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
Has anyone created a library for Altera parts that is usable with the Orcad schematic capture package. This would save me a lot of time. I especially need 7000 and 8000 series parts. Thanks! Dave ----------------------------------------------------- Dave Kingma 450 Canterbury St. Sr. Research Engineer Christiansburg, VA 24073 Fiber Optic Products Litton Poly-Scientific 1213 N. Main St. Blacksburg, VA 24060 USA e-mail: dkingma@vt.edu (703)-552-3011 x326 Amateur Radio: WA4RDIArticle: 401
Russell J. Petersen (peterser@utah.et.byu.edu) wrote: : In article <39b348$lrc@cat.cis.Brown.EDU>, mew@lems.brown.edu (Mike E. Wazlowski) writes: : |> : |> : |> --- : |> : |> : |> We're interested in partitioning logic among more than one Xilinx 4000 series : |> FPGA. What's the best way to get a file that has the logic partitioned into : |> function generators? In addition, we'd also like to retain some notion of : |> hierarchy so that we don't put half of a RPM for an adder in one chip and the : |> other half in a different chip. Any ideas? : |> : |> Thanks, : |> Mike : |> : |> : |> : I am afraid that you are on your own on this one. As far as : I know there is no automatic partitioner available for Xilinx chips : although I would like to be proven wrong. The only chips that I know : of that have a partitioner available for them are the Altera parts. : I plan to test this partitioner out soon but have not done so yet. : - Russell Petersen : petersr@fpga.ee.byu.edu Yes, there is automatic partitioner called PRISM, part of the NeoCAD FPGA development tool. Simon.Article: 402
In article <chasebCyw146.ICI@netcom.com> chaseb@netcom.com (Bryan Chase) writes: >Philip Freidin (fliptron@netcom.com) wrote: > ><some helpful info about bitstreams deleted> > >: There is a neat program on the Xilinx BBS called makesrc which uses a >: descriptor file to convert bitstream to source code for any language. >: i.e. it can create 'C' source code (an array initialization) or ASM86 >: source code (loads of .db instructions). then process the file with >: the rest of your program. > >: All the best, >: Philip Freidin > >What's this? It will make source-code out of a bitstream? How much >info do I have to give it along with the bitstream? > >-bryan chase > Makesrc needs a control file that describes the format of your required output. Xilinx supplies control files to generate x86 assembler source, and C source. Note that this is unsupported software. Philip FreidinArticle: 403
Steve: You asked about a test to grade your random number generator. Knuth has a bunch of tests for that in volume 2. After you code and run all the tests, however, you have to decide if your distribution is "random enough" based on the test results. Another technique may be to ask if your RNG is as good as another that everybody already thinks is good. Chapter 13, section 5 of "Numerical Recipes in C" lists several tests for determining if two distributions are different (code provided as well). If your FPGA-based RNG puts out a distribution that is indistinguishable from that of an accepted RNG, then you are doing at least as well as them. -- || Dave Van den Bout || || Xess Corporation ||Article: 404
---------------------------------------------------------------------- Message: 1 From: peterser@utah.et.byu.edu (Russell J. Petersen) Date: 4 Nov 94 11:03:51 Subject: Re: Xilinx chip partitioning In article <39b348$lrc@cat.cis.Brown.EDU>, mew@lems.brown.edu (Mike E. Wazlowski) writes: |> |> |> --- |> |> |> We're interested in partitioning logic among more than one Xilinx 4000 series |> FPGA. What's the best way to get a file that has the logic partitioned into |> function generators? In addition, we'd also like to retain some notion of |> hierarchy so that we don't put half of a RPM for an adder in one chip and the |> other half in a different chip. Any ideas? |> |> Thanks, |> Mike |> |> |> Quickturn Design Systems has been doing multi-chip partitioning (into anywhere from 2 chips to thousands of Xilinx chips) with correct timing across FPGA chip boundaries for many years with proven effectiveness. Hierarchy control, special clock extraction and mapping, and memory mapping are few examples of what it can do. Quickturn Design Systems 440 Clyde Ave., Mountain View, CA 415-967-3300 ------------------------------------------------------------------------Article: 405
