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
Does anyone know of a program that will calculate the checksum of a part from the programmer load file (binary and/or EXO format)? Thanks, Dan -- d@l.h.c | If you are a person you can read the vertical a s p o | email address to the left. No spam please. n i m | c d |Article: 7926
After back-annotate the design, you can change pin, logic cells in floorplan editor. The reason why you cannot modify this is you are in the view of last compilation. If you change it to current assignment(you can find this at layout menu in floorplan editor), you can modify it. Thanks D.H. Chung Songsong wrote: > Hi, everyone: > > I need your help. > > I use ALTERA's FPGA to design a controller. The first time, I assign > the > time for some pins, and fit successfully. But the pins are not > distributed as I wanted. When I assign the pins location, I found it > cann't fit or cann't get the specified time. > > I want to compile a design at first, when it success, back_annonate > the > design, but I find if I back_annonate the disign, I cann't modify the > pins and logic cells in Floorpan editor. I think it can work, I don't > know the answer, who can tell me. > > If you have used ALTERA's FPGA, please give me some advices. > > Thanx > > Songsong.Article: 7927
Songsong wrote > > Hi, everyone: > > I need your help. > > I use ALTERA's FPGA to design a controller. The first time, I assign the > time for some pins, and fit successfully. But the pins are not > distributed as I wanted. When I assign the pins location, I found it > cann't fit or cann't get the specified time. > > I want to compile a design at first, when it success, back_annonate the > design, but I find if I back_annonate the disign, I cann't modify the > pins and logic cells in Floorpan editor. I think it can work, I don't > know the answer, who can tell me. > Modifying the pin assignments in the floor plan editor make sure you are in the current assignments mode. Click on the icon which looks like a pencil pointing at the middle of a column of three squares. You cannot edit in the last compilation window. To try and fit a design, first clear your pin and logic cell assignments and then assign the signals whose placement is critical. Let Maxplus do all the work. If this doesn't work you might have look at your design to see if changes will make it fit better into the device. -- DerrickArticle: 7928
Hi all, after some days of RTFM and help files, ATLAs etc. I dare to ask the public: Is it possible to assign high slew rate mode to only some outputs in a design for a MAX9xxx (or FLEX). I only found the global style option for slow/high slew-rate, but in one application it is recommended (as I expected) to assign high slew rate only to some speed critical signals. But I found nowhere how to do so! Thank you, Andreas --------------------------------------------------------------- Andreas Doering Medizinische Universitaet zu Luebeck Institut fuer Technische Informatik Email: doering@iti.mu-luebeck.de ----------------------------------------------------------------Article: 7929
Andreas Doering wrote: > Is it possible to assign high slew rate mode to only some > outputs in a design for a MAX9xxx (or FLEX). > I only found the global style option for slow/high slew-rate, > but in one application it is recommended (as I expected) to > assign high slew rate only to some speed critical signals. > But I found nowhere how to do so! 1) Go to Assign->Logic Options 2) Enter your node's name, or name of the I/O pin. 3) Click Individual Logic Options 4) Select Turbo bit, or deselect Slow Slew Rate. Hope this helps. Regards, Rune BaeverrudArticle: 7930
Hi, I used SPW (signal processing worksystem) from Alta Group, Synopsys, and Xact software to design an 8bit input and 8bit output complex multiplier (4 multipliers and two adders). The final count of CLBS (using high mapping effort) was 416 CLBs on XC4013-4 part. Is there a better way (less CLBS) to design an 8 bit input complex multiplier ? Thanks, Robert Morawski -- Robert Morawski Department of Electrical and Computer Engineering Concordia University - 1455 de Maisonneuve West - Montreal, Que H3G 1M8 Phone: (514) 848-8784 e-mail:robert@ece.concordia.caArticle: 7931
