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
I'm thinking about using the NIOS processor from Altera in a new design and I am curious if anyone has had any good or bad experiences with it? Sent via Deja.com http://www.deja.com/Article: 28251
On 03 Jan 2001 16:17:19 +0000, Jon Schneider <jms@geriatrix.circlesquared.cXoXm> wrote: >OSFT Inc. <info@osft.com> wrote: >> Skematic Capture > >I did my first (1.2u) ASIC years ago with schematic capture when that >was the way to do it. Is this similar ? > >> Birth Trouble shooting > >Now I _know_ I'm completely out of touch with semiconductor technology. > > Jon I believe the current (or former) term is HW/SW integration. M. Simon Space-Time Productions http://www.spacetimepro.com Free CNC Machine Control Software Free Source Code Control the World From a Parallel PortArticle: 28252
You can assign your pins before the design is done, but you should apply at least a little thought to make success more likely. Some points to ponder: Try to register all the I/O in the design at the IOB. That isolates the internal timing from the external timing, and makes it more likely to get a design to work. The routing in the the Spartan II is more plentiful than the older 4K/spartan family, so a poor pin placement is not nearly as likely to give you a no-route than it had been previously. Some simple things can reduce the routing congestion greatly and accelerate the place and route times. FIrst, if it is a data flow design that will be using the carry chains, run your data across the chip (as viewed in the floorplanner) with the LSBs at the bottom to line up with the carry chain direction. If you have signals that go in and then back out iwth no or only one registers, then those should be placed nearby to keep the routing delays minimum. Mikhail Matusov wrote: > > Hi all, > > Egg-chicken kind of problem. I have to give my board design out to a layout > person but I haven't yet had chance to start my FPGA work. Usually I do some > draft FPGA design and run tools at least once to fix the pins before giving > it out to do a layout but this time the schedule is really tight and if I go > this route it will be too late. This is not a very demanding design neither > in terms of complexity nor in terms of speed and I am using Spartan II > family device. Scary part is that the pins utilization is almost 100%. > > Nonetheless, do you guys think that I can get away with pins fixed > beforehand without too much thought (I put my clocks on global clock lines)? > > Thanks in advance, > > -- > ============================ > Mikhail Matusov > Hardware Design Engineer > Square Peg Communications > Tel.: 1 (613) 271-0044 ext.231 > Fax: 1 (613) 271-3007 > http://www.squarepeg.ca -- -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 or http://www.fpga-guru.comArticle: 28253
Hi, I need to send and receive my data through serial interface for my FPGA project. I plan to use a NS16450 (UART) to do so. However I am not sure how to control and utilize it. Any reference / example / or suggestion? regards, chewArticle: 28254
On Wed, 03 Jan 2001 20:17:53 GMT, "Mikhail Matusov" <matusov@ANNTIsquarepegSPPAMM.ca> wrote: >Hi all, > >Egg-chicken kind of problem. I have to give my board design out to a layout >person but I haven't yet had chance to start my FPGA work. Usually I do some >draft FPGA design and run tools at least once to fix the pins before giving >it out to do a layout but this time the schedule is really tight and if I go >this route it will be too late. This is not a very demanding design neither >in terms of complexity nor in terms of speed and I am using Spartan II >family device. Scary part is that the pins utilization is almost 100%. > >Nonetheless, do you guys think that I can get away with pins fixed >beforehand without too much thought (I put my clocks on global clock lines)? > > >Thanks in advance, Mikhail, we always preassign pins so that we can do the PCB layout and the FPGA design concurrently. We try to pick a signal flow pattern that makes the PCB layout easy and seems to make FPGA routing happy, or at least make a rough guess at it. Seems to work. JohnArticle: 28255
Hi, I want to start with FPGA design. I will appreciate any suggestions for good starter kit. ValArticle: 28256
