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 5525

Article: 5525
Subject: Re: Xilinx or Altera?
From: s_clubb@netcomuk.co.uk (Stuart Clubb)
Date: Sat, 22 Feb 1997 10:05:25 GMT
Links: << >>  << T >>  << A >>
On Sat, 22 Feb 1997 02:00:58 -0500, Bill Harris <wmharris@cisco.com>
wrote:

>Lets talk about nonsense.  Lets look at the Xilinx app note located at
>http://www.xilinx.com/partinfo/3volt.pdf

<snip of valid problems with 4000E driving 3.3V devices>

Amen to all that. They also say it gets out of hand with industrial
temp due to +/- 10% supply swing. IMHO It's a kludge to get them off
the hook because 4000L is so comparatively slow.

What (if I may be so bold) do you use?

Stuart

Article: 5526
Subject: Re: Xilinx or Altera?
From: Mark Adams <us017033@mindspring.com>
Date: Sat, 22 Feb 1997 10:35:58 GMT
Links: << >>  << T >>  << A >>
On 20 Feb 1997 15:48:28 GMT, "Austin Franklin"
<#darkroom@ix.netcom.com#> wrote:

>Aaron,
< Snip >
>Austin Franklin
>..darkroom@ix.netcom.com.
>

Austin:
I'm the engineer that Aaron was referring to. Given your response to
his statements, I thought I would clarify the system and constraints
we were working with. The subsystem being implemented with the Xilinx
part performed the following functions:

900 ns A/D interface, 1.5 us Quad serial D/A interface and DMA engine
used to interface the A/D and D/A subsystems to a processor using two
memory resident table driven schedulers (Approx. 300 CLB's.)

6 center drive PWM modulators with anti-overlap, controlling 2
brushless DC motors. (Approx 200 CLB's)

The system target clock was 10 Mhz and the processors targeted were
the ADSP2181 and MC56301 (2 different DMA interfaces were developed
and simulated - The figures used are for the 56301 interface)

The FPGA design was entered as approx. 28 pages of schematic, which,
given the hierarchical and repetitive nature of the design would
expand to approx. 38 pages if flattened.

You would expect that this design would be fairly straight forward
and, that given the low clock speed, should have been fairly easy to
meet to design goals using a 4008A. It wasn't.
The principal problem with the XACT tools (Aside from being sloooow)
is that it's place and route stinks. To require ANY manual
intervention on a < 80% utilized design at 10 Mhz is absurd. The
Xilinx parts themselves are not bad, but the XACT tools are poor. By
my estimates, the job using the XACT + Orcad tools cost the job
approx. 2 mos. (over and above the expected learning curve).

Contrast the previous design with the current Altera design I am
doing: (BTW the Altera design took less than 3 mos. to design and
simulate, and was debugged and operating 3.5 weeks after power was
first applied)

The target chip is an 81500 (8K series because the customer didn't
have the 10K tools).

The design consists of:

16/20 bit micro coded special purpose processor for low level motor
control functions. (approx. 500 LE's)

Synch. serial communications interface (1Mb/sec) (speed limited by
opto interface) (approx. 100 LE's )

A/D interface and sequencer (approx. 100 LE's)

4x DAC interface (approx. 50 LE's)

Center drive PWM modulator, anti-overlap, and trap drive output decode
for three phase motor drive. (approx. 100 LE's)

This design is somewhat <Grin> bloated due to the Altera part's lack
of internal tri-state busses. This is exacerbated by the LE being
implemented as a 4 input LUT. This is fairly inefficient at
implementing the muxes (on the plus side it is helped by the cascade
logic chains) used to "emulate" the busses. The net effect of this is
to use approx. 25% of the logic to implement the busses. This lack
though, is more than made up by the tools WHICH SIMPLY WORK.

This design is approx. 42 pages of schematic and AHDL. At various
times in the design life this design reached 95% utilization (the
memory was being implemented on chip at the time). On a 100Mhz Pentium
W/40MB RAM this design compiles in approx. 15-20min and NEVER required
manual layout intervention to achieve >13Mhz performance. In spite of
the fact that the pins had been locked BEFORE the architecture was
completed, well before the design was even entered (required by the
parallel PWB generation). The Altera tools have their faults, the
parts have their limitations, but I spent <5% of my time fighting
Altera tool/part  limitations V.S.  >35% on the Xilinx tools (again,
exclusive of the learning curve).

From my point of view, in spite of the architectural advantages that
the Xilinx 4000 series has over the Altera 8K series (continued in the
4000EX /10K generation) the Altera tools are so much better than the
XACT tools that it is no contest.
                                       Mark Adams
                                        US017033@mindspring.com

BTW, I have been playing with the 10K parts, and without manual layout
intervention or tuning, I created a 24 bit shift and add 4 quadrant,
2's comp.  parallel frac. multiplier in 123 LE's (including local seq.
control logic and operand latches) which clocks at 74Mhz with a -3
part (64Mhz with a -4 part). Pretty fast for something requiring a
long carry chain, busses, and conditional operations.)
 





Article: 5527
Subject: Re: Xilinx or Altera?
From: Steve Wiseman <steve@sj.co.uk>
Date: Sat, 22 Feb 1997 13:38:31 +0000
Links: << >>  << T >>  << A >>
Peter Alfke wrote:

> Xilinx has a broader product offering in FPGAs or equivalent devices (
> XC2000, 3000, 3100, 4000, 4000EX, 5200, 6200 ) with more package options.
> Many of these families are pin-compatible within the family and between
> families. You can migrate a design without re-laying-out your pc-board.

Altera are, far too late, getting a lot better at this. An 84-pin PLCC
Xilinx device does get you a huge array of devices, if you can spare the
space. (Altera lost a design to Xilinx based on this last year, (a
network card with support for many, many line interfaces and protocols,
where the gate count varied by a factor of over 100))

> Xilinx devices tolerate pin-locking better than Altera's, an obvious
> result of the different interconnect architecture.

This has not been my experience. An almost-full Xilinx device gets very
sensitive to  pin positioning, I find. So do Altera devices, but I've
not been hit as hard (yet?)

> As pc-board lead times exceed FPGA design times, this becomes a critical issue.

I get my PCBs made on a 12-hour turn. Perhaps I'm a bit slow, but I
rarely get an FPGA design completed that fast....
 
