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
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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 16925

Article: 16925
Subject: Re: Virtex Boards
From: Daryl Bradley <dwb105@ohm.york.ac.uk>
Date: Thu, 17 Jun 1999 10:14:38 +0100
Links: << >>  << T >>  << A >>
Yep, received a reply directly from VCC  just after I posted this mail,
tried to cancel the message but my computer crashed!!!!! - oh the wonders of
service pack 4

Many thanks anyway

John Schewel wrote:

> The Virtual Workbench release had been delayed due to a delivery problem
> on BG352 Virtex FPGAs by Xilinx.
>
> The problem has been corrected according to our sources.  We have been
> informed production of this virtex package is on schedule now.
>
> The VW-300 is being assembled this week. We are sorry for the delay.
>
> --
>
> Best Regards,
> John Schewel, VP Marketing & Sales
> Virtual Computer Corp.

--

Bio-Inspired Architectures
Department of Electronics
University of York, UK

http://www-users.york.ac.uk/~dwb105


Article: 16926
Subject: Re: Virtex Boards
From: kip@remove.this.cs.tut.fi (Kalle Palomäki)
Date: Thu, 17 Jun 1999 13:38:23 +0300
Links: << >>  << T >>  << A >>
In article <376849CB.306CB265@vcc.com>, jas@vcc.com says...
> The Virtual Workbench release had been delayed due to a delivery problem
> on BG352 Virtex FPGAs by Xilinx. 

I would like to ask another question about VCC's products. VCC's web 
pages advertised in February/March that VCC is going to bring to market a 
daughter card for H.O.T. II that has Virtex XCV300. At the moment I 
cannot find any information on your web pages concerning this daughter 
card. What is the situation with this card at the moment? Does VCC have 
any plans to bring that Virtex daughter card to market? What will be the 
schedule?

I am just a little worried since the main reason for purchasing the 
H.O.T. II was the fact that this Virtex Card should have come to market 
soon...

	BR, 
		Kalle Palomäki, Tampere University of Technology
Article: 16927
Subject: Re: Virtex Boards
From: Daryl Bradley <dwb105@ohm.york.ac.uk>
Date: Thu, 17 Jun 1999 13:18:40 +0100
Links: << >>  << T >>  << A >>
that was our initial plan as well until the VW came out. Not too sure oth=
er
than I received a mail about a Hot III PCI board.

Best contact info@vcc.com  - they've been a great help to me before

"Kalle Palom=E4ki" wrote:

> In article <376849CB.306CB265@vcc.com>, jas@vcc.com says...
> > The Virtual Workbench release had been delayed due to a delivery prob=
lem
> > on BG352 Virtex FPGAs by Xilinx.
>
> I would like to ask another question about VCC's products. VCC's web
> pages advertised in February/March that VCC is going to bring to market=
 a
> daughter card for H.O.T. II that has Virtex XCV300. At the moment I
> cannot find any information on your web pages concerning this daughter
> card. What is the situation with this card at the moment? Does VCC have=

> any plans to bring that Virtex daughter card to market? What will be th=
e
> schedule?
>
> I am just a little worried since the main reason for purchasing the
> H.O.T. II was the fact that this Virtex Card should have come to market=

> soon...
>
>         BR,
>                 Kalle Palom=E4ki, Tampere University of Technology

--

Bio-Inspired Architectures
Department of Electronics
University of York, UK

http://www-users.york.ac.uk/~dwb105


Article: 16928
Subject: Re: Evolutionary computation
From: Tim Tyler <tim@uma.sublime.org>
Date: 17 Jun 1999 14:50:45 GMT
Links: << >>  << T >>  << A >>
Jack Greenbaum <jack@mesa.greenbaum.org> wrote:
> Tim Tyler <tt@cryogen.com> writes:

>> Say I am implementing a lattice-gas automata in order to simulate
>> turbulent fluid flow through a confined space.  What earthy use are
>> counters or addition and subtraction primitives to me? [...]

> Perhaps you'd be interested in some of the purpose-built CA machines
> that have been presented at FCCM. This one looks appropriate:

> Margolus, N., "An FPGA Architecture for DRAM-based Systolic
> Computation," FCCM 1997, pp 2-11.

http://www.fccm.org/preprog97.txt contained the only information I could
find about this - it was not sufficient for me to be able to say if the
approach would be useful or not.

I'm interested in CAM-style architectures, though.

"Unfortunately", my current designs require a small quantity of 
housekeeping to be performed by a serial program (currently written in
Java) which continuously monitors the automata, and is capable of reading 
from and writing to a small number of cells while these are in active
operation.

