51 -0600
Links: << >>  << T >>  << A >>
Mikhail Matusov wrote:
> I believe you are right and it is a timing problem. 

There is a problem with Foundation/Alliance and the FIFO cores 
in the Xilinx app notes (which I assume is the same as what's
in coregen).  This problem manifests itself as a very over-optimistic
timing report.

Look at the Xilinx "answer" #6450 at 
http://support.xilinx.com/techdocs/6450.htm

What this "answer" is saying is that the timing analyzer and the
timing stuff in PAR doesn't, by default, analyze a timing path 
through an async set/reset input to a flip flop (FF).  Let's say
that there are three flip flops, A B and C.  The output of A feeds
the async reset of B, which outputs to the D input of C.  If A and 
C use the same clock then the path through B is really just a 
combinatorial path-- however Foundation doesn't see it as a single
path.  Instead, it breaks it up into two paths, A-B and B-C.  The
net result is that the true path A-C isn't analyzed.  

If you look at some of the Xilinx app notes, you'll see this 
exact logic structure used to generate the full/empty flags.
So, when you run the FIFO fast, you will be violating setup
and hold time requirements and those flags will no longer
work properly.  The true speed is approximately 50% slower
than what's reported by the timing analyzer.

It also appears that the speeds reported in the Xilinx app
notes are based on the same bad timing data.  So, when they
are reporting speeds of 150 MHz, reality is closer to 100
MHz (adjust for your own design constraints).

I found this problem recently when debugging my own FIFO
design.  I used the same type of async reset system that
Xilinx used and had the same type of timing problems.  
I have recently rewritten my FIFO to not use async resets
while keeping the high speeds.  This FIFO will soon be
released as an upgrade to the Free-RAM core on the Free-IP
Project web site.  If this FIFO would be useful for you now,
email me directly.


Hope that helps!

David Kessner
davidk@free-ip.com
http://www.free-ip.com
Article: 24414
Subject: Re: Crossing Clock Domains.
From: rickman <spamgoeshere4@yahoo.com>
Date: Mon, 07 Aug 2000 11:11:55 -0400
Links: << >>  << T >>  << A >>
"K. Orthner" wrote:
> 
> I'm working on a design where I have two clock domains.
> 
> I've decided to implement everything crossing the two domains using FIFOs,
> both BlockRAM-based, and distributed RAM based.
> 
> I've looked at two application notes that deal with this, One aimed at the
> Spatan-II, and one for the 4000 series.
...snip...
> I'm somewhat new to working in two clock domains, and I would rather figure
> it all out now, as I'm doing my design, then later when I'm dabugging it.
> 
> Thanks a million.
> 
> -Kent

There is more than one way to skin a cat. Sorry fluffy... 

The FIFO described in the appnote is the one that is the hardest to
design and is used in the most stringent conditions. But you don't
always have to do it that way. 

I don't remember the details of the FIFO in the app note, but it used
some slightly tricky stuff to make sure that any timing "issues" with
the flags always erred on the safe side of telling you it was full when
it was not, etc... This circuitry made it a little harder to extend the
size of the FIFO and just was not my cup of tea if I didn't have to use
it.

If your design is not pumping data on every clock cycle, you can put the
FIFO entirely in one clock domain and cross the clock domains using a
data register and a pair of FFs for the "data avail" signal. The pair of
FFs are clocked by the "TO" domain and will keep the control signal from
being metastable (within the reliablity requirements). The data register
is clocked by the "FROM" domain but will not cause any problems because
of the control signal delay. 

If you need to bring the FULL signal or an ACK signal back to the "FROM"
domain, you will need to double FF it in the "FROM" clock domain. Keep
in mind that the FULL signal needs to compensate for the FF delay, so
set it high one or two samples early depending on your requirements. 

Because your counters are all in one domain, you can use regular
counters. I used three; read addr, write addr and count. I think this is
easier than comparing the addr counters. 

This can be a little complicated, but it is much simpler than the two
clock FIFO. You just have to make sure that the limitations of this
circuit fit your requirements. 


