Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
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
Hi All, After how many cycles the virtex clkdll reaches the stability state :) Sent via Deja.com http://www.deja.com/ Before you buy.Article: 21951
<!doctype html public "-//w3c//dtd html 4.0 transitional//en"> <html> Dear recruiting team, <p>- you are looking for a highly-motivated digital <br> design/systems/CAD engineer? <br>- you need an engineer who is fascinated of computer <br> technology, especially parallel/vector architectures, <br> System-On-a-Chip and high-speed datacommunications? <br>- you ask for an employee who's motivation to study <br> electrical engineering was generated by Seymour Cray <br> and his machines? <br>- you are searching for a creative mind combined with <br> strong problem solving capabilities? <br>- you want a new member for your team, who is committed <br> to excellence and success? <br>- you desire an open-minded person who is programmed to <br> reach the goal, increasing your efficiency and <br> pushing the limits? <br>Try this..... <p>=> <A HREF="http://www.asic.cc">http://www.asic.cc</A> <p>Best regards, <p>Andreas Schmidt <p>P.S.: Everything is possible, if you really want it... <br> <br> </html>Article: 21952
-- http://top50keywords.tripod.com/Article: 21953
Hi, I have seen Altera Max7192, downloaded via JTAG by byteblaster cable. But Xilinx XCS10 have CCLK,MODE,Din,Dout other than TCK,TMS,TDI,TDO, can I download it via JTAG port, or it can only be configured via Din,etc. Thanks in advance. AdamsArticle: 21954
My point is, in UNIX, multipass place and route option shares each PAR process to a processor. If one workstation has only one processor then it means one PAR process is assigned to one workstation. Since the PAR time is the most critical during design, my question is if one PAR process can be shared among more than one processor, in the same workstation or in more then one workstation. I think this is impossible in current tools. If you have more than one processor in a workstation, then you have to define one user account or session for each processor on the workstation, which is the same as having the number of workstation in which one processor has been plugged. The number of the workstations here is the number of the processors in the same workstation. That means two workstations with one processor each is exactly the same as one workstation with two processors, but as Ray stated, one workstation with multiple processors will be slower because of peripherals (RAM, harddisk access etc). Utku -- I feel better than James Brown.Article: 21955
That said, I have successfully run two separate PAR sessions on my dual processor machine concurrently. Granted, that doesn't accelerate the PAR for one, but where you might want to run PAR with more than one seed or have more than one FPGA in a design it can be helpful. If you run two copies of PAR on one design, I think you'll need to copy that design to a second directory to keep them from interfering. > -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 21956
In article <8cffgn$bi6$1@sulawesi-fi.lerc.nasa.gov>, "Geoffrey A. Landis" <geoffrey.landis@sff.net> wrote: > In article <38E3CE53.371D@pointecom.net> Lynn Killingbeck, > killbeck@pointecom.net writes: > }... > }It isn't that diverse a system (reading a lot into you undefined term). True, the term "diversity" has not been defined in this thread. I originally introduced this term into a discussion thread in comp.arch.fpga, where the posted question was not related to the space shuttle. > }.... The > }backup _software_ uses the same hardware, same specifications, same I/O > }system, same displays and controls, same compiler, same language(s), and > }same just-about-everything. > > Really? One of the first jobs I was offered out of college was one to > work on the software for the backup computer. The team leader who > interviewed me explained that one of the ground rules was that it was > completely independently-written code, running on a different platform > built by a different manufacturer. Did this original plan get dropped? Maybe in some systems, like the space shuttle. But true diversity uses both different hardware and different software. For example, I know of one system requiring distributed control where both LonWorks and DeviceNet were used, in order to provide as much diversity as possible. > > (that was about five years before the actual first shuttle launch, and > everything about the computer system could very well have changed.) > > -- > Geoffrey A. Landis > http://www.sff.net/people/geoffrey.landis > -- Greg Neff VP Engineering *Microsym* Computers Inc. greg@guesswhichwordgoeshere.com Sent via Deja.com http://www.deja.com/ Before you buy.Article: 21957