> Xilinx XC4000 and 4000EX offer BlockRAM, distributed RAM in the CLBs which
> offer very fast, synchronous and even true dual-port RAM. ( Altera's 10K
> offers a much smaller number of bigger and simpler RAMs ). 

I find the bigger blocks of RAM to be more convenient. That's probably
just my take on it, though. 

>Xilinx has app
> notes describing the use of these RAMs as very fast FIFOs, one of their
> most popular applications.

Bu I seem to remember that one ends up devoting a fair amount of the
silicon to this? If I want fast FIFOs, I'll buy them and hang them
outside, if my FPGA can't cut it. (I also tend to want much, much bigger
FIFOs than will fit in any sane FPGA)
 
> Xilinx power consumption is lower, everything else being equal. This is
> the result of the different interconnect structure, and no marketing
> posturing can defy the laws of physics.

Yet I find Altera 8K devices running cooler than Xilinx 3k1 series,
running the same VHDL at the same speed. It's no proof, just
interesting. Maybe everything else isn't equal?

> ( The Altera message that the "sum
> of internal power is proportional to the percentage of blocks toggling at
> the clock rate" is wrong. It's tough to respect people who publish this
> kind of nonsense ).

I've not cared enough about power consumption (yet?) to read either
side's hype.  

> Xilinx interconnect can be significantly faster than Altera's, for the
> same reason as above. It is true that a complicated concatenated
> interconnect can get slow in a Xilinx device, but the user can influence
> that by prescribing the max tolerable delay, and let the software do the
> work. 

Yes, short runs of interconnect can be very fast in Xilinx devices. It's
only when I get a few CLBs in series that the speed plummets. Modifying
VHDL to put more and more pipeline stages in to cut the path length lost
its appeal some time ago. For tiny, fast state machines, I'll take
Xilinx (or a CPLD solution off-chip). Bigger state machines seem to run
faster in Altera devices. I imagine the same effect cuts in for Altera
when I spill out from a single LAB, but that's an easier constraint for
me to meet. 

Try to capture 200-MHz data in an 8k or 10k device, but you can in
> XC3100. Try to build a 50 MHz fully asynchronous FIFO in 10K, but you can
> in XC4000E. The richness of the Xilinx interconnect structure offers extra
> flexibility to achieve highest performance.

But I _can't_ run a non-trivial machine at 200MHz in either
architecture. Maybe I'm missing the point again. 
 
> Regarding software, it is no secret that Altera's compiles faster 

Yup. No argument

>( that is an obvious result of their simpler interconnect structure),

Does that somehow demean it? 

> and traditionally Altera's software has been easier to use. But on the other
> hand, Altera's software can be extremely frustrating for the power user,
> because certain things just cannot be done. ( "...when she was good, she
> was really good, but when she was bad she was horrid..".). Xilinx software
> has always given the user full access to all the device features. Many
> users love that, others complain about the learning curve....

I'll admit, I like the Xilinx tools. I like the level of control I get.
If I want to build tiny, tight machines, there's (currently?) no better
environment. What I don't like is _having_ to use them at the lowest
level. I spent 3 weeks trying to get a design working in a 3164a. It was
a couple of k-lines of VHDL. Messing with the layout after each VHDL
change became a nightmare. Perhaps it's an effect of the Xilinx-supplied
VHDL tools,  so that all node names change per-compile, and seem to
consist soley of strings of the characters 1, l, i, I, $, 5 and S that
made debugging such a pain, or perhaps I'm only supposed to take
contracts from customers who never change the specs mid-project?
After the third wasted weekend, I gave in, compiled the VHDL (with an
afternoons worth of changes to the Xilinx-specific bits) into an Altera
8K, using a borrowed toolset. I never even bothered to take out the
hacks I'd made to the source to get a better fit to the Xilinx
archirecture, as the design was a) getting late and b) functional at
full speed. The silicon was cheaper, too, but that was secondary.
Result, delivery on-time. (just....), payment, continuation of
livelihood and a certain amount of re-evaluation. 

> It is interesting to note that Xilinx software is becoming easier and more
> user-friendly, and Altera's is becoming more complex, as time goes on...

Yes. Perhaps they'll converge. It was very pleasant to be able to port
VHDL between families as easily as I could. In a perfect world, I'll
choose silicon per-project, without the tools worries. 
 
> Lastly, I invite every user to form his/her own opinion about the quality
> of technical support and the level of "honesty in marketing" displayed by
> the two companies. I am too close to that issue to be objective, but I
> think it is very important. End of soapbox.

I believe you pushed it rather hard, here. Neither company has perfect
tools, silicon or marketing. When they do, I'll retire, as my customers
will be able to communicate telepathically with the silicon and save me
a lot of effort!

> I hope I did not start a flame.

Close, very close.

(For the record, I am not associated with either of the above companies,
except as a customer. I still use both manufacturers, making choices
based on my imperfect knowledge of an imperfect world)

  Steve

-- 
Steve Wiseman, Senior Systems Engineer, SJ Consulting Ltd, Cambridge, UK
Desk +44 1223 578524 (Fax 578524) Group +44 1223 578518   steve@sj.co.uk
Article: 5528
Subject: Re: State Diagram Tools
From: Steve Bird <steveb@vizef.demon.co.uk>
Date: Sat, 22 Feb 1997 14:14:08 +0000
Links: << >>  << T >>  << A >>
In article <330CCA1B.28FF@hns.com>, "Kevin D. Drucker"
<kdrucker@hns.com> writes
>Sorry about the psuedo-spam...
>
>I am looking for a tool to assist me in documenting state machines. 
>Something similar to Rational Rose (which is a C++ tool), but
>specifically geard towards digital logic design.  It would be nice if it
>allowed you to show the diagrams either as a mealy or moore type
>machine.
>
>Any one know of such a tool?  If there is a shareware program out there
>that'd be great.  Windoze 95 or HP-UX...

Try flowHDL from KBS. This is a great tool, runs on UNIX (Sun,HP,IBM).
Produces VHDL or Verilog from the same diagram and has a built in
simulator. Supports multiple state machines (parallel), sync/async
resets, seperate control and datapath clocks. It also handles test
benches graphically as well (splits them into seperate HDL files).
Supports all of the main stream synthesis tools too.