-- 

Rick Collins

rick.collins@XYarius.com

Ignore the reply address. To email me use the above address with the XY
removed.



Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design

Arius
4 King Ave
Frederick, MD 21701-3110
301-682-7772 Voice
301-682-7666 FAX

Internet URL http://www.arius.com
Article: 24415
Subject: Re: Xilinx Foundation 3.1i
From: rickman <spamgoeshere4@yahoo.com>
Date: Mon, 07 Aug 2000 11:18:12 -0400
Links: << >>  << T >>  << A >>
Nicolas Matringe wrote:
> 
> Rick Filipkiewicz a écrit :
> 
> > I don't actually mind paying a fair bit for Xilinx s/w - our original
> > total outlay up to XCV1000 was about UKP3500 - since in general its
> > pretty good. But I do strongly object to it timing out after a year.
> > I want to buy the s/w, not rent it.
> 
> Hi Rick
> From what I've heard, you don't rent it and it won't timeout, you just
> won't be able to use new chips if you don't buy the new version (in
> fact, you buy maintenance for one year)
> I hope I got it right...

They used to say... "We are selling you the routing and you get the
gates for free!" 

Now they say, "We are selling you the maintenance and you get the
software for free!"


-- 

Rick Collins

rick.collins@XYarius.com

Ignore the reply address. To email me use the above address with the XY
removed.



Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design

Arius
4 King Ave
Frederick, MD 21701-3110
301-682-7772 Voice
301-682-7666 FAX

Internet URL http://www.arius.com
Article: 24416
Subject: Re: Help! Troubles using async FIFO cores in Virtex
From: rickman <spamgoeshere4@yahoo.com>
Date: Mon, 07 Aug 2000 11:27:01 -0400
Links: << >>  << T >>  << A >>
erika_uk@my-deja.com wrote:
> 
> hey,
> 
> what are the advantages from using Johnston counters in FIFO design
> 
> --Erika

A Johnson ring counter uses N/2 FFs to encode N states. It is not nearly
as efficient as a binary counter, but can be more easily decoded and has
the Gray counter property of glitch free decoding. But the only way I
have seen a counter like this used in FIFOs is as the high order counter
which selected the bank of CLB rams. The Johnson counter would allow 2
input gates to decode the select signals for the banks. Or you could
just use another N/2 FFs and design a standard ring counter with *no*
decoding. 


-- 

Rick Collins

rick.collins@XYarius.com

Ignore the reply address. To email me use the above address with the XY
removed.



Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design

Arius
4 King Ave
Frederick, MD 21701-3110
301-682-7772 Voice
301-682-7666 FAX

Internet URL http://www.arius.com
Article: 24417
Subject: Re: Help!! Virtex system gate count.
From: rickman <spamgoeshere4@yahoo.com>
Date: Mon, 07 Aug 2000 11:41:12 -0400
Links: << >>  << T >>  << A >>
jimmy75@my-deja.com wrote:
> 
> > comparison is that the value of a LUT varies tremendously depending on
> > how it is used. As a logic element it's only worth aproximately 3
> gates,
> > in an ASIC a two wide AOI and a couple of inverters would do much the
> > same job. As memory it's worth a lot more, at least 16 ASIC gates, if
> > you compare it to an ASIC RAM, or maybe as many as 64 gates if the
> > equivalent structure was built out of random gates.
> 
> That's what I do not understand. How does Xilinx count the system
> gates? I understand the logic cell count (number of 4-LUTs). Now, what
> does 4-LUT give in terms of gates? I mean do they take it as <=> 3
> gates (simple LUT) or 16, 64 gates -as you said- for an equivalent
> RAM??? it seems it is ~12 gates. Anyone from Xilinx can help me here???

Why does this matter? The gate count is a very poor indicator of
anything useful. Even the LUT count is a marketing number which does not
even count LUTs! They count 4.5 LUTs per CLB when there are only 4 LUTs
in a CLB! So why would you expect the gate count to be anything other
than marketing fluff?