I need to know the physical size of a CLB and IOB for the .22um Xilinx Virtex CMOS process. Thanks in advance for your help... -Kevin MortimerArticle: 21958
--------------C682648B0DE78FBE50859F2B Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Call for Exhibitors -- Call for Participants FPL 2000 The 10th International Conference on Field Programmable Logic and Applications 28 - 30 August 2000 Villach, Carinthia, Austria (in the cheart of "silicon alps") High Growth Rate Conference: 50 % more submissions (from 1999 to 2000) 44% more participants (from '98 to '99) Also featuring keynotes and survey papers. Advance Program ready by end of May The eldest international conference on reconfigurable logic and systems (2 years elder than FPGA and FCCM), founded by Oxford University (UK) we welcome the submission of proposals like: exhibit proposals product demonstrations industrial papers and tutorials book exhibits others (please, propose!) Exhibitor Registration Deadline: July 28, 2000 Download Registration Form: PDF: http://xputers.informatik.uni-kl.de/FPL/fpl2000/ Postscript: http://xputers.informatik.uni-kl.de/FPL/fpl2000/CfP_FPL2000.ps for details on FPL 2000 see: http://xputers.informatik.uni-kl.de/FPL/FPL2000/detailed_fpl.html Regards, Reiner W. Hartenstein Program Chair FPL 2000 http://xputers.informatik.uni-kl.de/FPL/fpl2000/ --------------C682648B0DE78FBE50859F2B Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit <!doctype html public "-//w3c//dtd html 4.0 transitional//en"> <html> <tt></tt> <tt></tt> <p><tt>Call for Exhibitors -- Call for Participants</tt><tt></tt> <p><tt> <b><font size=+4>FPL 2000</font></b></tt><tt></tt> <p><tt> The 10th International Conference on</tt> <br><tt> Field Programmable Logic and Applications</tt><tt></tt> <p><tt> 28 - 30 August 2000</tt> <br><tt> Villach, Carinthia, Austria</tt> <br><tt> (in the cheart of "silicon alps")</tt><tt></tt> <p><tt>High Growth Rate Conference:</tt> <br><tt> 50 % more submissions (from 1999 to 2000)</tt> <br><tt> 44% more participants (from '98 to '99)</tt> <br><tt> Also featuring <b>keynotes</b> and <b>survey papers.</b></tt> <br><tt><b> </b><i> Advance Program ready by end of May</i></tt><i><tt></tt></i> <p><i><tt> The eldest international conference</tt></i> <br><i><tt> on reconfigurable logic and systems</tt></i> <br><i><tt> (2 years elder than FPGA and FCCM),</tt></i> <br><i><tt> founded by Oxford University (UK)</tt></i><tt></tt> <p><tt>we welcome the submission of proposals like:</tt><tt></tt> <p><tt> exhibit proposals</tt> <br><tt> product demonstrations</tt> <br><tt> industrial papers and tutorials</tt> <br><tt> book exhibits</tt> <br><tt> others (please, propose!)</tt><tt></tt> <p><b><tt>Exhibitor Registration Deadline: July 28, 2000</tt></b><tt></tt> <p><tt>Download Registration Form:</tt> <br><tt>PDF: <A HREF="http://xputers.informatik.uni-kl.de/FPL/fpl2000/">http://xputers.informatik.uni-kl.de/FPL/fpl2000/</A></tt> <br><tt>Postscript:</tt> <br><tt><A HREF="http://xputers.informatik.uni-kl.de/FPL/fpl2000/CfP_FPL2000.ps">http://xputers.informatik.uni-kl.de/FPL/fpl2000/CfP_FPL2000.ps</A></tt> <br><tt></tt> <tt></tt> <p><tt>for details on FPL 2000 see: <A HREF="http://xputers.informatik.uni-kl.de/FPL/FPL2000/detailed_fpl.html">http://xputers.informatik.uni-kl.de/FPL/FPL2000/detailed_fpl.html</A></tt><tt></tt> <p><tt>Regards,</tt><tt></tt> <p><tt>Reiner W. Hartenstein</tt> <br><tt>Program Chair FPL 2000</tt> <br><tt><A HREF="http://xputers.informatik.uni-kl.de/FPL/fpl2000/">http://xputers.informatik.uni-kl.de/FPL/fpl2000/</A></tt></html> --------------C682648B0DE78FBE50859F2B--Article: 21959
