Custom Search
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
HG wrote: > > Neurotech@t-online.de (Neurotech GmbH) wrote: > > >Has anybody out there already done a completely digital PLL inside an FPGA > >? > >I have to do an XC4028EX and one part of it shall be a programmable PLL > >with > >an input clock of about 300kHz and an output of about 10Mhz. But sometimes > >I need exactly 9.998 MHz, sometimes 10.011 and so on, as precise as > >possible. > > >Any idea ? > >Holger Schuppan > > >Neurotech@T-Online.de > > You might consider an fpga to implement the phase detector , digital > lead lag loop filter and control to an outside direct digital > synthesizer ic like the analog devices types . Those nco's have 32 bit > accumulators ,sine rom and dac included in the ic. They can control > frequency in 1part in 2**32 from the reference. You need to understand > how to compute the coefficients of your lead lag filter . > For the DPLL, he really doesn't need an NCO, as the desired output is simply digital. If you did use an NCO, moving it outside the FPGA offers little or no advantage over constructing one in the FPGA. You can build an NCO inside an FPGA with performance equal to or better than many of the dedicated NCO chips (such as the Harris chip) for a fraction of the cost. I recently did an NCO with a 32 bit phase accumulator, with ability for Amplitude and phase modulation. That little guy has quadrature sinsuoid outputs accurate to 12 bits in magnitude and 7 bits in phase angle and runs at a sample rate over 50 MHz in a Xilinx 4KE-2. Doing it in in a -1 or in an EX, or XV part gets even better performance. The precision can be extended with minimal effect on the performance. I fit two such NCOs, plus a pair of 12x12 bit multipliers and some other logic in a 4025E. You can get the NCO to fit in a 4008 with room to spare. Performance wise, this competes with some of the dedicated chips at a fraction of the cost. (Hint, the key to the performance and compact size is the use of the CORDIC algorithm to convert phase and amplitude to IQ samples). Now for the DPLL, you need to be concerned with jitter in the DPLL. To make the DPLL of any use for most applications, the master clock has to be much larger than the maximum output frequency. This is because as the output frequency approaches the clock frequency you get lots of jitter (due to quantum nature of the clock period). Operating the FPGA at 160 MHz gets just barely acceptable DPLL performance for a 10MHz output, and represents a sizable design challenge. I've done DPLLs in FPGAs, but the output frequencies have mostly been in the sub MHz region and the clocks have been much more than 10x the output frequencies. Unless you are prepared to accept alot of jitter, less precision on the output frequencies or both you would probably be better advised to use an external analog PLL. You can get away with a simple clock lock PLL like the IDT88951 and use a pair of loadable counters in the FPGA to turn it into a programmable frequency synthesizer (I've done this recently using lattice 1032E's to create a frequency synthesizer) -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: 9351
Has anyone else out there evaluated Active-VHDL and compared it to ModelSim? There's been a thread going on about a context-sensitive editor for ModelSim, which is integrated into Active-VHDL, along with a nice testbench shell generator. I've evaluated both products, but not to the extent where I've imported post-route delays, and only with designs under 15K gates. ModelSim didn't seem much faster. From what I can tell Active-VHDL seems to be a much better deal, and costs much less. Are there features of ModelSim that I'm missing that make it a much better tool, worth going out to find another tool for editing and test bench generation? Thanks for any input on these tools. Steve MitchellArticle: 9352
rk: : >don't know why people think you can't do good work with non unix/linux : >systems thomas: : A mere problem of interface. With Linux you can get the interface you want : (it can look like NT if you really want). With NT you have to be happy : with what Bill Gates has made for you. If you dislike the NT interface, : you are out of luck. rk: ok, i think i understand you. i assume you are talking about nt 4.0/win 95 interface. 