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 32000

Article: 32000
Subject: Re: one state machine
From: Rick Filipkiewicz <rick@algor.co.uk>
Date: Sun, 10 Jun 2001 17:08:20 +0100
Links: << >>  << T >>  << A >>


Austin Franklin wrote:

> > I don't agree that schematics substitute for a block diagram. They are
> > often so full of details that you can't see the flow for the gates!
>
> It all depends on how well you know how to use the tools!  Obvioulsy, you're
> not a schematic guy ;-)  You certainly shouldn't have many, if any at all,
> gates on your schematics, except at the lowest levels, or a random gate here
> and there...for simplicity of drawing...
>
> > Using heirarchical schematics start causing some of the same problems
> > that HDLs have. The signal names get complicated by the heirarchy path
> > and even renamed as one wire that traverses several levels will have
> > several names. The back end tool will want to pick just one name for the
> > whole net, but which one?
>
> This is not a problem.  The names are deterministic, unlike HDLs, which are
> NOT deterministic.  The signal names will be what ever is the topest level
> signal name is.  If you name your heirarchical elements, then the signal
> name is very easy to follow.

In another round of the, seemingly eternal, HDL vs. Schems debate no one seems
to have brought up what I consider one of the biggest drawbacks of schems. Since
they are usually binary files they are difficult to impossible to put under any
meaningful form of version control i.e. one that can do diffs between revisions.
One of the ++'s of text files is that its very easy to get something like CVS to
ignore whitespace changes so that diffs don't get drowned. Even if there were a
text file based schem package its a lot more difficult to extract the
connectivity "signal" from the graphical "noise".

The other missing issue, heavily related to the above, is portability.



Article: 32001
Subject: Re: Pentium 4 or AMD ?
From: Rick Collins <spamgoeshere4@yahoo.com>
Date: Sun, 10 Jun 2001 12:24:48 -0400
Links: << >>  << T >>  << A >>
Rick Filipkiewicz wrote:
> 
> Eric wrote:
> 
> (lots of snip)
> 
> Problem is the same as trying to make pizza delivery with an 18 weels truck.
> 
> > Sure you could deliver hundreds of them at a time, but who needs this kind
> > of "performance" ? what peoples want is just one fresh pizza, delivered ASAP,
> > not a zillion, delivered cold and somehow melted with the box, 1 hour later.
> > Here too, latency matters a lot more than volume/bandwidth.
> 
> That's the nicest analogy I've seen for what I usually paraphrase, following Bill
> Clinton's first campaign, as:
> 
> ``Its the latency, stupid''
> 
> No offence intended.
> 
> Note that DDR DRAM also falls into this category of solving the easy bandwidth
> problem instead of the hard latency one but at least it doesn't force you into
> insanely expensive controllers, PCB technology, and license fees. There is an
> indirect latency benefit in that high bandwidth stuff like PCI traffic uses less
> memory time.
> 
> <more snip>
> 
> > Well, maybe this goes a bit off topic, but since Rambus started to bully
> > SDRAM memory controllers makers (Hitachi H8 / PC chipset makers), it would
> > be wise to keep an eye on what's to come for FPGA based controllers.
> >
> > Eric.
> 
> They'll have to go after Xilinx first since the Virtex2 defines 2 ``DDR registers''
> in the IOB and there have been DDR DRAM controller apps notes on their web site for
> ages.
> 
> Great posting Eric. It should appear somewhere on  a DRAM FAQ but Rambus would
> probably sue the FAQ hoster. In fact if I were you I would consider getting a lawyer
> now & moving all your assets off-shore.
> 
> An interesting side note: When this came up I went I downloaded a RDRAM data sheet.
> What did I see on the Rambus home page ? A little picture with a hot spot marked
> ``New Litigation''. I think that just about says it all ...

All of this sounds good on paper, but what do the benchmarks say? Are
there any apples to apples comparisons? Or is the world split into
AMD/DDR and P4/RDRAM? 

-- 

Rick "rickman" 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      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Article: 32002
Subject: Re: Xilinx webpack annoyances (long and whiny)
From: Rick Collins <spamgoeshere4@yahoo.com>
Date: Sun, 10 Jun 2001 13:03:00 -0400
Links: << >>  << T >>  << A >>
Andy,

I won't argue that there is any excuse for the way FPGA development
tools are written. The standard of quality is low and I don't see where
it has improved much over the last few years. But I can tell you why. It
is simply a matter of quantity and the rapid rate of change of the
hardware they must support. 