Ray Andraka wrote in <38EE3920.E5EBBEAA@ids.net>... >If you go to a dos pox and run par without arguments you get the help screen. >From there I quote: >"-m = Multi task par run. File <node list file> ", > contains a list of node names on which to run the jobs. > (This option is not currently supported on WIN NT/WIN 95 > systems)." > >Also, I have not seen any evidence that Aldec 3.6c supports multiple processors >either (if you know different, then please educate me on how to set it up to do >so). I know that synplicity and modelsim also run just one processor. Note: If >you have a two processor system, the tasks will be spread out to the two >processors, but the combined processor load while running just synplicity or >xilinx (or substitute other EDA package) under NT only totals 50%, indicating >that teh programs are single threaded. The advantage two processors gets you is >that you can run two things at once even if they are both processor intensive >(where it slows down is on disk access). > For Aldec, there are 2 different versions : The base version and the FPGA Booster version. The second one increases the simulation speed for post-implemented netlists. Few months a go, i tried the 10 days evaluation version of the FPGA Booster on a design with 3 FPGA post-implemented netlists and i saw the CPU reaching 100%. The simulation was at least 10 times faster than with the base version. So, i suppose that in this case, the simulation is multi threaded, but i'm not sure of that, because as the Booster option is $10,000 more expensive than the base option and as i only make few post-implemented simulations, i didn't buy this option, and so, i can't retry. That's why i was asking if someone could confirm this point. Note that with the base version, the CPU never totals more than 50%. -- J-P GOGLIO GETRIS S.A. 13 Chemin des Prés 38240 Meylan Tel : (+33) 4 76 18 52 10 E-mail : goglio@getris.com Fax : (+33) 4 76 18 52 01Article: 21960
Bret Wade wrote: > > Hello Jonus, > > The "-m node_list" option of PAR is related to the Multi-Pass Place & Route > feature which farms out multiple PAR runs to the various machines in your > node list file, to be run concurrently. Each pass gets a different > placement seed. If your PAR command uses the -m switch but does not specify > multiple passes using the "-n #" switch, then you will see the message > "Multi-task PAR is not needed for this job", because you have only > specified a single pass. This has nothing to do with using multiple > processors on a single machine, and as others have said, it's only > supported on Solaris and HP-UX machines. > > Regards, > Bret Wade > Xilinx Product Applications OK, I have to try again. I did not set any value to the -n parameter. Have to look in the manuals for this flag. And, yes machine was a SUN running SunOS 5.5.1 Kind regards /JonasArticle: 21961
Hello, LavaLogic (http://www.lavalogic.com) is seeking experienced chip designers to download a FREE beta copy of our Java to Verilog compiler. The "Forge" allows chip designers to write high level algorithmic descriptions of their chip directly in pure Java, then translate that description to synthesizable Verilog. We are finding, through our early access program, that designers are able to capture the functionality of their chip more rapidly in Java than in Verilog, and the synthesis results of "Forge" produced Verilog are on-par with hand coded designs. Designs accomplished through use of the "Forge" have included complex algorithms such as a Floating Point Unit to more interface oriented designs like a DS3 serialization interface. The "Forge" is written in Java and will run on any platform running a Java2 virtual machine (available for free from Sun Microsystems). Automated install is supported for Solaris 2.6 and Linux (x86), however installation instructions are available for any platform. The current beta version of the "Forge" is available for FREE download right now. Simply let us know who you are at: http://www.lavalogic.com/join.html Then once you have had a chance to use the tool, let us know what you think (if you are so inclined!) Sincerely, Ian MillerArticle: 21962
