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
Kevin Brace <ihatespamkevinbraceusenet@ihatespamhotmail.com> writes: > Since there are almost no good books available on the implementation > side of a PCI IP core, Is that "almost no", or just "no"? If you know of one, by all means please let us know. :-)Article: 39851
Thanks for your efforts. I need to include the VHDL netlist (which defines only the I/Os of the core) to be able to connect my design to the core. This is according to the FPSC design flow as shown in the documentation. Next step is to export an EDIF netlist from Viewdraw. This EDIF netlist and the VHDL netlist of the core are the source files in FPGA Express that I use to synthesize the design. Following step is to export an EDIF netlist from FPGA Express and im port it into the Orca Foundry. rickman <spamgoeshere4@yahoo.com> wrote in news:3C72858F.520A6C49@yahoo.com: > I am not clear about what you are trying to do. If you want to create a > bit stream, then why do you need to include VHDL for the embedded > PCI-Core? This part is already in the silicon and should not be > represented in the bit stream. > > If you are trying to perform simulation, I expect you should do this > before you synthesize. You will need some form of model for the > PCi-Core. If you translate the schematic to VHDL, then your VHDL for > the PCI-Core should be a module in the VHDL that is produced. > > But the fact that you are trying to produce an EDIF netlist says to me > that you are doing the former rather than the later. So why are you > trying to include the PCI-Core in the bit stream?Article: 39852
Ray Andraka wrote: <snip> > load up the device programmer (in those days hardware debugger) without > <end snip> It appears that the hardware debugger has vanished from 4.1i. Have Xilinx abandoned it in favour of JBits and/or ChipScope ?Article: 39853
Kevin Brace wrote: > > Virtex-II/Virtex-E/Spartan-IIE don't support 5V PCI (3.3V PCI is > supported though, but almost no PC motherboards support it.) without an > external voltage level conversion chip. I shouldn't worry too much about the 3v3 vs. 5v PCI problem. Our Virtex-E board works fine with 5V PCI cards, all we added is a bunch of QuickSwitch type parts to protect the non-5V tolerant Viretx-E IOs [Xilinix recommends 100Rs for this but QS parts are less interfering]. I checked out the PCI signals coming from the FPGA looked like at various points along the PCI bus using a DSO and they easily made the 5V transition point within the 33MHz PCI timing spec. In fact there was margin to spare since IIRC the clock-to-final-crossing was about 13.5nsec. In fact there was no reflection step that I could see even though the PCI is an unterminated, first reflection, bus.Article: 39854
Tim wrote: > > Starting from a 10MHz clock, you can use Peter's circuit to double > to 20MHz, then run the faster clock through the DLL. Still outside > limits, but will almost certainly work on the bench - stick to > 25MHz in shipping product. It turns out that the ES parts will lock on less than 20 MHz, with a special "tweek" from Xilinx. Supposedly there will be a tactical patch for this. Also, look at the erratta sheets! Some of the ES parts have problems locking on the FX output of a DCM. -- NAME: David W. Bishop INTERNET: dbishop@vhdl.org ( \ ) US MAIL: Hilton NY A Long time ago, \__\/ PHYSICAL: 43:17:17N 77:47:37W 281' In a Galaxy far, far away... | | For Supernova info: http://www.RochesterAstronomy.org/snimages/ | | For VHDL/Synthesis info: http://www.vhdl.org/siwg _/___\_ All standard disclaimers apply. [_______]Article: 39855