The FPGA vendors only sell a few thousand seats of these tools every
year. With the limited revenue this brings in they just don't have the
resources to hone the tools to "perfection". (I know we are a long way
from perfection). 

With the hardware changing by coming out with a new family every year or
two, and adding new members nearly constantly, the software teams are
doing all they can to keep up. 

Whining may make you feel better and can serve as a  means of feedback
when posted to the newsgroup. But you need to temper your wants with
your needs. If the problem you listed was the biggest problem you have
with the tools, you will be very lucky. I would like a simulation tool
that does not crash every fourth or fifth time I reload the design and I
would really, really love a backend tool that lets me do a single mode
of timing constraints rather than making me enter them twice in two
formats; once for P&R and once for timing analysis (Altera MaxPlus+II). 

You might want to send the details of the VHDL issue to Xilinx support.
If they don't know about it, they will likely want to fix it. 



Andy Peters wrote:
> 
> OK, so I'm trying to synthesize some VHDL code that simulates fine.  Nothing fancy, just some generates to create some registers.
> 
> The synthesizer complains: ERROR :   (VHP__0163). Where's the list of errors? Hitting F1 is no help.  OK, look at the message: "Unexpected symbol read: BEGIN."
> 
> Wait, this tool doesn't seem to understand the LRM.  My generate statement looks like:
> 
>     foo: for ff in 0 to 2 generate
>     begin
>        bar: process (clk, rst) is
>        begin
>            if rst = '1' then
>                b(ff) <= '0';
>            elsif rising_edge(clk) then
>                b(ff) <= c(ff);
>            end if;
>         end process bar;
>     end generate foo;
> 
> Hmmm..Ashenden sez this is OK; so does ModelSim.  So, what's wrong?  Does XST support generates?
> 
> Soooo... what's with the online help?  Choosing Help->Online Documentation... brings up a web browser (IE only, can't change it) with the following error:
> 
> "WebPACK Online Help. Please invoke the online help system from the Help and Technical Support icon in the Xilinx WebPACK program group."  That's annoying.
> 
> So, Choosing the "Help and Technical Support icon" in my WebPACK program group launches the HTML-based Help system.  OK, hmmm..I'm doing a 9500XL design in VHDL, but I have no idea what XST supports and does not support.  Where are the synthesis tool details?
> 
> Oh, here they are: not under CPLD WebPACK ISE, but rather under FPGA WebPACK ISE->Tools->Synthesis.
> 
> But wait -- there's only one simple page here.  First line: "XST is a Xilinx tool that synthesizes HDL designs to create EDIF netlists. For detailed information about XST, refer to the XST User Guide." that's rather useless, so I click the "XST User Guide" hyperlink.
> 
> That apparently runs the web browser within the help browser, connects to the Xilinx web site's "software Manuals Online" section, where I am NOT brought right to the XST User Guide -- which is what any reasonable person would EXPECT -- but rather, the top level of the Xilinx docs.  Which still say 3.1i, but aren't they up to 3.3i SP8 or something?
> 
> I click on the Design Entry icon.  Wow -- looks like some outdated documentation.
> 
> Ah -- waaaay down on the bottom of the "Docscan" frame, there's an XST link.  Ah-ha!  There it is.  And there's VHDL Support, too.  Where's Generate?
> 
> Hmmm...it's under Combinatorial Circuits.  Your guess is as good as mine as to why.
> 
> Ah, here's something: I have the keyword BEGIN right after my generate statement (foo : for ff in 0 to 2 generate begin) which is required in VHDL'93 if we include any declarations, but can be omitted if there are none.  Emacs seems to put them there, and I like my ENDs balanced by BEGINs.
> 
> But the example in the docs does NOT have the BEGIN, and the synth is happy if I remove it from my code.  But I'm not happy, since I have to modify my code to satisfy the quirks of a non-compliant tool.
> 
> OK, so the tools are FREE.  What should I expect?
> 
> Well, I expect compliance with the language.  Whom does Xilinx think they are, anyway? Synopsys?
> 
> I also expect the documentation to be correct, and it and the tool should be OBVIOUS.  If a menu item says, "Online help..." it should bring me directly to online help, not to a static HTML page with no links that says, "find something somewhere else."
> 
> If a link says, "XST User's Guide," I expect it to take me DIRECTLY to the XST user's guide -- not to the top level of docs for everything.
> 
> Xilinx: are you listening?
> 
> --andy