I have the Aldec FPGA booster version. It accelerates simulation of the FPGA netlist, apparently by replacing the primitives with streamlined code. It is a very nice feature for doing back-annotated simulations for functional or timing simulation of the mapped or placed and routed FPGA design. It won't do anything for you before the FPGA is mapped. AFAIK, that does not take advantage of multiple processors. Jean-Paul GOGLIO wrote: > Ray Andraka wrote in <38EE3920.E5EBBEAA@ids.net>... > >If you go to a dos pox and run par without arguments you get the help > screen. > >From there I quote: > >"-m = Multi task par run. File <node list file> ", > > contains a list of node names on which to run the jobs. > > (This option is not currently supported on WIN NT/WIN 95 > > systems)." > > > >Also, I have not seen any evidence that Aldec 3.6c supports multiple > processors > >either (if you know different, then please educate me on how to set it up > to do > >so). I know that synplicity and modelsim also run just one processor. > Note: If > >you have a two processor system, the tasks will be spread out to the two > >processors, but the combined processor load while running just synplicity > or > >xilinx (or substitute other EDA package) under NT only totals 50%, > indicating > >that teh programs are single threaded. The advantage two processors gets > you is > >that you can run two things at once even if they are both processor > intensive > >(where it slows down is on disk access). > > > > For Aldec, there are 2 different versions : The base version and the FPGA > Booster version. The second one increases the simulation speed for > post-implemented netlists. > Few months a go, i tried the 10 days evaluation version of the FPGA Booster > on a design with 3 FPGA post-implemented netlists and i saw the CPU reaching > 100%. The simulation was at least 10 times faster than with the base > version. > So, i suppose that in this case, the simulation is multi threaded, but i'm > not sure of that, because as the Booster option is $10,000 more expensive > than the base option and as i only make few post-implemented simulations, i > didn't buy this option, and so, i can't retry. > > That's why i was asking if someone could confirm this point. > > Note that with the base version, the CPU never totals more than 50%. > > -- > J-P GOGLIO > GETRIS S.A. > 13 Chemin des Prés > 38240 Meylan > Tel : (+33) 4 76 18 52 10 > E-mail : goglio@getris.com > Fax : (+33) 4 76 18 52 01 -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 21963
Is there anyone know anything about distributed arithmetic and its uses in the DSP implementation on FPGAs? ThanksArticle: 21964
Ray Andraka <randraka@ids.net> wrote: > That said, I have successfully run two separate PAR sessions on my dual > processor machine concurrently. Granted, that doesn't accelerate the > PAR for one, but where you might want to run PAR with more than one seed > or have more than one FPGA in a design it can be helpful. If you run > two copies of PAR on one design, I think you'll need to copy that design > to a second directory to keep them from interfering. I do this with the Lucent (Neocad) software running 16 instances of PAR on a "farm" of 8 dual-cpu NT boxes. They can all run from the same directory as long as the command line specifies a different output directory. This is really useful on a tight design, where each of the 100 "cost tables" takes 3 hours to run. I can run all 100 cost tables over night, and try multiple hand-placements during the day. I see no degradation due to disk thrashing, since this is a CPU-bound task. It also helps that Lucent gives away their software for free (for all but the biggest chips.) -- Don Husby <husby@fnal.gov> http://www-ese.fnal.gov/people/husby Fermi National Accelerator Lab Phone: 630-840-3668 Batavia, IL 60510 Fax: 630-840-5406Article: 21965
Just use the search engine <google.com>, type in <distributed arithmetic>, and you get many hits. The best beginning may be Ray Andraka"s web page, followed by various Xilinx papers. Neat stuff. Have fun! Peter Alfke ==================== "Kevin T. Mortimer" wrote: > I need to know the physical size of a CLB and IOB for the .22um Xilinx > Virtex CMOS process. > > Thanks in advance for your help... > > -Kevin Mortimer > >Article: 21966
Just use the search engine <google.com>, type in <distributed arithmetic>, and you get many hits. The best beginning may be Ray Andraka"s web page, followed by various Xilinx papers. Neat stuff. Have fun! Peter Alfke ==================== Jamil Khatib wrote: > Is there anyone know anything about distributed arithmetic and its uses > in the DSP implementation on FPGAs? > > ThanksArticle: 21967