-- 
Steve Bird
Article: 5529
Subject: Re: replicating structure in MaxPlus
From: Bill Harris <wmharris@cisco.com>
Date: Sat, 22 Feb 1997 10:29:06 -0500
Links: << >>  << T >>  << A >>
Graham Seaman wrote:
> 
> Hi all,
> 
> I'm trying to implement a cellular structure using Altera's
> MaxPlus software; that is, I have a simple basic 'cell' of
> about 10 gates which I need to replicate many times across
> the FPGA (connections are mostly NEWS style; ie. to 4 nearest
> neighbours of each cell).
>  If I was doing this in VHDL, I'd just use the 'generate'
> statement in a nested for loop. But I don't have VHDL, only
> AHDL, and can't work out a good way of doing this. Or am
> I better off sticking to schematic entry (lots of cut-and-paste)?

Graham,

I don't know which version of Maxplus you have, but the latest versions have GENERATE 
capability in AHDL.  Altera made some major additions to AHDL when they came out with LPM 
"megafunctions" a year or so ago (Version 6.0 I think).  The additions were not documented 
very well at first, and the new functions were not listed the their AHDL book.  They are in 
the on-line help for Version 7.0 however. Search for Generate keyword (AHDL) and there is a 
nice example of how to use it.   If you have LPM support in your software, you can pull up the 
AHDL for one of their LPM designs, such as LPM_COUNTER, and see examples of how they use them 
in their own functions.

If you don't have the latest version you can try the old fashioned way...

Use the hierarchy of AHDL.  Do an AHDL implementation of the logic that you want to replicate, 
as if it were a stand alone design. Create a top level design and declare your replicated 
logic design as a function.  Now you can instantiate the function in the VARIABLES section of 
the top level design.  You can make multiple instantiations in a single line such as:

FRED[20..0]  : MY_FUNCTION;

In the body of the top level design, you can make common connections (i.e. a common clock, 
enable or reset) to all instantiations of your function in a single line such as:

FRED[].clock = 25MHZ;
FRED[].clear = SYSRESET;
and so on....

You can make connections between instantiations like this:

FRED[1].some_input = FRED[2].some_output;

The ".name" is the name of the signal as you declared it in the function statement.

It is more verbose than a VHDL generate statement, but a heck of a lot easier than schematics.

Look at Maxplus help for more info, it is quite good, although you can get lost pretty easily. 

The "NEWS" interconnection is Xilinx speak, and doesn't serve well in the Altera architecture. 
 It would be difficult, if not impossible, to force the interconnects exactly the way you want 
them.  The "floorplanner" in Maxplus is really just a graphical placement viewer which allows 
you to manually make assignments if you wanted to. It does not provide a means of manual 
interconnect, ala the XACT tool. Beware, however, that you can make manual assignments to 
individual logic cells in such a way that you guarantee a no-fit situation.  This is 
particularly bad if you also pre-assign the pinouts.  If you want to constrain the placement, 
stick to assignments where logic is grouped into rows or LABs.  It is really important to the 
router that it have the freedom to shuffle logic between the 8 logic cells in each LAB.  If 
you overconstrain, you wind up in trouble very quickly.  The best thing to start out with is 
without placement assignments, but assign timing requirements. Let the compiler do the hard 
work.

Hope that helps some...

Bill Harris
Cisco Systems
Article: 5530
Subject: Re: Reverse Engineering FPGAs
From: zx80@dgiserve.com (Peter)
Date: Sat, 22 Feb 1997 16:22:37 GMT
Links: << >>  << T >>  << A >>

>Reverse engineering is far more difficult. It is almost impossible to
>deduce the FPGA design from the bitstream.

Is this not what Neocad must obviously have done? Before Xilinx bought
them, I mean :)


Peter.

Return address is invalid to help stop junk mail.
E-mail replies to z80@digiserve.com.
Article: 5531
Subject: Re: Xilinx or Altera?
From: waynet@goodnet.com (Wayne Turner)
Date: Sat, 22 Feb 97 16:35:46 GMT
Links: << >>  << T >>  << A >>
In article <5ema4m$bc1$1@news.pacifier.com>, szamos@pacifier.com (szamos) wrote:
>Austin Franklin (#darkroom@ix.netcom.com#) wrote:
>: 
>: The Altera design environment is not any industry standard.  They have
>: their own schematic capture and HDL.  To me, this means the engineer has to
>: learn 'yet another proprietary schematic capture and HDL' to get the job
>: done. 
>
>While they do have their own HDL, Max+Plus II also takes .edf which 
>can generated from Verilog, VHDL, etc.

And Viewlogic, and OrCAD, and ..............

Wayne
Article: 5532
Subject: Re: Xilinx or Altera?
From: Bill Harris <wmharris@cisco.com>
Date: Sat, 22 Feb 1997 11:51:41 -0500
Links: << >>  << T >>  << A >>
Stuart Clubb wrote:
> 
> On Sat, 22 Feb 1997 02:00:58 -0500, Bill Harris <wmharris@cisco.com>
> wrote:
> 
> >Lets talk about nonsense.  Lets look at the Xilinx app note located at
> >http://www.xilinx.com/partinfo/3volt.pdf
> 
> <snip of valid problems with 4000E driving 3.3V devices>
> 
> Amen to all that. They also say it gets out of hand with industrial
> temp due to +/- 10% supply swing. IMHO It's a kludge to get them off
> the hook because 4000L is so comparatively slow.
> 
> What (if I may be so bold) do you use?
> 
> Stuart

Stuart,

I use whatever is the best device for the particular design in terms of 
suitability, price, performance, availability, etc.  I have done 
several Xilinx 3K, 4K and 7300 designs (one board had all three 
families on the same board), and more Altera 5K, 7K and 8K designs than 
I care to remember.  I have also done a couple of MACH parts and 
a few Lattice designs.  I do have the luxury of having access to all the 
different vendor's tools, which most engineers at small companies do not 
have.  I find that the Altera designs are generally the least 
troublesome with the smoothest development flow, but when things go bad, 
the Altera tools don't give you the flexibility to get into the part and 
really tweak it like the Xilinx tools do.  Fortunately, those situations 
are few and far between these days.

I recognize marketing hype when I see it, and consider the source when I 
see outrageous claims for any electronic component which are obviously 
not from the engineering group.  When I see bad engineering 
recommendations coming from a companies' applications group, it does put 
some doubt in my mind about their products.

Bill Harris
Cisco Systems

"My opinions are my own and do not necessarily reflect those of my 
employer"
Article: 5533
Subject: Re: Xilinx or Altera?
From: "Iswada Osumundli" <#io@galofzu.net#>
Date: 22 Feb 1997 16:59:14 GMT
Links: << >>  << T >>  << A >>
The Lattice parts do not provide enough resources to do a PCI interface,
except a simple target.  If you want burst target or master functionality,
the Xilinx parts are the only ones that can do it.

