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 32125

Article: 32125
Subject: Re: Xilinx PCI core location constraints
From: "Jamie Sanderson" <jamie@nortelnetworks.com>
Date: Thu, 14 Jun 2001 20:24:55 -0400
Links: << >>  << T >>  << A >>
Hi Matt;

Thanks for your reply. At the same time I posted here, I asked my local
Xilinx FAE about it. A correct version of the UCF has made it to my inbox.

In your opinion, am I taking a risk from a timing perspective by moving the
core to a single quadrant? I am too early in the design to run it through
place and route to check for myself.

Regards,
Jamie

"Matthieu Cossoul" <Matt.Cossoul@xilinx.com> wrote in message
news:3B257AEC.3932EFAB@xilinx.com...
> Jamie,
>
> That's a feature and a bug:
> . the feature: the TBUF placement algorithm is assuming that the datapath
(ADs
> and CBEs) is surrounding RDY signals, which makes some sense timing wise.
> . the bug: the test that was supposed to enforce this didn't report an
error as
> it was supposed to.
>
> Gathering all PCI signals on a single bank is freeing up a bank where a
> different IO standard could be used. The new version of the UCF Generator
to be
> released this Friday would now allow you to do that. Meanwhile, I could
> generate for you the UCF you need.
>
> Sorry for the inconvenience,
> - Matt




Article: 32126
Subject: Re: Xilinx Virtex 2: Configurations problems
From: Peter Alfke <peter.alfke@xilinx.com>
Date: Thu, 14 Jun 2001 17:36:10 -0700
Links: << >>  << T >>  << A >>
Martin, please tell us *when* INIT goes Low.

Peter Alfke



Martin Forest wrote:

> We observe a problem when programming a Virtex II XC2V1000 if the device is not at the end of the configuration chain.
>
> We used a concatenated configuration bitstream (generated by Xilinx PROM file formatter). During the configuration, the XC2V1000 FPGA receive the appropriate bitstream but don't transfer by DOUT pin the next bitstream to the next device of the chain. In more the INIT pin is drived low to indicate an error.
>
> -> Our configuration clock is 4MHZ
> -> We used the ServicePack #8
> -> The configuration EEPROM is XC18V04vq.
>
> If someone have an idea, could you share it.
> martin_forest@nmss.com
>
> Thanks


Article: 32127
Subject: Re: Xilinx PCI core location constraints
From: Matt Cossoul <Matt.Cossoul@xilinx.com>
Date: Thu, 14 Jun 2001 18:49:49 -0700
Links: << >>  << T >>  << A >>
Jamie,

In the Virtex-II part you're using, it does meet timing with the Ping reference
design (which is provided with the core) in a -4 speed grade. So that looks
pretty safe.

Regards,
- Matt

Jamie Sanderson wrote:

> Hi Matt;
>
> Thanks for your reply. At the same time I posted here, I asked my local
> Xilinx FAE about it. A correct version of the UCF has made it to my inbox.
>
> In your opinion, am I taking a risk from a timing perspective by moving the
> core to a single quadrant? I am too early in the design to run it through
> place and route to check for myself.
>
> Regards,
> Jamie


Article: 32128
Subject: Re: Pin locking in Maxplus2
From: Russell Shaw <rjshaw@iprimus.com.au>
Date: Fri, 15 Jun 2001 12:40:13 +1000
Links: << >>  << T >>  << A >>
Hi B.E,

I checked all the settings from your report file, and they seem to be
the defaults that match what i used. I've reinstalled maxplus2 and the
patch, but that didn't fix it.