A quick search turned up the following report which might be of some interest: http://infopad.eecs.berkeley.edu/~ian/ee241/proposal.html a quote: > For each complex multiply, for example (A + jB)*(C + jD), we must compute the result (R + jI) where R=AC - BD and I=AD + BC which, > brute force, requires four multiplies, an addition, and a subtraction. Some cleverness may be employed to reduce the number of multiplies > by one by noting that if we define M0=(A + B)*(C + D), M1=AC, and M2=BD, then we may calculate R=M1 - M2 and I=M0 - M1 - M2 > [Wei95][Oklobdzija94]. However, this still requires a fair amount of computation: three multiplies, two additions, and three > subtractions. If you can tolerate some truncation error from only computing the upper 8 product bits and doing only 8-bit adds and subs, this might also save some CLBs. Another recent paper: "Design of low-error fixed-width multiplier for DSP applications" Jou & Kuang, IEE Electronics Letters - 11 Sept./97 V33 #19 This looks at multiplier designs that minimize errors due to only computing uppermost product bits. regards, tom Robert Morawski wrote: > > Hi, I used SPW (signal processing worksystem) from Alta Group, Synopsys, and > Xact software to design an 8bit input and 8bit output complex multiplier > (4 multipliers and two adders). The final count of CLBS (using high mapping > effort) was 416 CLBs on XC4013-4 part. Is there a better way (less CLBS) to > design an 8 bit input complex multiplier ? > > Thanks, > Robert Morawski > > -- > Robert Morawski > Department of Electrical and Computer Engineering > Concordia University - 1455 de Maisonneuve West - Montreal, Que H3G 1M8 > Phone: (514) 848-8784 e-mail:robert@ece.concordia.caArticle: 7932
Hi. In my efforts to build an efficient divider (unsigned for the time being), I have used a method which separates the "y/x" quotient into a product of "1/x" and "y". The "1/x" part is implemented in a ROM (in and FPGA). So the input "x" addresses the ROM and the "1/x" value is outputed and then multiplied with the "y" factor. Neither "x" nor "y" are constants which makes implemtation a little more difficult (and less efficient). My synthesis procedure of the method I used is the following and yields less than acceptable results: 1. Block diagram design in SPW (Alta Group's (Cadence) Signal Processing Worksystem) 2. VHDL code generation from SPW 3. Synthesis using Synopsys 3.4b 4. Place and route using Xilinx Alliance M1.3 and a -1 speed grade part. Any kind of input about this method or another better method for implemeting division (or its sub-operations) would be greatly appreciated. VHDL algorithms for signed or unsigned division can also be useful to my project. Also, any info on where to obtain a good multiplication algorithm would be very time saving. Thanks in advance. -- Nestor Caouras nestor@ece.concordia.ca http://www.ece.concordia.ca/~nestor/addr.html |-------------------------------------------| | Dept. of Electrical and Computer Eng. | | Concordia University | | 1455 de Maisonneuve Blvd (West) | | Montreal, Quebec, Canada H3G 1M8. | | Tel: (514)848-8784 Fax: (514)848-2802 | |-------------------------------------------|Article: 7933
Just wonder how well this works. Xilinx sales reps don't know much about it. Peter. Return address is invalid to help stop junk mail. E-mail replies to z80@digiXYZserve.com but remove the XYZ.Article: 7934
Doug Any chance of posting the code for that 7000 part that you use to load the FLEX 10K ? Regards Steve In article <3458D8BB.6925@innocon.com> gibson@innocon.com writes: > > > > Hi Andrea, > > > > For FLEX 10k, the EPROM is called EPC1 (Altera reference). > > There are no EEPROMS. If you need a re-programmable solution for testing, > > you can use a Bitblaster download cable, that you can deconnect after the > > data has been transfered. This is what I'm using. > > > > Andy Negoi > > > We used one of the smaller 7000 series Altera parts in a socket that > controlled an EEPROM. New code is easily put into the EEPROM and a reset > to the 7000 part would reset the FLEX 10K and load it with the code. > Doug > -- Steve Dewey Steve@s-dewey.demon.co.uk Too boring to have an interesting or witty .sig file.Article: 7935
Partially inspired by Jan Gray's site on Homebrew RISCs in FPGAs, I'd like to start doing FPGA-based hardware design. But I'm not sure where to start. I'm interested in learning about FPGA hardware design, and then exploring processor architecture and micro-architecture. As I see it, I need: 1. Software design tools. I understand the software design tools the least. These tools take my design, in a standard HDL like VHDL or Verilog, or a proprietary HDL (?) and turn them into a bitstream for my FPGA device. I assume these tools also do HDL simulation. I've read parts of Peter Ashenden's _The Desginer's Guide to VHDL_. In general, I understand RTL-level description in VHDL. When programming a FPGA, do I also need to do layout in my design (ie., map components of my design to logical structures within the FPGA)? Are these mappings also specified in VHDL? What proprietary HDLs are out there? What are thier comparative advantages/disadvantages to VHDL/Verilog? Jan Gray mentioned using CNets HDL C++ class library for Xilinx FPGAs. Are there others? It seems most FPGA vendors have design tools. In addition, a number of 3rd party tool vendors exist. How do specific vendor's offerings (say, Xilinx and Altera) compare to 3rd party companies (like Exemplar and Mentor Graphics)? 