-- 

Rick "rickman" 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      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Article: 32003
Subject: Re: one state machine
From: Rick Collins <spamgoeshere4@yahoo.com>
Date: Sun, 10 Jun 2001 13:53:01 -0400
Links: << >>  << T >>  << A >>
Austin Franklin wrote:
> 
> > I don't agree that schematics substitute for a block diagram. They are
> > often so full of details that you can't see the flow for the gates!
> 
> It all depends on how well you know how to use the tools!  Obvioulsy, you're
> not a schematic guy ;-)  You certainly shouldn't have many, if any at all,
> gates on your schematics, except at the lowest levels, or a random gate here
> and there...for simplicity of drawing...
> 
> > Using heirarchical schematics start causing some of the same problems
> > that HDLs have. The signal names get complicated by the heirarchy path
> > and even renamed as one wire that traverses several levels will have
> > several names. The back end tool will want to pick just one name for the
> > whole net, but which one?
> 
> This is not a problem.  The names are deterministic, unlike HDLs, which are
> NOT deterministic.  The signal names will be what ever is the topest level
> signal name is.  If you name your heirarchical elements, then the signal
> name is very easy to follow.

I'm really not trying to continue the HDL vs Schematic debate. I am
addressing the one point that was made about block diagrams. My point is
that schematics are hard to use as block diagrams because they contain
far too much detail if done flat. If they are done as a heirarchy, which
makes them look more like a block diagram, they have some problems with
signal names. I have had to deal with this problem before, so I know it
exists. 

The fact that at any given level a net has a well defined name does not
change the fact that you don't know which of the several names will be
used in the end for this net. I seem to have noticed lately that the
backend tools are using the name that is highest in the heirarchy. In
the past I have seen the tool pick the name from the driver output or
sometimes it seems to be just random. Regardless, until the tool vendor
tells me what that name is, I have to guess from the many levels of the
heirarchy. 

-- 

Rick "rickman" 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      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Article: 32004
Subject: Re: FPGA based STN LCD Controller/Driver
From: "Carlo Kovacec" <dkova@xylon.hr>
Date: Sun, 10 Jun 2001 19:58:55 +0200
Links: << >>  << T >>  << A >>

"Chaudhry" <Imran@acronman.com> wrote in message
news:9fsb6h$o2n$1@dahlia.singnet.com.sg...
> Have anybody used FPGA device to control STN LCD panels either 16 bit or
24
> bit?

We have designed an IP core, logiCVC - Compact Video Controller for  Xilinx
FPGA.
Please visit http://www.logicbricks.com  , or contact me directly for more
info.

Best  Regards
Gordan Galic
gordan.galic@xylon.hr

>
> What is the possibility to do it ? we have a lots of stock of 6" STN
panels
> so we want to make standalone monitor for Video or VGA interface or both.
> All the controllers for STN LCD's are either ISA or PCI based which is not
> suitable for our application.
>
> Any comments will be appreciated.
>
> Regards,
>
> --
> ***********************
> Chaudhry
> ICQ# 11502664
> Singapore
> ***********************
>
>




Article: 32005
(removed)


Article: 32006
Subject: Re: one state machine
From: Kolja Sulimma <kolja@sulimma.de>
Date: Sun, 10 Jun 2001 23:10:49 +0200
Links: << >>  << T >>  << A >>
>
> > This is not a problem.  The names are deterministic, unlike HDLs, which are
> > NOT deterministic.  The signal names will be what ever is the topest level
> > signal name is.  If you name your heirarchical elements, then the signal
> > name is very easy to follow.

If I were doing schematic entry, I would still want my tools to perform logic
optimizations.
Doing something like running a SIS script over the netlist wont leave many of your
signals intakt except in datapaths.

Kolja Sulimma




Article: 32007
Subject: Re: Force tristate enable register into IOB
From: Ken McElvain <ken@synplicity.com>
Date: Sun, 10 Jun 2001 14:58:26 -0700
Links: << >>  << T >>  << A >>
Set syn_useioff=1 as a global attribute in your constraint file.  This
will lead to replication of tristate enable FF's automatically.  It will
also put IOB=TRUE on your registers for you.

hamish@cloud.net.au wrote:
> 
> fred <x@y.z> wrote:
> > With my toolset you also have top tell the synth tool _and_ the placement
> > tool to set "use IOB regs = true" or "push regs into IOBs" or whatever
> > equivalent to make this work.
> 
> Isn't "-pr b" on MAP supposed to do this?
> 
> BTW, Synplify always warns me that the option to pack FFs into IOBs
> hasn't been specified. I could never work out how to specify it though.
> I see to get IOB=TRUE on the relevant flip flops anyway.
> 
> Hamish
> --
> Hamish Moffatt VK3SB <hamish@debian.org> <hamish@cloud.net.au>