Hello, I am currently attempting to put together a design to be implemented in a Xilinx 4010. I would like to use the the wide edge decoders, to reduce the clb count of my design. As I understand it, there are only four decoders available per side. I have 6 control signals, which I would like to decode to a minimum of 6 signals if possible. There are a total of 34 decoded signals thus far, however only 6 of them are critical. Does anyone know of any tricks to getting around this. I imagine one would be to duplicate a set of input pins on another side of the chip, however any other ideas/suggestions would be greatly appreciated. Thanks Richard Wieler rwieler@ee.umanitoba.ca University of ManitobaArticle: 406
John Forrest (jf@ap.co.umist.ac.uk) wrote: (a few paragraphs deleted ...) : > : >for the first time controll every step of the programming sequence: : > : >lower prog, lower reset, wait for init low, rise reset, wait for init : high, ... etc : Of course for 4000s the programming sequence is different and don't : forget you often have to send an extra byte than seems really necessary. The spec for the 3000 series is an additional 12 clocks MINIMUM, and we've always used 24 Just-In-Case. I'd guess the 4000 series is similar or identical. Sending 3 extra bytes does the trick.Article: 407
In article <395u4p$84t@news.csie.nctu.edu.tw> Ching-Yau Jong (yau@theda18) wrote: >I am now writing a assembly program to configure a FPGA >through AT-bus. But I don't know what the actual configuration >bitstreams are. The .bit file seems to have some unnecessities >just like "3020PC6894/10/3120:49:43demo3020.lca:". >Should I delete it or keep it? >Please help me. Thanks. I gather from the file header fragment that you are trying to download a .BIT file to a Xilinx XC3020 device. My first suggestion is that you read your Xilinx manual carefully, as it describes the bit stream format in more detail than you will need (or, probably, will want) to know. It will state that the bit stream begins with the preamble FF 20. What it probably won't state -- at least I've never seen this in any Xilinx book -- is that all of the stuff in the .BIT file prior to the preamble MUST be discarded and MUST NOT be written into the device. If you don't pay attention to this 4000 series devices will reject the configuration attempt by asserting the INIT# pin. I don't know exactly what 2000 and 3000 series devices do; it's been a long time since I had to program them. But I DO know that all the Xilinx LCA devices require that the .BIT file header be stripped om the manner I describe. In order to give you an idea how to proceed I've included a sanitized fragment of a program I wrote to configure an XC4005 device. Its INIT#, BUSY#, PROG#, and DONE pins were connected to a four-bit parallel port ("PPI_PORT") which had an inverting data path. Data was written to the Xilinx using parallel peripheral slave mode. Note that the MSB of each byte is the first one transmitted and corresponds to bit D0 on the Xilinx. You may use a different arrangement -- for instance, I have programmed XC2064 devices serially (a bit at a time). What's important for your purposes is to find the pattern FF 20 and START PROGRAMMING THE DEVICE STARTING WITH FF. My "NEXT_CHAR" function below in effect reads the next byte from a previously opened bit file and provides one character of lookahead to allow me to do this. If you can just put the bit file in memory that will be easier. Hope this helps, and let me know if you have any other questions. Mike CONFIG_XILINX PROC NEAR ; ------------- Assert the PROG# pin to put devices in configuration mode MOV AL, MASK PROG ;Set bit than asserts PROG# OUT PPI_PORT, AL ;Activate PROG# pin ; ------------- Wait for INIT active, then clear PROG MOV ECX, 7 ;Set 7 usec minimum time limit CHK_INIT_ACTV: IN AL, PPI_PORT ;Check AND AL, MASK INIT ; that INIT# pin CMP AL, MASK INIT ; is active LOOPNE CHK_INIT_ACTV ;loop until actv or cnt expired MOV AH, 0FFH ;set AH=0FFH in case of error JNE XLX_CFG_EXIT ;and exit if such is the case XOR AL, AL ;else deactivate OUT