This suggests to me a hybrid architecture, with an ordinary serial
machine (e.g. a PC) connected to the programmable logic via a PCI bridge -
rather than a "pure" CAM approach, with a custom machine devoted purely
to automata acceleration.
-- 
__________
 |im |yler  The Mandala Centre  http://www.mandala.co.uk/  tt@cryogen.com

Inertia makes the world go 'round.
Article: 16929
Subject: Logic factoring program
From: Joshua Lamorie <jpl@xiphos.ca>
Date: Thu, 17 Jun 1999 11:41:01 -0400
Links: << >>  << T >>  << A >>
Gidday there,

	I'm trying some old fashioned digital design, rather than this VHDL
business, and I'm trying to find some programs to do factoring of
logic.  In university I used 'cnd' from a guy at Michigan State. 
However, I can't find that any more.  Any hints would be appreciated. 
Oh, and I can't seem to get 'espresso' to compile on linux.

Joshua Lamorie
Systems Designer
Xiphos Technologies Inc.
Article: 16930
Subject: DS2 and E2 Framer???
From: sw[remove]@cellware.de (Stefan Wimmer)
Date: Thu, 17 Jun 1999 16:02:37 GMT
Links: << >>  << T >>  << A >>
Hi everybody,

sorry for spreading this message into several newsgroups, but I'm looking for 
DS2 and E2 (yes, that's _2_!) framer chips (or FPGA cores) and don't really 
know where to start.
Search engines didn't come up with something useful, but maybe someone out 
there in usenet land has a good tip (besides Transwich)?

Stefan


--
Stefan Wimmer                             Cellware Broadband
Email   sw@cellware.de                    Rudower Chaussee 5
WWW     http://www.cellware.de/           12489 Berlin, Germany

Visit my private Homepage:  Love, Electronics, Rockets, Fireworks!
http://www.geocities.com/CapeCanaveral/6368/
Article: 16931
Subject: Actel ActGen and Desktop problem
From: "Sung H. Lim" <Sung.Lim@jhuapl.edu>
Date: Thu, 17 Jun 1999 12:42:58 -0400
Links: << >>  << T >>  << A >>
Hi everyone,

I am trying out the free version of the Actel Desktop (my main tool is
Orcad Express).  Top level design is schematic.  I followed the
recommended procedure for adding an ActGen macro, but the CDB compiler
gives the warning
"Warning block INV has no view specified and will be ignored"  for all
the hard macros (the name INV could be any other hard macro used in the
module) inside the ActGen module.  And when I try to compile the design
via Synplicity, the ActGen macros are essentially ignored.  It is as if
the hard macros inside the ActGen module are treated as hierarchical
modules.  Any suggestions?  Thanks in advance.

Sung.Lim@jhuapl.edu



Article: 16932
Subject: Re: aobut analog
From: Tom Burgess <tom.burgess@hia.nrc.ca>
Date: Thu, 17 Jun 1999 10:23:48 -0700
Links: << >>  << T >>  << A >>
Zetex offers some interesting Field Programmable Analog Arrays:
http://www.fas.co.uk/ 
Motorola had an switched-capacitor FPAA effort going for a while,
but it got canned along with their FPGA family.
There's an FPAA FAQ and other info available at:
http://www.eecg.toronto.edu/~vgaudet/fpaa/faq.html

regards, tom

Sandro Wefel wrote:
> 
> Ray Andraka wrote:
> 
> > FPGAs don't have analog elements, but there is no reason you can't use
> > them in conjunction with external ADC's, DAC's, PLL's and RF
> > components.  The FPGA itself is a strictly digital circuit.
> 
> There are some mixed mode FPGA's from the Fraunhofer Institute
> of Microelectronic Circuits and Systems Dresden Germany.  They
> have an analog and digital core and also analog and digital ports.
> The problem: there aren't synthesis tools. A simulation tool
> for such AFPGAs was part of my diplom thesis and solved
> in conjunction with SPICE (not VHDL).
> 
> To contact the FHG watch
> http://www.imsdd.fhg.de/ims-dd-e.html
> 
> Regards
> Sandro Wefel

Tom Burgess
-- 
Digital Engineer
Dominion Radio Astrophysical Observatory
P.O. Box 248, Penticton, B.C.
Canada V2A 6K3
Email:        tom.burgess@hia.nrc.ca
Article: 16933
Subject: Actel's proASIC
From: "Martin Duffy" <martin_duffy@penarth31.freeserve.co.uk>
Date: Thu, 17 Jun 1999 20:52:00 +0100
Links: << >>  << T >>  << A >>
I see Actel have announced their new proASIC family. This is billed as a
cheap (programmable) alternative to using gate arrays.
This sounds familiar - Xilinx/Altera have similar product offerings.

But it's different in that it's 1) flash-based, and 2) a fine-grained
sea-of-gates architecture.

Does anyone know why this choice of technology & architecture should prove
superior - or even the equal - of those of the 2 big boys ?
Because to me "flash-based" means CPLDs, "sea-of-gates FPGA" means Motorola,
Toshiba, Xilinx 8K (and what happened to them?).
So this product is definitely going against the grain (sorry! couldn't
resist!).




Article: 16934
Subject: Re: aobut analog
From: Steven Casselman <sc@vcc.com>
Date: Thu, 17 Jun 1999 13:20:41 -0700
Links: << >>  << T >>  << A >>

http://www.sidsa.com/fipsoc.htm
This place something you might want to look at.