Article: 32008
Subject: Re: one state machine
From: "Austin Franklin" <austin@darkroom76.com>
Date: Sun, 10 Jun 2001 21:03:32 -0400
Links: << >>  << T >>  << A >>
> This
> points out that all FPGA's (from different companies) are not created
equal and
> logic elements in FPGA's are non portable. If one tries to use features
common
> to all FPGA's you are 50% slower and 50% less dense in most designs.

Not true.  Differences can be made up for in libraries, such as, say
counters.  You can impalement a generic symbol for an N bit counter of a
particular flavor, and change the underlying schematic to switch to an
entirely different technology.  I've been doing this for 10 years!

You can easily isolate the low level implementation from the high level
functionality, and with a well developed library for each technology, this
is not an issue.




Article: 32009
Subject: Re: one state machine
From: "Austin Franklin" <austin@darkroom76.com>
Date: Sun, 10 Jun 2001 21:07:33 -0400
Links: << >>  << T >>  << A >>

> I'm really not trying to continue the HDL vs Schematic debate. I am
> addressing the one point that was made about block diagrams. My point is
> that schematics are hard to use as block diagrams because they contain
> far too much detail if done flat. If they are done as a heirarchy, which
> makes them look more like a block diagram, they have some problems with
> signal names. I have had to deal with this problem before, so I know it
> exists.
>
> The fact that at any given level a net has a well defined name does not
> change the fact that you don't know which of the several names will be
> used in the end for this net. I seem to have noticed lately that the
> backend tools are using the name that is highest in the heirarchy. In
> the past I have seen the tool pick the name from the driver output or
> sometimes it seems to be just random. Regardless, until the tool vendor
> tells me what that name is, I have to guess from the many levels of the
> heirarchy.

It is entirely deterministic, unlike HDLs, it is absolutely not random.

I really don't know what problem you are talking about, and I've been doing
schematic based designs for over a decade...  Perhaps you could elaborate on
what you think the problem is???




Article: 32010
Subject: Re: one state machine
From: "Austin Franklin" <austin@darkroom76.com>
Date: Sun, 10 Jun 2001 21:13:47 -0400
Links: << >>  << T >>  << A >>
> The other missing issue, heavily related to the above, is portability.

Schematics are pretty much just as portable as a language.  You pick your
schematic tool, just like you pick your language.  VHDL or Verilog, ViewDraw
or OrBAD.  I haven't tried it, but isn't there EDIF?

 I also believe "portability" is entirely over-rated.  Even though it is
touted as one of the reasons to use HDLs, I know of almost no one who has
actually takes advantage of this purported "portability".  Some do, so I'm
not poo-pooing it, but I just don't believe it is as prolific as the
proponents of "portability" make the claim it is.




Article: 32011
Subject: Re: Async FIFO in maxplus2
From: Russell Shaw <rjshaw@iprimus.com.au>
Date: Mon, 11 Jun 2001 11:29:38 +1000
Links: << >>  << T >>  << A >>
Hi,

I found all the information by looking at the help file for the scfifo,
and the notes in the scfifo implementation.

Russell Shaw wrote:
> 
> Hi,
> 
> What other signal gates the clock in a 'gated clock' design?
> 
> For LPM_FIFO_DC, to write data, is this correct:
> 
> 1) Apply data to input port
> 2) Assert wrreq if not full
> 3) Apply a low-to-high edge at wrclock
> 
> How much delay should there be between 2 and 3? A small delay? or whole
> clock cycles?
> 
> bob elkind wrote:
> >
> > Many FPGA/ASIC designers (myself included) avoid gated clocks like the plague.
> > The preferred alternative is what is commonly called the "clock enable".
> >
> > For the LPM_FIFO_DC (we're talkin' MAX+2 here), the rdreq and wrreq are the
> > clock enables for the read clock and the write clock, respectively.  In the case
> > of the LPM_FIFO_DC, these inputs are not optional.  If you prefer/insist on using
> > gated clocks for read and write, then tie the rdreq and wrreq inputs to 'VCC'.
> >
> > Hope this helps, Russell!
> >
> > -- Bob Elkind, the e-team (fpga consulting, etc.)
> >
> > Russell Shaw wrote:
> >
> > > To clock data in and out, you just need inclock, outclock, empty, full.
> > > Why then rdreq, wrreq?
> > >