Austin Franklin
..darkroom@ix.netcom.com.

Article: 5534
Subject: Re: Xilinx or Altera?
From: "Iswada Osumundli" <#io@galofzu.net#>
Date: 22 Feb 1997 17:28:56 GMT
Links: << >>  << T >>  << A >>
Mark,

In the original post, it was stated the problem was with the Orcad/Xilinx
interface, in reading your e-mail, that doesn't sound like that was the
problem at all.  I'm glad you elaborated on what the problem was.....

> To require ANY manual
> intervention on a < 80% utilized design at 10 Mhz is absurd. The
> Xilinx parts themselves are not bad, but the XACT tools are poor.

It is always to your advantage to do manual placement if the design has any
regular structures in it, i.e. registers, counters, tri-state busses (or
even logic-muxed busses) etc.  The 10MHz I agree with, should have been
real easy, but not the %80 utilization, especially if you have regular
structures.  I have used both the Altera tools and the Xilinx tools, and
the Xilinx tools have a much better router than the Altera tools.  No tool
set today for either FPGAs, ASICs or PCBs can do placement on regular
structures anywhere near as well as we humans can.

> By
> my estimates, the job using the XACT + Orcad tools cost the job
> approx. 2 mos. (over and above the expected learning curve).
> 
> Contrast the previous design with the current Altera design I am
> doing: (BTW the Altera design took less than 3 mos. to design and
> simulate, and was debugged and operating 3.5 weeks after power was
> first applied)

It's always easier to do things faster and better the second time, so that
data is not really relevant.

> 
> The target chip is an 81500 (8K series because the customer didn't
> have the 10K tools).
> 
> The design consists of:
> 
> 16/20 bit micro coded special purpose processor for low level motor
> control functions. (approx. 500 LE's)
> 
> Synch. serial communications interface (1Mb/sec) (speed limited by
> opto interface) (approx. 100 LE's )
> 
> A/D interface and sequencer (approx. 100 LE's)
> 
> 4x DAC interface (approx. 50 LE's)
> 
> Center drive PWM modulator, anti-overlap, and trap drive output decode
> for three phase motor drive. (approx. 100 LE's)
> 
> This design is somewhat <Grin> bloated due to the Altera part's lack
> of internal tri-state busses.

Which, in a Xilinx, when placed (which is real easy to do...) take up no
routing or logic resources.  When not placed, they can really screw things
up, which is what happened to you.  If you spend a week doing placement on
the design, it would have saved you weeks and weeks of frustration with
routing.  It also speeds up the routing from hours to minutes, so it pays
for it self to expect to do this up front.

The Xilinx tools also have a GUI for doing floorplanning which makes it
really easy to floorplan the regular structures.  I almost never floorplan
the random logic, because the tools do a better job than I can with these
in almost every case.

I'm glad you got your stuff to work.  It might have been in your best
interest to hire a consultant who was knowlwdgeable with the tools/parts
you were using, and ask them to either do the placement for you (which I
have done for many satisfied customers ;-)  ) or give you a short lesson on
how to do it, and set you up with a template so you could build from that.

If you want to e-mail me your .xff file, I would be interested in seeing
what could have been done with it...  I would hate to see you throw the
Xilinx parts (which can be a far better part for a lot of applications) out
of your bag of tools if there was something that could have been easily
done to solve your problem.

Did you contact your Xilinx local FAE, or the Xilinx tech support and tell
them your problem?  If you did, and they didn't tell you about the
floorplanner, or how to do placement, then they did you a dis-service. 
Could you please e-mail the name of your local Xilinx rep?

Thanks!

Austin Franklin
..darkroom@ix.netcom.com.
 
Article: 5535
Subject: Re: Using FPGA for PCI interface
From: "Iswada Osumundli" <#io@galofzu.net#>
Date: 22 Feb 1997 18:01:59 GMT
Links: << >>  << T >>  << A >>
That would be a tight squeeze, if able to be done at all.  It certainly
would leave no room for much of a back end interface....  Could you be so
kind as to elaborate on how you did this?  It seems like the resources
needed to implement a target are barely there, much less a master?

Thanks,

Austin Franklin
darkroom@ix.netcom.com


Hack-Man <sahack@ccgate.hac.com> wrote in article
<330A2E36.781B@ccgate.hac.com>...
> We have used Altera MAX7000 parts to implement PCI masters and targets. 
> At first we target Xilinx, but couldn't get the claimed speed.  We had
> Xilinx Apps help, and they did worse than we did.  So.....Altera seems
> to be the way to go.
> 
> Scott
> 
Article: 5536
Subject: Re: Xilinx or Altera?
From: eteam@aracnet.com (bob elkind)
Date: Sun, 23 Feb 1997 02:28:46 -0000
Links: << >>  << T >>  << A >>
In article <330EF6D7.2311@sj.co.uk>, steve@sj.co.uk says...

It's always fun to watch civilians beat on Peter Alfke [Xilinx] <g>.
In a recent episode, the following excerpt is replayed...

> > Xilinx devices tolerate pin-locking better than Altera's, an obvious
> > result of the different interconnect architecture.
> 
> This has not been my experience. An almost-full Xilinx device gets very
> sensitive to  pin positioning, I find. So do Altera devices, but I've
> not been hit as hard (yet?)
> 
> > As pc-board lead times exceed FPGA design times, this becomes a critical issue.
> 
> I get my PCBs made on a 12-hour turn. Perhaps I'm a bit slow, but I
> rarely get an FPGA design completed that fast....

Steve, I'm deeply envious of your access to miraculous PC board
turns.  However, Peter has a good point here, and your comments
shouldn't be allowed to stand without some gratuitous followup!

1.  For the rest of us, PC board layout, drcs, checking,
"tapeout", tooling, fab, incoming inspection, build, cleaning,
test, etc. represents a *way* longer "turn" than most FPGA
design changes that *may* affect pinout in an FPGA.

The ability of an FPGA to accommodate unanticipated
pinout changes is arguably a very valuable attribute.

2.  Once a board is built, and is in the hands of several
development/eval/test engineers, design changes in the
form of FPGA "code" are *much* easier to distribute and
install than a "board turn".