The gate counting that they do is a little complicated since they have
to make some assumptions for how much of the block ram to count as well
as what modes you will use the other features in. So I don't think you
will be able to figure it out without contacting them directly. 


-- 

Rick Collins

rick.collins@XYarius.com

Ignore the reply address. To email me use the above address with the XY
removed.



Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design

Arius
4 King Ave
Frederick, MD 21701-3110
301-682-7772 Voice
301-682-7666 FAX

Internet URL http://www.arius.com
Article: 24418
Subject: FPGA
From: Jeff Carter <jeff.carter@xilinx.com>
Date: Mon, 07 Aug 2000 15:42:16 +0000
Links: << >>  << T >>  << A >>
This is a multi-part message in MIME format.
--------------53812157A4FB52AEF0F98E11
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit



--------------53812157A4FB52AEF0F98E11
Content-Type: text/x-vcard; charset=us-ascii;
 name="jeff.carter.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Jeff Carter
Content-Disposition: attachment;
 filename="jeff.carter.vcf"

begin:vcard 
n:Carter;Jeff
tel;work:408.879.6961
x-mozilla-html:FALSE
url:http://www.xilinx.com
org:Xilinx Inc;W2C    web | workflow | collaboration
adr:;;2100 Logic Drive;San Jose;CA;95124;
version:2.1
email;internet:jeff.carter@xilinx.com
title:Sr. Information Systems Analyst
fn:Jeff Carter
end:vcard

--------------53812157A4FB52AEF0F98E11--

Article: 24419
Subject: Re: models of digital ICs
From: "Andy Peters" <apeters.Nospam@nospam.noao.edu.nospam>
Date: Mon, 7 Aug 2000 09:42:02 -0700
Links: << >>  << T >>  << A >>
Christian Mautner wrote in message ...

>Btw, I was looking for memory chip models some weeks again, and had to
learn
>that most vendors stopped giving them away for free, now you have to pay
for
>a software license of some denali (?) memory model software. I wouldn't be
>surprised if this is going to happen with the 74xx ICs too.

AMD's models are all in the Denali format, and the Denali stuff is
EXPENSIVE.  Hence, a Real Good Reason not to use AMD parts.

AMD -- are you listening?  Free models beget parts sales.  Expensive models
beget alternate sources.


-- a
-----------------------------------------
Andy Peters
Sr Electrical Engineer
National Optical Astronomy Observatories
950 N Cherry Ave
Tucson, AZ 85719
apeters (at) noao \dot\ edu

"A sufficiently advanced technology is indistinguishable from magic"
     --Arthur C. Clarke



Article: 24420
Subject: Re: Xilinx Alliance Base
From: "Andy Peters" <apeters.Nospam@nospam.noao.edu.nospam>
Date: Mon, 7 Aug 2000 09:46:29 -0700
Links: << >>  << T >>  << A >>
peter dudley wrote in message <8mk3gg$9gg$1@abq-news-01.ihighway.net>...
>Last December I bought the Xilinx Alliance Base software for some home
>projects that I had to do. Since then 3.1i has come out.
>
>Does anyone know if I should receive a free update on this software?

If you've paid the maintenance ransom, you should receive the 3.1i software.

At least, that's what I'm told.  I'm still waiting for the software.


-- a
-----------------------------------------
Andy Peters
Sr Electrical Engineer
National Optical Astronomy Observatories
950 N Cherry Ave
Tucson, AZ 85719
apeters (at) noao \dot\ edu

"A sufficiently advanced technology is indistinguishable from magic"
     --Arthur C. Clarke



Article: 24421
Subject: Re: Xilinx Foundation 3.1i
From: Philip Freidin <philip@fliptronics.com>
Date: Mon, 07 Aug 2000 10:09:02 -0700
Links: << >>  << T >>  << A >>
On Mon, 07 Aug 2000 13:36:30 +0100, Rick Filipkiewicz <rick@algor.co.uk> wrote:

>Is it worth upgrading from 2.1iSP6 to the new s/w ? Any experiences good
>or bad ?