--
   ___                                           ___
  /  /\                                         /  /\
 /  /__\                                       /  /\/\
/__/   / Russell Shaw, B.Eng, M.Eng(Research) /__/\/\/
\  \  /  Victoria, Australia, Down-Under      \  \/\/
 \__\/                                         \__\/


Article: 32012
Subject: Re: XC4005XL is it a modern chip?
From: "Austin Franklin" <austin@darkroom76.com>
Date: Sun, 10 Jun 2001 21:32:26 -0400
Links: << >>  << T >>  << A >>
> I suggest that you don't purchase an old technology
> development board, when you can buy a brand new
> one for < US$120!
>
> The B3-SPARTAN2+ board from Burch Electronic
> Designs has a modern 200K gate Spartan2 device.
> http://www.burched.com.au/bedspartan2.html

Shameless plug ;-)




Article: 32013
Subject: Re: [Xilinx] Spartan II Devices ..internal tristate busses ...
From: Ray Andraka <ray@andraka.com>
Date: Mon, 11 Jun 2001 01:37:34 GMT
Links: << >>  << T >>  << A >>
The answer is yes, but you'll want to floorplan the tristate drivers. 
The automatic placement does an exceptionally poor job placing the 
tristates.  There is a limitation, in that only half of the tristates in 
a particular row can drive a give tristate bus line because of the chip 
architecture.  The tristate bus is 4 bits wide, and is rotated by one 
bit as it passes each CLB.  Each CLB can only get onto the two of the 
wires in the bus, so you end up having to pattern your drivers to 
account for this.  If you don't floorplan, it makes a ratsnest that is 
both slow and unnecessarily uses extra resources.

Simon Bacon wrote:

> Yes.  Look at the HDL design guide on the Xilinx Web site.
> 
> 
> 
>> is it possible within the Spartan-II devices  to instantiate - Verilog -
>> tristate busses, as it was possible in earlier Xilinx FPGA's. Actually
>> we are near 100% of the largest Spartan-II device, and maybe we could
>> get rid off the readback multiplexer for the internal registers using this
>> appraoch ...
>> 
>> Any hints help would be appreciated ...
>> 


Article: 32014
Subject: Re: Force tristate enable register into IOB
From: "Tobias Stumber" <tobias.stumber@de.bosch.com>
Date: Mon, 11 Jun 2001 08:15:43 +0200
Links: << >>  << T >>  << A >>
> How can I force tristate enable registers into IOBs (xilinx Spartan II)
> ?
>
> The synthesis tool (XST) makes one register out of the following 16
> registers. But I want 16 registers directly in the IOBs ! How can I code
> this without using the IOBUFs in the VHDL code !
>
> -- generate 16 tristate enable register
> gen1: for i in 0 to 16-1 generate
> process(clk, reset)
> begin
> if reset='1' then
> ge_mem_oe_reg(i) <= '1';
> elsif rising_edge(clk) then
> ge_mem_oe_reg(i) <= ge_mem_oe;
> end if;
> end process;
>       end generate;
>
> -- 16 tristate enable buffer
> gen2: for i in 0 to 16-1 generate
> GE_MEM_D(i) <= ge_mem_out_reg(i) when ge_mem_oe_reg(i)='0' else 'Z';
>    end generate;

Use:
  attribute IOB: string;
  attribute IOB of ge_mem_oe_reg : signal is "true";
This makes XST to replicate the FFs again after (!) having merged them into
one in the final optimization phase.
Works great !

Tobias




Article: 32015
Subject: Re: Pin locking in Maxplus2
From: Marcin E. Hamerla <mehamerla@pro.onet.pl>
Date: Mon, 11 Jun 2001 08:32:26 +0200
Links: << >>  << T >>  << A >>
Russell Shaw wrote:

>Using the Quartus fitter in maxplus2 (for acex1k), it crashes at
>28% when doing a simple 32x8 lpm fifo function and not much else
>(pc has 256MB ram). By setting the option not to use quartus, the
>compilation completes without error. Have you found the same bugginess?

I switched off Quartus. After making some minor changes in an existing
and working design compiled under older Max (without Quartus) some
parts of the design stoped working.