FPGA 2002 Final Program and Information ACM/SIGDA Tenth International Symposium on Field Programmable Gate Arrays Sponsored by ACM/SIGDA with support from Actel, Cypress Semiconductor, Altera and Xilinx Monterey Beach Hotel, Monterey, California February 24-26, 2002 web site: http://www.ecs.umass.edu/ece/fpga2002/ Please join us THIS COMING SUNDAY for the annual FPGA symposium. Highlights include: - 26 top technical papers covering cutting-edge research - numerous poster sessions for discussion and meeting with colleagues - new, scenic symposium location at the Monterey Beach Hotel - Monday night banquet at the Monterey Aquarium Hope to see you there! ---------------------------------------------------------------- PROGRAM -------------------- SUNDAY, February 24, 2002 6:00PM Registration 7:00PM Welcoming Reception MONDAY, February 25, 2002 7:30AM Continental Breakfast and Registration 8:20AM Opening Remarks, Martine Schlag, Steve Trimberger Session 1. Interconnect Architecture Chair: Steve Wilton, University of British Columbia 8:30AM Interconnect Enhancements for a High-Speed PLD Architecture, Michael Hutton, Vinson Chan, Peter Kazarian, Victor Maruri, Tony Ngai, Jim Park, Rakesh Patel, Bruce Pedersen, Jay Schleicher, Sergey Shumarayev, Altera 8:50AM FPGA Switch Block Layout and Evaluation, Herman Schmit, Vikas Chandra, Carnegie Mellon University 9:10AM Circuit Design of FPGA Routing Switches, Guy G. Lemieux, David M. Lewis, University of Toronto 9:30AM Coffee Break and Poster Presentations Session 2. Arithmetic Chair: Tom Kean, Algotronix 10:30AM A Faster Distributed Arithmetic Architecture for FPGAs, Radhika S. Grover, Weijia Shang, Qiang Li, Santa Clara University 10:50AM Efficient Architectures for Implementing Montgomery Modular Multiplication and RSA Modular Exponentiation on Reconfigurable Logic, Alan Daly, Liam Marnane, University College, Cork 11:10AM A Flexible Floating-Point Format for Optimizing Data-Paths and Operators in FPGA Based DSPs, J. Dido, N. Geraudie, L. Loiseau, O. Payeur, Y. Savaria, École Polytechnique de Montréal; D. Poirier, Miranda Technologies, Inc. 11:30 AM Poster Presentations 12:00 N Lunch Session 3. Physical Design Chair: Russ Tessier, University of Massachusetts - Amherst 1:30PM Efficient Circuit Clustering for Area and Power Reduction in FPGAs, Amit Singh, Malgorzata Marek-Sadowska, University of California, Santa Barbara 1:50PM Integrated Retiming and Placement for Field Programmable Gate Arrays, Deshanand P. Singh, Stephen D. Brown, University of Toronto 2:10PM SPFD-Based Global Rewiring, Jason Cong, Yizhou Lin, Wangning Long, University of California, Los Angeles 2:30PM Eve: A CAD Tool for Manual Placement and Pipelining Assistance of FPGA Circuits, William Chow, Jonathan Rose, University of Toronto 2:50PM Coffee Break and Poster Presentations Session 4. Cellular and Cryptographic Applications Chair: Scott Hauck, University of Washington 3:50PM Application of FPGA Technology to Accelerate the Finite-Difference Time-Domain (FDTD) Method, Ryan N. Schneider, Laurence E. Turner, Michal M. Okoniewski, University of Calgary 4:10PM FPGA Implementation of Neighborhood-of-Four Cellular Automata Random Number Generators, Barry Shackleford, Motoo Tanaka, Richard J. Carter, Greg Snider, Hewlett-Packard 4:30PM Cryptographic Rights Management of FPGA Intellectual Property Cores, Tom Kean, Algotronix 4:50PM Poster Presentations 6:30PM Busses Depart for Monterey Bay Aquarium 7:00PM-11:00PM Dinner at the Monterey Bay Aquarium TUESDAY, February 26, 2002 7:30 AM Breakfast Session 5. Synthesis, Verification and Test Chair: Jason Cong, UCLA 8:30AM Constrained Clock Shifting for Field Programmable Gate Arrays, Deshanand P. Singh, Stephen D. Brown, University of Toronto 8:50AM Timing Verification of Dynamically Reconfigurable Logic for the Xilinx Virtex FPGA Series, Ian Robertson, University of Strathclyde; David Robinson, The Alba Centre; James Irvine, University of Strathclyde 9:10AM FPGA Test Time Reduction Through a Novel Interconnect Testing Scheme, Stuart McCracken, Zeljko Zelic, McGill University 9:30AM Coffee Break and Poster Presentations Session 6. Architecture Analysis and Automation Chair: Vaughn Betz, Altera 10:30AM On the Sensitivity of FPGA Architectural Conclusions to Experimental Assumptions, Tools and Techniques, Andy Yan, Rebecca Cheng, Steven Wilton, University of British Columbia 10:50AM Dynamic Power Consumption in Virtex-II FPGA Family, Li Shang, Princeton University; Alireza S. Kaviani, Kusuma Bathala, Xilinx, Inc. 11:10AM Automatic Layout of Domain-Specific Reconfigurable Subsystems for System-on-a-Chip, Shawn Phillips, Scott Hauck, University of Washington 11:30AM Poster Presentations 12:00 N Lunch Session 7. Software for Reconfigurable Systems Chair: Miriam Leeser, Northeastern University 1:30PM Performance-Constrained