I understand your enthusiasm in presenting the "other side" to
Peter's assertions, but in this one case you may have outrun
the lot of us.

>   Steve

> Steve Wiseman, Senior Systems Engineer, SJ Consulting Ltd, Cambridge, UK


****************************************************************
Bob Elkind                              mailto:eteam@aracnet.com 
7118 SW Lee Road               part-time fax number:503.357.9001
Gaston, OR 97119           cell:503.709.1985   home:503.359.4903
****** Video processing, R&D, ASIC, FPGA design consulting *****
Article: 5537
Subject: Re: Xilinx or Altera?
From: Steve Wiseman <steve@sj.co.uk>
Date: Sun, 23 Feb 1997 10:38:54 +0000
Links: << >>  << T >>  << A >>
Austin Franklin wrote:
> 
> Paul,
> 
> Given what you say, then it is impossible to meet the PCI requirements for
> a master in an Altera part?  The CEs are required in the IOB in case the
> master disconnects, or stalls.
> 
> Well, then the only answer for a PCI interface is Xilinx...
> 
> Thanks for the info...
> 
> Austin

No, he stated (correctly) that the 8K series had no explicit CLKen
facility. 10K has them, with a bus of up to 6 clock enables running
round the IOEs. 10K parts are hardly new. 
The decision as to which manufacturer to use for PCI is not at all
clear-cut. (but Actel's devices with a PCI core on a corner are going to
be well worth a look.)
I'm about to start building a slave-only PCI card (to run a dawn'o'time
non-PC network to PCs, just for fun). That ought to prove things one way
or the other. (Anyone out there want to trade some PCI knowlege for some
ATM experience?)

  Steve

-- 
Steve Wiseman, Senior Systems Engineer, SJ Consulting Ltd, Cambridge, UK
Desk +44 1223 578524 (Fax 578524) Group +44 1223 578518   steve@sj.co.uk
Article: 5538
Subject: COURSE: High Level Design Using VHDL, Beaverton, Oregon
From: lindab@qualis.qualis.com (Linda Boyd)
Date: 23 Feb 1997 16:24:14 GMT
Links: << >>  << T >>  << A >>

                            A N N O U N C E

Qualis Design Corporation is offering another session of our popular course
"High Level Design Using VHDL" at our Beaverton, Oregon, Training Center.

This course presents a comprehensive introduction to the VHDL
language while teaching you how to approach complex design tasks using
High Level Design methods.  Through our advanced presentation methods,
we teach cutting-edge knowledge in a way that sticks -- graduates of the
course are immediately ready to tackle large scale VHDL-based designs
including ASICs and FPGAs.

For additional information about the material covered in this leading-edge
course, see the course description below.

This course can also be held at your facilities in a private, one-on-one
setting -- contact us for more information.

Schedule
--------
The March 1997 schedule for this class follows:

       Course Title                        Course Date       Status
       -------------------------------------------------------------------
       High Level Design Using VHDL:       Mar 31 - Apr 4    Open


The Qualis Difference
---------------------
We know what it's like to work under the pressure of aggressive schedules and
immense technical challenges.  We believe that High Level Design methods
and technology, such as HDL-based verification and synthesis, are the key
to tackling those challenges and conquering today's design problems.
Our courses can make a real difference in your day-to-day work life
by showing you the high leverage points of VHDL and High Level Design.
Here's how we do it:

 -- The Qualis "Best In Class" Instructor Team draws upon the absolute best
    VHDL and Verilog consultants in the industry -- our people!  Your
    instructor regularly tackles 500,000+ gate ASICs, and is ready to
    share his in-depth knowledge with you.  Synthesis, verification,
    design methodology -- it's all there.  You won't find instructors
    of this caliber anywhere else.

 -- Our courses are intense, hands-on events using the latest EDA tools
    and hardware.  Everything you need to learn quickly and efficiently is
    provided -- you supply the brain, we'll supply everything else.

 -- Our courses are like no other in the EDA industry.  Engineers and
    Managers who attend our courses will learn what's important and why, and
    where to focus their time and resources for maximum leverage from HDLs
    and design tools.  And, unlike other vendor courses, our courses are
    *dynamic* -- we constantly update our material with the latest in
    High Level Design techniques and information, so you're assured of
    learning the latest in the field.

--  Our courses are respected in the industry.  We have taught our
    High Level Design courses to many Fortune 500 companies and
    hundreds of Engineers and Managers.  Our student references attest
    to the outstanding quality and real-world usefulness of our classes.
    Just ask and we'll prove it.

About Qualis Design Corporation
-------------------------------
Founded in 1992, Qualis Design Corporation has quickly become the
leading independent provider of Elite Consulting and Training
Services.  The company provides services to leading-edge
high technology firms worldwide, including Intel, Hewlett-Packard,
Tektronix, Xerox, TRW, and Northern Telecom.  Qualis' corporate
headquarters are located in Beaverton, Oregon.

Don't miss this opportunity to learn the latest in High Level Design
from the best in the industry.  For course syllabi and registration
information, contact us at:

                        Linda Boyd, Training Registrar
                           Qualis Design Corporation
                       8705 SW Nimbus Avenue, Suite 118
                          Beaverton, Oregon 97008 USA

                            Phone: +1-503-644-9700
                             FAX: +1-503-643-1583
                          mailto:training@qualis.com

Brief Course Description
--------------------------

                    High Level Design Using VHDL

                          Course Overview

        Copyright (c) 1995, 1996, 1997 Qualis Design Corporation

"High Level Design Using VHDL" is a fast paced, five-day hands-on,
multimedia course designed to make class participants immediately productive
in a VHDL-based design environment using state-of-the-art simulation and
synthesis tools.

After an introduction to VHDL, the course deviates from the traditional
bottom-up, gates-to-behavioral modeling presentation of other VHDL courses
and reverses the flow, teaching top-down design practices, with early
special emphasis on coding guidelines, efficient testbench generation and
advanced design verification techniques. These skills are reinforced
throughout the week while teaching VHDL from a top-down perspective.

The course labs are designed to accommodate the learning aptitudes of
a wide range of students with diverse design experiences.  All students
complete the main part of the lab and an optional part is for students
who finish early and want to learn additional material.  This lab structure
caters to all student skill levels and provides excellent opportunities
to expand one's knowledge of VHDL simulation and modeling techniques.