--
Steve Casselman, President
Virtual Computer Corporation
http://www.vcc.com


Article: 16935
Subject: Re: Virtex Boards
From: Steven Casselman <sc@vcc.com>
Date: Thu, 17 Jun 1999 13:39:44 -0700
Links: << >>  << T >>  << A >>


Daryl Bradley wrote:

> that was our initial plan as well until the VW came out. Not too sure other
> than I received a mail about a Hot III PCI board.
>
>

Yes we are working on a HOTIII, board prices
will range from $995 to $4995.  The Virtex daughter
board was taken off the site because we were thinking
of just offering the daughter board with the HOTIII.
We have PCBs in house for the daughter boards and we will build some just
as soon as we get some !@#$%^&* parts.  The Virtex
daughter board will be $4995 for a V800. The pricing on the
HOTIII will be _something_ like $995 for V100
$1995 for V300 and $4,995 for V800.  Of course
all these boards have one meg configuration caches and use
the high speed Virtex parallel configuration port (PCP).
Each board has 4 meg of fast SRAM (15ns or faster).


All the above is advanced information and subject to change
at any time.

Thanks
--
Steve Casselman, President
Virtual Computer Corporation
http://www.vcc.com


Article: 16936
Subject: *.bit, *.ubt, *.rbt
From: "bob" <bob_kimberly75@hotmail.com>
Date: Fri, 18 Jun 1999 00:51:49 +0100
Links: << >>  << T >>  << A >>
Hi All,

What is the difference between *.bit, *.ubt and *.rbt files. Which one to
use to configure the FPGA?

Cheers.


Article: 16937
Subject: WISHBONE Interconnection Standard for IP Core Reuse
From: "Wade D. Peterson" <peter299@maroon.tc.umn.edu>
Date: Thu, 17 Jun 1999 22:50:25 -0500
Links: << >>  << T >>  << A >>
PRESS RELEASE - FOR IMMEDIATE RELEASE

Silicore Corporation Announces New IP Core Standard for System-On-A-Chip

June 21, 1999 - 36th DAC / New Orleans, LA  USA.  Silicore Corporation announced
today a design reuse standard for off-the-shelf IP cores.  The new standard is
called the WISHBONE INTERCONNECTION FOR IP CORES, and is intended to make IP
core integration easier for system-on-a-chip users.

The WISHBONE interconnect is a portable and flexible interface for use with
semiconductor IP cores.  It defines a common, logical interface between IP
cores.  This improves the portability and reliability of the system components,
and results in faster time-to-market for the end user.

Previously, IP cores used non-standard interconnection schemes that made them
difficult to integrate.  They required the end user to create substantial
amounts of custom glue logic to connect each of the cores together.  By adopting
a standard interconnection scheme, they can be integrated more quickly and
easily by the end user.

The WISHBONE standard can be used with soft, firm and hard IP cores.
Furthermore, it can be used with any target architecture (such as FPGA or ASIC
devices), and is independent of any hardware description, synthesis, router,
layout or other development tools.  It is a high performance interconnect where
speed is limited only by the IC semiconductor technology.

Wade Peterson, President and CEO of Silicore, stated that: "Our research has
shown that a major impediment to IP core integration is the lack of a standard,
logical interface between cores.  The WISHBONE interconnect solves this problem
by forming common data exchange formats and bus cycles.  We were heavily
influenced by the traditional microcomputer buses such as VMEbus and PCI bus,
which have served the industry very well for many years.  Essentially, WISHBONE
forms a microcomputer bus that is tailored to the needs of system-on-a-chip."

Peterson also stated that: "We believe that the industry needs this, and have
gone ahead and made WISHBONE available on a no cost basis.  We are doing this as
a public service to our customers and the IP core industry as a whole.
Furthermore, it will improve the quality of our own products, as well as to
foster cooperation among the users and suppliers of other IP cores."

Silicore will be releasing products based on WISHBONE in 3Q/99.  For more
information, and copies of the WISHBONE specification, see the Silicore web site
at http://www.silicore.net, or contact: Wade Peterson at 612-722-3815 or by
e-mail at: peter299@maroon.tc.umn.edu.


Article: 16938
Subject: Re: Actel ActGen and Desktop problem
From: "zootsuit" <leem1@2xtreme.net>
Date: Fri, 18 Jun 1999 03:59:38 GMT
Links: << >>  << T >>  << A >>
Hi Sung,

Your problem is you probably did not quite follow the directions for placing
an ACTgen block.  If you run ACTgen from inside of DesignView it will
automatically create a symbol under the Place menu Symbol command.  When the
dialog box comes up you will see VBACTgen and the symbol will be under this
heading.  Place the symbol from this menu as it is now part of your
components to select from.