Sorry, that was an answer to a different posting. I had answered Kevin already privately. Peter ================== Peter Alfke wrote: > Just use the search engine <google.com>, type in <distributed > arithmetic>, and you get many hits. > The best beginning may be Ray Andraka"s web page, followed by various > Xilinx papers. > Neat stuff. Have fun! > > Peter Alfke > ==================== > "Kevin T. Mortimer" wrote: > > > I need to know the physical size of a CLB and IOB for the .22um Xilinx > > Virtex CMOS process. > > > > Thanks in advance for your help... > > > > -Kevin Mortimer > > > >Article: 21968
Hello, I have a little design (a Logiblox updown counter plus a vhdl based macro to give the updown signal to the counter). The functional simulation works fine but when I use the timing simulation I have a setup and hold time violation with a clock frequency of 30 MHz. When I use a clock of 20Mhz the design works fine! I know from the data book that the fpga I am targeting (XC4000E) can be clocked up to 80MHz. So what is the problem here and what can I do to resolve it. Thank you in advance for your answer. Red. Sent via Deja.com http://www.deja.com/ Before you buy.Article: 21969
This is a multi-part message in MIME format. --------------37A3AB7D837387977E36224F Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hello, When I try to implement a simple design (completely HDL), I get following error message: "logical blok xxxx is unexpanded". I already checked support.xilinx.com for an answer but no solution there. Any advice? Thanx!!! --------------37A3AB7D837387977E36224F Content-Type: text/x-vcard; charset=us-ascii; name="sanders.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Steven Sanders Content-Disposition: attachment; filename="sanders.vcf" begin:vcard n:Sanders;Steven x-mozilla-html:FALSE org:imec ;DTS adr:;;Flanders Language Valley 44;Ieper;;8900;Belgium version:2.1 email;internet:sanders@imec.be fn:Steven Sanders end:vcard --------------37A3AB7D837387977E36224F--Article: 21970
Sorry for the late reply, but here is something I had a while ago with Foundation 1.5, which was an actual SW-problem (shouldn't be in later versions AFAIK) ; maybe someone can get some help from it (part from a reply to Neil Jacobson at Xilinx who got me on the right track) >>> <snip> A little more info on the old question, your answer, and how it got me where I had to be <snip> The problem I had, was that I weren't able to program a Xilinx XC95216-CPLD with the parallel-JTAG-cable from Xilinx, and JTAG-programmer M.1.5.25, once the chain was bigger than 1 element, and I needed to program the CPLD as the 1st element in a chain of 8 elements (the chain consisted of the XC95216-CPLD, one 68360-Quicc processor from Motorola, 5 DSP-processors from TI, and a XC4028-FPGA). You gave me the hint that I should use the XC4003E-bsdl files for the other elements, but it didn't work for the complete chain. Anyway, it was already an improvement, as a 2 element-chain worked (CPLD+Quicc). SO I needed to know why the BSDL-file of the XC4003E had more than the one of the Quicc that made that chain work, so maybe it would bring me to the solution for the whole chain. After a 'game of cut&paste' I brought the original quicc-bsdl in a state that made the jtag-programmer happy, and the programming possible. The problem was that I had changed already that much that it wasn't clear what brought the break-through. In the second try I changed the thing I changed the last in the first try as first, but that wasn't enough, but I knew I was on the right track, and it didn't take me long to 'break the code'. It seemed I only needed to change 2 things: 1. Change the instance-name from m68360 to XC4003E_PC84 (although the last part of that name didn't matter) 2. Change the used standard from 1990 to 1994 (not really rewrite the file to the new standard, just tell that the 1994-standard is used - line 'use std_1049_1_1990' - are something close to that, I just haven't a bsdl file around) If a changed the standard bsdl-files from the FPGA and the DSP's with the 1st change (in both the 1994-standard was already in use) I got an other error than before, but it wasn't ok yet. Then I saw that the DSP's had instruction-lenghts of 4 instead of 3 (like the quicc, xc4028 and xc4003e). So I searched if there was a xilinx-bsdl-file around that had instruction-lenghts of 4 as well, but it seemed that only the CPLD's had more