PPI_PORT, AL ; PROG pin ; ------------- Vfy INIT#/BSY#/DONE go inactv within ~3 msec MOV ECX, 2800 ;Set 3 msec minimum time limit CHK_INIT_DONE: IN AL, PPI_PORT ;INIT#/BSY#/DONE inactve? CMP AL, MASK DONE LOOPNE CHK_INIT_DONE ;loop until true or cnt expired MOV AH, 0FFH ;set AH=0FFH in case of error JNE XLX_CFG_EXIT ;and exit if such is the case ; ------------- Scan bit file for the frame header (skip file name & rev info) FIND_HEADER: CALL NEXT_CHAR ;AL=byte at current file posn MOV AH, LOOKAHEAD_BUF ;AH=byte to be read next JS XLX_CFG_EXIT ;exit if none avail (EAX=-1) CMP AX, 020FFH ;else check for FF 20 JE SHORT PROG_NXT_BYTE ;start programming if found JMP SHORT FIND_HEADER ;otherwise advance one byte ; ------------- Write data to dev, wait for BSY# inactive or DONE actv W_NXT_CFG_DATA: CALL NEXT_CHAR ;read one byte of cfg data TEST EAX, EAX ;check for end-of-file (EAX=-1) JS XLX_CFG_EXIT ;exit if such is the case PROG_NXT_BYTE: OUT XILINX_PROG_PORT, AL ;output data to XLX device MOV ECX, 16 ;wait about 16 microseconds CHK_LCA_STATUS: IN AL, PPI_PORT ;read LCA status TEST AL, MASK DONE ;if DONE is active LOOPNZ CHK_LCA_STATUS ;then operation is complete JZ DONE_CHECK_OK ;else if time limit expired TEST AL, MASK BSY OR MASK INIT JZ SHORT W_NXT_CFG_DATA ;verify BSY#/INIT# inactive MOV AH, 0FFH ;set AH=0FFH in case of error JMP SHORT XLX_CFG_EXIT ;and exit if such is the case ; ------------- Clear error flag if device returns DONE active DONE_CHECK_OK: XOR AH, AH ;otherwise clear error flag ; ------------- Set return value & exit XLX_CFG_EXIT: MOVSX EAX, AH ;set EAX=-1 on error RET ;and exit CONFIG_XILINX ENDP -- C. M. Heard VVNET, Inc. voice: (408) 247-9375 4040 Moorpark Ave. Suite 206 fax: (408) 244-3651 San Jose, CA 95117 e-mail: heard@btr.comArticle: 408
In article <39omeh$4re@canopus.cc.umanitoba.ca> rwieler@ee.umanitoba.ca writes: > > >Hello, > >I am currently attempting to put together a design to be implemented >in a Xilinx 4010. I would like to use the the wide edge decoders, >to reduce the clb count of my design. As I understand it, there are >only four decoders available per side. I have 6 control signals, The decoders have 'splitters' in the middle of each side, so theoretically you can have 8 decoders per side of the chip (chips are limited to 4 sides in the current versions, so the max is 32, half length decoders). The association of the decoders is 'strong' with I/O, and 'less strong', but 'weak' with internal logic. >which I would like to decode to a minimum of 6 signals if possible. >There are a total of 34 decoded signals thus far, however only 6 of >them are critical. since you are decoding only 6 control signals, I think you will be able to decode them faster if you just run them into a single CLB. The use of FMAP and HMAP control primitives can help in forcing this packing. This is particularly true if your signal origins are within the chip. If they are coming in from outside, a CLB is still faster: The input delay (in a -5) is 3 nS, and the decoder is 12 ns, total of 15nS. ...OR... The input delay is 3nS, 1.5 to 2.5 nS for routing to a single CLB, then 7 nS thru the CLB, total of 12.5 Since you want 6 of these, this might be difficult (to have 6 CLBs all doing a decode of 6 bits, and all clustered together, and all getting all their inputs in under 2.5 nS. The decoder lines would almost be better (assuming external sources, except their performance is based on signals getting to them directly form adjacent I/O pins. After selecting your input pins (which for optimization sake, will be all on one side of the chip, and restricted to one half of that side), you will be left withthe problem that there are only 4 half length decoders there. So if the signals are external, try this: Decode 4 of them on decoder lines, and do the other 2 with CLBs. > >Does anyone know of any tricks to getting around this. I imagine >one would be to duplicate a set of input pins on another side of the >chip, however any other ideas/suggestions would be greatly appreciated. > Is the above enough tricks?? > >Thanks >Richard Wieler rwieler@ee.umanitoba.ca >University of Manitoba All the best Philip Freidin.Article: 409