If you created the ACTgen macro outside of DesignView you can still add it
as VHDL source it is just that the macro is made from primitive elements
from the Actel libraries thus they have "no view", ie. they have no lower
level schematic or anything to describe them (this is probably what you
did).  When this gets passed to the synthesis tool the macro will be treated
as regular structural VHDL and be synthesized along with the rest of your
schematic (if you are doing a mixed design with schematics and VHDL).
Should you be doing a design that is only composed of schematic elements
then you should use the "Generate EDIF" command from the Tools menu.  There
is no need to synthesize your design if it is purely schematic to get to an
EDIF for place and route.

Hope this helps.

Regards,
Mike Lee


Sung H. Lim wrote in message <37692592.E8D289CC@jhuapl.edu>...
>Hi everyone,
>
>I am trying out the free version of the Actel Desktop (my main tool is
>Orcad Express).  Top level design is schematic.  I followed the
>recommended procedure for adding an ActGen macro, but the CDB compiler
>gives the warning
>"Warning block INV has no view specified and will be ignored"  for all
>the hard macros (the name INV could be any other hard macro used in the
>module) inside the ActGen module.  And when I try to compile the design
>via Synplicity, the ActGen macros are essentially ignored.  It is as if
>the hard macros inside the ActGen module are treated as hierarchical
>modules.  Any suggestions?  Thanks in advance.
>
>Sung.Lim@jhuapl.edu
>
>
>


Article: 16939
Subject: Re: Die size of XILINX fpga's
From: NOJUNK@gecm.com (Dave Storrar)
Date: Fri, 18 Jun 1999 07:57:51 GMT
Links: << >>  << T >>  << A >>
On Thu, 17 Jun 1999 02:09:42 -0700, Eugene Grayver
<egrayver@janet.ucla.edu> wrote:
>I am repeating my post since no one had the answer last time.  Does
>anyone know
>the die sizes of the xilinx XC4000 series FPGAs?  I need them for a
>comparative
>study.

I don't know myself, but you might want to have a look at
http://www.chipsupply.com.  They will let you look at die maps online,
but you have to fill in an online form for them to send you a
password.

Hope this helps

Dave.

-- 
REPLACE "NOJUNK" in address with "david.storrar" to reply
Development Engineer       |
Marconi Electronic Systems | Tel: +44 (0)131 343 4282
RCS                        | Fax: +44 (0)131 343 4091

Article: 16940
Subject: Re: Virtex Boards
From: Daryl Bradley <dwb105@ohm.york.ac.uk>
Date: Fri, 18 Jun 1999 10:42:04 +0100
Links: << >>  << T >>  << A >>
Yeah, I heard that there was a minor?!?!?!?!?! well oky, bit of a nightmare
problem getting parts from Xilinx at the moment.



Steven Casselman wrote:

> Daryl Bradley wrote:
>
> > that was our initial plan as well until the VW came out. Not too sure other
> > than I received a mail about a Hot III PCI board.
> >
> >
>
> Yes we are working on a HOTIII, board prices
> will range from $995 to $4995.  The Virtex daughter
> board was taken off the site because we were thinking
> of just offering the daughter board with the HOTIII.
> We have PCBs in house for the daughter boards and we will build some just
> as soon as we get some !@#$%^&* parts.  The Virtex
> daughter board will be $4995 for a V800. The pricing on the
> HOTIII will be _something_ like $995 for V100
> $1995 for V300 and $4,995 for V800.  Of course
> all these boards have one meg configuration caches and use
> the high speed Virtex parallel configuration port (PCP).
> Each board has 4 meg of fast SRAM (15ns or faster).
>
> All the above is advanced information and subject to change
> at any time.
>
> Thanks
> --
> Steve Casselman, President
> Virtual Computer Corporation
> http://www.vcc.com

--

Bio-Inspired Architectures
Department of Electronics
University of York, UK
http://www-users.york.ac.uk/~dwb105


Article: 16941
Subject: Re: vhdl and viewlogic problem
From: Jim Kipps <jkipps@viewlogic.com>
Date: Fri, 18 Jun 1999 08:56:58 -0400
Links: << >>  << T >>  << A >>
Ingo-

Sorry, but vhdl2sym.exe does not support the creation of bars over pin names
in the symbols it creates.  You have to edit the symbol after generation.

Regards,
-Jim

Ingo Purnhagen wrote:

> Sorry, forgot my real name; damned Outlock ;-) !!!
>
> Hi newsgroup,
> I have a simple (?) problem using VHDL and Viewlogic. After declaration of
> an entity I use VHDL2SYM.EXE to create a symbol for ViewDraw. It works fine,
> but I want to have an active low signal (signalname with overline) in my
> schematic.
> I have no idea how to realize it in VHDL . Any suggestions?
>
> --
> Ingo Purnhagen
> OHB-System GmbH
> Universitaetsallee 27-29
> D-28359 Bremen
> Telefon: 0421-2020-702
> Telefax: 0421-2020-610
> mailto:Purnhagen@ohb-system.de
> http://www.ohb-system.de