Do you have just the 10.0-10.01 patch applied, or are there other
patches?
TIA

bob elkind wrote:
> 
> As I mentioned in previous followup, your source code compiled without
> problems on my machine.  One place to look for clues is the logic synthesis
> settings section of the report file, and another place to check would be
> pin assignments (in the .ACF file).
> 
> -- Bob Elkind, the e-team  fpga design, consulting
> 
> Russell Shaw wrote:
> 
> > bob elkind wrote:
> > >
> > > Hmmm....  I had no probs with 10.0 (no patch).  I wonder if the patch
> > > may be a problem.  Maybe someone else listening in on this thread might
> > > have some data to help correlate our experiences.  Also, I was targeting
> > > only 6K and Acex1K devices.
> >
> > Hi,
> >
> > I had the problem with maxplus2 10.0. I applied the latest patch to
> > get 10.01, but still get the same problem. The file that causes the
> > quartus fitter to crash when doing an auto acex1k device is:
> >
> > INCLUDE "lpm_fifo_dc.inc";
> >
> > SUBDESIGN fifo
> > (
> >         data[7..0]       : INPUT;
> >         wrreq    : INPUT;
> >         rdreq    : INPUT;
> >         rdclock  : INPUT;
> >         wrclock  : INPUT;
> >         q[7..0]  : OUTPUT;
> >         rdempty  : OUTPUT;
> > )
> >
> > -- rest of design source code snipped

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

Article: 32129
Subject: Re: Pin locking in Maxplus2
From: bob elkind <eteam@aracnet.com>
Date: Thu, 14 Jun 2001 23:41:02 -0700
Links: << >>  << T >>  << A >>
Russell,

I've not installed *any* patches to Max+2 10.0.

We could be here forever.  Why don't you just email me your .ACF file and
your .TDF file(s).  I'll see if it compiles on my machine.  If it does compile,
then you may want to re-install MAX+2 without any patches, see if it runs,
then install a patch and see if it still runs, etc. etc.

-- Bob Elkind, the e-team, fpga/design consulting.

Russell Shaw wrote:

> Hi B.E,
>
> I checked all the settings from your report file, and they seem to be
> the defaults that match what i used. I've reinstalled maxplus2 and the
> patch, but that didn't fix it.
>
> Do you have just the 10.0-10.01 patch applied, or are there other
> patches?
> TIA


Article: 32130
Subject: Re: From EDF to VHDL?
From: "Ulises Hernandez" <ulisesh@ecs-telecom.removeplease.co.uk.invalid>
Date: Fri, 15 Jun 2001 09:11:15 +0100
Links: << >>  << T >>  << A >>
Hello,

Thank you for your answer.
I think this is a time an effort problem, I would like to convert all the
gates to 'comprehensible' VHDL, maybe the fastest way is to understand the
design and redesign it in VHDL without copying the schematics sheet by sheet
and gate by gate.
Anyway we'll see.

Thanks all of you.

"Ulises Hernandez" <ulisesh@ecs-telecom.removeplease.co.uk.invalid> wrote in
message news:992445896.9765.0.nnrp-08.c2de5a16@news.demon.co.uk...
> Hello to the group,
>
> I am currently working in FPGA designs. We are now investigating if is
> possible to upgrade a design. This is an old design which was implemented
in
> a XC5215 FPGA and was built using schematics :-((, the schematics were
done
> using Mentor Graphics tools. This tool generates the EDF file which you
use
> for the build process, we will probably migrate this design to the Spartan
> II family. But my personal challenge will be to pass them to VHDL. It is a
> nightmare I know, os I would like to know if someone knows some software
to
> convert to VHDL a EDF file o convert to VHDL a schematic using Mentor
tools.
> There must be something, to pass it to Verilog will do but it much better
to
> VHDL. Maybe the VHDL generated is quite poor but easier to modify than the
> schematics.
>
> Thank you to all in advance.
>
> Ulises Hernandez
> Design Engineer
>
>
>
>



Article: 32131
Subject: NIOS users ?
From: stephane <sjulhes@free.fr>
Date: Fri, 15 Jun 2001 09:28:18 GMT
Links: << >>  << T >>  << A >>
Hi,

We are about to use a NIOS solution in an APEX 200E.

Are there any known problems with the NIOS ?
Is it really as performant as ALTERA says ?
Are the developpement tools easy to use ?
Are they powerfull ?

In fact is it really worth using it rather than a small microcontroller
with flash and ISP ???

Thanks.

Stephane.
Thales Microelectronics.



