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
Where can I obtain a FAQ for this newsgroup? I would like to advise subscribers of opportunities I am working on, but do not want to upset or get flamed. Any help with the rules will be appreciated! Also, any suggestions on other newsgroups will be welcomed. -- rcp@frontiernet.net F-O-R-T-U-N-E Personnel Consultants of Worcester (978) 466 - 6800Article: 16426
FPGA 2000 Call for Papers Eighth ACM International Symposium on Field-Programmable Gate Arrays Monterey, California February 10-11, 2000 The annual ACM/SIGDA International Symposium on Field-Programmable Gate Arrays is the premier conference for presentation of advances in all areas related to FPGA technology. For FPGA 2000, we are soliciting submissions describing novel research and development in the following (and related) areas of interest: * FPGA Architecture: Logic block & routing architectures, I/O structures and circuits, new commercial architectures, Field-Programmable Interconnect Chips and Devices (FPIC/FPID), Field-Programmable Analog Arrays (FPAA). * CAD for FPGAs: Placement, routing, logic optimization, technology mapping, system-level partitioning, logic generators, testing and verification. CAD for FPGA-based accelerators. * Applications: Innovative use of FPGAs, exploitation of FPGA features, novel circuits, high-performance and low-power/mission-critical applications, DSP techniques, uses of reconfiguration, FPGA-based cores. * FPGA-based computing engines: Compiled accelerators, reconfigurable computing, adaptive computing devices, systems and software. * Rapid-prototyping: Fast prototyping for system level design, Multi-Chip Modules (MCMs), logic emulation. Authors are invited to submit PDF of their paper (12 pages maximum) by October 1, 1999 via E-mail to hauck@ece.nwu.edu. Notification of acceptance will be sent by November 17, 1999. The authors of the accepted papers will be required to submit the final camera-ready copy by December 1, 1999. A proceedings of the accepted papers will be published by ACM, and included in the Annual ACM/SIGDA CD-ROM Compendium publication. Address questions to: Scott Hauck Program Chair, FPGA 2000 Dept. of ECE, Northwestern University 2145 Sheridan Rd Evanston, IL 60208 Phone: (847) 467-1849 Fax: (847) 467-4144 hauck@ece.nwu.edu General Chair: Steve Trimberger, Xilinx Program Chair: Scott Hauck, Northwestern U. Finance Chair: Sinan Kaptanoglu Publicity Chair: Herman Schmit, CMU Program Committee: Miron Abramovici, Lucent David Lewis, U. of Toronto Ray Andraka, Andraka Consulting Fabrizio Lombardi, Northeastern U. Mike Bershteyn, Quickturn Wayne Luk, Imperial College Michael Butts, Synopsys Margaret Marek-Sadowska, UCSB Jason Cong, UCLA Jan Rabaey, UCB Eugene Ding, Lucent Jonathan Rose, U. of Toronto Carl Ebeling, U. of Washington Martine Schlag, UCSC Reiner Hartenstein, U. Kaiserslautern Herman Schmit, CMU Scott Hauck, Northwestern U. Tim Southgate, Altera Brad Hutchings, BYU Russ Tessier, U. Mass. - Amherst Sinan Kaptanoglu, Actel Steve Trimberger, Xilinx Tom Kean, Algotronix John Wawrzynek, UCB Martin Wong, UT at Austin Sponsored by ACM SIGDA, with support from Altera, Lucent, Actel and Xilinx. Please visit the web site <http://www.ece.cmu.edu/~fpga2000> for more information.Article: 16427
David Pashley <David@edasource.com> writes: > > How does the fact that Synopsys goes from strength to strength, and is > unarguably both successful and competitive (see their website, Q2 sales > 21% up at $190m) square with your comments? > Do you sell Synopsys products? Homann -- Magnus Homann Email: d0asta@dtek.chalmers.se URL : http://www.dtek.chalmers.se/DCIG/d0asta.html The Climbing Archive!: http://www.dtek.chalmers.se/Climbing/index.htmlArticle: 16428