2. A FPGA programmer After I've developed, simulated and verified a design, I need to download it to an FPGA and run it. It sounds like for testing purposes I can use a hardware programmer driven by my desktop computer to program and debug the actual hardware. The Xilinx XChecker programmer sounds like a good option here, since it can examine the running state of an FPGA. Are there other 3rd party programmers that support XChecker-like features? FPGAs must be configured on power-up (usually by a boostrap ROM), right? Or are most FPGAs configured with non-volatile memory? 3. An FPGA Choosing a device supported by the tools I want to use with the capabilities required by my application seems simple enough. Xilinx 4k series, Altera Flex10k, and Lucent Orca FPGAs seem particularly attractive. There are two things I'm confused about. 1. What do the speed ratings mean? -1, -4, etc? 2. What do the various package types mean? I'll probably wire wrap any larger hardware design I use an FPGA in. Any advice or experiences which could help me get started would be most appreciated. - CottonArticle: 7936
Go to the Xilinx web site (www.xilinx.com) and get some of their data sheets. Same for Altera and Lucent if you're so inclined. This will give you your best starting point, understanding the architecture. You might want to just get an entire Xilinx data book, it explains all the available parts, packaging, tools etc. Austin Franklin darkroom@ix.netcom.comArticle: 7937
Peter Alfke <peter@xilinx.com> wrote in article <34551D1A.B5DB5A29@xilinx.com>... <snip> > Nobody makes pin- and functional compatible substitutes, and it is > unlikely that there will be much second-sourcing in the future. At most, > there might be devices that happen to have their supply pins in the same > location, whatever significance that has. > > Peter Alfke, Xilinx Applications hi pete, i think there is some significance for getting pins to line up and i'll add clock and other special pins to the supply pins. i have had feedback from customers that they like the flexibility to potentially expand capabilities in the future w/out having to make new boards. and it helps keep costs down to the bare minimum since i don't have to select a device w/ reasonable spare resources for changes and i can push things closer to the edge and keep unit costs low. i did this within actel by selecting a a3265dx with the fall back of going to the a1280xl series if i got in trouble on logic resources. or back to a larger dx model if i wanted to add some internal dual-port sram for new features, for example. another approach was taken by sei (space electronics inc.) in san diego. they took some quick logic die and packaged them to fit into 1020 and 1280 foot prints and called them the 1020yp and 1280yp if my memory is correct. this gave users more choice without changing their layout with the faster quick logic devices. lastly, i think this was the motivation by atmel; quickly going through their recent announcements shows that they appear to have made their new fpga family xilinx compatible and not atmel compatible. here's what they say on their www site: The AT40K FPGAs are pin-compatible with Xilinx's XC4000 and XC5200 family, allowing users to upgrade their existing designs to higher speed, lower cost and power, without having to relayout their boards. it might be interesting to see just how compatible these devices are from the pin-compatible perspective. now i do not intend to be in a flame war over this, these are just some observations. ------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) -------------------------------------------------------------Article: 7938
Robert Morawski wrote: > > Hi, I used SPW (signal processing worksystem) from Alta Group, Synopsys, and > Xact software to design an 8bit input and 8bit output complex multiplier > (4 multipliers and two adders). The final count of CLBS (using high mapping > effort) was 416 CLBs on XC4013-4 part. Is there a better way (less CLBS) to > design an 8 bit input complex multiplier ? > > Thanks, > Robert Morawski > > -- > Robert Morawski > Department of Electrical and Computer Engineering > Concordia University - 1455 de Maisonneuve West - Montreal, Que H3G 1M8 > Phone: (514) 848-8784 e-mail:robert@ece.concordia.ca What is the data rate? If the data rate low enough, you can use a 4x clock and do all 4 multiplies using a single multiplier and adder. Also, depending on what the multiplicands are, there may be some short cuts that can be taken there. For instance, if one multiplicand is a unit vector with some arbitrary phase angle, A cordic rotator will run faster and take less logic than using multipliers. If a multiplicand is a limited set of complex constants, then a modified LUT approach may be the best. -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 7939