Article: 32132
Subject: Re: Fpga tutorial
From: Srinivasan Venkataramanan <srini@realchip.co.in>
Date: Fri, 15 Jun 2001 17:25:14 +0530
Links: << >>  << T >>  << A >>
Try,

http://www.fpga-guru.com/links.htm

HTH,
Srini

juanjo wrote:
> 
> I´m very interested in FPGAs. Where I can find a tutorial about FPGAs?
> 
> Thank you

-- 
Srinivasan Venkataramanan (Srini)
ASIC Design Engineer,
Chennai (Madras), India

Article: 32133
Subject: Re: From EDF to VHDL?
From: "S. Ramirez" <sramirez@cfl.rr.com>
Date: Fri, 15 Jun 2001 12:36:52 GMT
Links: << >>  << T >>  << A >>
> Thank you for your answer.
> I think this is a time an effort problem, I would like to convert all the
> gates to 'comprehensible' VHDL, maybe the fastest way is to understand the
> design and redesign it in VHDL without copying the schematics sheet by
sheet
> and gate by gate.
> Anyway we'll see.
>
> Thanks all of you.
>
> "Ulises Hernandez" <ulisesh@ecs-telecom.removeplease.co.uk.invalid> wrote
in

Ulises,
    I agree with you.  I think it would be worth the effort to convert an
old design to VHDL (or Verilog), especially since old designs tend to be
smaller than new designs.  I have done several conversions with great
success.
    One problem you might watch out for, though, is timing.  If the old
design is completely synchronous, then a new design will work better,
because it is typically faster and will meet the clock to clock timings
easier.  However, if the old design is asynchronous, which could involve
race conditions, the faster speeds and the randomness of an FPGA place and
route can cause problems.  I suggest that you look for anything that is
asynchronous in nature and try to redesign it into a synchronous function.
Another problem area is internally generated clocks.  They can generally be
designed out and replaced by more FPGA-friendly designs.
     Good luck.
Simon Ramirez, Consultant
Synchronous Design, Inc.
Oviedo, FL  USA




Article: 32134
(removed)


Article: 32135
Subject: Re: Xilinx webpack annoyances (long and whiny)
From: Brian Gogan <briang@xilinx.com>
Date: Fri, 15 Jun 2001 15:42:09 +0200
Links: << >>  << T >>  << A >>
Here are some of the language support improvements in XST for the 4.1i
release:

Support generate statement (example: processes under generate) including
"begin" key word in "if...generate" statement
Support multi-dimensional constants
Support multi-dimensional arrays
Support "exit" statement
Support "next" statement
Support conditional signal assignment without ELSE
Support functions in vector declarations
Support such expressions like "funct(ND)(2 downto 0)", where funct is a
function name
Support "type'predifined_attribute" expressions
Support description of control signals of FF when only one bit of the
vector is used (ex: falling_edge(signal(i)))
Configuration statement, generally used in simulation.
Support "+", "-", etc. in the "use" statement
etc.

We are constantly working on improving the language coverage of XST and
appreciate any feedback regarding such and indeed any other aspect of the
tool.

Brian.



Ian Young wrote:

> 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: 32136
Subject: Re: From EDF to VHDL?
From: Brian Drummond <brian@shapes.demon.co.uk>
Date: Fri, 15 Jun 2001 15:00:07 +0100
Links: << >>  << T >>  << A >>
On Thu, 14 Jun 2001 08:39:41 +0100, "Ulises Hernandez"
<ulisesh@ecs-telecom.removeplease.co.uk.invalid> wrote:

>Hello Rick,
>
>Thank you for your answer.
>
>I have tried to convert the NGD file to VHD file.
>I have now a nice and incomprehensible file of 29000 lines of code with all
>the simprims in it (and this is the smallest LCA).
>
>I tried to synthesize this file with Leonardo without converting the
>primitives to VHDL RTL mode (Just trying). It didn't work as you probably
>knew.

If you're using Leonardo, can't it read the EDF file directly? (And
write VHDL from Leonardo using the inbuilt Xilinx library, if you so
want?)

You may need to load the XILINX primitive library first...