Pipelining of Software Loops onto Reconfigurable Hardware, Greg Snider, Hewlett-Packard 1:50PM Configuration Prefetching Techniques for Partial Reconfigurable Coprocessor with Relocation and Defragmentation, Zhiyuan Li, Northwestern University; Scott Hauck, University of Washington 2:10PM Analysis of Quasi-Static Scheduling Techniques in a Virtualized Reconfigurable Machine, Yury Markovskiy, Eylon Caspi, Randy Huang, Joseph Yeh, Michael Chu, University of California Berkeley; André DeHon, Caltech; John Wawrzynek, University of California, Berkeley 2:30PM Incremental Reconfiguration of Multi-FPGA Systems, K.K. Lee, Synopsys; D.F. Wong, University of Texas at Austin 2:50PM Coffee Break and Poster Presentations Session 8. Innovative Applications Chair: Ray Andraka, Andraka Consulting Group 3:50PM Parallel-Beam Backprojection: An FPGA Implementation Optimized for Medical Imaging, Srdjan Coric, Miriam Leeser, Eric Miller, Northeastern University; Marc Trepanier, Mercury Computer Systems 4:10PM A Dynamically Reconfigurable Adaptive Viterbi Decoder, Sriram Swaminathan, Russell Tessier, Dennis Goeckel, Wayne Burleson, University of Massachusetts, Amherst 4:30PM Data Reorganization Engines for the Next Generation of System-On-a-Chip FPGAs, Pedro Diniz, Joonseok Park, University of Southern California/ISI 4:50PM Closing Remarks, Martine Schlag, Steve TrimbergerArticle: 39856
They replaced it with something called "iMPACT". I haven't used it yet, so I can't tell you how good/bad it is. It does still require you to do a full xilinx install to be able to download devices :-( Rick Filipkiewicz wrote: > Ray Andraka wrote: > > <snip> > > > load up the device programmer (in those days hardware debugger) without > > > > <end snip> > > It appears that the hardware debugger has vanished from 4.1i. Have Xilinx > abandoned it in favour of JBits and/or ChipScope ? -- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759Article: 39857
Jay wrote: > I'm not sure if I understand what you are trying to accomplish but > I'll take a SWAG at answering your question: > > Step 1) Use the BRAM in synchronous mode You have no choice in this matter. BRAM is synchronous regardless of what you do to it. The input setup times are relative to the BRAM clock. The BRAM clock should generally be the same free-running clock that is used in the logic associated with the BRAM port (eg, the address counters, enable logic etc). If you do that, even the relatively slow ena and we setup times are easily met at a 77 MHz system clock speed. > > Step 2) Drive the inputs to the BRAM with F/F's clocked with the same > clock as the BRAM > Step 3) Let the P&R tool worry about metting set/hold issues after > you've selected a topology that is easily handled by the tool set. > > Regards! > > "H.L" <alphaboran@yahoo.com> wrote in message news:<a4t9p2$hog$1@ulysses.noc.ntua.gr>... > > Hello all, > > I know that this issue is well discussed in this newsgroup but still I have > > some questions. > > In a Virtex-E FPGA the setup time for the EN,WE signals are 2.5 ns and 2.2ns > > (plz confirm) but the setup time for the address,data is about 1.1ns (plz > > confirm). I want to use 3 BRAMs 128x32 in a Virtex-E 600 (speed grade 6) @ > > 77 MHz (or clock period=12ns), do I have to keep the "input" signals stable > > for 2 clock periods to succeed a correct timing in the read/write operations > > or is it totally wrong (except for too slow)? > > > > Best Regards, > > Harris -- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759Article: 39858
You need to be careful attempting this. Peter's circuit can easily have jitter which is well outside of the acceptable jitter spec for the DLL. In order to have even a small chance of success, the 10 MHz clock has to be exactly 50% duty cycle, which is very hard to achieve with a single ended clock input. David Bishop wrote: > Tim wrote: > > > > Starting from a 10MHz clock, you can use Peter's circuit to double > > to 20MHz, then run the faster clock through the DLL. Still outside > > limits, but will almost certainly work on the bench - stick to > > 25MHz in shipping product. > > It turns out that the ES parts will lock on less than 20 MHz, with > a special "tweek" from Xilinx. Supposedly there will be a tactical > patch for this. Also, look at the erratta sheets! Some of the ES > parts have problems locking on the FX output of a DCM. > > -- > NAME: David W. Bishop INTERNET: dbishop@vhdl.org ( \ ) > US MAIL: Hilton NY A Long time ago, \__\/ > PHYSICAL: 43:17:17N 77:47:37W 281' In a Galaxy far, far away... | | > For Supernova info: http://www.RochesterAstronomy.org/snimages/ | | > For VHDL/Synthesis info: http://www.vhdl.org/siwg _/___\_ > All standard disclaimers apply. [_______] -- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759Article: 39859