Each day of class includes interactive lectures with four or five lab
sessions.  Students will have access to individual Sun Sparcstations,
the Synopsys VSS and Model Technology V-System / Workstation simulation
environments, and the Synopsys DC Expert synthesis environment for use
during the laboratory sessions.  The instructor presents the material
using a projection system that allows 30% more material to be presented
in a given amount of time with vivid, interest-grabbing color slides.

Full Course Syllabus Available
------------------------------
A full course syllabus listing all topics covered in this course
is available.  Contact us for more information.

------------------------------------------------------------------------------

                           Qualis Design Corporation
                       8705 SW Nimbus Avenue, Suite 118
                          Beaverton, Oregon 97008 USA

                            Phone: +1-503-644-9700
                             FAX: +1-503-643-1583
                          mailto:training@qualis.com

"DC Expert" is a trademark of Synopsys, Inc.
"Verilog" is a registered trademark of Cadence Design Systems, Inc.

Copyright (c) 1995, 1996, 1997, Qualis Design Corporation.  All Rights Reserved.

Article: 5539
Subject: COURSE: High Level Design Using Verilog, Beaverton, Oregon
From: lindab@qualis.qualis.com (Linda Boyd)
Date: 23 Feb 1997 16:26:41 GMT
Links: << >>  << T >>  << A >>

                            A N N O U N C E

Qualis Design Corporation is offering another session of our popular course
"High Level Design Using Verilog" at our Beaverton, Oregon, Training Center.

This course presents a comprehensive introduction to the Verilog
language while teaching you how to approach complex design tasks using
High Level Design methods.  Through our advanced presentation methods,
we teach cutting-edge knowledge in a way that sticks -- graduates of the
course are immediately ready to tackle large scale Verilog-based designs
including ASICs and FPGAs.

For additional information about the material covered in this leading-edge
course, see the course description below.

This course can also be held at your facilities in a private, one-on-one
setting -- contact us for more information.

Schedule
--------
The March 1997 schedule for this class follows:

       Course Title                        Course Date       Status
       -------------------------------------------------------------------
       High Level Design Using Verilog:    Mar 10 - Mar 14   Open


The Qualis Difference
---------------------
We know what it's like to work under the pressure of aggressive schedules and
immense technical challenges.  We believe that High Level Design methods
and technology, such as HDL-based verification and synthesis, are the key
to tackling those challenges and conquering today's design problems.
Our courses can make a real difference in your day-to-day work life
by showing you the high leverage points of Verilog and High Level Design.
Here's how we do it:

 -- The Qualis "Best In Class" Instructor Team draws upon the absolute best
    Verilog and VHDL consultants in the industry -- our people!  Your
    instructor regularly tackles 500,000+ gate ASICs, and is ready to
    share his in-depth knowledge with you.  Synthesis, verification,
    design methodology -- it's all there.  You won't find instructors
    of this caliber anywhere else.

 -- Our courses are intense, hands-on events using the latest EDA tools
    and hardware.  Everything you need to learn quickly and efficiently is
    provided -- you supply the brain, we'll supply everything else.

 -- Our courses are like no other in the EDA industry.  Engineers and
    Managers who attend our courses will learn what's important and why, and
    where to focus their time and resources for maximum leverage from HDLs
    and design tools.  And, unlike other vendor courses, our courses are
    *dynamic* -- we constantly update our material with the latest in
    High Level Design techniques and information, so you're assured of
    learning the latest in the field.

--  Our courses are respected in the industry.  We have taught our
    High Level Design courses to many Fortune 500 companies and
    hundreds of Engineers and Managers.  Our student references attest
    to the outstanding quality and real-world usefulness of our classes.
    Just ask and we'll prove it.

About Qualis Design Corporation
-------------------------------
Founded in 1992, Qualis Design Corporation has quickly become the
leading independent provider of Elite Consulting and Training
Services.  The company provides services to leading-edge
high technology firms worldwide, including Intel, Hewlett-Packard,
Tektronix, Xerox, TRW, and Northern Telecom.  Qualis' corporate
headquarters are located in Beaverton, Oregon.

Don't miss this opportunity to learn the latest in High Level Design
from the best in the industry.  For course syllabi and registration
information, contact us at:

                        Linda Boyd, Training Registrar
                           Qualis Design Corporation
                       8705 SW Nimbus Avenue, Suite 118
                          Beaverton, Oregon 97008 USA

                            Phone: +1-503-644-9700
                             FAX: +1-503-643-1583
                          mailto:training@qualis.com

Brief Course Description
--------------------------

                   High Level Design Using Verilog

                          Course Overview

        Copyright (c) 1995, 1996, 1997 Qualis Design Corporation


"High Level Design Using Verilog" is a fast paced, five-day hands-on,
multimedia course designed to make class participants immediately productive
in a Verilog-based design environment using state-of-the-art simulation,
waveform viewing and synthesis tools.

After an introduction to Verilog, the course deviates from the
traditional bottom-up, gates-to-behavioral modeling presentation of
other Verilog courses and reverses the flow, teaching top-down design
practices, with early special emphasis on coding guidelines, efficient
testbench generation and advanced design verification techniques.
These skills are reinforced throughout the week while teaching Verilog
from a top-down perspective.

The course labs are designed to accommodate the learning aptitudes of
a wide range of students with diverse design experiences.  All students
complete the main part of the lab and an optional part is for students who
finish early and want to learn additional material.  This lab structure
caters to all student skill levels and provides excellent opportunities
to expand one's knowledge of Verilog simulation and modeling techniques.

Each day of class includes interactive lectures and several lab sessions.
Students will have access to individual Sun Sparcstations, the Verilog-XL
simulation environment, and the Synopsys DC Expert synthesis environment
for use during the laboratory sessions.  The instructor presents the material
using a projection system that allows 30% more material to be presented
in a given amount of time with vivid, interest-grabbing color slides.

Full Course Syllabus Available
------------------------------
A full course syllabus listing all topics covered in this course
is available.  Contact us for more information.

------------------------------------------------------------------------------

                           Qualis Design Corporation
                       8705 SW Nimbus Avenue, Suite 118
                          Beaverton, Oregon 97008 USA

                            Phone: +1-503-644-9700
                             FAX: +1-503-643-1583
                          mailto:training@qualis.com

"DC Expert" is a trademark of Synopsys, Inc.
"Verilog" is a registered trademark of Cadence Design Systems, Inc.

Copyright (c) 1995, 1996, 1997, Qualis Design Corporation.  All Rights Reserved.