1. now, i know lots of people who do good work with non unix/linux. interface doesn't seem to stop them. 2. with nt (4.0), you get the mac-like i/f which is not bad. and very nice for sharing between 'puters. the earlier nt, using the win 3.1 i/f was far more klutzy. or you can go to a command line prompt if you want. or use another shell, either gui or text. they give you the text-based interface where you can start any program you want, either text-based or gui, for those who like to type those long command lines or have batch/command/script/whatever files. 3. i think there's a lot of things that unix does that nt/95 do not do. but i don't think it's a mere problem of the interface. and we discussed some of the real differences between the system that are critical for certain applications of the system. 4. yes, i've worked in a unix shop. and i know some unix guys were very, very picky about their interfaces. so much, there was a crowd of them who each designed their own character font. couldn't get a product out, but they had beautiful fonts. 5. i think the real problem w/ linux that the applications aren't there. like i say, "it's the software, stupid." if my software doesn't run on linux, i'm out of business, no matter how nice the i/f is. [posted list earlier]. so i gotta stick w/ 95-nt/pc for most apps. unix for others. linux doesn't seem to help. of course, this is not a technical argument. but it is the reality. thomas: : The "keyboard guys" who do not use the mouse, neither the arrow keys (they : are too far on the right), have real difficulties to use efficiently : mouse-driven interfaces. About the same difficulties as what a Mac user : feels when he sits in front of a DOS prompt. rk: i use the keyboard to navigate through the gui a lot and find it, for many situations, faster than using the mouse. they have a lot of keyboard shortcuts in there. me, i'm more of a keyboard guy. learned to type on an old, manual type writer. result: i break a lot of keyboards. of course, for certain apps, like drawing schematics or doing other graphics, the mouse/gui works better. can't do everything through the classic vt100 i/f. rk: : >and i just don't have the time to become unix wiz. friends i know says : >it takes 1 year of pain to get there. thomas: : It has been true. It still is with some unixes. But not with modern Linux : distributions. Actually, they seem to be simpler to install than Windows. rk: bought big unix box last summer (sun). took quite a while and a couple of gurus to get it configured and going. and still not totally happy. gotta find the right guru for the right part of the machine. aaaarrrrgggghhhh!!!! and i don't have time to learn sysadmin of it. for linux, zero first-hand experience. but have read a lot of reviews saying that installing and configuring was not for the "faint-hearted." perhaps that has changed, recently. installing windows stuff (nt, 95) easy, really. only a tad tough if there's a lot of odd-ball hardware and a lot of crap left over from old win 3.1 kludges. and i have a bunch of systems running that go back many, many years from their initial boot. even the tougher ones aren't that bad (except for my father-in-law's; sweated that one out). thomas: : The reason why the average PC user has Windows and not Linux is that he : got it on his harddisk when he bought his computer, and does not want : to (re)install any OS. rk: does microsoft office run on linux? need that to communicate these days. does aol run on linux? very popular in america. visual basic run on linux? national instruments, do they supply ieee-488 drivers for linux? labview run on linux? matlab run on linux? orcad run on linux? viewlogic run on linux? actel designer run on linux? xilinx tools run on linux (pete)? altera tools run on linux? orca tools run on linux (stu)? synplicity run on linux? examplar run on linux? i think this is a yes, do they still support? aldec run on linux? does it network directly with win 95/nt machines with drag and drop? delphi run on linux? sigma plot run on linux? statecad run on linux? mentor run on linux? they did move *to* unix and c. cadence run on linux? pads-pcb run on linux? ms-dos apps run on linux? (still running lots of apps we wrote - and even link them into '95 programs). eudora run on linux? modeltech run on linux? quicken run on linux? can i play the cd-rom interactive disks for the kids on linux? can i read the cd-rom databases from manufacturers on linux? print shop deluxe run on linux? turbo tax run on linux? big seller this time of year, gotta pay the man. these are mostly pretty common apps. some for general, some for eda, some for engineering. don't know answers, perhaps you can feel us in. don't think it's re-installing the os. not at all. even linux guy (fan) at work says hardly any s/w. like i say, "it's the s/w, stupid." [paraphrased from the pres.] who cares about the os or cpu. it's the apps. -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9353