You need to initialize these twice. Once with a generic for the simulation, and once with an INIT= attribute for the back end tools. The generic has to be ignored during synthesis, which means it must be surrounded by translate_off/on pragmas. Note that the format for the generic is different than that for the attribute. In the example below, itoh is an integer to hex string conversion, and int2bit_vec is an integer to bit_vector conversion. These are used to get the correct formats for the two init strings. attribute INIT of U1 : label is itoh(lut_init);--INIT= attribute to pass to PAR through synplicity begin U1: SRL16E --synthesis translate_off generic map ( -- init generic is for simulation model, not seen by Synplicity or PAR INIT => int2bit_vec(lut_init,16)) --synthesis translate_on port map ( Q => y, A0 => tap(0), A1 => tap(1), A2 => tap(2), A3 => tap(3), D => ldi(i), CLK => clk, CE => we ); AT wrote: > Hallo, > > I would like to make divider by 32 from SRL16E of SpartanII, but > translate occurs this error (WebPack 4.1): > "logical block 'divog' with type 'srl16e0' is unexpanded" > > divG: SRL16E > generic map(INIT => X"8000") > port map ( D => inter(1), CE => Enable, CLK => Clock, > A0 => ADDR(0), A1 => ADDR(1), A2 => ADDR(2), A3 => ADDR(3), > Q => inter(0)); > > divoG: SRL16E > generic map(INIT => X"0000") > port map ( D => inter(0), CE => Enable, CLK => Clock, > A0 => ADDR(0), A1 => ADDR(1), A2 => ADDR(2), A3 => ADDR(3), > Q => inter(1)); > > Is it impossible initialize SRL16Es by different values? > Thanks. > AT -- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759Article: 39860
Since my name was used, I must state that I strongly agree with Ray: "My" frequency doubler differentiates the two clock edges, and will thus be the victim of any non-50% duty cycle variation on the input. And the DLL or DCM does not like that... Peter Alfke, Xilinx Applications ================================= Ray Andraka wrote: > You need to be careful attempting this. Peter's circuit can easily have > jitter which is well outside of the acceptable jitter spec for the DLL. In > order to have even a small chance of success, the 10 MHz clock has to be > exactly 50% duty cycle, which is very hard to achieve with a single ended > clock input. > > David Bishop wrote: > > > Tim wrote: > > > > > > Starting from a 10MHz clock, you can use Peter's circuit to double > > > to 20MHz, then run the faster clock through the DLL. Still outside > > > limits, but will almost certainly work on the bench - stick to > > > 25MHz in shipping product. > > > > It turns out that the ES parts will lock on less than 20 MHz, with > > a special "tweek" from Xilinx. Supposedly there will be a tactical > > patch for this. Also, look at the erratta sheets! Some of the ES > > parts have problems locking on the FX output of a DCM. > > > > -- > > NAME: David W. Bishop INTERNET: dbishop@vhdl.org ( \ ) > > US MAIL: Hilton NY A Long time ago, \__\/ > > PHYSICAL: 43:17:17N 77:47:37W 281' In a Galaxy far, far away... | | > > For Supernova info: http://www.RochesterAstronomy.org/snimages/ | | > > For VHDL/Synthesis info: http://www.vhdl.org/siwg _/___\_ > > All standard disclaimers apply. [_______] > > -- > --Ray Andraka, P.E. > President, the Andraka Consulting Group, Inc. > 401/884-7930 Fax 401/884-7950 > email ray@andraka.com > http://www.andraka.com > > "They that give up essential liberty to obtain a little > temporary safety deserve neither liberty nor safety." > -Benjamin Franklin, 1759Article: 39861