Article: 5540
Subject: Re: 2nd try: What kind of functions mostly implemented using FPGAs?
From: Peter Alfke <peter@xilinx.com>
Date: Sun, 23 Feb 1997 14:47:34 -0700
Links: << >>  << T >>  << A >>
Erik de Castro Lopo wrote:

> I'd agree with what Robert has to say. FPGAs and CPLD are just means of
> replacing 74XX series logic and small PALs (ie Lattice 22v10).
> 
I would not agree with this statement. I worked at Fairchild in 1968
when we introduced one TTL-SSI and -MSI circuit per week. That kind of
effort lasted until the early 'seventies. TTL-SSI and -MSI innovation
has stagnated since that time, when many of today's designers were still
in diapers.

It's like saying: an automobile is nothing but a substitute for a
horse-drawn carriage, or a TV is a radio with pictures.

I find it flattering that designers think of logic functions in TTL-MSI
terms. We must have done a good job back then, if designers, 25 years
later, still think of counters as 74161s and decoders as 74138s and ALUs
as 74181s. That's o.k., as long as that does not stifle the imagination. 
TTL circuits were defined with one overriding constraint: the 14- or
16-pin package, later stretched to 18 to 24 pins. Today, it would be
silly to consider the 7490 or the -much better- 74161 as the "standard"
counter. FPGAs and CPLDs offer much more exciting flexibility: wide
decoding, small RAMs, abundance of flip-flops, internal busses,
efficient identity comparators, novel ways to perform the
multiplications needed in DSP, etc.

Yes, you can use FPGAs and CPLD as TTL-equivalent "garbage collectors".
But you can also use them for much more productive things

Peter Alfke, Xilinx Applications
Article: 5541
Subject: Re: Xilinx or Altera?
From: Brad Taylor <blt@emf.net>
Date: Sun, 23 Feb 1997 16:35:47 -0800
Links: << >>  << T >>  << A >>
Bill Harris wrote:

> .....
> Lets talk about nonsense.  Lets look at the Xilinx app note located at
> 
> http://www.xilinx.com/partinfo/3volt.pdf
> 


Obviously! If you can't get the manufacturer to guarantee 
safe input currents, you can't use it.( except maybe on the bench, if
you monitor your supply voltages). Sometimes I read these app notes and
I feel like Xilinx is saying, 'yeah we know its technically illegal, but
it'll probably work, so go ahead and do it but don't tell your boss'.  I
mean, if it works, why don't they guarantee it?
 
I find it very irritating that while they won't guarantee tracking
delays, or min delays, there are many references to 'reasonable'
tracking (70%) and min delays(20%). It is desireable to make use of
tracking, especially when confronted with interfacing to memories, and I
never quite know where the line is. 

Let me give you a simple example of what I mean. You may need to
guarantee an output pulse width, say 10ns. There are 4 delays of
interest, the clkin to pin register delay for both low and high
transitions, and the clk to output delay for both types of 
transitions.

clkin   __/--\__/--\__/--\__/--\__/--\__/--\__/--\__/--\
clk       __/--\__/--\__/--\__/--\__/--\__/--\__/--\__/--\
pulse    ------\_____/-----------------------------
                
                    

Without using tracking delays:

clkin->clk  7ns   (use max delay )
clk  ->lo   7ns   (use max delay )
Tpw        10ns   ( the spec )
clkin->clk  0ns   (use min delay )
clk  ->hi   0ns   (use min delay )     
----------------------------------
Total      24ns   (without use of tracking)


With using tracking delays:
clkin->clk  7ns   (use max delay )
clk  ->lo   7ns   (use max delay )
Tpw         10ns  ( the spec )
clkin->clk -5ns   (use max*70% delay )
clk  ->hi  -5ns   (use max*70% delay )     
----------------------------------------------
Total      14ns   (with use of tracking)


There is considerable difference in the max cycle time.


_
Brad
Article: 5542
Subject: Re: Reverse Engineering FPGAs
From: "Andrew Metcalfe" <metcalfe@iaccess.com.au>
Date: 24 Feb 1997 01:32:22 GMT
Links: << >>  << T >>  << A >>
One way to secure your FPGA (with Xilinx anyway) is to do a "hardwire"
conversion. This is basically a mask version of the programmable product.
Secure as far as external probing is concerned. 

For reconfigurable systems, I have seen systems where an external CPLD is
used as a "dongle" for the FPGA. A unique handshake occurs each time the
system boots. This stops the system being copied but does not actually
secure the FPGA design.

-Andrew Metcalfe, ACD Australia


John A. Harding <harding@pelican.cs.ucla.edu> wrote in article
<5eii65$399@pelican.cs.ucla.edu>...
> 
> Hello-
>  
> Are there any ways to prevent SRAM FPGAs from being
> reversed engineered. Or copied. I have heard about
> the approach using a battery and loading the config.
> in the Factory, but what about upgrades, and reconfigurable
> systems.
>  
> It seems that many applications cannot use FPGAs because
> of this reason.
>  
> Thanks,
>  
> John
> 
> 
Article: 5543
Subject: Market share - synthesis tools?
From: P Nibbs <pnibbs@icd.com.au>
Date: Mon, 24 Feb 1997 13:59:29 +1100
Links: << >>  << T >>  << A >>
Hi All,

I would be very greatful if anybody was aware of a site on the WWW where
I could find information on the market share of the major FPGA EDA
vendors - in particular in relation to their synthesis tools.

Thanks in advance for any help,

Cheers,

Phil.

Article: 5544
Subject: Re: Reverse Engineering FPGAs
From: "Austin Franklin" <#darkroom@ix.netcom.com#>
Date: 24 Feb 1997 03:50:38 GMT
Links: << >>  << T >>  << A >>
If the application has enough volume, or security is manditory, use a
hardwire after debugging the design.

Austin Franklin
..darkroom@ix.netcom.com.


John A. Harding <harding@pelican.cs.ucla.edu> wrote in article
<5eii65$399@pelican.cs.ucla.edu>...
> 
> Hello-
>  
> Are there any ways to prevent SRAM FPGAs from being
> reversed engineered. Or copied. I have heard about
> the approach using a battery and loading the config.
> in the Factory, but what about upgrades, and reconfigurable
> systems.
>  
> It seems that many applications cannot use FPGAs because
> of this reason.
>  
> Thanks,
>  
> John
> 
> 
Article: 5545
Subject: Q: Xilinx PPR Strategy Tips?
From: no@spam.net (Tom Vrankar)
Date: 24 Feb 1997 04:48:26 GMT
Links: << >>  << T >>  << A >>
We're designing several larger XC4KE devices under XACT 5.2.1 with XABEL as 
our HDL (long story; absence of timing-driven mapping is annoying, but not 
terminal). Most of the designs are proceeding reasonably. One is giving us 
hell.