- Brian


Article: 32137
Subject: Re: Force tristate enable register into IOB
From: Sergei Storojev <Sergei.Storojev@xilinx.com>
Date: Fri, 15 Jun 2001 16:27:06 +0200
Links: << >>  << T >>  << A >>
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
Dear Mr. Fuchs,
<p>Please find enclosed 2 VHDL files which whows 2 different was to describe
output FFs and tristate buffers:
<ul>
<li>
In the first case (iobs_1.vhd) the 8-bit data register <b>data_reg</b>
and the FF controlling tri-state buffers <b>control_ff</b> are described
separately. I applied IOB attribute to the both of them and XST will replicate
<b>control_ff</b>
7 times and attach IOB attributes in NCF file. You do not have to manually
replicate <b>control_ff</b> in the HDL code. In the current release XST
will remove these FFs. In the next release this has been fixed and there
is an additional switch to prevent the removal of equivalent FFs.</li>

<li>
In the second case (iobs_2.vhd), to describe the same functionality I used
a Synchronous OE template. In the current release this method will not
work but will be available in the 4.1i release.</li>
</ul>

<p><br>Here are some additional notes concerning IOB constraint:
<ul>
<li>
IOB constraint must be applied:</li>

<ul>
<li>
to the FF instance, if the FF is directly instantiated in the HDL code</li>

<li>
to the output signal of the FF, if the behavioral description is used.</li>

<br>NOTE:&nbsp; if FF is an output FF, then today XST does not allow you
to apply this constraint directly to the output port.&nbsp; The user must:
<ul>
<li>
declare the signal representing the FF as “signal” and not as an output
port, and</li>

<li>
apply the IOB attribute to this signal.</li>
</ul>
The fix for this issue is done in 4.1i.</ul>
</ul>

<p><br>Please let me know if you have further questions.
<br>Have a nice week-end
<p>Sergei.
<p>-------------------------------------
<br>Sergei Storojev,
<br>XST Technical Marketing Engineer
<br>Grenoble, France
<br>tel (direct): +33 4 76 70 51 04
<br>&nbsp;
<p>========================== IOBS_1.VHD =============================
<br><tt>library ieee;</tt>
<br><tt>use ieee.std_logic_1164.all;</tt>
<br><tt>use ieee.std_logic_unsigned.all;</tt>
<p><tt>entity design is</tt>
<br><tt>&nbsp; port(clk&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; SEL&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; A,B&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic_vector (7 downto 0);</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; REZ&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: out std_logic_vector (7 downto 0)</tt>
<br><tt>&nbsp; );</tt>
<br><tt>end design;</tt>
<p><tt>architecture archi of design is</tt>
<p><tt>&nbsp; signal arith_res, data_reg : std_logic_vector (7 downto 0);</tt>
<br><tt>&nbsp; signal control_ff&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: std_logic;</tt>
<br>&nbsp;
<p><tt>&nbsp; attribute IOB: string;</tt>
<br><tt>&nbsp; attribute IOB of control_ff: signal is "true";</tt>
<br><tt>&nbsp; attribute IOB of data_reg: signal is "true";</tt>
<br><tt>begin</tt>
<p><tt>&nbsp;&nbsp; arith_res &lt;= A + B;</tt>
<br>&nbsp;
<p><tt>&nbsp;&nbsp; process(clk)</tt>
<br><tt>&nbsp;&nbsp; begin</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; if (clk'event and clk='1')
then</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; control_ff
&lt;= SEL;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; data_reg
&lt;= arith_res;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end if;</tt>
<br><tt>&nbsp;&nbsp; end process;</tt>
<br>&nbsp;
<p><tt>&nbsp;&nbsp; REZ &lt;= data_reg when control_ff = '0'</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
else "ZZZZZZZZ";</tt>
<p>end archi;
<br>&nbsp;
<p>========================== IOBS_2.VHD =============================
<br><tt>library ieee;</tt>
<br><tt>use ieee.std_logic_1164.all;</tt>
<br><tt>use ieee.std_logic_unsigned.all;</tt>
<p><tt>entity design is</tt>
<br><tt>&nbsp; port(clk&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; SEL&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; A,B&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: in&nbsp; std_logic_vector (7 downto 0);</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; REZ&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
: out std_logic_vector (7 downto 0)</tt>
<br><tt>&nbsp; );</tt>
<br><tt>end design;</tt>
<p><tt>architecture archi of design is</tt>
<p><tt>&nbsp; signal arith_res, data_reg: std_logic_vector (7 downto 0);</tt>
<p><tt>&nbsp; attribute IOB: string;</tt>
<br><tt>&nbsp; attribute IOB of data_reg: signal is "true";</tt>
<br><tt>begin</tt>
<p><tt>&nbsp;&nbsp; arith_res &lt;= A + B;</tt>
<br>&nbsp;
<p><tt>&nbsp;&nbsp; process(clk)</tt>
<br><tt>&nbsp;&nbsp; begin</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; if (clk'event and clk='1')
then</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; if
(SEL = '0') then</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
data_reg &lt;= arith_res;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; else</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
data_reg &lt;= "ZZZZZZZZ";</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end
if;</tt>
<br><tt>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end if;</tt>
<br><tt>&nbsp;&nbsp; end process;</tt>
<p><tt>&nbsp;&nbsp; REZ &lt;= data_reg;</tt>
<br>&nbsp;
<p><tt>end archi;</tt>
<br>&nbsp;
<p>Matthias Fuchs wrote:
<blockquote TYPE=CITE>Hi !
<p>How can I force tristate enable registers into IOBs (xilinx Spartan
II)
<br>?
<p>The synthesis tool (XST) makes one register out of the following 16
<br>registers. But I want 16 registers directly in the IOBs ! How can I
code
<br>this without using the IOBUFs in the VHDL code !
<p>-- generate 16 tristate enable register
<br>gen1: for i in 0 to 16-1 generate
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; process(clk, reset)
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
begin
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
if reset='1' then
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
ge_mem_oe_reg(i) &lt;= '1';
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
elsif rising_edge(clk) then
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
ge_mem_oe_reg(i) &lt;= ge_mem_oe;
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
end if;
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end process;
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end generate;
<p>-- 16 tristate enable buffer
<br>gen2:&nbsp;&nbsp; for i in 0 to 16-1 generate
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
GE_MEM_D(i) &lt;= ge_mem_out_reg(i) when ge_mem_oe_reg(i)='0' else 'Z';
<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; end generate;
<p>Matthias</blockquote>
</html>