Cotton Seed wrote: > Partially inspired by Jan Gray's site on Homebrew RISCs in FPGAs, I'd > like to start doing FPGA-based hardware design. But I'm not sure > where to start. check our website at http://www.associatedpro.com/aps look at the VHDL tutorialpages. They use the XILINX FOUNDATION software and the APS X84 FPGA board and step you through VHDL code, synthesis, routing, downloading designs, and even in circuit testing. > > > I'm interested in learning about FPGA hardware design, and then > exploring processor architecture and micro-architecture. > > As I see it, I need: > > 1. Software design tools. I understand the software design tools the > least. These tools take my design, in a standard HDL like VHDL or > Verilog, or a proprietary HDL (?) and turn them into a bitstream for > my FPGA device. I assume these tools also do HDL simulation. Some do some don't. We offer a stand alone Windows VHDL simulatorwhich sells for around 700.00. (That's very low). We also offer synthesis and router options for various FPGA vendors, as well as various levels of debug boards. > > > I've read parts of Peter Ashenden's _The Desginer's Guide to VHDL_. > In general, I understand RTL-level description in VHDL. > > When programming a FPGA, do I also need to do layout in my design > (ie., map components of my design to logical structures within the > FPGA)? Are these mappings also specified in VHDL? No, you don't have to, since the PLACE and ROUTE tools do this for you. Theytake the synthesized netlist (XNF, EDIF etc) and first place and then route your design in the target FPGA. It is always best however to understand the FPGA hardware that you are using. This means that you will have to analyze the basic building blocks (CLBs,PLCs,LUTs,IOBs..etc) which the target technology. You actually don't have to know this, but the appropriate use of that knowledge leads to optimized designs. > > > What proprietary HDLs are out there? What are thier comparative > advantages/disadvantages to VHDL/Verilog? Jan Gray mentioned using > CNets HDL C++ class library for Xilinx FPGAs. Are there others? I haven't tried any C++ stuff, but highly reccomend VHDL. It is more andmore becoming the standard. XILINX Foundation Series for example ships with a VHDL compiler, not a Verilog compiler. Also the new PeakLucent Kits ship with VHDL as does the Altera suite I believe. > > > It seems most FPGA vendors have design tools. In addition, a number > of 3rd party tool vendors exist. How do specific vendor's offerings > (say, Xilinx and Altera) compare to 3rd party companies (like Exemplar > > and Mentor Graphics)? My experiences with all of the above have been pretty positive. I have used Exemplar, Synplicity, Accolade, XILINX Foundation (Metamor), and will soon be receiving SYNOPSIS FPGA Express (as part of my FOUNDATION upgrade). Each has advantages. Exemplar and Synplicity are comporably priced. Exemplar allows for hierarchial synthesis (many seperate synthisized blocks resolved at router), while Synplicity does not. Synplicity, has an advanced diagraming tool which lets you view all the way down into the FPGA architecture components. Other tools like Accolade's have an integrated VHDL Synthesis and VHDL Simulator at lower costs than the EXEMPLAR/SYNPLICITY tools. We sell the Accolade tools under our Synth-All brand and under the PeakLucent kits, as well as XILINX Foundation tools which use the ALDEC tools. > > > 2. A FPGA programmer > > After I've developed, simulated and verified a design, I need to > download it to an FPGA and run it. > > It sounds like for testing purposes I can use a hardware programmer > driven by my desktop computer to program and debug the actual > hardware. The Xilinx XChecker programmer sounds like a good option > here, since it can examine the running state of an FPGA. Are there > other 3rd party programmers that support XChecker-like features? > If you buy our FPGA harware cards (APS-X84 for example) you can downloadthe bit stream right from the PC ISA BUS into the card, and can control the design right in the circuit using the boards facilities to stimulate and read back pins right through the PC IO bus. This allows you write test and control code in C/C++ DELPHI/Visual BASIC etc. and do in circuit tests or develop real time aplication code. We also sell a PC instrument for use with our card which is basically a 100 Mhz POD LOGIC ANALYZER which connects to the PC serial port and runs a logic analysis window on your PC, to allow for debugging right down to the pin. BTW- the PC FPGA cards alos have the capability to use PROMS or XCHECKER CABLES also if desired. But after using the ISA BUS as a downloader I think that you will stay with that option. > FPGAs must be configured on power-up (usually by a boostrap ROM), > right? Or are most FPGAs configured with non-volatile memory? SRAM