--
--------------------------------------------------------
James R. Kipps                  FPGA Marketing Manager
jkipps@viewlogic.com            Phone: (508) 303-5246
--------------------------------------------------------


Article: 16942
Subject: Re: Virtex Boards
From: "Bill" <bb@alphadata.co.uk>
Date: Fri, 18 Jun 1999 15:35:25 +0100
Links: << >>  << T >>  << A >>
Suprisingly, we have no problem getting hold of V1000/BG560. They cost a bit
though.
>============
Bill Blyth
http://www.alphadata.co.uk
============

Daryl Bradley wrote in message <376A146C.D582D32E@ohm.york.ac.uk>...
>Yeah, I heard that there was a minor?!?!?!?!?! well oky, bit of a nightmare
>problem getting parts from Xilinx at the moment.
>
>
>
>Steven Casselman wrote:
>
>> Daryl Bradley wrote:
>>
>> > that was our initial plan as well until the VW came out. Not too sure
other
>> > than I received a mail about a Hot III PCI board.
>> >
>> >
>>
>> Yes we are working on a HOTIII, board prices
>> will range from $995 to $4995.  The Virtex daughter
>> board was taken off the site because we were thinking
>> of just offering the daughter board with the HOTIII.
>> We have PCBs in house for the daughter boards and we will build some just
>> as soon as we get some !@#$%^&* parts.  The Virtex
>> daughter board will be $4995 for a V800. The pricing on the
>> HOTIII will be _something_ like $995 for V100
>> $1995 for V300 and $4,995 for V800.  Of course
>> all these boards have one meg configuration caches and use
>> the high speed Virtex parallel configuration port (PCP).
>> Each board has 4 meg of fast SRAM (15ns or faster).
>>
>> All the above is advanced information and subject to change
>> at any time.
>>
>> Thanks
>> --
>> Steve Casselman, President
>> Virtual Computer Corporation
>> http://www.vcc.com
>
>--
>
>Bio-Inspired Architectures
>Department of Electronics
>University of York, UK
>http://www-users.york.ac.uk/~dwb105
>



Article: 16943
Subject: Re: vhdl and viewlogic problem
From: "Austin Franklin" <austin@darkr9oom.com>
Date: 18 Jun 1999 14:50:26 GMT
Links: << >>  << T >>  << A >>
In fact, I'd suggest NOT using the overbar convention.  Some tools may not
process the name correctly.  The overbar has to be translated into 'some'
other character, per se, because one can not have an overbar in a text file
;-)  They used to translate it to a tilde.  After YEARS of fighting with
tool vendors to all come up with ONE convention for active low signals, and
getting no where, I got smart and came up with a convention that doesn't
use the overbar.

It can be any textual convention you want, but I personally use "SIGNAL_L",
and it seems to work fine thru all my tools.

Austin Franklin
austin@darkroom.com


Jim Kipps <jkipps@viewlogic.com> wrote in article
<376A421A.737B7038@viewlogic.com>...
> Ingo-
> 
> Sorry, but vhdl2sym.exe does not support the creation of bars over pin
names
> in the symbols it creates.  You have to edit the symbol after generation.
> 
> Regards,
> -Jim
> 
> Ingo Purnhagen wrote:
> 
> > Sorry, forgot my real name; damned Outlock ;-) !!!
> >
> > Hi newsgroup,
> > I have a simple (?) problem using VHDL and Viewlogic. After declaration
of
> > an entity I use VHDL2SYM.EXE to create a symbol for ViewDraw. It works
fine,
> > but I want to have an active low signal (signalname with overline) in
my
> > schematic.
> > I have no idea how to realize it in VHDL . Any suggestions?
> >
> > --
> > Ingo Purnhagen
> > OHB-System GmbH
> > Universitaetsallee 27-29
> > D-28359 Bremen
> > Telefon: 0421-2020-702
> > Telefax: 0421-2020-610
> > mailto:Purnhagen@ohb-system.de
> > http://www.ohb-system.de
> 
> --
> --------------------------------------------------------
> James R. Kipps                  FPGA Marketing Manager
> jkipps@viewlogic.com            Phone: (508) 303-5246
> --------------------------------------------------------
> 
> 
> 
Article: 16944
Subject: Read/Writes to memories/register files for PIC core
From: tcoonan@mindspring.com (Thomas A. Coonan)
Date: Fri, 18 Jun 1999 15:10:05 GMT
Links: << >>  << T >>  << A >>
Hi,

I've asked this question once before, but it is important (to me) so
here goes again.

I have this freeware Verilog PIC CPU core
(www.mindspring.com/~tcoonan) that I like to play with.  I have a
nagging issue related to memories and clocks/instruction.  If you know
about the PIC, then you know that it has seperate program and data
memories.  My current model uses 2 clocks per instruction because I'm
using "standard" synchronous memories and I can't seem to figure out
how to do everything that needs doing in terms of memory cycles with 1
clock per instruction.  I'd like to offer 1 clock/instruction!  The
real PIC actually has 4 internal "phases".  wheew..  that was a lot to
say.  For example, the PIC allows you to execute the following code:

   incf counter1, counter1
   incf foo, foo
   incf bar, bar

The above means that the increment instruction fetches the data memory
locations indicated (these "registers" are really from the internal
data memory or "register file" that is potentially big, on the order
of 1024 8-bit words), increments them (in my ALU) and rewrites the new
value back into the data memory.  Seems like I need 2 cycles, yes?