Article: 32138
Subject: Re: Xilinx Virtex 2: Configurations problems
From: "Eric Lewis" <eric_lewis@nmss.com>
Date: Fri, 15 Jun 2001 07:45:14 -0700
Links: << >>  << T >>  << A >>
I am going to describe the case because Martin is temporarily out of the office.

The exact number of bits has not been counted but by the approximative time, it seems to go low at the end of the second bitstream.

I don't know if it is related but we have 2 other different boards for which the VirtexII is the last in the chain and the people had to disable the CRC checking to be able to make the board work.

In the current case, changing the CRC checking don't change anything.

Thanks.

Eric
eric_lewis@nmss.com

Article: 32139
Subject: FPGA programing via the parallel port
From: Mikael <mikael@imsys.se>
Date: Fri, 15 Jun 2001 07:49:59 -0700
Links: << >>  << T >>  << A >>
Hi,

I would like to program the FPGA via the parallel port in slave parallel mode. I'm using a spartan-II. I'm thinking to conecct D0~D7 to parallelport D0~D7. Strobe to CCLK, Busy/Dout to Busy, nSelect to nWrite, nPinit to nCs. But I don't know what to do with nProgram. I vill use the parallel port after the FPGA is programed to boot a yprocessor so all parallel conections vill be conected all time. Is it ok to have CCLK moving when the device is programed? 
Is this somthing anyone of yoh have an idea of?

/Mikael

Article: 32140
Subject: Using the Triscend A7 UART
From: Michael Dales <michael@dcs.gla.ac.uk>
Date: 15 Jun 2001 16:21:29 +0100
Links: << >>  << T >>  << A >>
Hi there,

I just got a Triscend A7 starter kit, and wanted to use one of the
UARTs to talk to another machine. I'm a little confused by the
documentation, and just wondered if someone could let me know if I
have the right idea.

As far as I can tell from the documentation, there are two UARTs on
the A7, but they don't have any pins connected to them by default. To
use one of the UARTs I'd need to take the side-band signals into the
CSL and connect them to pins (as specified in the board doc) using the
I/O->Input and I/O->Output modules in Fastchip.