In article <CyzK3K.2up@indirect.com> Stan Eker, seeker@indirect.com writes: >: Of course for 4000s the programming sequence is different and don't >: forget you often have to send an extra byte than seems really necessary. > >The spec for the 3000 series is an additional 12 clocks MINIMUM, and we've >always used 24 Just-In-Case. I'd guess the 4000 series is similar or >identical. Sending 3 extra bytes does the trick If you take my advise you send one extra byte *and only one extra byte* for 4000s. After this point all the flags used in the handshaking process (INIT, RDY etc) become invalid DONE should go high at this point. _____________________________________________________________ Dr John Forrest Tel: +44-161-200-3315 Dept of Computation Fax: +44-161-200-3321 UMIST E-mail: jf@ap.co.umist.ac.uk MANCHESTER M60 1QD UKArticle: 410
Seong-Woon Kim (ksw@kiwi.etri.re.kr) asked about toggle rates, I thought that you find the following info interesting: Production Parts: Performance: Information source: ------------------------------------------------------------------------ ---------------------- Xilinx 3195A-2, 484+ FF, not inc I/O 322 MHz Xilinx Flex 8452A-3 336+ FF, not inc I/O 185 MHz Paul Brown - GEC Marconi. ------------------------------------------------------------------------ ---------------------- Note that the only production Flex 8000A part is the 8452 claimed 4000 gates, and is DLM not TLM as often mis-reported. Note also that the A-6 spec is == Non A -3. Altera's Q? 1995 A-2 speed grade for Flex is quoted at ~2/3 performance of todays production 3100 silicon.Article: 411
In article <Cyq5MH.1Ks@indirect.com> seeker@indirect.com (Stan Eker) writes: > short the VCC bus to GND through improperly programmed gates. If it gets > hotter than warm, cut the power INSTANTLY. Don't wait, or it's a goner. I try to use a current-limited bench supply to avoid damage from this sort of problem. If you know how much the rest of your circuit draws, and about how much the FPGA *should* draw, you can set the current limit. Then if there is a problem the supply will go into constant current mode. If the part really has close to a dead short between VCC and ground, the supply voltage will drop dramatically, and you won't smoke the part. EricArticle: 412
In article <39jbik$gv0@yama.mcc.ac.uk> John Forrest <jf@ap.co.umist.ac.uk> writes: > Someone said there is a Xilinx utility but we wrote our own for > converting a program into a C array. This is bound into the program (we I've been thinking about trying to use Xilinx part with a parallel EPROM, and having it hold my CPU in reset until it is configured. The configuration loaded into the Xilinx would be designed to let the CPU access the EPROM for its code. The code (and reset vector) obviously have to be in a part of the EPROM not used by the Xilinx configuration. This would let me make my board completely soft-configurable, particularly if I use EEPROM or Flash instead of EPROM. Has anyone tried this sort of thing? EricArticle: 413
In article <ERIC.94Nov10222023@marble.telebit.com> eric@telebit.com (Eric Smith) writes: >In article <39jbik$gv0@yama.mcc.ac.uk> John Forrest <jf@ap.co.umist.ac.uk> writes: >> Someone said there is a Xilinx utility but we wrote our own for >> converting a program into a C array. This is bound into the program (we > >I've been thinking about trying to use Xilinx part with a parallel EPROM, >and having it hold my CPU in reset until it is configured. The configuration >loaded into the Xilinx would be designed to let the CPU access the EPROM for >its code. The code (and reset vector) obviously have to be in a part of the >EPROM not used by the Xilinx configuration. This would let me make my board >completely soft-configurable, particularly if I use EEPROM or Flash instead of >EPROM. > >Has anyone tried this sort of thing? > >Eric The Xilinx chips have 2 specific configurations just for this type of application: Parallel Master UP, and DOWN mode. Most uPs start fetching their instructions or initial jump vector from either the very top of memory, or the bottom. If your uP starts at the high addresses, use the UP mode which starts fetching config data from address 0, and increments the address. If your uP starts at location 0 (or there abouts), then use the DOWN mode, which starts with addresses at FFFF, and