-- 
Pozdrowienia, Marcin E. Hamerla

"The value of achievement lies in the achieving"

Article: 32016
Subject: FPGA design tips in new monthly e-newsletter
From: "Tony Burch" <tony@BurchED.com.au>
Date: Mon, 11 Jun 2001 16:38:59 +1000
Links: << >>  << T >>  << A >>
Burch Electronic Designs announces new monthly e-newsletter,
featuring FPGA design tips.

Subscribe at http://www.BurchED.com.au
...just type in your email address and
press the Subscribe button. It's free.

Contents this month:
1. Website news
2. Feature link: Koay Kah Hoe, 8051 Core
3. Feature kit: B3-FSET
4. FPGA tips (1): Synchronous Design - Why?


Don't miss out on the FPGA tips!  Subscribe now!
http://www.BurchED.com.au

Tony Burch
http://www.BurchED.com.au




Article: 32017
Subject: Re: Force tristate enable register into IOB
From: Matthias Fuchs <matthias.fuchs@esd-electronics.com>
Date: Mon, 11 Jun 2001 09:30:30 +0200
Links: << >>  << T >>  << A >>
syn_useioff - Is this an XST option ?

Ken McElvain wrote:
> 
> Set syn_useioff=1 as a global attribute in your constraint file.  This
> will lead to replication of tristate enable FF's automatically.  It will
> also put IOB=TRUE on your registers for you.
> 

-- 
-------------------------------------------------
\ Matthias Fuchs                                 \
 \ esd electronic system design Gmbh              \
  \ Vahrenwalder Stra▀e 205                        \
   \ D-30165 Hannover                               \
    \ email: matthias.fuchs@esd-electronics.com      \
     \ phone: +49-511-37298-0                         \
      \ fax:   +49-511-37298-68                        \
       --------------------------------------------------

Article: 32018
Subject: Re: one state machine
From: Rick Filipkiewicz <rick@algor.co.uk>
Date: Mon, 11 Jun 2001 08:56:07 +0100
Links: << >>  << T >>  << A >>


Austin Franklin wrote:

> > The other missing issue, heavily related to the above, is portability.
>
> Schematics are pretty much just as portable as a language.  You pick your
> schematic tool, just like you pick your language.  VHDL or Verilog, ViewDraw
> or OrBAD.  I haven't tried it, but isn't there EDIF?
>

The difference here is that ViewDraw et al are proprietary whereas the 2 major
HDLs are public standards.
Hence if you stick to the defined synthesisable subset you know that all the
tools have to support your code.

As to EDIF:  Although its useful as a netlist exchange format for tools to pass
information its opaque nature makes it no use as a design language [except for
that small band of brothers who think LISP is the best thing since ASCII]. Its
also *not* a schematic format.

>
>  I also believe "portability" is entirely over-rated.  Even though it is
> touted as one of the reasons to use HDLs, I know of almost no one who has
> actually takes advantage of this purported "portability".  Some do, so I'm
> not poo-pooing it, but I just don't believe it is as prolific as the
> proponents of "portability" make the claim it is.

One of the big uses of modern, large, FPGAs is for ASIC prototyping. This is
either set explicitly at the start of a project or implicitly when your company
suddenly finds its shipping 20K WonderWidgets per month instead of an assumed
couple of K. Its an awful lot easier at that point to just throw the HDL at your
Vendor & tell them to get on with it than to have to do the technology
retargeting yourself.

The second place where portability comes in is the entire IP arena.

Portability takes some work but with effort its possible to reduce the
technology changes to this list:

o IO buffer selection.
o Clock trees.
o DP RAMs.
o DLL/PLLs.
o Choosing a cell for any clock domain synchronisers.

I also notice that you haven't taken up the question re version control & the
ability to do a diff between revisions.


Article: 32019
Subject: Re: Flash programming via FPGA's JTAG ????
From: Jens Hildebrandt <jens.hildebrandt@etechnik.uni-rostock.de>
Date: Mon, 11 Jun 2001 10:03:57 +0200
Links: << >>  << T >>  << A >>
This is a multi-part message in MIME format.
--------------C77319C808A54D704FF0D5BE
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit

Stephane wrote:
> 
> Hi,
> 
> I'm studying the following architecture :
> 
> I have an APEX200E with 1 embbeded NIOS microcontroller which runs his
> program from an external flash.
> The APEX's configuration file beeing contained in his 2 EPC2
> configuration flash PROM.
> 
> My main problem is NIOS and FPGA software update.
> 
> I intend to use JTAG.
> 
> The FPGA's EPC2 would be programmed by their JTAG pins.
> The flash memory has no JTAG pins.
> So I was wondering if I could program it by driving the FPGA's pins (
> the address, data and control bus pins )  using the FPGA's JTAG
> capabilities.
> 
> Does anyone has already done such a thing ?
> 
> Does it works ?
> What is the consequence on the JTAG software on the host PC ?
> 
> Thanks in advance.
> 
> Stephane.
> Thales Microelectronics.