FPGAs (the most popular) use ROMs or are downloaded from a microcontrolleror PC interface. > > > 3. An FPGA > > Choosing a device supported by the tools I want to use with the > capabilities required by my application seems simple enough. Xilinx > 4k series, Altera Flex10k, and Lucent Orca FPGAs seem particularly > attractive. > I like the XILINX and Lucent parts in particular because of the on boardtristates. > There are two things I'm confused about. > > 1. What do the speed ratings mean? -1, -4, etc? Your guess are as good as ours. The vendors don't stick to any real standard.In XILINX parts for example I always interpreted the -4 to be 4ns combinatorial delay thorough one level of logic. A -1 would be faster at 1ns. But all vendors don't use the same scheme and some are actually slower as the numbers get smaller. > > > 2. What do the various package types mean? I'll probably wire wrap > any larger hardware design I use an FPGA in. THere are various packages from PLCC sockets to surface mount QFP (Quad Flat packs) and now Ball Grid Arrays (MONSTERS) > > > Any advice or experiences which could help me get started would be > most appreciated. > > - Cotton For your first design I would try using the APS-X84 board kits with a VHDL synthesi system. Depending on cost and features the prices can go from 800.00 for our APS-X84-FBV (BASE VHDL) which limits the size of the FPGA to 8K gates to our APS-X84-FSV which goes into the thousands. Or our Peak Lucent kits come with a board which has some extra features (SRAM, 30 Mhz Dual DtoA and Direct Digital Syntheized programmbale clock) plus a VHDL Simulator and Synthesis package. Both systems come with everything you need to get started and are expandable to allow other FPGA vendors to be added on. Let me know more about you specific needs and I'll be happy to help you out in selecting the best system for you. -- __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ Richard Schwarz, President EDA & Engineering Tools Associated Professional Systems (APS) http://www.associatedpro.com 3003 Latrobe Court richard@associatedpro.com Abingdon, Maryland 21009 Phone: 410.569.5897 Fax:410.661.2760 __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/Article: 7940
In article <34577319.DB6@verdon.imag.fr>, Theodor Calin <calin@verdon.imag.fr> wrote: I need to implement some polynomial division algorithms for test >patterns generated >for ISCAS benckmarks. Could anybody suggest me some useful hints and >references for the >case I must do it myself ? #define POLY 0x20400021 /* for instance, x^32+x^29+x^22+x^5+1 */ unsigned crc,k; /* assuming 32-bit machine here... */ k = 32; crc = [[init-value]]; for([[each word]]){ crc ^= [[next-input-word]]; do{ /* one shift of a CRC register */ if( crc & 0x80000000 ){ crc = (crc<<1)^POLY; }else{ crc = (crc<<1); } }while(--k); } Is that any help? You can do it faster with lookup tables.Article: 7941
In article <345A27E7.3FB6@ece.concordia.ca>, Nestor Caouras <nestor@ece.concordia.ca> wrote: >Hi. > > In my efforts to build an efficient divider (unsigned for the time >being), I have used a method which separates the "y/x" quotient into a ... > Any kind of input about this method or another better method for >implemeting division (or its sub-operations) would be greatly >appreciated. VHDL algorithms for signed or unsigned division can also be >useful to my project. > I assume your application can't tolerate the time required for a 'bit-by-bit' long division? can you put several of these in skew-parallel to reduce the effective time? Nestor, it makes a huge difference what kinds of errors you can tolerate. Your 1/x table, unless x is always close to some center value, will become very inaccurate for large x. Have you considered a 'log' approach? use tables to approximate log2(x) and log2(y), and then another table for 2^(y-x). The choice of 2 is important; for instance, 2^z can be done by looking-up the fractional part of z and then barrel shifting by the integer part. With this method the errors are uniform relative to the magnitude of x and y. For log x: rewrite x=2^k*(1+d), where k is an integer such that (sqrt(0.5)-1) < d < (sqrt(2)-1) about -0.293 .. 0.414 'k' can be found by a 'normalize' function which finds the MSB; then compare the normalized result to sqrt(2) and shift again if larger. Now d is in the range above, and 'k' is the integer part of your log. The fractional part (which may be negative) is log2(1+d); Since 1+d is close to 1, this can be approximated efficiently by log2(1+d) = (l/ln(2)) d + lookup[d] where the lookup table doesn't have to be very big (i.e. only use the first few MSB's of d). You may not have to go to such lengths, but this will reduce the table size at the expense of additional logic. The result above (log2(1+d)) will always fall in the range [-0.5,0.5]. Of course you can just normalize so 'd' is in [0..0.999] and lookup the fractional log from there. log2(1+d) will also be in [0..0.9999]. > Also, any info on where to obtain a good multiplication algorithm would >be very time saving. Many VLSI texts cover this. Look for 'Booth Coding' or 'Booth Multiplier'; that will at least bring you to the right place.Article: 7942