Val, check out the DIGILAB picoMAX, 10K10 and 1Kx208 from El Camino (http://www.elca.de/prode.html ). All you need, to get started with one of these kits, is a PC. The included tools support Altera Hardware Description Language (AHDL) synthesis, schematic design entry, full featured timing analysis and simulation, floorplanning and more. You can also download free LeonardoSpectrum and Synopsys FPGA Express versions from the Altera web (http://www.altera.com/html/tools/download_2.html), that can be used together with these kits for VHDL and Verilog synthesis. Regards Wolfgang valeri <vassenov@dsl.ca> wrote in message news:978591339.623677@sai.velocet.net... > Hi, > I want to start with FPGA design. > I will appreciate any suggestions for good starter kit. > > Val > > >Article: 28257
We've implemented NIOS on three of our prototyping platforms with a fourth to come and had only good experiences so far. For some of these systems we use NIOS with software to do board tests and for others its used for on-chip test pattern generation and checking when doing ASIC prototyping. It worked very well with APEX 20K/Quartus and we just had a little hassle using MAX+plus II for FLEX 10KE and ACEX 1K which are not officially supported yet. The problem with MAX+plus II is that it doesn't support node names longer than 32 characters. With Synopsys DC you can automatically limit the length of node names in the EDIF netlist passed to MAX+plus II. For other synthesis tools we've written a little utility that shortens the names in an EDIF netlist. The problem will go away, once ACEX 1K and FLEX 10KE are supported in Quartus II. We're supporting a customer here in Germany that decided to completely switch to NIOS for future designs and they use a lot embedded processors. The best thing about NIOS is that once you generate a system, including a core and it's peripherals you get VHDL or Verilog source code. You can easily re-target this code. Never again you'll have to worry about discontinued processors or peripherals and the associated hassle with last time buys, major re-designs or re-writing software. Regards Wolfgang El Camino http://www.elca.de <eriks@avidyne.com> wrote in message news:930771$rds$1@nnrp1.deja.com... > I'm thinking about using the NIOS processor from Altera in a new design > and I am curious if anyone has had any good or bad experiences with it? > > > Sent via Deja.com > http://www.deja.com/Article: 28258
NS16450 - a big archaic chip without fifo and without input/output drivers. A UART inside FPGA (or CPLD) plus MAX232 derivate could be better variant.Article: 28259
I'm using the example listed at XILINX web site for implementing XILINX SRL16E as a FIFO. But in order to use the SRL 16E as a FIFO, I've added Write_Enable and ADDRESS_rd signals. The problem that I'm having is that once the signal addr_rd is added to the design, the LEONARDO EXEMPLAR (Version: v20001b.106) does not "recognize" that I'm referring to SRL16E. Please look at the VHD file listed below, and you can notice at the end of the architecture that I have 2 line codes for the DOUT signal where one line is remarked If I switch between those 2 line codes, everything is OK, but It's not what I want to do. In other words: If I use the line: dout <= zigmond(cycle-1); LEONARDO recognises SEL16E But if I use: dout <= zigmond(conv_integer(addr_rd)); LEONARDO makes it all using FLIP-FLOPS, which I do not want. Can you assist? ---------------------------------------------------- ---------------------------------------------------- -- -- Library Name : srl_fifo -- Unit Name : srl16e -- Unit Type : Text Unit -- ------------------------------------------------------ ------------------------------------------ ------------------------------------------ -- Date : Wed Jan 03 17:34:14 2001 -- -- Author : Yoav Hirsch -- -- Company : ECI - ACCESS SOLUTIONS - ASIC group -- -- Description : XILINX SRL 16E basic element for FIFO -- Synthisizable by Exemplar ------------------------------------------ ------------------------------------------ Library IEEE; use ieee.std_logic_1164.all; use ieee.std_logic_arith.all; use ieee.std_logic_unsigned.all; entity srl16e is Generic (cycle : integer := 16; width : integer := 16); port ( din : IN std_logic_vector(width-1 downto 0); addr_rd : IN std_logic_vector(3 downto 0); clk_w : IN std_logic; we : IN std_logic; dout : OUT std_logic_vector(width-1 downto 0) ); attribute clock_node : boolean; attribute clock_node of clk_w : signal is TRUE; end; ------------------------------------------ ------------------------------------------ architecture srl16e_arch of srl16e is type freud is array(0 to cycle -1) of std_logic_vector(width-1 downto 0); signal zigmond : freud; begin process(clk_w) begin IF clk_w'event AND clk_w = '1' THEN if we = '1' THEN zigmond <= din & zigmond(0 to cycle - 2); END IF; END IF; end process; -- dout <= zigmond(cycle-1); dout <= zigmond(conv_integer(addr_rd)); end; Sent via Deja.com http://www.deja.com/Article: 28260