Hi Stephane,

here at Rostock University we use such a system to update the
configuration PROM of a Virtex-FPGA. We use the boundary scan cells of a
CPLD normally used to boot the FPGA from the parallel FLASH to program
the FLASH with a new configuration bitstream. Currently we use bit
banging on the PC's parallel port as JTAG interface and, of course, it
is awfully slow as it takes one LPT access per signal change on the JTAG
lines and some fourty or so JTAG clock cycles for one signal change on
the FLASH pins (depends on FLASH size). But if you have a more
sophisticated JTAG-Interface (e.g. ÁC, special hardware device like
National's SCANPSC100 or homegrown FPGA/CPLD solution) you can
compensate for the high number of JTAG clocks per signal change on the
FLASH pins with a higher JTAG clock rate. We are currently building such
a parallel port - JTAG interface that should give us a 6 MHz JTAG clock,
but IIRC higher clock rates up to 10...20 MHz should be possible with
most JTAG devices (check the APEX data sheet).

So, you see, what you want is possible an , according to the number of
replies to your post, has been done many times before.

Regards 
	Jens
--------------C77319C808A54D704FF0D5BE
Content-Type: text/x-vcard; charset=us-ascii;
 name="jens.hildebrandt.vcf"
Content-Transfer-Encoding: 7bit
Content-Description: Card for Jens Hildebrandt
Content-Disposition: attachment;
 filename="jens.hildebrandt.vcf"

begin:vcard 
n:Hildebrandt;Jens 
tel;fax:+49 381 4983601
tel;work:+49 381 4983537
x-mozilla-html:FALSE
org:University of Rostock;Dept. of Applied Microelectronics and Computer Science
adr:;;Richard-Wagner-Str. 31;Rostock-Warnemuende;;18119;Germany
version:2.1
email;internet:jens.hildebrandt@etechnik.uni-rostock.de
x-mozilla-cpt:;23728
fn:Jens  Hildebrandt
end:vcard

--------------C77319C808A54D704FF0D5BE--


Article: 32020
Subject: Flash programming via FPGA's JTAG ?
From: stephane <sjulhes@free.fr>
Date: Mon, 11 Jun 2001 09:37:47 GMT
Links: << >>  << T >>  << A >>
First of all, thanks to you all for answering.

So yes it's possible !!

This time the question comes from the software team.
As this will be controled by a PC, they would like to know what will be
the amount of work to realize a software for programming the system
flash, the EPC2 and for testing the APEX which will come in BGA.

They argu that all the last experiences in other services were not
successful because of the difficulty to create the JTAG sequences.

So is it so complex to implement ?

Stephane.
Thales microelectronics.



Article: 32021
Subject: Re: Xilinx webpack annoyances (long and whiny)
From: Brian Gogan <briang@xilinx.com>
Date: Mon, 11 Jun 2001 12:09:33 +0200
Links: << >>  << T >>  << A >>
Andy -

Full support of VHDL'93 generate statements will be available in XST in the next release of the Xilinx tools, due the end of August.

Brian.

Andy Peters wrote:

> OK, so I'm trying to synthesize some VHDL code that simulates fine.  Nothing fancy, just some generates to create some registers.
>
> The synthesizer complains: ERROR :   (VHP__0163). Where's the list of errors? Hitting F1 is no help.  OK, look at the message: "Unexpected symbol read: BEGIN."
>
> Wait, this tool doesn't seem to understand the LRM.  My generate statement looks like:
>
>     foo: for ff in 0 to 2 generate
>     begin
>        bar: process (clk, rst) is
>        begin
>            if rst = '1' then
>                b(ff) <= '0';
>            elsif rising_edge(clk) then
>                b(ff) <= c(ff);
>            end if;
>         end process bar;
>     end generate foo;
>
> Hmmm..Ashenden sez this is OK; so does ModelSim.  So, what's wrong?  Does XST support generates?
>
> Soooo... what's with the online help?  Choosing Help->Online Documentation... brings up a web browser (IE only, can't change it) with the following error:
>
> "WebPACK Online Help. Please invoke the online help system from the Help and Technical Support icon in the Xilinx WebPACK program group."  That's annoying.
>
> So, Choosing the "Help and Technical Support icon" in my WebPACK program group launches the HTML-based Help system.  OK, hmmm..I'm doing a 9500XL design in VHDL, but I have no idea what XST supports and does not support.  Where are the synthesis tool details?
>
> Oh, here they are: not under CPLD WebPACK ISE, but rather under FPGA WebPACK ISE->Tools->Synthesis.
>
> But wait -- there's only one simple page here.  First line: "XST is a Xilinx tool that synthesizes HDL designs to create EDIF netlists. For detailed information about XST, refer to the XST User Guide." that's rather useless, so I click the "XST User Guide" hyperlink.
>
> That apparently runs the web browser within the help browser, connects to the Xilinx web site's "software Manuals Online" section, where I am NOT brought right to the XST User Guide -- which is what any reasonable person would EXPECT -- but rather, the top level of the Xilinx docs.  Which still say 3.1i, but aren't they up to 3.3i SP8 or something?
>
> I click on the Design Entry icon.  Wow -- looks like some outdated documentation.
>
> Ah -- waaaay down on the bottom of the "Docscan" frame, there's an XST link.  Ah-ha!  There it is.  And there's VHDL Support, too.  Where's Generate?
>
> Hmmm...it's under Combinatorial Circuits.  Your guess is as good as mine as to why.
>
> Ah, here's something: I have the keyword BEGIN right after my generate statement (foo : for ff in 0 to 2 generate begin) which is required in VHDL'93 if we include any declarations, but can be omitted if there are none.  Emacs seems to put them there, and I like my ENDs balanced by BEGINs.
>
> But the example in the docs does NOT have the BEGIN, and the synth is happy if I remove it from my code.  But I'm not happy, since I have to modify my code to satisfy the quirks of a non-compliant tool.
>
> OK, so the tools are FREE.  What should I expect?
>
> Well, I expect compliance with the language.  Whom does Xilinx think they are, anyway? Synopsys?
>
> I also expect the documentation to be correct, and it and the tool should be OBVIOUS.  If a menu item says, "Online help..." it should bring me directly to online help, not to a static HTML page with no links that says, "find something somewhere else."
>
> If a link says, "XST User's Guide," I expect it to take me DIRECTLY to the XST user's guide -- not to the top level of docs for everything.
>
> Xilinx: are you listening?
>
> --andy


Article: 32022
Subject: Re: Xilinx webpack annoyances (long and whiny)
From: Ian Young <ian@iay.org.uk>
Date: Mon, 11 Jun 2001 12:41:24 GMT
Links: << >>  << T >>  << A >>
Brian Gogan <briang@xilinx.com> wrote:

>Full support of VHDL'93 generate statements will be available in XST
>in the next release of the Xilinx tools, due the end of August.

Interesting (and welcome).  Can you say anything about any other parts
of VHDL'93 you're planning to tackle in the same release?

	-- Ian


Article: 32023
(removed)


Article: 32024
Subject: Re: Triscend A5: can it reconfigure itself?
From: "Ulf Samuelsson" <ulf@atmel.dot.com>
Date: Mon, 11 Jun 2001 15:24:38 +0200
Links: << >>  << T >>  << A >>
The Atmel FPSLIC can definitely reconfigure itself.
Contains AVR (nicer core) and 40k FPGA.
5k and 10 k FPGAs are out of the mask shop
will be here "real soon".
System Designer 2.0 with Win2k support will be shipping withing weeks.

--
Best regards,
ulf at atmel dot com
The contents of this message is intended to be my private opinion and
may or may not be shared by my employer Atmel Sweden

"Miguel Arias" <ariasm@inaoep.mx> skrev i meddelandet
news:ad5d52be.0106081408.63db47b7@posting.google.com...
> Hello,
>
> I am planning an application where I need a remote FPGA based
> architecture which I want to reconfigure using a microcontroller. The
> Triscend E5 CSoc (8051 + FPGA) specifications seem very interesting
> since all resides in a single chip but, is it possible to reconfigure
> the FPGA part of the chip using a program in the internal
> microcontroller?
>
> Based on the specifications, reconfiguration goes through the JTAG
> input, which I could control using the PIOs of the E5, but for
> security reasons, I would prefer to find out an internal way to do
> this, is it possible?
>
> Thanks for any comments with a copy to my email.
>
> Miguel
> FPGAs for Computer Vision
> INAOE - Mexico





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