I've used it for a few minutes, and it didn't crash. It gave poorer results for
a 4085XLA that is 95% full and totally floorplanned. The design still meets all
timespecs, but the safety margin ( -d 5) is not as good. Of course this might
also be explained by new speed files that are not as optimistic. Since we dont
have detailed routing timing information in the data sheet, how would we know?
Run time was insignificantly faster. Certainly nothing like the claimed halving
of runtime promised by marketing for 3.1i . Maybe the improvements are only for
virtex designs. 

>Considering that I now have to pay the ``time based license'' fee aka
>`the shareholders are complaining and chip prices are competitive so
>lets make more money out of the s/w' I want to know what the benefits
>are, if any.

No you dont. The second paragraph of the license is:

"IF YOU HAVE OBTAINED THIS SOFTWARE AS AN UPDATE
TO SOFTWARE FOR WHICH YOU HAVE PREVIOUSLY
OBTAINED A LICENSE,  THE TERMS OF THAT PRIOR
LICENSE WILL CONTINUE TO CONTROL YOUR USE OF
THE SOFTWARE.  IF YOU ARE A QUALIFIED UNIVERSITY
USER, YOU MAY OBTAIN AN EXTENSION OF THIS LICENSE
BY REGISTERING WITH THE XILINX UNIVERSITY PROGRAM."

>I don't actually mind paying a fair bit for Xilinx s/w - our original
>total outlay up to XCV1000 was about UKP3500 - since in general its
>pretty good. But I do strongly object to it timing out after a year. I
>want to buy the s/w, not rent it.

The software does not time out. You just wont be able to get updates after a
year unless you pay more maintainance. I suspect that access to the service
packs will be controlled by your license registration (and its date), rather
than the current system where you can download service packs without a check of
whether you are in maintainance. Basically, just enforcing the terms of the
license

>Note for the lawyers: Normally if you are renting something and it goes
>wrong its the responsibility of the owner to get it fixed or replaced.
>Does this mean I can demand an instant fix for any serious bug  I trip
>over ?

Sure you can demand. The results I suspect will be no different.


Philip Freidin

Mindspring that acquired Earthlink that acquired Netcom has
decided to kill off all Shell accounts, including mine.

My new primary email address is    philip@fliptronics.com

I'm sure the inconvenience to you will be less than it is for me.
Article: 24422
Subject: Re: Verilog multiplier in Xilinx...
From: Mike Peattie <mpeattie@xilinx.com>
Date: Mon, 07 Aug 2000 11:18:04 -0700
Links: << >>  << T >>  << A >>
Austin,

    You can run ngdbuild on your design, which will merge all your netlists
together (including the coregen multiplier), then run ngd2ver to get a verilog
simulation file.  This simulation will not have routing delays- it's purely
functional.

Mike

Austin Franklin wrote:

> Torbjörn Stabo <etxstbo@kk.ericsson.se> wrote in article
> <398A8153.6E265E32@kk.ericsson.se>...
> > > Anyone have any Verilog code they'd be willing to share for a 'decent'
> > > multiplier?  I am looking for something that can do a 24 x 24
> multiply...it
> > > can take quite a few cycles, and it's for a Virtex architecture.
> >
> > Have you tried Coregen? No source, but it sounds like it's the result
> that matters to you..
> >
>
> Thanks for the suggestion, I did look at it, but it doesn't appear to have
> a Verilog output mode.
> If I don't get the Verilog source (or a protected model) then how do I unit
> delay simulate it?

Article: 24423
Subject: Re: Help!! Virtex system gate count.
From: Peter Alfke <peter@xilinx.com>
Date: Mon, 07 Aug 2000 11:21:10 -0700
Links: << >>  << T >>  << A >>


rickman wrote:
<The gate counting that they do is a little complicated since they have

> to make some assumptions for how much of the block ram to count as well
> as what modes you will use the other features in. So I don't think you
> will be able to figure it out without contacting them directly.