In article <ltpv3utpor.fsf@palver.dtek.chalmers.se>, Magnus Homann <d0asta@palver.dtek.chalmers.se> writes >David Pashley <David@edasource.com> writes: >> >> How does the fact that Synopsys goes from strength to strength, and is >> unarguably both successful and competitive (see their website, Q2 sales >> 21% up at $190m) square with your comments? >> > >Do you sell Synopsys products? > Indirectly, yes. Among other activities, we sell and support Viewlogic products which include an OEM of FPGA Express. However, I was trying to substantiate a point (that while you might debate FPGA Express, DC is *clearly* competitive), not promote Synopsys. What I'm trying to work out is why the criticisms levelled at FPGA Express in terms of language don't also affect DC (or maybe they do?). -- David Pashley < --------------------------- < < < --- mailto:david@edasource.com | Direct Insight Ltd < < < < > Tel: +44 1280 700262 | | http://www.edasource.com < < < Fax: +44 1280 700577 | ------------------------------ < ---------------------------------Article: 16429
sc@vcc.com wrote: > > Software engineers are starting to > say "you can do what?" ... The goal, of course, is > to not even know whether you have a reconfigurable > computer or not and just be able to code away as usual. So which is it? Heads up to the software developers, or not? --== Sent via Deja.com http://www.deja.com/ ==-- ---Share what you know. Learn what you don't.---Article: 16430
rolandpj@bigfoot.com wrote: > > I like to to view the problem as an extension of HotSpot/JIT. > ... Why not do the same thing, but right down to the hardware? Reliability. If an FPGA fails to reconfigure itself properly, then how to recover? --== Sent via Deja.com http://www.deja.com/ ==-- ---Share what you know. Learn what you don't.---Article: 16431
David Pashley wrote in message <4PYc7IAHxRR3MA7i@edasource.com>... >In article <37446aa8.4459222@news.dial.pipex.com>, ems@riverside- >machines.com.NOSPAM writes >>there is absolutely *no way* that fpga express is competitive in terms >>of vhdl language support, and i'm sure that you know this as well as >>the rest of us do. true, there have been major advances in both DC and >>express recently, but you've still got some way to go. >> >>on the other hand, there have been some really nice additions in 3.1 - >>FST/scripting and attribute passing primarily, which would make this a >>very competitive tool if you got the language support right. >> >>evan > >As I understand it, and as you suggest, the language support in DC and >FPGA Express is the same. > >How does the fact that Synopsys goes from strength to strength, and is >unarguably both successful and competitive (see their website, Q2 sales >21% up at $190m) square with your comments? They're the Microsoft of the synthesis market. >Do you mean that the language requirements for FPGA are totally >different from those for ASIC? No, he means that Synopsys' support of VHDL differs from the standard. For instance, their whole idea of putting the std_logic_arith/unsigned/signed stuff into the ieee library, rather than calling it the synopsys library. And the lack of real VHDL'93 support (in 1999, no less). -- a ------------------------------------------ Andy Peters Sr. Electrical Engineer National Optical Astronomy Observatories 950 N Cherry Ave Tucson, AZ 85719 apeters@noao.edu "Space, reconnaissance, weather, communications - you name it. We use space a lot today." -- Vice President Dan QuayleArticle: 16432
This is a multi-part message in MIME format. --------------156FB254376CE33DF2CF18D4 Content-Type: text/plain; charset=us-ascii Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Content-Transfer-Encoding: 7bit Adam J. Elbirt wrote: > Terry, > > That's interesting. I downloaded SP1 and 2 from the web site > specifically > from the 1.5 area, not the 1.5i area. When I talked to Xilinx support > they > claimed that the service packs were fine for 1.5 and that they had > never heard > of the page faulting problem. They're still working on it so > hopefully we'll > here from them soon. > > Adam > Hello Adam, The Service Packs for version 1.5i are not compatible with version 1.5. Version 1.5 did have an SP1, but not an SP2, so it appears that you may have installed 1.5i SP2 into a 1.5 environment which will not work. I suggest installing 1.5i and applying the 1.5i SP2 update. BTW, I was unable to find an open hotline case for you in the call tracking system. Please send me the case number and I'll check into it. Regards, Bret Wade Xilinx Product Applications --------------156FB254376CE33DF2CF18D4 Content-Type: text/x-vcard; charset=us-ascii; name="vcard.vcf" Content-Transfer-Encoding: 7bit Content-Description: Card for Bret Wade Content-Disposition: attachment; filename="vcard.vcf" begin: vcard fn: Bret Wade n: Wade;Bret org: Xilinx email;internet: bret.wade@xilinx.com title: Senior Engineer Product Applications x-mozilla-cpt: ;0 x-mozilla-html: TRUE end: vcard --------------156FB254376CE33DF2CF18D4--Article: 16433