>Neil Glenn Jacobson <neil.jacobson@xilinx.com> schrieb in im Newsbeitrag: 3A539955.58A802CE@xilinx.com... >Please check this app. note for a Xilinx-supplied solution: > Xilinx XAPP Note Summaries >The accompanying software is available from here: > eisp_pc.zip I tried the software from the eisp_pc.zip (xapp058) and my system crashes after/while the ISPEN command. Has anybody succesfully used this software for ISP purposes ? I found that the xsvf-download software (from the micro.c file) always goes to the Run-Test/Idle State after a XSIR command. But the EZTAG programmer does not go to the Run-Test/Idle state but continues with the Select-DR-Scan state to shift in the 6bit value 0x0f before it goes to Reset_test/Idle state. Maybe someone has a modified/corrected micro.c sourcecode ? TIA Leopold Faschalek SkiData AG http://www.skidata.com mailto:fale@skidata.comArticle: 28261
Hi Pascal, See whether the following article of XILINX would be of any help to you: http://www.xilinx.com/support/techxclusives/signals-techX5.htm Yoav Hirsch. Sent via Deja.com http://www.deja.com/Article: 28262
> Hi, > I want to start with FPGA design. > I will appreciate any suggestions for good starter kit. > > Val I would suggest you go to http://www.deja.com and do a search for the term "FPGA starter kit" within the comp.arch.fpga news archive. I got 40 hits on this subject within the past year. You should be able to find references to all the popular introductory kits in this way. -- || Dr. Dave Van den Bout XESS Corp. (919) 387-0076 || || devb@xess.com 2608 Sweetgum Dr. (800) 549-9377 || || http://www.xess.com Apex, NC 27502 USA FAX:(919) 387-1302 ||Article: 28263
Hello everybody, I am working on an FPGA with the surrounded hardware being given, so that I have to find a solution, even if it is not optimal. I want to quadruple a clock of 24MHz, so that I finally have a clock with 96MHz frequency. I used two DLLs to do so and- it seems to work fine. However, the datasheet gives 25MHz as the lowest DLL input frequency, so that I am definitly out of spec. There seem to be two solutions: 1) There is some safety margin, so that I can feed 24MHz into the DLL. As the DLL's resolution is limited to 60ps I assume that it is built from 60ps-delay items- which will have some temperature or voltage dependency. Is there some detailed information on the behavior of the DLLs??? 2) I first double the clock by using combinational logic, basically a small delay followed by an XOR. This results in a 48MHz clock with a very small duty cycle which is fed into a single DLL. This DLL will correct my duty cycle to 50% and at the same time doubles my clock to 96MHz. If this works, has anybody done so? I am working with VHDL, how do I prevent the tools from "optimizing" my combinational path? Any comments very much appreciated, best regards Felix Bertram Sent via Deja.com http://www.deja.com/Article: 28264
Hello, I'm trying to generate a FPGA compiler2 script. What I want to do is create user defined VHDL libraries. From the help menu the command to do this is: create_library -name my_library This gives the following error: Error: extra positional option 'my_library' (CMD-012) Does anyone know what I'm doing wrong? Lee -- Lee Weston, Philips Semiconductors, CS-DM Southampton, SO15 0DJ, UK mailto:lee.weston@philips.com seri:weston@ukpsshp1 phone: +44 (0) 23 80316471 fax +44 (0) 23 80316303Article: 28265
In the one that works, the delay tap is constant, while in the second it is dynamic. The synthesizers until very recently won't infer an SRL16 with a dynamically changed delay. I think the latest release of Synplicity is supposed to infer the dynamic SRL16, although I have not tried it. For this case, instantiate yoru SRL16's and it should work for you. hirsch_yoav@hotmail.com wrote: > > I'm using the example listed at XILINX web site for implementing XILINX > SRL16E as a FIFO. But in order to use the SRL 16E as a FIFO, I've added > Write_Enable and ADDRESS_rd signals. > > The problem that I'm having is that once the signal addr_rd is added to the > design, the LEONARDO EXEMPLAR (Version: v20001b.106) does not "recognize" > that I'm referring to SRL16E. > > Please look at the VHD file