than 3, to be more specific 8. I was afraid it wouldn't work with that (double instruction lenght, and if would try with two bsdl-files for 4 elements, the problem that not enough bypass-registers were counted). So, remembering maths, I thought of trying with defining 4 dsp's as xc4003e, and one as xc9536, giving me an instruction-length of 20 as well, and 5 bypass-cells too. Mathematically OK, but not for the JTAG-programmer. However the error changed on the placement of the ''xc9536-dsp''. Last try : Define them all as xc9536. Finally I got 'Checking Integrity...OK', but now I had an other problem. The DSP-elements were questioned for their ID-code, which they happened to have, and the programmer saw they weren't xilinx at all. Leaving the ID-instruction out of the bsdl-file made it even worse, than the ID-code was still checked, but than it came to complete chaos, as he wanted to know the ID for one or another reason, but he didn't know how to ask it in a nice way. So I had the feeling I was close to victory - close, but no cigar - but I didn't have a clue how to go further, just until I wanted to shutdown the computer I thought of one more thing. As the name XC4003E could be followed by things other than '_PC84', e.g. '_PC85' it had to do something with first part of the name. So I searched for all files having the text 'XC4003E' in them, hoping it made me any wiser. One of the files brought me to the solution, a definition-file. It had all the supported names in them, and as the last line there stood 'IDCODE USERCODE' for the CPLD's and 'NO_IDCODE NO_USERCODE' for the FPGA's. So I changed that line to 'NO_IDCODE USERCODE' for the XC9536, and than, finally : I could program the CPLD... (After that I changed the instance-name for the DSP's to XC9572, and changed the definition file for that element, and restored the definition of the XC9536, as we use the CPLD as well, and there the ID-code checking is maybe better allowed) I don't know the complete name of the file anymore, but if you want to, I can look it up for you (it's in the data-directory, where also the bsdl-files are located, and the name starts with 'isp' and I guess the extention is '.def') <<< Hope this hlps someone, AlainArticle: 21971
See the March 2000 issue of Embedded Systems Programming for a good article. http://www.embedded.com/2000/0003/ Tom Jamil Khatib wrote: > > Is there anyone know anything about distributed arithmetic and its uses > in the DSP implementation on FPGAs? > > ThanksArticle: 21972
Steven Sanders <sanders@imec.be> writes: > > When I try to implement a simple design (completely HDL), I get > following error message: > "logical blok xxxx is unexpanded". I already checked support.xilinx.com > for an answer but no solution there. Any advice? > We might be able to help you if you would post your code and the full error message. That "... is unexpanded" messages is usually due to library problems. chm. -- cmautner@ - Christian Mautner utanet.at - Vienna/Austria/EuropeArticle: 21973
redbens@my-deja.com writes: > I have a little design (a Logiblox updown counter plus a vhdl based > macro to give the updown signal to the counter). > The functional simulation works fine but when I use the timing > simulation I have a setup and hold time violation with a clock > frequency of 30 MHz. When I use a clock of 20Mhz the design works fine! > > I know from the data book that the fpga I am targeting (XC4000E) can be > clocked up to 80MHz. So what is the problem here and what can I do to > resolve it. You should not depend on simulation only. What does the static timing analysis say? What timing constraints did you specify? chm. -- cmautner@ - Christian Mautner utanet.at - Vienna/Austria/EuropeArticle: 21974
For starters, try my web page. You'll find distributed arithmetic under the DSP page. Basically, it is a rearrangement of the bitwise multiplications and adds to take advantage of the small tables in the FPGA. Jamil Khatib wrote: > Is there anyone know anything about distributed arithmetic and its uses > in the DSP implementation on FPGAs? > > Thanks -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randraka
Site Home Archive Home FAQ Home How to search the Archive How to Navigate the Archive
Compare FPGA features and resources
Threads starting:
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