Peter answers:

Here is what I explained more than a year ago. The numbers refer to an
XC4000-XL part, but the concept is similar for Virtex. Gate count is a
borderline meningless number when applied to LUT-based FPGAs, with large
amounts of RAM on-chip, plus dedicated ( hidden and un-counted ) carry logic,
sophisticated clock manipulation, and multi-standard I/Os.
But as amanufacturer we have to come up with a number, and market the devices
in a competitive environment.
Anyhow, here is the old explanation from early 1999:

Let me explain the 10,982 Logic Cells, 147 k bits of RAM and 265000
 max logic gates claimed for the XC40125XV.

 This is a mixture of engineering specifications and marketing.

 Engineering specifications:

 The XC40125XV has a 68 x 68 array of CLBs = 4624 CLBs
 Each of these CLBs has 32 RAM bits = 147 968 RAM bits total.
 No ifs and no buts.


 Now we get into marketing:

 Since this is a competitive world, our users want to compare different
 manufacturers, but the structures are not the same.  Altera puts eight
 LUTs in a block, Lucent puts four, and we put two LUTs into a CLB, but
 everybody's LUTs are more or less identical.

 So Xilinx decided to standardize the nomenclature and emphasize not
 CLBs but rather Logic Cells, as a lingua franca for FPGAs.  Good idea!

 Then marketing saw the third LUT in our CLB and gave it a value of 3/8
 of a Logic Cell, so the whole CLB is worth 2.375 LCs.  Obviously, we can
 argue about this addition, but it is true that one can use the third LUT
 for some really nice and efficient solutions. :-)

 Multiply the 4624 CLBs by 2.375 and you get 10 982 Logic Cells.


 What is that in ASIC gates?

 There is no scientific answer, because it depends on the design.  What
 is the LUT being used for, is the flip-flop being used at all, and what if
 you use the LUT as RAM ?

 Assume that every LUT is worth 6 gates and every flip-flop is worth 6
 gates, then every Logic Cell is worth 12 gates (sometimes more,
 sometimes less).  12 gates times 10 982 LCs makes a bare-bone
 gate-count of 131 784, and that is reflected in the name of the device.
 We are conservative and round it down a little. :-)

 But there is the RAM in the LUT, and if only 25% of them are really
 used as RAM, these LUTs are not worth 6 gates, but rather 64 gates (16
 bits with 4 bits per RAM cell, again, very conservative, ignoring the
 select structure and the read/write circuitry).  That means, we must add
 another 58 gates times 25% of 9248 LUTs, which means another 134 096
 gates, for a total of 265 880 logic gates. And we say explicitly that this
 assumes 20 to 30 % of the CLBs being used as RAM.

 That's how the XC40125XV got its name and its gate-count range of
 125,000 to 265,000 gates.  It is an attempt to bring some sanity to the
 gate-count confusion.

 Users will never agree about these assumption, and if you don't want to
 compare different manufacturers, then you can forget the whole thing
 and just stick with CLBs and RAM bits, for they are physical and thus
 non-controversial.

Peter Alfke
 February 27, 1999

Article: 24424
Subject: Place&Route report of spartan2
From: "Michael Randelzhofer" <mrandelzhofer@compuserve.com>
Date: Tue, 8 Aug 2000 00:40:08 +0200
Links: << >>  << T >>  << A >>
Hi Xilinx Freaks,

I'm doing my first Spartan2(XC2S50) design with Foundation 2.1i_sp6.
In my incremental design-process in vhdl, i want to check up the left
resources in the fpga after each design-step.

The Place&route report only informs me about the used I/O's, 'slices' and
tbufs.

In my design are lots of shiftregisters which do not need any lut, but if
only 1
FF is used in a slice, the report seems to mark it as used.
So the slice-count could not help me to estimate the free resources.

In former Xilinx fpga families, the report gave me the exact information
about
used flipflops, 4input-luts, 3input-luts and so on.

Is there any other simple way to check the resources ?
I don't want to count the used luts in the fpga-editor, or in a netlist.
Is Foundation 3.1 a solution ?

Thanks in advance
MIKE




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