On the instruction side, I think I can operate on 1 clock per
instruction by concurrently loading my PC register *AND* the Address
port of my synchronous ram.  So, the problem appears to be reduced to:

   How can I provide a portable memory to do a read/modify/write in
   only one cycle?

I don't think a "standard" synchronous memory can do this.  Here are
some alternatives I know of, but I would like to hear any insights
others can offer.  Here's my thinking and comments to date:

   1)  First; Ideally, I want to stay in standard Verilog - no custom
        memory designs or floorplanning can be used.  No
        non-synchronous multiple clock edge techniques, no
        latches, can be used  (i'm trying to do something very
        portable and flexible).
   2)  Synchronous memories are the convenient, area-efficient cells
        that people will have access to either in ASICs or in FPGAs,
        so perhaps, the price to pay for this is, sadly, to 
        require 2 clocks/instruction.  This is my current status.
   3)  Use a "register file" and just synthsize the flip-flops.  This
        *can* offer the read/modify/write feature.  Obviously, this
        is only practical for a dozen or two data memory locations.
        yes?  I have always assumed that delcaring something like
        reg [7:0] data_memory[0:511] and synthesizing, would
        be crazy, whereas reg[7:0] data_memory[0:31] might be
        a reasonable thing to do if it buys us 1 clock/instruction.

If I go with #3, then one suggestion that I have heard, is to use
something like Module Compiler.  I don't like this because of point
#1.  Has anyone done a register file and come up with RTL techniques
that lead to decent speed area?  Or, should I simply declare the 2-D
memory, synthesize it and move on.  OR, am I missing something
altogether!

That's it.  I hope someone out there has had a similar issue and can
help me!  (I'll be at DAC next week if anyone would actually like to
discuss this sort of thing..)

tom coonan
tcoonan@mindspring.com
www.mindspring.com/~tcoonan
Article: 16945
Subject: Re: Read/Writes to memories/register files for PIC core
From: Rickman <spamgoeshere4@yahoo.com>
Date: Fri, 18 Jun 1999 13:54:56 -0400
Links: << >>  << T >>  << A >>
"Thomas A. Coonan" wrote:
...SNIP...
> say.  For example, the PIC allows you to execute the following code:
> 
>    incf counter1, counter1
>    incf foo, foo
>    incf bar, bar
> 
> The above means that the increment instruction fetches the data memory
> locations indicated (these "registers" are really from the internal
> data memory or "register file" that is potentially big, on the order
> of 1024 8-bit words), increments them (in my ALU) and rewrites the new
> value back into the data memory.  Seems like I need 2 cycles, yes?
> 
> On the instruction side, I think I can operate on 1 clock per
> instruction by concurrently loading my PC register *AND* the Address
> port of my synchronous ram.  So, the problem appears to be reduced to:
> 
>    How can I provide a portable memory to do a read/modify/write in
>    only one cycle?
> 
> I don't think a "standard" synchronous memory can do this.  Here are
> some alternatives I know of, but I would like to hear any insights
> others can offer.  Here's my thinking and comments to date:

If you are referring to synchronous memory inside the FPGA or ASIC, then
I believe you are wrong about the read/modify/write capability. In a
Xilinx part, for example, an address can be presented which will cause
the data to be read from a memory location. This data can then be
processed and fed back into the data in port. If you have the write
signal asserted on the next rising edge of the clock the new data will
be written into the same location. You could even use different
addresses on the read and write by using a dual port ram. Or if you
multiplex the address, you can still use different addresses on read and
write by clocking the read data into an output register on the falling
edge of the clock and changing the address with the clock as well. You
will need to be very careful about timing of the multiplexer in this
case. If it changes too quickly, you will not meet the address
setup/hold times on the ram.

 
>    1)  First; Ideally, I want to stay in standard Verilog - no custom
>         memory designs or floorplanning can be used.  No
>         non-synchronous multiple clock edge techniques, no
>         latches, can be used  (i'm trying to do something very
>         portable and flexible).

As I outlined above, you can do a very portable design that will work
with many (but not all) manufacturers. I know that you can do this with
Xilinx and Lucent parts. I would be willing to bet that there are many
ASIC devices that provide such a single or dual port ram as well. 


>    2)  Synchronous memories are the convenient, area-efficient cells
>         that people will have access to either in ASICs or in FPGAs,
>         so perhaps, the price to pay for this is, sadly, to
>         require 2 clocks/instruction.  This is my current status.