sorry for the extra post, left out the most important comment! rk rk: : >don't know why people think you can't do good work with non unix/linux : >systems thomas: <snip> : The "keyboard guys" who do not use the mouse, neither the arrow keys (they : are too far on the right), rk: well, congrats. i use lots of s/w, 'puters, os, etc. worked in lots of environments. and it's fun looking at the different systems, comparing them, perhaps jousting a tad. but having to move your fingers, like three inches to the right to get to an arrow key as being just too, well, inconvenient, well, that's the most imaginative argument i've heard yet for the superiority of unix. perhaps a pc guru knows of a short cut to for the arrow keys. thanks! -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------Article: 9354
I would like to know the approximate high output voltage of the Altera EPF8452A with light load, say 1 milliamp or less, using 5V power supply. The data book implies it is about 3.7 volts, but its chart gives no good information about it for light load. Altera tech support knows nothing more than what is in the data book and cannot, or will not, give any more information.Article: 9355
I am looking for ideas why version 8.1 of Altera's MaxPlus II has a delay of about 0.5 to 1 second each time a new window is selected, either automatically by the program or by clicking on a different window. Does anyone else experience this? I have been using version 8.1 at home on my Pentium 233 MMX system, which has a 64 Mb of EDO memory and a large fast disk drive, and version 8.0 at work on a classic Pentium 133 system, which has 32 Mb of FPM memory and a slower disk drive. My home system runs the Altera compiler nearly twice as fast as the one at work, but has these annoying and continual pauses. The system at work does not have the pauses. Altera tech support has been no help, only insisting that I must not have sufficient memory or something else must be running. The pauses happen as soon as the program starts, and I have not compiled anything which takes over about 9 Mb of working memory. Nothing else (except clock) is visibly running. My system at work is connected to a 3COM network, the one at home is not connected to a network (except thru DUN to my Internet provider).Article: 9356
Interesting topic. However, I can't help recalling the old times
when power pc claimed a die size advantage over the pentium chip.
The results proved sillicon size is an important but small part of
the whole equation.
Other aspects to heaviliy consider are process technology, gate density,
software, support, IP, etc.
In article <34fe96e2.1302583@nntp.netcruiser>,
Stuart Clubb <s_clubb@die.spammer.netcomuk.co.uk> wrote:
>On Tue, 03 Mar 1998 12:28:12 -0800, Peter Alfke
><peter.alfke@xilinx.com> wrote:
>
>>Don't believe what Marketing says about its competitor. You can be sure
>>that the facts have been massaged. There are hundreds of ways of
>>manipulating the truth without necessary creating outright lies,
>>although even that happens. ( Would you be so naive to believe what
>>Ford says about Chevrolet, or Toyota about Honda, or McDonald about
>>Burger King or vice versa ? )
>
>Amen to that, but a sticky wicket approaches...
>
>>Be careful when you evaluate "equivalent" devices. One observer's
>>equivalence is another one's big difference.
>
>You could always just cut to the chase and count the number of 4ip
>LUTs. Or maybe the number of registers? Imperfect, but better than the
>marketing "specmanship" that has been going on recently.
>
>Let's try it shall we?....
>
>Xilinx XC4062 - 4608 4ip LUTs, and a total of 5376 registers, of which
>768 are internal. I/O is 384 maximum.
>
>Altera FLEX10K100A - 4992 4ip LUTs, and a total of 5398 registers, of
>which 406 are in the I/O blocks. I/O is 406 maximum.
>
>Similar devices yes?
>
>Courtesy of http://www.arrowsemi.com/
>
>EPF10K100ARC2403 is $165 for 100-499 quantity.
>EPF10K100ARC2401 is $335 for 100-499 quantity.
>
>Courtesy of http://www.marshall.com/
>
>XC4062XL-3HQ240 is $472 for 100+ quantity.
>XC4062XL-09HQ304 is $1218 for 100+ quantity.
>
>Trying smaller (5V this time) devices again such as the 10K20, 4013E,
>and OR2C12A:
>
>EPF10K20RC2084 is $42.50 for 100-499 quantity.
>EPF10K20RC2083 is $63.50 for 100-499 quantity.
>
>OR2C12A2S208DB is $51.50 for 26-100 quantity. 100+ not listed.
>OR2C12A4S208DB is $74.00 for 26-100 quantity. 100+ not listed.