-- 
Michael Dales --- email: michael@dcs.gla.ac.uk --- tel: +44 141 330 6297
Department of Computing Science, University of Glasgow, Glasgow, G12 8QQ

Article: 32141
Subject: Re: NIOS users ?
From: "Victor Schutte" <victors@mweb.co.za>
Date: Fri, 15 Jun 2001 18:08:55 +0200
Links: << >>  << T >>  << A >>
It is a great product but keep in mind it is  work in progress.


I also thought that a small microcontroller would do the job but after I had
to change my last 3 designs the flexilibility and speed of NIOS made my life
a lot easier.

With NIOS you get the following (at least):
    Quartus II (which you can use for any 100k and 200k APEX device,  I
think)
    NIOS development board. After playing with the board I reconfigured in
and installed it in to one of my proto products.
    NIOS HDK ver 1.1.
    NIOS SDK (you can also use the Cygwin compiler to generate a toolset for
your own custom CPU).

You can use APEX20K, -20KE,-20KC and ACEX1K100 devices.

I recently designed a NIOS core with a 32bit RISC CPU, 2x 32 bit timers, 40x
I/O lines including 8 dual edge IRQs, 6x Uarts with setable prescalers. The
entire design was targeted for a APEX20k100 -240pin QFP device. Using a CPLD
(3128 device) as a config device and additional I/O the PCB now boasts with
80 I/O lines. The entire PCB is only 80mm x 110mm and contains a Real Time
Clock (with onboard CR2032 battery), WDT, 1 MB Flash, 256KB SRAM,  ports for
4MB external SRAM and a option to use either a crystal or oscillator.  If
you need more speed (e.g. MUL command ) you can always drop a peripheral or
2 and add the new functionalilty.

The compiler is free (within bounds) which means that you can supply you
NIOS board with a compiler to a client.

Known problems?
    C++ support is stil dodgy, but since version 1.01 to 1.1 they sorted out
many bugs.
    Interrupts feature from PIO peripheral is still very buggy. It tends to
skip interrupts (1% loss but bad enough) but I added my own VHDL patch to
the inputs (nice feature when your CPU is in a FPGA) to filter these
glitches. I have tested IRQs at about 20 000 per second without any
problems.


Cost?  Looking at my new PCB the total cost actually dropped compared to my
multi-processor 8051 card. Also I can reuse the same PCB for other FPGA
designs reducing later development cycles.

Performance? If you compare it to something like a Pentium it will come last
if you look at features such as true 32bit code (NIOS is 16 bit CPU with
32bit datapath) and floating point (FP routines slows NIOS down
considerably). If you want to replace a big 8051 design or even leaning
towards a CPU like a 386EX   NIOS might be for you.

When not to get involved: If you don't have any serious Altera FPGA
background be carefull!! To get the best of NIOS you must know how to add
you own functionality, otherwise a standard off the shelf CPU is still the
best option.

My new NIOS PCB will be ready middle of July and will cost about $600. Those
interested can reply to me directly for more info.


Victor Schutte
Zerksus Engineering CC


"stephane" <sjulhes@free.fr> wrote in message
news:3B29D349.8475700E@free.fr...
> Hi,
>
> We are about to use a NIOS solution in an APEX 200E.
>
> Are there any known problems with the NIOS ?
> Is it really as performant as ALTERA says ?
> Are the developpement tools easy to use ?
> Are they powerfull ?
>
> In fact is it really worth using it rather than a small microcontroller
> with flash and ISP ???
>
> Thanks.
>
> Stephane.
> Thales Microelectronics.
>
>



Article: 32142
Subject: Re: Using the Triscend A7 UART
From: "Dion Kriel" <dionHateSpam@triscendHateSpamToo.com>
Date: Fri, 15 Jun 2001 10:15:11 -0700
Links: << >>  << T >>  << A >>
Hi Michael,

Yes, there are 2 UARTS and you have to connect the sideband signals to the
pins as you describe.
I will forward your posting to our support guys, so you should expect a call
soon.
PLease let us know what part of the documentation you found confusing, so we
can improve it.

from the Triscend trenches... :-)



"Michael Dales" <michael@dcs.gla.ac.uk> wrote in message
news:n23d91kbuu.fsf@kettle.dcs.gla.ac.uk...
Hi there,

I just got a Triscend A7 starter kit, and wanted to use one of the
UARTs to talk to another machine. I'm a little confused by the
documentation, and just wondered if someone could let me know if I
have the right idea.