Roland Paterson-Jones <rpjones@hursley.ibm.com> wrote: [HotSpot...] : Now, why not do the same thing, but right down to the hardware, rather than : down to machine code. What you need, however, is a general compiler from a : high-level language (Java bytecode?) to fpga gates. [...] This would probably wind up being a rather poor way of exploiting the power of the available hardware ;-/ While it's /possible/ to write multi-threaded Java code that's capable of executing concurrently, threaded code is less common than it might be, partly due to it being harder to write. Apart from it's thread support, Java's pretty inherently serial. The thought of attempting to extract parallelism from such a serial stream of instructions usually makes me feel nauseous. While I'm all in favour of a portable, universal means of describing algorithms that enables them to be implemented efficiently on parallel hardware, unfortunately, Java doesn't look /remotely/ like what I envisage. Java's ability to exploit parallelism (aside from threading code) really isn't very good. You even typically have to initialise an array of objects using a "for" loop. Writing a smart compiler to examine loops to see if this sort of thing is occurring seems to me to be a backwards approach: a /sensible/ concurrent language should allow this type of parallelism to be explicitly indicated, rather than expecting advanced AI techniques to extract the information about when it can and can't be done from the code. Extending Java's collection frameworks to allow operations to be performed on all elements simultaneously, and some other more "functional" bits-and-pieces might help with concurrency issues - but picking something else entirely, something /without/ such a strong serial legacy would appear to be a much more sensible idea to me. I have other concerns, apart from speeding up Java's execution speed - I want something capable of reasonably efficient exploitation of the available hardware. Personally, I'd like an approach based fairly directly on circuit design, /preferably/ something that would allow construction of an n-dimensional circuit from an original, higher-level representation. I can't really see how a written, language-based, serial HDL would be very suitable as a basis for this. I envisage something like a "rubber circuit" which 'stretches' to fit the characteristics of the target hardware, while retaining the relevant inter-component distances and ratios, for purposes of retaining correct synchronous operation. -- __________ |im |yler The Mandala Centre http://www.mandala.co.uk/ tt@cryogen.com A clean disc is a sign of a sick computer.Article: 16434
Roland Paterson-Jones wrote: > > > I like to to view the problem as an extension of HotSpot/JIT technologies > in virtual machine implementation, most notably lately in Java Virtual > Machines. What these technologies do is profile a program on-the-fly (with > embedded profiling instructions, or through interpretation, or whatever). > When they determine that a certain portion of a program is heavily > exercised, then that portion is (re-)compiled with heavy optimisation. Hey thats what my patent is about. The patent examiner actually quoted "generation" from the dictionary. http://www.patents.ibm.com/details?pn=US05684980__&s_clms=1#20?pn=US05684980__&s_clms=1 Look at claim 26. > > > Now, why not do the same thing, but right down to the hardware, rather than > down to machine code. What you need, however, is a general compiler from a > high-level language (Java bytecode?) to fpga gates. According to the empty > response to a previous posting, nobody is interested in such a thing. Macro-Based Hardware Compilation of Java Bytecodes into a Dynamic Reconfigurable Computing System. Cardoso et al. FCCM 99. > > > What you also need is a sense of when this would be more useful than merely > compiling to the machine-code level. I'd be inclined to think that it's > almost always useful, since you can parallelize as much as the source code > will allow for each specific case, which