DECREMENTS the address register while fetching configuration data. Some uPs need their reset held high, and some need it held low to sustain a reset state. The xilinx chips support this too: There are two output pins on the Xilinx chips, LDC and HDC which are respectively: Low During Configuration, and High During Configuration. Choose the appropriate one to hold the uP in reset while the FPGA is sucking its configuration out of the EPROM. It is a good idea to use the 8 pins that were used for configuration as the pins to transfer byte data between the uP and the FPGA, as they will already be wired that way. All the best, Philip Freidin.Article: 414
In article <39jbik$gv0@yama.mcc.ac.uk>, John Forrest <jf@ap.co.umist.ac.uk> wrote: >[ ... ] The other problem is that these arrays can grow in size. >I wondered if anyone had tried compression something like >the gzip algorithm. When I worked at Telecommunications Techniques Corporation a chap named Aaron Baranoff developed a program to turn Xilinx rawbit files into Huffman encoding trees. In fact the encoder emitted in-line assembler instructions to directly decompress the trees and program the Xilinx device. I do not know the details as I never used this technology personally. Aaron, if you are listening, would you please say a few words about this? Mike -- C. M. Heard VVNET, Inc. voice: (408) 247-9376 4040 Moorpark Ave. Suite 206 fax: (408) 244-3651 San Jose, CA 95117 e-mail: heard@btr.comArticle: 415
Dave, > Has anyone created a library for Altera parts > that is usable with the Orcad schematic capture > package? This would save me a lot of time. I > especially need 7000 and 8000 series parts. MAX+PLUS II supports most functions in the OrCAD Library Files pldgates.lib and ttl.lib. The file orcad.lmf in the MAX+PLUS II installation directory lists the 342 OrCAD symbols supported automatically. If you wish to use a function that is not mapped in orcad.lmf, you must add a mapping. You can map OrCAD functions to Altera-provided functions or to any design file created with MAX+PLUS II. For more information, search for "OrCAD" in MAX+PLUS II on-line help, and go to the topic "OrCAD Interface Information". Kerry Veenstra kerry@altera.comArticle: 416
Hi all, We're designing a prototype memory subsystem, and would like to plug in a FPGA-compatible Rambus interface. So far, we haven't been able to find any products that exactly match our needs, though Xilinx representatives have hinted that such a device may be forthcoming. Do any of you have any suggestions? (What's available now? What will be available soon? How soon?) Thanks Kelvin Nilsen/Dept. of Computer Science/Iowa State University/Ames, IA 50011 (515) 294-2259 kelvin@cs.iastate.edu uunet!cs.iastate.edu!kelvinArticle: 417
Any leads? All kinds of places have 22V10's but what if I want if I want a GAL26V12, a FlexLOGIC, or maybe an Aletra part? With the exception of Digikey's selection of Xylinx parts, I have come up empty. Where *does* a hobbyist get large PLD's and FPGA's? ---- Eric Edwards: Bang= cg57.esnet.com!wolf359!eric Domain= eric@exile.org Remember the home hobbyist computer: Born 1975, died April 29, 1994Article: 418
November 14 - The Plastic Computer - Porting Hardware to Software During the past few months The PARALLEL Processing Connection has enlarged its agenda so that it now encompasses High Performance Computing in general with the emphasis still on parallel processing. We've talked about run-time hardware reconfigurability as a way to affordably accelerate, in hardware, those tasks that are usually accomplished in software; of course, part of the speed-up comes about because hardware is intrinsically parallel. On November 14th Shey Hakusui of the Harminix Corporation will tell us about Parthenon - the C to gates complier that they have developed under contract to NTT, Japan and the "plastic computer" they are developing in which the entire system is reconfigurable. A discussion of member projects currently underway and other issues of interest to entrepreneurs begins at 7:15PM and the main meeting starts promptly at 7:45PM at Sun Microsystems at 901 San Antonio Road in Palo Alto. This is just off the southbound San Antonio exit of 101. Northbound travelers also exit at San Antonio and take the overpass to the other side of 101. Please be prompt; as usual, we expect a large attendance; don't be left out or left standing. There is a $10 fee for non-members and members will be admitted free. Yearly membership fee is $50; corporate membership fee is $400/year. -- B. Mitchell Loebel parallel@netcom.com Director - Strategic Alliances and Partnering 408 732-9869 PARALLEL Processing ConnectionArticle: 419