We had gotten PPR to produce results that _almost_ met our timing constraints 
several times. After complicating the timing constraints to exclude multicycle 
paths, and a few design adjustments indicated by functional simulation, the 
design now no longer routes.

We have tried to add placement hints in the form of "bounding boxes" (i.e. 
[clb_r1c1 clb_r5c5] kinds of things where the available number of cells in the 
box is about twice what is needed), and more carefully placing bussed items, 
but PPR is producing more and more unroutes with each run.

Both placement and routing efforts are maxed. The design used to complete in 
about 6 hours, but now seems to want to go for days. As part of trying to 
diagnose this problem, we'd been waiting till the ppr.log file indicated 
placement was complete and then issuing a kill -2 to review the placement, but 
have begun to just set route=false.

A symptom seems to be that when the solutions were _almost_ meeting timing, 
the occupied CLBs reported by Floorplanner seemed to be evenly distributed 
across the device (as were the unoccupied CLBs), but now PPR keeps packing the 
CLBs in a tight knot near the center of the device, so it runs out of routing 
channels. It also used to sequence bus bits pretty well, but now scrambles 
them hopelessly (so we now explicitly place key register bits as anchors).

The hotline suggested that we noplace every third row or so to force better 
distribution of CLBs. This sounds like a hack, and we didn't need it before 
(nor on the other designs that use timing constraints). Has anyone else seen 
this kind of behaviour, or have any PPR or constraint tips that might get us 
out of this hole? Can anyone provide a better description of how PPR works, 
what it looks for, what it tries, to give us some insight on how to work 
around it? Any help is appreciated.

-- 
                         ---------------------------------
                                               Tom Vrankar
                                               twv@ici.net
                         http://www.ici.net/customers/twv/
                                         Rhode Island, USA

Article: 5546
Subject: Re: Xilinx or Altera?
From: CoxJA@augustsl.demon.co.uk (Julian Cox)
Date: Mon, 24 Feb 1997 11:30:28 GMT
Links: << >>  << T >>  << A >>
"Iswada Osumundli" <#io@galofzu.net#> wrote:

>The Lattice parts do not provide enough resources to do a PCI interface,
>except a simple target.  If you want burst target or master functionality,
>the Xilinx parts are the only ones that can do it.
>

IYHBO (In Your Heavily Biassed Opinion)  ;-)



Article: 5547
Subject: Re: State Diagram Tools
From: Volker Hetzer <hetzer.abg@sni.de>
Date: Mon, 24 Feb 1997 13:05:10 +0100
Links: << >>  << T >>  << A >>
Kevin D. Drucker wrote:
> 
> Sorry about the psuedo-spam...
> 
> I am looking for a tool to assist me in documenting state machines.
> Something similar to Rational Rose (which is a C++ tool), but
> specifically geard towards digital logic design.  It would be nice if it
> allowed you to show the diagrams either as a mealy or moore type
> machine.
We've a tool called SME (State Machine Editor).
It allows the drawing of state-machines, but
doesn't care about moore/mealy. Instead you can specify for each output
whether it should change whenever the input changes or only at the clock
edge.
The diagrams look like program flowcharts, consisting of
state/switch/condi-
tion/assignment symbols. Variables that keep their values over several
clocks
exist as well. The tool can generate VHDL and Verilog (synthesizeable).

It is available vor HPUX and Sunos/Solaris (sparc).
The tool is usually not used to document state machines afterwards but
to design them in the first place. Basically, our ASIC-designers use it
as a tool for implementing program flowcharts.

> Any one know of such a tool?  If there is a shareware program out there
> that'd be great.  Windoze 95 or HP-UX...
Sorry, it costs money.

Volker
Article: 5548
Subject: How are states changed in ALTERA
From: Yourname@somwhere.COM (Your Name)
Date: Mon, 24 Feb 1997 14:19:06 GMT
Links: << >>  << T >>  << A >>
Help,

We are trying to convert a AHDL file into an ABEL file. The problems that 
we have, have to do with the default section of the file and how it is 
defined during the state transitions. He defaults a signal to VCC (all of 
the circuit was designed to be active low). Then during the first state 
he assigns a logic 1 to that signal defined in the default section. Then 
for the next few states he does not mention the state of that output, so 
we assume the output will take on that default (VCC or unchanged). After 
that he will again assign that signal to a logic level 1. From what we 
can gather, this signal will never change, hence has no real need to be a 
part of this state machine.

Are we right ? From some of the examples of state machines in AHDL and on 
a handout we had on AHDL basics tells us that we are right, but it does 
seem strange that the programmer did this. We want to make sure who is 
right. The state machine has 82 states, and we don't want to guess. 
(Note: We can not get in contact with the orginal programmer)

Also while I am at it, we have another problem from this same state 
machine. Has has named two state the same names, but assigned different 
outputs in the two states (i.e. s11 has one set of outputs and s11 has 
another set of outputs). I would say this would be a problem, but to get 
to and from each of those states has nothing to do with inputs, hence I 
would guess this would not effect the status of the state machine. Does 
anybody know for sure ?

Thank you for your time,
Jonathan Montag, SSD Raytheon

Article: 5549
Subject: Re: replicating structure in MaxPlus
From: Woody Johnson <woodyj@transition.com>
Date: Mon, 24 Feb 1997 09:05:33 -0600
Links: << >>  << T >>  << A >>
Use the AHDL FOR/GENERATE statement.

Woody
-----------------------------------
Graham Seaman wrote:
> 
> Hi all,
> 
> I'm trying to implement a cellular structure using Altera's
> MaxPlus software; that is, I have a simple basic 'cell' of
> about 10 gates which I need to replicate many times across
> the FPGA (connections are mostly NEWS style; ie. to 4 nearest
> neighbours of each cell).
>  If I was doing this in VHDL, I'd just use the 'generate'
> statement in a nested for loop. But I don't have VHDL, only
> AHDL, and can't work out a good way of doing this. Or am
> I better off sticking to schematic entry (lots of cut-and-paste)?
> 
> Anyone got any ideas?
> Thanks for any help
> Graham
>


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