Hi all, here is an argument with one of my friend who is a mechanical engineer and writing S/W. He argues that if the input and output waveforms are given he can write a tool which generates a digital circuit for those inputs and outputs. He knows about truth tables and gates. He is not convinced when I told all digital circuits are not just combinational. Then I asked him whether he can generate a C program if the inputs and outputs are given. But he says that it is obsurd that I am comparing a language with digital design. Still he believes in what he says. So friends, please do respond. Thank you Madhu -- Posted via Mailgate.ORG Server - http://www.Mailgate.ORGArticle: 39862
Yes. But someone posted that he tried it and it does work well enough for benchtop debugging. As I recall, he used this arrangement in a production design ;-) For a more reliable patch, find a space on the PCB to epoxy a PLL chip from ICS (www.icst.com). The ICS502 will multiply by 2, 2.5, 3, 3.5, 4, 5. Peter Alfke wrote > Since my name was used, I must state that I strongly agree with Ray: "My" > frequency doubler differentiates the two clock edges, and will thus be the > victim of any non-50% duty cycle variation on the input. And the DLL or DCM > does not like that... > > Peter Alfke, Xilinx Applications > ================================= > Ray Andraka wrote: > > > You need to be careful attempting this. Peter's circuit can easily have > > jitter which is well outside of the acceptable jitter spec for the DLL. In > > order to have even a small chance of success, the 10 MHz clock has to be > > exactly 50% duty cycle, which is very hard to achieve with a single ended > > clock input. > > > > David Bishop wrote: > > > > > Tim wrote: > > > > > > > > Starting from a 10MHz clock, you can use Peter's circuit to double > > > > to 20MHz, then run the faster clock through the DLL. Still outside > > > > limits, but will almost certainly work on the bench - stick to > > > > 25MHz in shipping product. > > > > > > It turns out that the ES parts will lock on less than 20 MHz, with > > > a special "tweek" from Xilinx. Supposedly there will be a tactical > > > patch for this. Also, look at the erratta sheets! Some of the ES > > > parts have problems locking on the FX output of a DCM. > > >Article: 39863
Madhu, test his skills with the following: Input = clock outputs = several outputs from a binary counter driven by that clock. This is admittedly trivial, but it gets him to think in terms of state machines. BTW: Given a finite sequence of inputs and outputs, one can obviously generate the combinatorial and/or sequential logic that generates these outputs. The brute-force solution may not be the most elegant one, but there always is a solution. Peter Alfke ===================== Madhu wrote: > Hi all, > here is an argument with one of my friend who is a mechanical engineer > and writing S/W. > He argues that if the input and output waveforms are given he can > write a tool which generates a digital circuit for those inputs and > outputs. He knows about truth tables and gates. > He is not convinced when I told all digital circuits are not just > combinational. > Then I asked him whether he can generate a C program if the inputs and > outputs are given. But he says that it is obsurd that I am comparing a > language with digital design. > Still he believes in what he says. > So friends, please do respond. > Thank you > Madhu > > -- > Posted via Mailgate.ORG Server - http://www.Mailgate.ORGArticle: 39864
Madhu, Ask your friend just what he thinks runs that C program. It is a digital circuit with known inputs and outputs. (for example, either a series of numbers, or a keyboard input, etc.) All of these are ultimately digital circuits. There are no analog computers in use anymore (to the best of my knowledge.) Theron "Madhu" <pp_madhavi@yahoo.com> wrote in message news:c074a7b1c70d79729ac318bbc52aa174.67011@mygate.mailgate.org... > Hi all, > here is an argument with one of my friend who is a mechanical engineer > and writing S/W. > He argues that if the input and output waveforms are given he can > write a tool which generates a digital circuit for those inputs and > outputs. He knows about truth tables and gates. > He is not convinced when I told all digital circuits are not just > combinational. > Then I asked him whether he can generate a C program if the inputs and > outputs are given. But he says that it is obsurd that I am comparing a > language with digital design. > Still he believes in what he says. > So friends, please do respond. > Thank you > Madhu > > > -- > Posted via Mailgate.ORG Server - http://www.Mailgate.ORGArticle: 39865