will always be better than a fixed > super-scalar processor architecture. It all depends on the bandwidth at which you can supply the RC unit (in most cases). > > > Now, each thread of your program will have a different hotspot footprint, > so when you do a context switch at the software (thread) level, you switch > your gate array for the hardware-implemented hotspots of the new thread. That is what most companies do. Take the bit stream and convert it into a static array. It can then be complied directly into byte codes or the software image. > > > I believe this approach simplifies things, and also truly unifies hardware > and software, since the programmer is entirely unaware of what's going on > under the hood. That's what we want in the end, isn't it - the software is > the machine. > > I have dreams of a single multitasked fpga doing all of the stuff that each > separate component of a motherboard + cards currently does (or an array of > fpga's multi-tasking this). Cheap and fast and simple (once you've > implemented the JIT technology!). > > Roland One day (I know it is heresy) all computer will be made of FPGA like devices and there will be no CPUs or any of the current day structures. When quantum dots or whatever comes along and you have a tera gate system with giga bytes of memory diffused through out the system you won't need a pentium 23. In fact it will be too hard to design a CPU make it testable and get it into production. It will take 20 yr just to do the test vectors. No, in the end everything will be reconfigurable and advances in computing will come at the reconfigurable cell/architecture level, the algorithm level and the algorithm expression (HLL) level. -- Steve Casselman, President Virtual Computer Corporation http://www.vcc.comArticle: 16435
George wrote: > Hi Steve. > > Any idea if this is available on-line anywhere? I looked > around on www.fccm.org but did not see it. > > Macro-Based Hardware Compilation of Java > > Bytecodes into a Dynamic Reconfigurable > > Computing System. > > Cardoso et al. FCCM 99. I can't find it anywhere (except in the pre-proceedings) Joao M P Cardoso INSEC/University of Algarve and Horacio C Neto {Joao.Cardoso, hcn}@inesc.pt -- Steve Casselman, President Virtual Computer Corporation http://www.vcc.comArticle: 16436
Roland Paterson-Jones <rpjones@hursley.ibm.com> wrote: > > What you also need is a sense of when this would be more useful than merely > > compiling to the machine-code level. I'd be inclined to think that it's > > almost always useful, since you can parallelize as much as the source code > > will allow for each specific case, which will always be better than a fixed > > super-scalar processor architecture. > > > > Now, each thread of your program will have a different hotspot footprint, > > so when you do a context switch at the software (thread) level, you switch > > your gate array for the hardware-implemented hotspots of the new thread. I have no clue about FPGAs, but, I'm wondering, aren't the switching times prohibitive? Remember, you not only have to switch the hardware around for thread switches in your own process, but also for context switches among all the processes running on the hardware. Remember, you don't own the processor. Obviously, the switching circuitry takes up space on the chip. Maybe it's more efficient to spend this space on more processing logic. Another thing is, and here I'm indeed curious, what are the basic hardware building blocks you want to arrange in a thread-specific fashion? The finer the granularity, the costlier the reprogramming. Michael -- Michael Schuerig mailto:schuerig@acm.org http://www.schuerig.de/michael/Article: 16437
Hi, I have put the paper (postscript format) in: http://esda.inesc.pt/~jmpc If you have any problem with it, just send me an email and I will send you the paper. I am working on a Web page describing the project. Soon, more information will be available. Let me know if you want that I send you an email when the page is completed. Thanks for the interest. Regards, Joao Cardoso Steven Casselman wrote: > George wrote: > > > Hi Steve. > > > > Any idea if this is available on-line anywhere? I looked > > around on www.fccm.org but did not see it. > > > Macro-Based Hardware Compilation of Java > > > Bytecodes into a Dynamic Reconfigurable > > > Computing System. > > > Cardoso et al. FCCM 99. > > I can't find it anywhere (except in the > pre-proceedings) Joao M P Cardoso > INSEC/University of Algarve > and Horacio C Neto > {Joao.Cardoso, hcn}@inesc.pt > > -- > Steve Casselman, President > Virtual Computer Corporation > http://www.vcc.com -- ******************************************* Joao Manuel Paiva Cardoso INESC/ESDA Group phone: +351 1 3100288 Email: Joao.Cardoso@inesc.pt --------------987CE1625DEE2DF93A5DCAD0--Article: 16438