As far as I can tell from the documentation, there are two UARTs on
the A7, but they don't have any pins connected to them by default. To
use one of the UARTs I'd need to take the side-band signals into the
CSL and connect them to pins (as specified in the board doc) using the
I/O->Input and I/O->Output modules in Fastchip.

--
Michael Dales --- email: michael@dcs.gla.ac.uk --- tel: +44 141 330 6297
Department of Computing Science, University of Glasgow, Glasgow, G12 8QQ




Article: 32143
Subject: Re: looking for work
From: "Kevin Timmons" <kevin@hwhp.freeserve.co.uk>
Date: Fri, 15 Jun 2001 19:31:24 +0100
Links: << >>  << T >>  << A >>
Hi

My name is Kevin Timmons from NEC in Milton Keynes

Yes we are looking for 1 more Engineer. Can you please answer the following
questions too see if you are what we are looking for.

1)  What ASIC Design experience do you have?(describe breifly)

2)   Do you know about the following:

Unix
VHDL
Verilog
Modelsim
Synopsys
 DC & Formality

3)  What degree do you have?. Type and Class?

4) What salary are you expecting?

We provide a very good package(including company car for non graduates(and
graduates after 2 years) The work is also very interesting.

If this does not suit you could get your hands on Electronics Weekly. All
the best agencies are advertised here.

I hope this helps good luck

Kevin Timmons




cyber_spook <pjc@cyberspook.freeserve.co.uk> wrote in message
news:3B1FEE33.C2D5BF06@cyberspook.freeserve.co.uk...
> Is anyone looking for an Hardware engineer? Or can anyone recommend a
> good agent?
>
> I live in the South East of the UK
>
> Regards
>
> Cyber_spook_man
>  pjc@cyberspook.freeserve.co.uk
>



Article: 32144
Subject: Fpga tutorial
From: "juanjo" <jalonsosuarez@wanadoo.es>
Date: Fri, 15 Jun 2001 20:41:37 +0200
Links: << >>  << T >>  << A >>
I´m very interested in FPGAs. Where I can find a tutorial about FPGAs?

Thank you



Article: 32145
Subject: efficient CAM in Virtex or Spartan II?
From: Eric Smith <eric-no-spam-for-me@brouhaha.com>
Date: 15 Jun 2001 12:36:46 -0700
Links: << >>  << T >>  << A >>
Is there any particularly efficient way to implement a small fast CAM,
perhaps eight entries, in a Xilinx Virtex or Spartan II FPGA?

The obvious approach is to use eight registers, eight comparators, and
an eight-input priority encoder.  This uses a lot of CLBs.

The slower approach is to use a RAM, a single comparator, and have a
counter cycle through the addresses.  But I think this is too slow for
my current needs.

But is there another approach that I've overlooked?

Thanks!
Eric Smith

Article: 32146
Subject: Re: efficient CAM in Virtex or Spartan II?
From: Vikram Pasham <Vikram.Pasham@xilinx.com>
Date: Fri, 15 Jun 2001 13:02:51 -0700
Links: << >>  << T >>  << A >>
Eric,

There are three different ways of implementing CAMs in Virtex/Spartan-II
devices. You either use LUT (distributed memory), SRL16s or BlockRAMs to
implement CAMs depending on how you want the read/write cycles. SRL16 and
BlockRAM based CAMs will be faster and may meet your performance
requirements.

We have three app. notes describing these CAM implementations at
http://support.xilinx.com/apps/virtexapp.htm

Refer XAPP201, XAPP201, XAPP202 and XAPP203.