The problem I initially posted about was the fact that the FIFO from CoreGen has a major bug in it. I have since found a work-around using a dual port ram. However, in general, I am greatly displeased with ISE4.1. The only good thing I can say is that it does handle CoreGen instantiations seamlessly. The PAR algorithm does a really poor job of placing parts. The manual placer has a major bug. In short, it appears that Xilinx is using the customer for beta testing and then making us pay for the software. I hope that the next service pack will solve most of these problems. What about ISE4.2? Is it any better? I am using the Spartan2 series parts. Perhaps if enough of us scream _really_ loud, Xilinx will get their act together. Theron Hicks "Jonas Weiss" <jweiss@kontronmedical.ch> wrote in message news:3C6CBFDF.EC64A773@kontronmedical.ch... > Hello, > I was just wondering if you got your asynchronous FIFO to work by now. For some > days I'm also trying to get my design using one of these to work, without > success yet (ISE 4.1 & Model Sim XE Starter). I am using continuous read and > write clocks with synchronous enables. > Behavioral simulation works fine, but as soon as I proceed to a post translate > simulation I get 'open' select signals of internal fifo-muxes in the translate > file. > P&R works, just the post P&R simulation is not exactly what I expected it to be, > as if the behavioral model and the coregen netlist were not consistent. > Any new ideas yet? > > Thanks > > Jonas > >Article: 39866
I have a project written in VHDL that I'm trying to synthesize using XST and ISE 4. Part of the project uses a dual port ram block that was created using CoreGen. When I try and place and route the design, I get errors that say the dual port ram block is unexpanded. I have included the .xco file for the ram block inb the project files, what more do I need to do to get place and route to work? The component that uses the RAM has the correct component declartions from the CoreGen generated template file. Any help would be great. Thanks, DaveArticle: 39867
Theron Hicks (hicksthe@egr.msu.edu) wrote: : Madhu, : Ask your friend just what he thinks runs that C program. It is a : digital circuit with known inputs and outputs. (for example, either a : series of numbers, or a keyboard input, etc.) All of these are ultimately : digital circuits. There are no analog computers in use anymore (to the best : of my knowledge.) I read this and it set me thinking: I am sure I have come across references to a NASA Analogue computer that calculates balistics etc. for space missions, as it is a somewhat more cost effective solution than solving analytically all the equations. I tried to find a link to this and failed, but I found another one instead... http://www.simlabs.arc.nasa.gov/vms/controls.html I guess the undergrad lab session I had on analogue computers wasn't wasted after all... Chris SaunterArticle: 39868
"Ray Andraka" <ray@andraka.com> schrieb im Newsbeitrag news:3C750BE8.DE42CE4C@andraka.com... > They replaced it with something called "iMPACT". I haven't used it yet, so I > can't tell you how good/bad it is. It does still require you to do a full > xilinx install to be able to download devices :-( If you want to use only the commercial version, then YES. But there is also the WEBPACK version of IMPACT, which is AFAIK identical to the commercial version. So you jst need to install IMPACT (WEBPACK) -- MfG FalkArticle: 39869