WATKINS BIZ-OPP! **Since 1868, Watkins Products have stood for quality, from generation-to-generation! **The Watkins Product Line features 375+ items. Gourmet Food Items, Medicinal, Health & Nutrition, Household, Personal Care, Dietary Supplements, Gift Items and MORE! **Founder J.R. WATKINS, introduced the first ever SATISFACTION GUARANTEE! **WATKINS REPRESENTATIVES can save/ earn 28-61% discounts/profits on the sug- gested retail prices. **WATKINS REPRESENTATIVES have NO monthly sales quotas or required purchase amounts. **WATKINS REPRESENTATIVES are not required to maintain inventories, attend meetings of hype or folklore, deliver product or make collections. **THE WATKINS HOME OFFICE is now set up to process and ship all retail customer and/or representative orders. **YOUR RETAIL CUSTOMERS can order direct via Watkins by calling Toll Free, by Direct Mail and/or via the Internet. Watkins will accept personal checks, money orders, Visa, Discover or Master- Cards. **THE WATKINS BUSINESS OPPORTUNITY was chosen as one of the top eight by MLM INSIDER MAGAZINE! **THE WATKINS BUSINESS OPPORTUNITY was chosen as one of the top ten by SUCCESS MAGAZINE! **THE WATKINS BUSINESS OPPORTUNITY was chosen as the NUMBER ONE BUSI- NESS OPPORTUNITY in the USA and CANADA by WAVE 3. **WATKINS will soon be expanding abroad by opening distribution centers in NEW ZEALAND and AUSTRALIA. **A PRESENT DAY WATKINS BUSINESS is no longer a horse & buggy, door-to- door solicitation mode. You can build an extremely successful WATKINS BUSI- NESS via mailorder or the internet. GET YOUR FREE NO OBLIGATION WATKINS PROSPECT PACK & CATALOG! JUST SEND YOUR USPS SNAIL MAIL ADDR- ESS TO: cauleys@aol.com CAULEY MARKETING GROUP P.O. BOX 10 PANACEA, FLORIDA 32346-0010 Tel No: (850)926-9350 or (850)926-9330 E-mail: cauleys@aol.com cauleys@freeyellow.com VISIT OUR WEBSITE FOR ADDITIONAL INFO! http://MasterNetworker.com/watkins/cauley.htmlArticle: 7943
In article <19971101174200.MAA23866@ladder01.news.aol.com>, cauleys@aol.com (Cauleys) wrote: [nothing of consequence] Oh, piss off. -- Pete FraserArticle: 7944
I think your example application (and its results) are an excelent example of why I am still an old stick-in-the-mud schematics-type-of-guy. Synthesis design promotes architecture neutral design, which unfortunately decouples you from the ability of taking advantage of facilities and advantages of a specific target architecture. With probably little more effort than you need to do your synthesis design, I would have drawn a schematic withe the following symbols from my data path libraries: 4 copies of an 8 by 8 multiplier 2 copies of a 16 bit in, 17 bit out adder. Because the libraries I am using are XC4000 specific (and also work for XC4000E and XC4000EX and XC4000XL and XC4000A and XC4000H and XC4000D (so not too specific)) ther is no ambiguity about how this will compile and the resources it will take. The multipliers are 54 CLBs each, are 4 pipe stage deep, run at 80MHz in the -1 speed grade of the XC4000E ( so probably more like 25 to 30 MHz in the 4013-4 parts you are using) , and are floor planned for minimal area. The adders will be 18 CLBs each, including an output pipeline stage. Total is 252 CLBs. 5 Pipe stages. 25 to 30 MHz clock rate. This actually gives two 17 bit numbers R+I. If you only want 8 bit results, throw away the low order result bits. You may also be able to throw away the low order logic of the adders, depending on your tollerance to errors in the LSB of your remaining results. A good demonstration that design re-use with schematics is easy to do if the libraries are designed correctly. You asked for a better way. I think the above is close to optimal. You could use even less CLBs by time sharing fewer multipliers (minimum of one, time sgared 4 ways), and similarly with the adders. There are diminishing returns down this path, as the extra muxes and registers required for timesharing start to approach the logic they are replacing. The multipliers are a derivative of the Xilinx LogiCore DSP libraries (available from their web site I believe), and the adders come from a library that Fliptronics (my company) will soon be making available. Similar adders are available in the standard Xilinx libraries. Philip Freidin. In article <63d514$r9v$1@newsflash.concordia.ca> robert@ece.concordia.ca (Robert Morawski) writes: >Hi, I used SPW (signal processing worksystem) from Alta Group, Synopsys, and >Xact software to design an 8bit input and 8bit output complex multiplier >(4 multipliers and two adders). The final count of CLBS (using high mapping >effort) was 416 CLBs on XC4013-4 part. Is there a better way (less CLBS) to >design an 8 bit input complex multiplier ? > >Thanks, >Robert Morawski > > >-- >Robert Morawski >Department of Electrical and Computer Engineering >Concordia University - 1455 de Maisonneuve West - Montreal, Que H3G 1M8 >Phone: (514) 848-8784 e-mail:robert@ece.concordia.caArticle: 7945