The PARALLEL Processing Connection is an entrepreneurial association; we mean to assist our members in spawning very successful new businesses involving high performance computing, in general, with specific emphasis on parallel processing. Our meetings take place on the second Monday of each month at 7:15 PM at Sun Microsystems at 901 South San Antonio Road in Palo Alto, California. Southbound travelers exit 101 at San Antonio ; northbound attendees also exit at San Antonio and take the overpass to the other side of 101. There is an $10 visitor fee for non- members and members ($50 per year) are admitted free. Our phone number is (408) 732-9869 for a recorded message about upcoming meetings; recordings are available for those who can't attend - please inquire. Since the PPC was formed in late 1989 many people have sampled it, found it to be very valuable, and even understand what we're up to. Nonetheless, certain questions persist. Now, as we approach our fifth year of operation, perhaps we can and should clarify some of the issues. For example: Q. What is PPC's raison d'etre? A. The PARALLEL Processing Connection is an entrepreneurial organization intent on facilitating the emergence of new businesses. PPC does not become an active member of any such new entities, ie: is not itself a profit center. Q. The issue of 'why' is perhaps the most perplexing. After all, a $50 annual membership fee is essentially free and how can anything be free in 1994? What's the payoff? For whom? A. That's actually the easiest question of all. Those of us who are active members hope to be a part of new companies that get spun off; the payoff is for all of us -- this is an easy win-win! Since nothing else exists to facilitate hands-on entrepreneurship, we decided to put it together ourselves. Q. How can PPC assist its members? A. PPC is a large technically credible organization. We have close to 100 paid members and a large group of less regular visitors; we mail to approximately 500 engineers and scientists (primarily in Silicon Valley). Major companies need to maintain visibility in the community and connection with it; that makes us an important conduit. PPC's strategy is to trade on that value by collaborating with important companies for the benefit of its members. Thus, as an organization, we have been able to obtain donated hardware, software, and training and we've put together a small development lab for hands-on use of members at our Sunnyvale office. Further, we've been able to negotiate discounts on seminars and hardware/software purchases by members. Most important, alliances such as we described give us an inside opportunity to JOINT VENTURE SITUATIONS. Q. As an attendee, what should I do to enhance my opportunities? A. Participate, participate, participate. Many important industry principals and capital people are in our audience looking for the 'movers'! For further information contact: -- B. Mitchell Loebel parallel@netcom.com Director - Strategic Alliances and Partnering 408 732-9869 PARALLEL Processing ConnectionArticle: 420
eric@telebit.com (Eric Smith) writes: >I've been thinking about trying to use Xilinx part with a parallel EPROM, >and having it hold my CPU in reset until it is configured. The configuration >loaded into the Xilinx would be designed to let the CPU access the EPROM for >its code. The code (and reset vector) obviously have to be in a part of the >EPROM not used by the Xilinx configuration. This would let me make my board >completely soft-configurable, particularly if I use EEPROM or Flash instead of >EPROM. > >Has anyone tried this sort of thing? > >Eric We've recently designed two products which have solved the same problem in two slightly different ways. The first does exactly what you are suggesting: Use the LDC or HDC pin to hold the uP in