Coregen also offers a CAM core based on SRL16 implementation which takes
one cycle to read and 16 cycles to write. For further details, check the
CAM datasheet at
http://support.xilinx.com/ipcenter/coregen/updates.htm


-Vikram Pasham
Xilinx Applications


Eric Smith wrote:

> Is there any particularly efficient way to implement a small fast CAM,
> perhaps eight entries, in a Xilinx Virtex or Spartan II FPGA?
>
> The obvious approach is to use eight registers, eight comparators, and
> an eight-input priority encoder.  This uses a lot of CLBs.
>
> The slower approach is to use a RAM, a single comparator, and have a
> counter cycle through the addresses.  But I think this is too slow for
> my current needs.
>
> But is there another approach that I've overlooked?
>
> Thanks!
> Eric Smith


Article: 32147
Subject: Re: efficient CAM in Virtex or Spartan II?
From: John_H <johnhandwork@mail.com>
Date: Fri, 15 Jun 2001 20:17:32 GMT
Links: << >>  << T >>  << A >>
Check out

  http://support.xilinx.com/xapp/xapp201.pdf

and you'll find a neat little way to use SRL16 elements to build up a cam
among other suggestions.


Eric Smith wrote:

> Is there any particularly efficient way to implement a small fast CAM,
> perhaps eight entries, in a Xilinx Virtex or Spartan II FPGA?
>
> The obvious approach is to use eight registers, eight comparators, and
> an eight-input priority encoder.  This uses a lot of CLBs.
>
> The slower approach is to use a RAM, a single comparator, and have a
> counter cycle through the addresses.  But I think this is too slow for
> my current needs.
>
> But is there another approach that I've overlooked?
>
> Thanks!
> Eric Smith


Article: 32148
Subject: Virtex II multiplier question
From: awebb@clara.net (Alistair Webb)
Date: Fri, 15 Jun 2001 21:34:39 GMT
Links: << >>  << T >>  << A >>
I have a question regarding the Virtex II on-chip multipliers:

I would like to use the Virtex II to implement some long FIR filters
(512 or 1024 taps) with 24 bit coefficients stored in block RAM (it's
an audio application). 

From what I can gather in the data sheets, the dedicated mults can be
configured either as 18x18, or possibly as multiple 9x9's.

Cascading 4 mults gives me 36x36 which seems a bit wasteful of
resources considering data is only 24 bits wide.

How can I get the most efficienct use out of these embedded
multipliers?  Since the data rate is slow (96kHz) I'll be using just a
few multipliers per filter, with coeffs fetched sequentially from
block RAM.

I was thinking of perhaps sacrificing a bit of precision by limiting
filter coeffs to 18 bits, and then the FIR stage consists of a 24x18
multiplier / adder, which could then be realised by a 18x18 cascaded
with several 9x9's to get a 27x18 mult, and ignoring the last 3
bits...

All suggestions welcome!  I would really like to keep all filter
coeffs on-chip as this cuts down the overall cost and need for
external I/O.

Maybe someone at Xilinx would care to comment?

regards,

Alistair Webb.

p.s:  Will probably be targetting the XC2V250 device. What is the
target price for this device and when will it be available?  Until
then I may prototype on the smaller XC2V40 part and scale up later...

Article: 32149
Subject: Re: Xilinx Virtex 2: Configurations problems
From: "Eric Lewis" <eric_lewis@nmss.com>
Date: Fri, 15 Jun 2001 14:49:32 -0700
Links: << >>  << T >>  << A >>
Here are some new informations regarding our problem of configuration with the VirtexII in the midlle of the chain.

Unfortunately, the done pins for the 3 FPGAs are tied together and cannot be separated. The 3 FPGAs used are BGAs.

By looking at the sequence of the bits before the moment where init goes low when there are 3 bitstreams in the concatenated file stored in the serial eeprom, we have determined that the last thing written is the CRC itself. Init going low is probably a CRC error detected by the VirtexII.

We have tried to concatenate 2 bitstreams instead of 3 with the Prom File Formatter and the init signal don't go low at the end of the second bitstream.
By comparing the files that contain 2 concatenated bitstreams and 3 concatenated bitstreams, we have seen that only the length after the write to the LOUT register was different. All other data up to the end of the second bitstream
was the same (including the CRC).
Do you have any idea of what can make a difference if the concatenated files are almost the same?

We have verified the write to the COR register and the bit that define the CRC has not to be verified is 1. Why in this
case the CRC verified?

Thank.

Eric

eric_lewis@nmss.com



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