Well, that's one of the reasons why we called our chip PoliMorph: I was inspired by Morph, a particular JIT profiling technology that does more or less what you described. It's not easy, though, not at all, overall when you're talking about hardware reconfiguration. There are a number of serious hardware hurdles that will probably prevent FPGA coupled processors from conquering the general purpose market: first and foremost, exceptions and interrupts' handling is a *real* problem, secondly handling a multitasking system would present an array of trade off. But I'm pretty sure that the artificial intelligence my grandchildren will be quite accustomed to will be able to rewire itself (what about reconfigurable neural nets? We've always talked about them, now we could implement them). Guido Roland Paterson-Jones <rpjones@hursley.ibm.com> wrote in message 374517F1.6649758F@hursley.ibm.com... > Jonathan Feifarek wrote: > > > the problem is difficult > > because there are too many degrees of freedom right now. > > I like to to view the problem as an extension of HotSpot/JIT technologies > in virtual machine implementation, most notably lately in Java Virtual > Machines. What these technologies do is profile a program on-the-fly (with > embedded profiling instructions, or through interpretation, or whatever). > When they determine that a certain portion of a program is heavily > exercised, then that portion is (re-)compiled with heavy optimisation. > > Now, why not do the same thing, but right down to the hardware, rather than > down to machine code. What you need, however, is a general compiler from a > high-level language (Java bytecode?) to fpga gates. According to the empty > response to a previous posting, nobody is interested in such a thing. > > What you also need is a sense of when this would be more useful than merely > compiling to the machine-code level. I'd be inclined to think that it's > almost always useful, since you can parallelize as much as the source code > will allow for each specific case, which will always be better than a fixed > super-scalar processor architecture. > > Now, each thread of your program will have a different hotspot footprint, > so when you do a context switch at the software (thread) level, you switch > your gate array for the hardware-implemented hotspots of the new thread. > > I believe this approach simplifies things, and also truly unifies hardware > and software, since the programmer is entirely unaware of what's going on > under the hood. That's what we want in the end, isn't it - the software is > the machine. > > I have dreams of a single multitasked fpga doing all of the stuff that each > separate component of a motherboard + cards currently does (or an array of > fpga's multi-tasking this). Cheap and fast and simple (once you've > implemented the JIT technology!). > > Roland > >Article: 16439
This is a multi-part message in MIME format. --------------987CE1625DEE2DF93A5DCAD0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit -- Steve Casselman, President Virtual Computer Corporation http://www.vcc.com --------------987CE1625DEE2DF93A5DCAD0 Content-Type: message/rfc822 Content-Transfer-Encoding: 7bit Content-Disposition: inline Return-Path: <Joao.Cardoso@inesc.pt> Received: from mail.vcc.com (mail.vcc.com [192.168.0.21]) by vcc.com (8.8.5/8.8.5) with ESMTP id OAA16412 for <sc@mail.vcc.com>; Fri, 21 May 1999 14:01:28 -0700 Received: from popd.vcc.com by mail.vcc.com (fetchmail-4.5.0 POP3) for <sc/mail.vcc.com> (single-drop); Fri, 21 May 1999 14:01:38 PDT Received: by multi10.netcomi.com for sc (with Netcom Interactive pop3d (v1.21.1 1998/05/07) Fri May 21 20:54:12 1999) X-From_: jmpc@esda.inesc.pt Fri May 21 15:48:16 1999 Received: from inesc.inesc.pt (inesc.inesc.pt [146.193.0.1]) by multi10.netcomi.com (8.8.5/8.7.4) with SMTP id PAA25473 for <sc@vcc.com>; Fri, 21 May 1999 15:48:16 -0500 Received: from esda.inesc.pt (hobbes.inesc.pt) by inesc.inesc.pt with SMTP; id AA22095 (/); Fri, 21 May 1999 21:48:02 +0100 Received: from hobbes.inesc.pt (localhost) by esda.inesc.pt (4.1/OSversion) id AA18885; Fri, 21 May 99 21:48:01 +0100 Sender: Joao.Cardoso@inesc.ptArticle: 16440