stasis until the Xilinx (4K) "boots up" from the common ROM in Master Parallel mode. Then the uP awakens and runs normally. This design works fine. The second design allows the uP to be the "boss" of the ROM. It boots up normally and, as part of the booting process, it programs two Xilinx 4K devices. The Xilinx FPGAs are configured in Slave Peripheral Parallel mode, they sit on the uP address and data bus much the same as the first design example, above. However, by allowing the uP to configure the FPGAs (they appear to be memory-mapped ports/devices), the uP can configure and reconfigure the FPGAs at will. We've found this approach to be more flexible. Neither one of these variants is an outstanding technical accomplishment, mind you, as they are more or less straight out of the data book or app note texts. I mention them to confirm that both approaches work well. Bob Elkind, Tektronix TVArticle: 421
The IDA Supercomputing Research Center (SRC) maintains a World Wide Web page for the comp.arch.fpga newsgroup that contains: 1) The group charter; 2) An archive of posted messages with various retrieval mechanisms; 3) A list of upcoming events of interest to the custom computing community; 4) Pointers to organizations engaged in FPGA based computing research; 5) Several other interesting links. This page can be found at: http://www.super.org:8000/FPGA/caf.html Note the new URL: we've finally started running htpd. The old ftp URL should still work, but may go away soon. This page is still very much under construction, so any comments or additions are welcome. We would especially like to solicit contributions to a bibliography and a FAQ list. -jeff ------ Jeffrey Arnold IDA Supercomputing Research Center 17100 Science Dr. Bowie, MD 20715 email: jma@super.orgArticle: 422
The IDA Supercomputing Research Center (SRC) maintains a World Wide Web page for the comp.arch.fpga newsgroup that contains: 1) The group charter; 2) An archive of posted messages with various retrieval mechanisms; 3) A list of upcoming events of interest to the custom computing community; 4) Pointers to organizations engaged in FPGA based computing research; 5) Several other interesting links. This page can be found at: http://www.super.org:8000/FPGA/caf.html Note the new URL: we are finally running htpd. The old 'ftp' URL should still work, but may go away soon. This page is still very much under construction, so any comments or additions are welcome. We would especially like to solicit contributions to a bibliography and a FAQ list. -jeff ------ Jeffrey Arnold IDA Supercomputing Research Center 17100 Science Dr. Bowie, MD 20715 email: jma@super.orgArticle: 423
Eric Edwards (Eric@wolf359.exile.org) wrote: : Any leads? All kinds of places have 22V10's but what if I want if I want : a GAL26V12, a FlexLOGIC, or maybe an Aletra part? With the exception of : Digikey's selection of Xylinx parts, I have come up empty. Where *does* a : hobbyist get large PLD's and FPGA's? Arrow *used* to do catalog sales, but as usual you're limited to the lines they deal with. Minimum was $25. The OLD number I have is 1-800-932-7769. If you're working for a company that buys from normal distribution, just do the same thing I do - make a deal with the purchasing department. I call and place the order, verbal PO #, COD, and it's no problem here. It depends TOTALLY on the purchasing folks, though. If they're against it, you're SOL.Article: 424
An updated list of FPGA-based Computing machines is available via WWW. The URL for this page is: List of FPGA-based Computing Machines http://www.utexas.edu/~guccione/HW_list.html Please note the change in the machine name of the previously posted URL from "bongo.cc" to "www". Both names should work, but the "www" URL is preferred. The list currently contains over 40 research and commercial systems, with a few links to other appropriate Web pages. As always, new systems and corrections to existing entries are welcome. Since this list has become quite large, I will not be posting it directly to this comp.arch.fpga. I can email out text versions, or (if demmand warrants it) find an anonymous FTP site. -- Steve Guccione -- guccione@ccwf.cc.utexas.edu -- 11/14/94
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