"X. Q." <qijun@okigrp.com.sg> schrieb im Newsbeitrag news:3c746c0c@news.starhub.net.sg... > I built my design with Spartan-II. Now the research centre in another > country > needs my design. Let's pressume they don't have any expertise in FPGA and So why dont you send a) the HDL code b) the netlist ?? So they can use it. > such, while they want my design. How can I transport my design to them? My > design is 10K gates in ASIC. Can I get any CPLD which can hold this design? AFAIK there is no such big CPLD around. 1024 macrocells is the maximum. > I understand FPGA dies after I unplug the power. They dont ;-), They just lose their configuration, because it is stored in SRAM. But ACTEL sells FLASH and Antifuse based devices, they keep their programming after power OFF. Anyway, I dont quite see your problem. IF the "other" gus have a board with a FPGA, (maybe same familie as you have) then they can easyly recompile the code and are done. -- MfG FalkArticle: 39870
"Kumar" <yatiks@yahoo.com> schrieb im Newsbeitrag news:b479edf8.0202202254.58dbda7e@posting.google.com... > Then I connected the same Board to other PC with less speed( config > --366 Mhz Celeron ) I have not faced any problem in programming the > FPGA ( XC2s200-5PQ208 ) through JTAG with same cable (length = appr 5 > feet). > > Can anybody tell what is causing the problem (is the speed of the PC > or something else) and also how can I solve this problem ???????? Looks like a driver problem. -- MfG FalkArticle: 39871
Dave, Make sure the netlist for the core (.edn) is in the project directory or macro search path, and that the core is named correctly. Regards, Kamal Dave Brown wrote: > I have a project written in VHDL that I'm trying to synthesize using XST and > ISE 4. Part of the project uses a dual port ram block that was created using > CoreGen. When I try and place and route the design, I get errors that say > the dual port ram block is unexpanded. I have included the .xco file for the > ram block inb the project files, what more do I need to do to get place and > route to work? The component that uses the RAM has the correct component > declartions from the CoreGen generated template file. Any help would be > great. > Thanks, > DaveArticle: 39872
"Christopher Saunter" <Christopher.Saunter@durham.ac.uk> wrote in message news:a539vr$32i$1@sirius.dur.ac.uk... > Theron Hicks (hicksthe@egr.msu.edu) wrote: > : Madhu, > : Ask your friend just what he thinks runs that C program. It is a > : digital circuit with known inputs and outputs. (for example, either a > : series of numbers, or a keyboard input, etc.) All of these are ultimately > : digital circuits. There are no analog computers in use anymore (to the best > : of my knowledge.) > > I read this and it set me thinking: > I am sure I have come across references to a NASA Analogue computer that > calculates balistics etc. for space missions, as it is a somewhat more > cost effective solution than solving analytically all the equations. I > tried to find a link to this and failed, but I found another one > instead... > > http://www.simlabs.arc.nasa.gov/vms/controls.html > > I guess the undergrad lab session I had on analogue computers wasn't > wasted after all... > > Chris Saunter Chris, I think I had the same lab back in the 70's. I guess we are both showing our ages. TheronArticle: 39873
I am trying to use the KCPSM microcontroller in a VHDL design in ISE 4.1. 4.1 does not appear to recognize the EDIF file in a VHDL project. (ISE3.3 does seem to work OK.) I get the following messages... EDIF source are not supported for the selected device and flow. Import file kcpsm.EDN? This design contains sources that are not supported by the current device and flow. You may view these sources, but not process them. These are : kcpsm.EDN Am I missing something obvious?Article: 39874
"Dave Brown" <dbrown12@shaw.ca> wrote in message news:a539he$9b4$1@pallas.novatel.ca... > I have a project written in VHDL that I'm trying to synthesize using XST and > ISE 4. Part of the project uses a dual port ram block that was created using > CoreGen. When I try and place and route the design, I get errors that say > the dual port ram block is unexpanded. I have included the .xco file for the > ram block inb the project files, what more do I need to do to get place and > route to work? The component that uses the RAM has the correct component > declartions from the CoreGen generated template file. Any help would be > great. > Thanks, > Dave > > > Dave, Did you include the library? -- synopsys translate_off library UNISIM; use UNISIM.VCOMPONENTS.ALL; -- synopsys translate_on Theron
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