"Computer chips that can rewire themselves to perform different functions are starting to take the “hard” out of hardware" See http://www.economist.com/editorial/freeforall/22-5-99/index_st9188.html. Jan GrayArticle: 16441
I was wondering if anyone here attempted to make a device that would perform the readback of Xilinx 4K FPGAs through s parallel port. I know about Xchecker cable, but it seems to be good for smaller devices only. Another question about Virtex readback. I haven't managed to find any info on the site. Does anyone know if they have some documentation on the topic and where can I get it? Thanks, Ivan HamerArticle: 16442
Michael Schuerig wrote in message <1ds6977.1ng0im2yreirkN@[192.168.1.2]>... > >I have no clue about FPGAs, but, I'm wondering, aren't the switching >times prohibitive? > Watch this space;-) If there's a good argument for fast-switching hardware, then fast switching hardware is what there'll be! > >Obviously, the switching circuitry takes up space on the chip. Maybe >it's more efficient to spend this space on more processing logic. > It's more difficult to get static-configuration hardware to use all of its resources than to reconfigure an exact hardware match to a problem. I'll bet that throughput figures for super-scalar processors are well below the theoretical maximum - with sharply diminishing returns with each additional parallel unit. > >Another thing is, and here I'm indeed curious, what are the basic >hardware building blocks you want to arrange in a thread-specific >fashion? The finer the granularity, the costlier the reprogramming. > I'd personally go for the lowest possible, to allow the compiler the maximum opportunity for optimization (including sharing etc.). This approach is analagous to the realisation of the RISC revolution that its better to have a simple instruction set and let the compiler do the hard work. RolandArticle: 16443
Hello! I've got the Xilinx FPGA/JTAG parallel port cable that comes as part of the XC9500 demo board, and wanted to know if it it was at all possible to use this with Altera devices. If not, is there any freeware schematic so one could make an Altera JTAG programmer? Is it possible to program a JTAG device independent of software (MAXPLUS II vs. Foundation, etc)? Also, can one actually use JTAG to apply stimulus to the EPLD/observe signals? I've always been a bit confused about that. Thanks for the help. Vasant remove nospam to e-mail.Article: 16444
-----Original Message----- From: Tim Tyler <tt@cryogen.com> Newsgroups: comp.arch.fpga Date: 21 May 1999 18:55 Subject: Re: High Speed Reconfigurability >Roland Paterson-Jones <rpjones@hursley.ibm.com> wrote: > >[HotSpot...] > >: Now, why not do the same thing, but right down to the hardware, rather than >: down to machine code. What you need, however, is a general compiler from a >: high-level language (Java bytecode?) to fpga gates. [...] > >This would probably wind up being a rather poor way of exploiting the >power of the available hardware ;-/ In the same way that a pentium/<paste your least favourite cpu> is a poor way of exploiting 16m(?) transitors ;-) >Apart from it's thread support, Java's pretty inherently serial. The >thought of attempting to extract parallelism from such a serial stream of >instructions usually makes me feel nauseous. > ...but this is exactly what any optimising compiler does - evaluating dependencies is quite close to extracting paralellism - you just need to schedule for an infinite-scalar machine, plus conditionals and both their branches can be evaluated simultaneously, plus loops can be unrolled to the limit of the hardware... The HANDEL-C examples where the explicit 'par' keyword is used to parallelise loops, didn't look too hard to auto-detect. I imagine any detection in C code is going to be more difficult than