listed below, and you can notice at the end of > the architecture that I have 2 line codes for the DOUT signal where one line > is remarked If I switch between those 2 line codes, everything is OK, but > It's not what I want to do. > > In other words: > If I use the line: dout <= zigmond(cycle-1); > LEONARDO recognises SEL16E > > But if I use: dout <= zigmond(conv_integer(addr_rd)); > LEONARDO makes it all using FLIP-FLOPS, which I do not want. > > Can you assist? > > ---------------------------------------------------- > ---------------------------------------------------- > -- > -- Library Name : srl_fifo > -- Unit Name : srl16e > -- Unit Type : Text Unit > -- > ------------------------------------------------------ > ------------------------------------------ > ------------------------------------------ > -- Date : Wed Jan 03 17:34:14 2001 > -- > -- Author : Yoav Hirsch > -- > -- Company : ECI - ACCESS SOLUTIONS - ASIC group > -- > -- Description : XILINX SRL 16E basic element for FIFO > -- Synthisizable by Exemplar > ------------------------------------------ > ------------------------------------------ > Library IEEE; > use ieee.std_logic_1164.all; > use ieee.std_logic_arith.all; > use ieee.std_logic_unsigned.all; > > entity srl16e is > Generic (cycle : integer := 16; > width : integer := 16); > port ( > din : IN std_logic_vector(width-1 downto 0); > addr_rd : IN std_logic_vector(3 downto 0); > clk_w : IN std_logic; > we : IN std_logic; > dout : OUT std_logic_vector(width-1 downto 0) > ); > attribute clock_node : boolean; > attribute clock_node of clk_w : signal is TRUE; > end; > ------------------------------------------ > ------------------------------------------ > architecture srl16e_arch of srl16e is > type freud is array(0 to cycle -1) of > std_logic_vector(width-1 downto 0); > signal zigmond : freud; > begin > process(clk_w) > begin > IF clk_w'event AND clk_w = '1' THEN > if we = '1' THEN > zigmond <= din & zigmond(0 to cycle - 2); > END IF; > END IF; > end process; > -- dout <= zigmond(cycle-1); > dout <= zigmond(conv_integer(addr_rd)); > end; > > Sent via Deja.com > http://www.deja.com/ -- -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 or http://www.fpga-guru.comArticle: 28266
Felix, The low frequency limit of the DLL is process, voltage and temperature dependent (as you might guess!). The delay lines run off their own regulated supply, so the voltage variations are extremely small (don't worry about them). The temperature variation is important, because if it locks at hot, and then it gets cold, it could run out of taps at a low frequency (line not long enough as it gets faster as it got colder). There is definitely a safety margin, and it is at least 10%. But we may choose to ship product that is much much faster (everyone loves faster), and reduce this margin in the future. Now, I am assuming you are not going below 0C, so there will be some margin there. But if not, then definitely don't do it. In process, the delay lines use techniques to make them less sensitive to transistor speed, but a fast lot will still have slightly faster delay lines, which will not work reliably at lower frequencies. All of that said, I would not recommend its use out of specification. You would probably get away with it until that one lot came along that was just a little too fast when your board got just a little too cold.....(been there, done that: not good for your career!). Doubling with a delay and an XOR is a common technique, and if you can take the delay element and place it off chip, you will not have to worry about it getting too fast due to the process/voltage/temperature variations in the FPGA. I used to use a small RC off chip that added to the delay and made this reliable (used in 100K+ units of a fiber optic transmission system). I lived through three process changes of Xilinx fpga's with this method, even though it is an asynchronous "no-no". The DLL only cares about the rising edge of the input, and is incredibly tolerant to input jitter, so it can easily use the XOR doubled input signal. The resulting 4X output, however achieved, will have more jitter. I would expect less than 200ps peak to peak period jitter, worst case, when everything is up and running (all ff's clocking, IOB's toggling, etc). A single DLL (not doubling) in the same scenario would have less than 100 ps. Jitter is not strictly additive, and the DLL does no filtering of any input jitter. Austin felix_bertram@my-deja.com wrote: > Hello