Philip Freidin wrote in article ... ...snip.. > >With probably little more effort than you need to do your synthesis >design, I would have drawn a schematic withe the following symbols from >my data path libraries: > >4 copies of an 8 by 8 multiplier >2 copies of a 16 bit in, 17 bit out adder. > >Because the libraries I am using are XC4000 specific (and also work for >XC4000E and XC4000EX and XC4000XL and XC4000A and XC4000H and XC4000D (so >not too specific)) ther is no ambiguity about how this will compile and >the resources it will take. > >The multipliers are 54 CLBs each, are 4 pipe stage deep, run at 80MHz in the >-1 speed grade of the XC4000E ( so probably more like 25 to 30 MHz in the >4013-4 parts you are using) , and are floor planned for minimal area. > Using Atmel's AT40K FPGA architecture which is optimised for array multiplier problems such as this, each pipelined 8x8 multiplier will take 72 Atmel AT40K core cells (each half the size of a Xilinx CLB) and will achieve 70MHz+ operation with only one stage of pipelining (-1 worst case commercial). The non-pipelined 8x8 multiplier is 64 cells and 40MHz (-1 w.c. commercial). The AT40K Macro Generators, which ship as standard with our IDS 5.0 s/w, will allow you to generate the multiplier in less than 5 mins (including symbol and schematic) and it will be architecture optimised (hard macro). If you would like more info. check out http://www.atmel.com and follow the AT40K FPGA link on the home page splash. >The adders will be 18 CLBs each, including an output pipeline stage. > >Total is 252 CLBs. >5 Pipe stages. >25 to 30 MHz clock rate. > >This actually gives two 17 bit numbers R+I. If you only want 8 bit >results, throw away the low order result bits. You may also be able to >throw away the low order logic of the adders, depending on your >tollerance to errors in the LSB of your remaining results. > >You asked for a better way. I think the above is close to optimal. >You could use even less CLBs by time sharing fewer multipliers (minimum >of one, time sgared 4 ways), and similarly with the adders. There are >diminishing returns down this path, as the extra muxes and registers >required for timesharing start to approach the logic they are replacing. > We achieve similar results for our adders too.... Martin Mason. (Atmel Corp.) ...snip..Article: 7946
I too am very interested in how these new ATMEL parts will pan out. I have seen the structure in the data sheets. The architecture seems to be smaller than XILINXs CLB with only one FF in the CLB. They do have diagonal routing capabilities between their CLBs which they claim will allow for better utilization in a samller area. They also calim to be coming out with an XNF netlist translater which will translate XNF designs into the new ATMEL designs.(not the easiest of jobs after looking at the two architecure differences) They also have distributed RAM throughout the chip which do not take up LUTs which is a nice feature. I am convinced that the key for ATMEL will be getting low cost tools out and getting the parts accessible. They will have a 900.00 development system and will of course cater to the well financed campanies and probably give away those tools. But I am convinced that getting low cost tools out to the masses is the way for this FPGA family to take off. We are looking closely at the parts and will be looking to see what kind of support ATMEL is going to give to facilitate the acceptance of their parts. I would be interested to hear XILINX's evaluations on the ATMEL parts and architecture. Perhaps Peter will enlighten us as to his thoughts on the parts. I always enjoy reading his informative posts. Richard B. Katz wrote: > Peter Alfke <peter@xilinx.com> wrote in article > <34551D1A.B5DB5A29@xilinx.com>... > > <snip> > > > Nobody makes pin- and functional compatible substitutes, and it is > > unlikely that there will be much second-sourcing in the future. At > most, > > there might be devices that happen to have their supply pins in the > same > > location, whatever significance that has. > > > > Peter Alfke, Xilinx Applications > > hi pete, > > i think there is some significance for getting pins to line up and > i'll add > clock and other special pins to the supply pins. i have had feedback > from > customers that they like the flexibility to potentially expand > capabilities > in the future w/out having to make new boards. and it helps keep > costs > down to the bare minimum since i don't have to select a device w/ > reasonable spare resources for changes and i can push things closer to > the > edge and keep unit costs low. i did this within actel by selecting a > a3265dx with the fall back of going to the a1280xl series if i got in > trouble on logic resources. or back to a larger dx model if i wanted > to > add some internal dual-port sram for new features, for example. > > another approach was taken by sei (space electronics inc.) in san > diego. > they took some quick logic die and packaged them to fit into 1020 and > 1280 foot prints and called them the 1020yp and 1280yp if my memory is > > correct. this gave users more choice without changing their > layout with the faster quick logic devices. > > lastly, i think this was the motivation by atmel; quickly going > through > their recent announcements shows that they appear to have made their > new > fpga family xilinx compatible and not atmel compatible. here's what > they > say > on their www site: > > The AT40K FPGAs are pin-compatible with Xilinx's XC4000 and > XC5200 > family, allowing users to upgrade their existing designs to > higher > speed, lower cost and power, without having to relayout their > boards. > > it might be interesting to see just how compatible these devices are > from > the pin-compatible perspective. now i do not intend to be in a flame > war > over this, these are just some observations. > > ------------------------------------------------------------- > rk > > "there's nothing like real data to screw up a great theory" > - me (modified from original, slightly more colorful version) > ------------------------------------------------------------- -- __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ Richard Schwarz, President EDA & Engineering Tools Associated Professional Systems (APS) http://www.associatedpro.com 3003 Latrobe Court richard@associatedpro.com Abingdon, Maryland 21009 Phone: 410.569.5897 Fax:410.661.2760 __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/Article: 7947
Gregory Smith wrote: > > In article <345A27E7.3FB6@ece.concordia.ca>, Nestor Caouras <nestor@ece.concordia.ca> wrote: > >Hi. > > > > In my efforts to build an efficient divider ... I have some new algorithms for high-radix division that don't use complicated table look-up. The most recent paper is available at http://math.isu.edu/~vkantabu/radix8.pdf, and can be viewed with Adobe Acrobat Reader 3.0 or later.Article: 7948
Ray Andraka wrote: > Mr Barry Tso wrote: > > > > May be the best substitute is the E part from Xilinx :> > > They deliver better performance and lower cost. > > > > Although Atmel claims they have pin compatible part, they > > can't have all the architecture feature from Xilinx. > > The Atmel AT40K is pin compatible in that it can be used in a Xilinx > socket (dedicated pin assignments are the same, xilinx configuration > modes and protocol are supported). The internal architecture is quite > > different however. Which part is the better one really depends on the > > application. Atmel's pricing is very aggressive (I think it is about > $40 for a 20K gate part in small quantities) so it is worth a look, > especially in cost sensitive projects. > > -Ray Andraka, P.E. > President, the Andraka Consulting Group, Inc. > 401/884-7930 Fax 401/884-7950 > email randraka@ids.net > http://users.ids.net/~randraka The concern I have is the flip flop count of the ATMEL FPGAs when compared to the flip flop count of the XILINX 20Kish range is not as high. It does appear that using the FFs in the XILINX CLB, that at least one require the use of an LUT, while all the ATMEL FFs are available without using LUTs. One big plus which ATMEL has is that their distributed RAM is not using their on board LUTs as XILINX does. Also the diagonal routing resources look interesting. How all these facts will pan out in actual implementation will be interesting to see.As it stands now, ATMEL is just sending out the 20K parts. They have tweaked my interest, but lets see what else they do now. -- __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ Richard Schwarz, President EDA & Engineering Tools Associated Professional Systems (APS) http://www.associatedpro.com 3003 Latrobe Court richard@associatedpro.com Abingdon, Maryland 21009 Phone: 410.569.5897 Fax:410.661.2760 __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/ __/Article: 7949
hi, I am having trouble instantiating IO cells (INBUF, OUTBUF and BUFT) in a verilog design with fpga express. FE doesn't seem to know about these 10K IO cells. It doesn't put any of them into the EDIF file and maxplus2 complains about unconnected ports etc. Any ideas how I can convince FE to use these cells ? thanks muzo WDM & NT Kernel Driver Development Consulting <muzok@pacbell.net>
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