They are convenient and should work in one clock cycle. Perhaps you
think that it can only do one thing at a time. But the ones I have seen
will always do a read of whatever address in on the address pins and
will also do a write when the write enable pin is asserted. 


>    3)  Use a "register file" and just synthsize the flip-flops.  This
>         *can* offer the read/modify/write feature.  Obviously, this
>         is only practical for a dozen or two data memory locations.
>         yes?  I have always assumed that delcaring something like
>         reg [7:0] data_memory[0:511] and synthesizing, would
>         be crazy, whereas reg[7:0] data_memory[0:31] might be
>         a reasonable thing to do if it buys us 1 clock/instruction.

I aggre that this is the least efficient approach. But I am pretty sure
that you don't need to do this.
 
I am not a big user of VHDL or Verilog since I don't like being isolated
from the hardware. But I don't see any reason why this won't work. I
know that the hardware supports it. 

-- 

Rick Collins

rick.collins@XYarius.com

remove the XY to email me.



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: 16946
Subject: Re: Read/Writes to memories/register files for PIC core
From: "Andy Peters" <apeters@noao.edu.NOSPAM>
Date: Fri, 18 Jun 1999 11:48:46 -0700
Links: << >>  << T >>  << A >>
Thomas A. Coonan wrote in message
<376a5a2c.694032145@news.mindspring.com>...

> The real PIC actually has 4 internal "phases".
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>On the instruction side, I think I can operate on 1 clock per
>instruction by concurrently loading my PC register *AND* the Address
>port of my synchronous ram.  So, the problem appears to be reduced to:
>
>   How can I provide a portable memory to do a read/modify/write in
>   only one cycle?

You mention that the PIC has four internal phases - that means the
instruction cycle is not one clock cycle.   Yes, the PIC can do a
read-modify-write in one instruction execution cycle.  But they divide the
clock by four (those four phases) to do all of the work, so it really takes
four clock cycles to do anything.  What happens inside the PIC is that data
is read during phase Q2 and written back during phase Q4.

So the memory does NOT have to do a R/M/W in only one cycle.

-- a
------------------------------------------
Andy Peters
Sr. Electrical Engineer
National Optical Astronomy Observatories
950 N Cherry Ave
Tucson, AZ 85719
apeters@noao.edu

NY Knicks in '99:
"Ya gotta believe!"



Article: 16947
Subject: FW: Xilinx Acquisition of CoolRunners
From: Brian Boorman <XZY.bboorman@harris.com>
Date: Fri, 18 Jun 1999 14:59:42 -0400
Links: << >>  << T >>  << A >>
Dear Valued CoolRunner Customer,

We are pleased to inform you of a change in ownership of the CoolRunner
product
family.  Philips Semiconductors has elected to sell its CoolRunner
Programmable
Logic business to Xilinx Inc. This sale includes the CoolRunner  XPLA 
and  the
CoolRunner  22V10  devices.  The  CoolRunner  product  family will
benefit from
dedicated and expanding support due to the new arrangement.

Xilinx intends to keep and enhance the CoolRunner product  line.  In 
the  near
term,  the  primary goal during the transition is maintaining the same
level of
service and support you have come to expect of the CoolRunner product
group.

Philips Semiconductors and Xilinx  are  grateful  for  your 
understanding  and
patience  in  this  time  of transition. If you have any questions,
please feel
free   to   call   us   at   1-888-COOLPLD   or   email   your  
questions   to
coolpld@abq.sc.philips.com.

Sincerely,

Xilinx Inc. and the CoolRunner product group
Article: 16948
Subject: Re: Read/Writes to memories/register files for PIC core
From: tcoonan@mindspring.com (Thomas A. Coonan)
Date: Fri, 18 Jun 1999 20:30:16 GMT
Links: << >>  << T >>  << A >>
Rickman <spamgoeshere4@yahoo.com> wrote:

>"Thomas A. Coonan" wrote:
>...SNIP...
>> say.  For example, the PIC allows you to execute the following code:
>> 
>>    incf counter1, counter1
>>    incf foo, foo
>>    incf bar, bar
>> 
>> The above means that the increment instruction fetches the data memory
>> locations indicated (these "registers" are really from the internal
>> data memory or "register file" that is potentially big, on the order
>> of 1024 8-bit words), increments them (in my ALU) and rewrites the new
>> value back into the data memory.  Seems like I need 2 cycles, yes?
>> 
>> On the instruction side, I think I can operate on 1 clock per
>> instruction by concurrently loading my PC register *AND* the Address
>> port of my synchronous ram.  So, the problem appears to be reduced to:
>> 
>>    How can I provide a portable memory to do a read/modify/write in
>>    only one cycle?
>> 
>> I don't think a "standard" synchronous memory can do this.  Here are
>> some alternatives I know of, but I would like to hear any insights
>> others can offer.  Here's my thinking and comments to date:
>
>If you are referring to synchronous memory inside the FPGA or ASIC, then
>I believe you are wrong about the read/modify/write capability. In a
>Xilinx part, for example, an address can be presented which will cause
>the data to be read from a memory location. This data can then be
>processed and fed back into the data in port. If you have the write
>signal asserted on the next rising edge of the clock the new data will
>be written into the same location. You could even use different
>addresses on the read and write by using a dual port ram. Or if you
>multiplex the address, you can still use different addresses on read and
>write by clocking the read data into an output register on the falling
>edge of the clock and changing the address with the clock as well. You
>will need to be very careful about timing of the multiplexer in this
>case. If it changes too quickly, you will not meet the address
>setup/hold times on the ram.
That would be good news.  I thought that the synchronous RAMs
register the address and the rw control on the clock ege.  So that if
you drive the address and rw in one cycle, you won't get the result
of your read until the next cycle.  Remember my example that had
3 read/modify/writes of 3 different addresses in 3 cycles.  I'll
review the synchronous RAM data sheets for my ASIC vendor and also for
XILINX Virtex.