everybody, > > I am working on an FPGA with the surrounded hardware being given, so > that I have to find a solution, even if it is not optimal. > > I want to quadruple a clock of 24MHz, so that I finally have a clock > with 96MHz frequency. I used two DLLs to do so and- it seems to work > fine. However, the datasheet gives 25MHz as the lowest DLL input > frequency, so that I am definitly out of spec. > > There seem to be two solutions: > 1) There is some safety margin, so that I can feed 24MHz into the DLL. > As the DLL's resolution is limited to 60ps I assume that it is built > from 60ps-delay items- which will have some temperature or voltage > dependency. Is there some detailed information on the behavior of the > DLLs??? > > 2) I first double the clock by using combinational logic, basically a > small delay followed by an XOR. This results in a 48MHz clock with a > very small duty cycle which is fed into a single DLL. This DLL will > correct my duty cycle to 50% and at the same time doubles my clock to > 96MHz. If this works, has anybody done so? I am working with VHDL, how > do I prevent the tools from "optimizing" my combinational path? > > Any comments very much appreciated, > best regards > > Felix Bertram > > Sent via Deja.com > http://www.deja.com/Article: 28267
************************************************************************** DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 ************************************************************************** ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ + + + CALL FOR DEMOS AT THE UNIVERSITY BOOTH + + -------------------------------------- + + + + DESIGN AUTOMATION AND TEST IN EUROPE + + DATE 2001 + + + + Munich, Germany + + 13-16 March 2001 + + + + http://www.date-conference.com/conference/university-booth.htm + + + ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Sponsored by the European Design and Automation Association, the EDA Consortium, the IEEE Computer Society - TTTC, ECSI and RAS. In cooperation with ACM SIGDA. A university booth will be organized for CAD software and hardware demonstrations during DATE'2001 in Munich, Germany. The conference and the exhibition including the university booth will be held in the superb new facilities of the International Congress Center (ICM), (Info: http://www.icm-muenchen.de) All demonstrations will take place on Tuesday 13th until Friday 16th March 2001. The demonstrations have to be accompanied by a one-page abstract which will be published and distributed to all conference participants. -------------------------------------------------------------------------- CAD Software Demonstrations: Several machines (SPARC Ultra running Solaris 7) will be provided. The demonstration software has to be installed on the SPARC Ultra machines BEFORE the exhibition itself takes place: Demonstrators will be given a machine name and an user-id valid from February 19th, 2001, to March 2nd, 2001, for an installation via Internet. Each CAD demonstration should be supported by a poster (not larger than 0.9m width by 1.4m height). CAD demonstrations on other computer architectures are possible on the condition that demonstrators provide their own equipment and set up the demonstration well before the assigned presentation time slot. -------------------------------------------------------------------------- Working Silicon Hardware Demonstrations: All circuits must be demonstrated within a working environment. Stand alone circuits will not be exhibited. Each exhibitor is responsible for bringing and installing the hardware and software resources necessary for showing the entire functionality of the circuit. Each hardware demonstration should be supported by a poster (not larger than 0.9m width by 1.4m height). Each poster should contain a plot of the circuit, complemented by any information on the functionality of the circuit, test results and working environment. -------------------------------------------------------------------------- Deadline and addresses: The deadline for submission of a CAD software or a working silicon demonstration is January 31st, 2001. The appropriate application form (CAD software or working silicon) should be sent together with a one page description (in PostScript or PDF) via email to DATE2001@informatik.uni-tuebingen.de For any further information or requests feel free to contact Juergen Ruf or Carsten Schulz-Key DATE2001@informatik.uni-tuebingen.de ************************************************************************** DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 ************************************************************************** -------------------------------------------------------------------------- REGISTRATION FORM FOR CAD SOFTWARE DEMOS AT THE UNIVERSITY BOOTH -------------------------------------------------------------------------- 1/ University or Institution: 2/ Contact person Name: Address: Phone: Fax: E-mail: 3/ Software demonstrated (name): 4/ Abstract of the functionality of the software (please attach a one page description as a separate PostScript or PDF file) 5/ Machine specifications: SPARC: [ ] Yes [ ] No other computer (hardware and operating system requirements): In this case, can you bring your own machine? [ ] Yes [ ] No 6/ Configuration requirements: Hard Disk:....... Internal Memory:........ Swap Space:........... 7/ Number of one hour slots wanted for demonstrations: -------------------------------------------------------------------------- REGISTRATION FORM FOR WORKING SILICON DEMOS AT THE UNIVERSITY BOOTH -------------------------------------------------------------------------- 1/ University or Institution: 2/ Contact Person Name: Address: Phone: Fax: E-mail: 3/ Circuit demonstrated (Name): 4/ Abstract of the functionality of the circuit, test results, working environment (please attach a one page description as a separate PostScript or PDF file) 5/ Necessary hardware and software resources (provided by the demonstrator): 6/ Number of one hour slots wanted for demos: ************************************************************************** DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 DATE'2001 **************************************************************************Article: 28268
Hi, I have a hierarchical design, with the entities stored in individual files and all the architectures stored in 1 file, some of the archs use synopsys Design Ware components, in the past I have always designed with the entity and architecture for each module the same file, is there any advantage in doing it this way? Where do I put the the Library and Use statements for the libraries, in the entity, architecture or somewhere else? and how do I go about compiling these files in Synopsys DC and ModelSim? Thanks jonArticle: 28269
This is a multi-part message in MIME format. --------------E70F5D543D1AEA87E45EC4CD Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hey jon, jon wrote: > Hi, > > I have a hierarchical design, with the entities stored in individual files > and all the architectures stored in 1 file, some of the archs use synopsys > Design Ware components, in the past I have always designed with the entity > and architecture for each module the same file, is there any advantage in > doing it this way? The main advantage I see for this methodlogy is the fact you can have multiple architectures (if you desire) for each entity. A side advantage, if you are targeting Xilinx, could be that you would be able to specify generics (for bus widths for instance) within the entity, then generate an architecture only structural netlist from the place and route tools. You can then re-use the RTL entity with the structural architecture and most likely not have to do any modification to your testbench for back-end simulation. You can still do this with both in the same file but this is a bit cleaner this way. The disadvantage is now you have more files to deal with and an associated compilation order to deal with. > Where do I put the the Library and Use statements for the libraries, in the > entity, architecture or somewhere else? Both. You need library statements in the entity to specify port typing (i.e. std_logic requires the ieee libraries). You need libraries specifications in the architecture file as well for signal typing as well as any other simulation libraries such as in your case, the DW simulation library models. I usually just mimic the library and use specification for both files for clarity. > and how do I go about compiling > these files in Synopsys DC and ModelSim? There will be an associated compilation order now. You need to compile the entity before the architecture for both DC and ModelSim. If you are to use multiple architectures, it is generally a good idea to write a configuration (possibly in yet another seperate file) to at minimum bind the right architecture to the right entity. This way you know what you get when you simulate or synthesize. Otherwise, the general rule is the last architecture that was compiled is the one that is bound. -- Brian > > > Thanks jon --------------E70F5D543D1AEA87E45EC4CD Content-Type: text/x-vcard; charset=us-ascii; name="brian.philofsky.