that for Java (due to C's dirty semantics). I'm not too interested in extracting 100% from an architecture (the answer to this is handcoding to the lowest level). What I'm interested in is the fastest way to run software that is easily written. I'm convinced that the compiler needs to do the hard work, not the programmer. This becomes crucial as the scale of a program grows. >Java's ability to exploit parallelism (aside from threading code) >really isn't very good. [...] Writing a smart compiler to examine loops to >see if this sort of thing is occurring seems to me to be a backwards >approach: a /sensible/ concurrent language should allow this type of >parallelism to be explicitly indicated, rather than expecting advanced AI >techniques to extract the information about when it can and can't be done >from the code. Good point. Are you suggesting a HANDEL-C-like approach? I'd like to be convinced that a compiler is really unable to detect these things. Do you have an example that would demonstrate this to me? (on the other hand, what if you're wrong, and it's not parallelisable? Should the compiler verify all such parallelisable assertions? This is likely to be as difficult as auto-detecting them in the first place. If this stuff is difficult for a compiler to detect, then it's sure to be difficult for a human to decide for all but the smallest examples.) Regards RolandArticle: 16445
On Fri, 21 May 1999 17:55:57 +0100, David Pashley <David@edasource.com> wrote: <snip> >However, I was trying to substantiate a point (that while you might >debate FPGA Express, DC is *clearly* competitive), not promote Synopsys. DC is "competitive" because: Effectively first in market Huge installed base Library support Numerous "add-ons" DC does not typically compete on conventional points set in the FPGA market such as QoR, speed, ease of use, and price. This is to some extent because the cost dynamics of the market are very different. >What I'm trying to work out is why the criticisms levelled at FPGA >Express in terms of language don't also affect DC (or maybe they do?). They do, but having fantastic language support is bugger all use when the ASIC vendor says "we sign off and only support DC". The amount of scripting, code changing, and general dicking about that most DC users seem to accept in an ASIC flow is indicative that some improvement may be welcome. By being effectively an "Industry Standard" in a very significant market, Synopsys can (to some extent) do what it likes in terms of language support. This is not the case with FPGA synthesis. IMHO, Express exists as an OEM purely to queer the pitch for other FPGA synthesis vendors who either are, or (as the FPGA/ASIC line blurs even more) will, try to "muscle in" on what is considered Synopsys turf. It's certainly a strange business model for a high-value, high-touch EDA company like Synopsys. Closer examination and interpolation of Synopsys financial figures from '98 reveals that their revenue stream is probably now close to being 45% maintenance. Synthesis and associated product accounts for 45-50% of total revenue so the SEAT sales of Synthesis AND "Design creation" products might soon account for less than a quarter of revenue. Take a look at the Synopsys line card and there are quite a few products that could be termed "design creation". With a seat of DC costing some 100K, the actual number of DC seats being sold might not be as impressive as the headline revenue figure might suggest. I heard a rumour that Synopsys sales people are no longer remunerated on DC sales as their job is to "sell-up" on the back of the headline product. Could be just a rumour though... Cheers Stuart An employee of Saros Technology: Model Technology, Exemplar Logic, TransEDA, Renoir. www.saros.co.ukArticle: 16446