The DPRAM is an idea..  I know both my ASIC vendor and XILINX Virtex
have those.
>
> 
>>    1)  First; Ideally, I want to stay in standard Verilog - no custom
>>         memory designs or floorplanning can be used.  No
>>         non-synchronous multiple clock edge techniques, no
>>         latches, can be used  (i'm trying to do something very
>>         portable and flexible).
>
>As I outlined above, you can do a very portable design that will work
>with many (but not all) manufacturers. I know that you can do this with
>Xilinx and Lucent parts. I would be willing to bet that there are many
>ASIC devices that provide such a single or dual port ram as well. 
>
>
>>    2)  Synchronous memories are the convenient, area-efficient cells
>>         that people will have access to either in ASICs or in FPGAs,
>>         so perhaps, the price to pay for this is, sadly, to
>>         require 2 clocks/instruction.  This is my current status.
>
>They are convenient and should work in one clock cycle. Perhaps you
>think that it can only do one thing at a time. But the ones I have seen
>will always do a read of whatever address in on the address pins and
>will also do a write when the write enable pin is asserted. 
>
>
>>    3)  Use a "register file" and just synthsize the flip-flops.  This
>>         *can* offer the read/modify/write feature.  Obviously, this
>>         is only practical for a dozen or two data memory locations.
>>         yes?  I have always assumed that delcaring something like
>>         reg [7:0] data_memory[0:511] and synthesizing, would
>>         be crazy, whereas reg[7:0] data_memory[0:31] might be
>>         a reasonable thing to do if it buys us 1 clock/instruction.
>
>I aggre that this is the least efficient approach. But I am pretty sure
>that you don't need to do this.
> 
>I am not a big user of VHDL or Verilog since I don't like being isolated
>from the hardware. But I don't see any reason why this won't work. I
>know that the hardware supports it. 
>
>-- 
>
>Rick Collins
>
>rick.collins@XYarius.com
>
>remove the XY to email me.
>
>
>
>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: 16949
Subject: Re: Read/Writes to memories/register files for PIC core
From: "Wade D. Peterson" <peter299@maroon.tc.umn.edu>
Date: Fri, 18 Jun 1999 15:45:11 -0500
Links: << >>  << T >>  << A >>
Thomas A. Coonan <tcoonan@mindspring.com> wrote in message
news:376a5a2c.694032145@news.mindspring.com...
> I've asked this question once before, but it is important (to me) so
> here goes again.
>
> I have this freeware Verilog PIC CPU core
> (www.mindspring.com/~tcoonan) that I like to play with.  I have a
> nagging issue related to memories and clocks/instruction.  If you know
> about the PIC, then you know that it has seperate program and data
> memories.  My current model uses 2 clocks per instruction because I'm
> using "standard" synchronous memories and I can't seem to figure out
> how to do everything that needs doing in terms of memory cycles with 1
> clock per instruction.  I'd like to offer 1 clock/instruction!  The
> real PIC actually has 4 internal "phases".  wheew..  that was a lot to
> say.
[SNIP]

We've solved all of the problems in our VHDL PIC processor.  It runs at 1 clock
per instruction, is 100% PIC code compatible, and uses a synchronous memory that
allows the read-modify-write operations.  We've completely documented how the
core works (including complete descriptions of the internal components and the
portable memory interface) in our SLC1655 technical reference manual.  We did
our implementation in VHDL, but the circuit descriptions in the manual are
pretty generic, and will apply to Verilog(tm) as well.  If you want to see how
we did it you can download the from our website at http://www.silicore.net.

--
Wade D. Peterson
Silicore Corporation
3525 E. 27th St. No. 301, Minneapolis, MN USA 55406
TEL: (612) 722-3815, FAX: (612) 722-5841
URL: http://www.silicore.net/  E-MAIL: peter299@maroon.tc.umn.edu




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
2017JanFebMarAprMayJunJulAugSepOctNovDec2017
2018JanFebMarAprMayJunJulAugSepOctNovDec2018
2019JanFebMarAprMayJunJulAugSepOctNovDec2019
2020JanFebMarAprMay2020

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