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Brian Philofsky Content-Disposition: attachment; filename="brian.philofsky.vcf" begin:vcard n:Philofsky;Brian tel;work:1-800-255-7778 x-mozilla-html:TRUE url:http://www.xilinx.com org:Xilinx, Inc.;Software Marketing adr:;;2300 55th Street;Boulder;CO;80301;USA version:2.1 email;internet:brianp@xilinx.com title:Sr. Technical Marketing Engineer fn:Brian Philofsky end:vcard --------------E70F5D543D1AEA87E45EC4CD--Article: 28270
Hi all I'm a PhD student whose current research focuses on nondeterministic FSMs. It seems to me that there are interesting ways of implementing them efficiently in hardware. I was wondering if nondeterministic FSMs have previously been implemented in VLSI/programmable logic and if so, for what applications. I would be grateful for any info/pointers regarding above. Thanks in advance. RegardsArticle: 28271
Hi, tell us more about non-deterministic finite state machines. What makes them non-deterministic? Do they roll dice? ;-) What's good about them? Peter Alfke, Xilinx ============================== "Reetinder P. S. Sidhu" wrote: > Hi all > > I'm a PhD student whose current research focuses on nondeterministic > FSMs. It seems to me that there are interesting ways of implementing > them efficiently in hardware. I was wondering if nondeterministic FSMs > have previously been implemented in VLSI/programmable logic and if so, > for what applications. > > I would be grateful for any info/pointers regarding above. Thanks in > advance. > > RegardsArticle: 28272
"Reetinder P. S. Sidhu" wrote: > > Hi all > > I'm a PhD student whose current research focuses on nondeterministic > FSMs. It seems to me that there are interesting ways of implementing > them efficiently in hardware. I was wondering if nondeterministic FSMs > have previously been implemented in VLSI/programmable logic and if so, > for what applications. > > I would be grateful for any info/pointers regarding above. Thanks in > advance. > > Regards and I would be grateful if someone would simply explain what these things are... i.e., in what sense deterministic? thanks, -eric -- Eric C. Fromm efromm@sgi.com Principal Engineer Scalable Systems Division SGI - Silicon Graphics, Inc. Chippewa Falls, Wi.Article: 28273
In article <3A54FB53.832ACB93@sgi.com>, Eric C. Fromm wrote: >"Reetinder P. S. Sidhu" wrote: >> >> Hi all >> >> I'm a PhD student whose current research focuses on nondeterministic >> FSMs. It seems to me that there are interesting ways of implementing >> them efficiently in hardware. I was wondering if nondeterministic FSMs >> have previously been implemented in VLSI/programmable logic and if so, >> for what applications. >> >> I would be grateful for any info/pointers regarding above. Thanks in >> advance. >> >> Regards > >and I would be grateful if someone would simply explain what these things >are... > >i.e., in what sense deterministic? > >thanks, > >-eric > >-- >Eric C. Fromm efromm@sgi.com >Principal Engineer Scalable Systems Division >SGI - Silicon Graphics, Inc. Chippewa Falls, Wi. Given a state and an input symbol, a deterministic finite automaton (DFA) can "go to" only one next state. A nondeterministic finite state automaton (NFA), however, may have multiple candidates for the next state. DFAs and NFAs have equivalent expressive power, namely, that of the regular expressions. -- George Coulouris - http://www.tc.cornell.edu/~glc5/Article: 28274
George Coulouris wrote: > > Given a state and an input symbol, a deterministic finite automaton (DFA) > can "go to" only one next state. A nondeterministic finite state automaton > (NFA), however, may have multiple candidates for the next state. > > DFAs and NFAs have equivalent expressive power, namely, that of the regular > expressions. > > -- > George Coulouris - http://www.tc.cornell.edu/~glc5/ OK, I get the 'multiple candidates for next state' bit. You lost me with the 'expressive power' assertion though. What is this characteristic and how do DFAs and NFAs compare to other constructs (and what are those constructs)? thanks, -eric -- Eric C. Fromm efromm@sgi.com Principal Engineer Scalable Systems Division SGI - Silicon Graphics, Inc. Chippewa Falls, Wi.
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