On Wed, 19 May 1999 16:51:04 -0500, Tom McLaughlin <tomm@arl.wustl.edu> wrote: >I've looked at all of the manuals and finally decided to ask the list. >I want to assign the pads for my XCV1000 devices and assign I/O >standards to those pads such as LVTTL, LVCMOS, drive strength and such. >I am using Leonardo and cannot find an attribute to do this. I can find >attributes to assign slew rate and such, but not what kind of pad it >is!!! Tom, The GUI is probably staring you in the face :-) Load the library for Virtex and then load your design by either reading, or analysing and elaborating. When you read/elaborate your design you will find that the constraint tab extracts your clocks, inputs, outputs, Bidi's, signals, modules, and paths etc. The clock "power-tab" will enable explicit buffering with a BUFGP, while the input and output tabs will allow you to assign through the BUFFER pull-down menu, all the various types of Virtex I/O. You can assign to busses, or individual pins. The constraint file or script command is quite simply "PAD" so to pad a single signal "ena" in the "work" library of an entity called "counter" with an "rtl" architecture, you would produce: PAD IBUF_SSTL2_II .work.counter.rtl.enable For busses, you can use wild cards and bus selection in the form: PAD OBUF_HSTL_III .work.counter.rtl.q(*) or PAD OBUF_HSTL_III .work.counter.rtl.q(7:4) PAD OBUF_HSTL_I .work.counter.rtl.q(3:0) etc. Cheers Stuart For Email remove "NOSPAM" from the addressArticle: 16447
Has anyone successfully used the iob tristate registers in a 4000-XLA device ?. I am trying to decrease the clock to tristate time of some outputs connected to ZBT rams. I've read the relevant xilinx app-note on how to use the iob .nmc hard-macro, and i can't see any sensible way of doing a front-end simulation of the xilinx device and the rams, since the iob macro doesn't have a PAD pin. It seems i would need to have one design for simulation, using a model of the iob register connected to the rams, and another design for P&R which doesn't have the i/o pins. Is this correct ?. I was hoping for Unified Library support for the iob tristate flip-flop, but Xilinx seems to be moving away from using iob library primitives (they don't seem to be used for Virtex's). So has anyone heard of when the P&R software will be able to utilize the tristate registers automatically ? Edward Moore PS : you can use a -XL bitstream in a -XLA device, and suprisingly vice-versa.Article: 16448
Vasant, Attached is a link to ByteBlaster MV datasheet on the Altera website. http://www.altera.com/document/ds/dsbytemv.pdf It contains the schematic for the byteblaster cable (I would suggest the mv version, rather than the standard Byteblaster, as it supports the 3.3V parts as well as the 5V ones). To answer your question on s/w independent programming, check out Altera's JAM initiative ... http://www.altera.com/html/mktg/isp-jam.html Hope this helps, Dave Vasant Ram wrote: > Hello! I've got the Xilinx FPGA/JTAG parallel port cable that comes as > part of the XC9500 demo board, and wanted to know if it it was at all > possible to use this with Altera devices. If not, is there any freeware > schematic so one could make an Altera JTAG programmer? Is it possible to > program a JTAG device independent of software (MAXPLUS II vs. Foundation, > etc)? > > Also, can one actually use JTAG to apply stimulus to the EPLD/observe > signals? I've always been a bit confused about that. > > Thanks for the help. > Vasant > > remove nospam to e-mail. -- "The Jerry Springer show is positive proof that there is insufficient Chlorine in the gene pool" - Dave D'Aurelio, 1/99Article: 16449
sc@vcc.com wrote: > > When ... you have a tera gate system with giga bytes > of memory diffused through out, ... it will be too hard > to design a CPU make it testable and get it into production. > It will take 20 yr just to do the test vectors. Between symmetric multiprocessing and distributed computing, we shouldn't need great processors to make great systems. If the network is to be the computer. Everything I hear about scalability regards more CPUs, not fancier ones. IIRC, even Semour Cray finally chose x86s. > Everything will be reconfigurable and advances in computing > will come at the reconfigurable cell/architecture level. And the motivation for such a paradigm shift would be? Speed? It certainly can't be for testability. If anything, a reconfigurable machine is going to be the hardest to test. You mentioned test vectors. Doesn't a reconfigurable machine require more test patterns than a fixed machine of identical transistor count? --== Sent via Deja.com http://www.deja.com/ ==-- ---Share what you know. Learn what you don't.---
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