>
>XC4013E-4PQ208C is $110 for 100+ quantity.
>XC4013E-1PQ208C is $270 for 100+ quantity.
>
>if I let an XL creep in :
>
>XC4013XL-3PQ208C is $64.60 for 100+ quantity.
>XC4013XL-1PQ208C is $113 for 100+ quantity.
>
>I tried for Spartan :
>
>XCS30-3PQ208C is $57.45 for 100+ quantity.
>
>Oh, and a EPF6016 is $22.95 in 100-599 quantity.
>
>Now, I know there is some speed grade difference, and your mileage may
>vary, but there is, shall we say, a "difference".
>
>>Don't equate chip size with cost. There are many other factors affecting
>>cost, some of them technology-oriented, some not. Cost is what matters,
>>not square microns.
>
>Agree. However, when a company runs on NET margins of 60%+, the
>product must be costing them buttons to manufacture. Perhaps some
>people are just prepared to make a lower margin? I wonder if the
>reverse applies where Xilinx is attacking the CPLD marketspace? Now if
>only people published revenue by density and product family...
>
>>Don't compare devices on the basis of today's single-quantity price, but
>>don't blindly accept high-volume futures either.
>
>Agree, although 100 up generally gives a reasonable finger in the air
>in distribution land.
>
>Stuart
Article: 9357In article <6dkmus$hff@engnews1.Eng.Sun.COM>, Robert Walsh <rjwalsh@Eng.Sun.COM> wrote: > 1) How is Linux's MP support these days? Linux supports up to 16 cpus (and this may be trivally enhanced -- some people made it run on a 1024-cpus computer). There are a few things I would like to clarify: -- "Unix" is a trademark from AT&T and denotes an Operating System they wrote -- many variants of this OS have been written; those OS may be called "unix" because the system calls are much similar to the "Unix" ones. The small differences between Unix flavours are a bit annoying but there is some effort of standardisation (posix); moreover, there exists the GNU autoconf package which allows to build architecture-independent code. Some code written with this free package is much likely to compile cleanly on many known unixes (SunOS 4, Solaris, OSF, Ultrix, Linux, FreeBSD, Irix, HP/UX,...) -- Linux is one flavour of unix, developped primarily on the PC architecture; but, nowadays, it runs also on DEC Alpha, Sun sparcstation, Sun ultrasparc, PowerMac, Atari, Amiga, Mips, ARM. The PC and the Alpha ports are as efficient as commercial unixes, sometimes better. The Sun ports will soon achieve this state, and are already quite usable. -- Windows NT is some variant of VMS (which is a unix-cousin) with the Microsoft interface; it will run on several architectures, sometimes quite inefficiently (on the Alpha, which is 64 bits, the high 32 bits of the registers are ignored, and clobbered -- this is how Microsoft turns a 64-bits machine into a 32-bits one). -- Linux will provide the same functionalities as NT, but with less resources (mainly memory -- NT eats much ram). It handles SMP much better than NT (with NT, all inputs/ouputs go through the first cpu, which is a bottleneck with more than 2 cpus). -- A code developped on Linux is likely to be easily ported to many architectures, whereas NT is a closed architecture. To sum up: Linux runs better than NT on the same hardware, and code developped for Linux will be easily ported to other architectures; and Linux is cheaper. The NT interface has been for long a bit more user-friendly than what you could find on Linux, but this is not true anymore (think about KDE for instance), and Linux still has more flexibility (if you do not like the interface, there are many others to try). So the pity is that companies do not sell Linux binaries of their tools (well, some do), although many develop on some unix (sometimes Linux: for instance, Quake has been developped on Linux and has been ported to DOS). --Thomas PorninArticle: 9358
In article <01bd4973$5a8d0560$aa85accf@homepc>, rk <stellare@erols.comNOSPAM> wrote: >does microsoft office run on linux? need that to communicate these days. There is StarOffice, that understand the Microsoft "standards" and is free for personnal use. Works pretty well. Really looks like Office. >does aol run on linux? very popular in america. Dunno. I know several people connected at home with a modem and a Linux-only machine, so this might be possible. >visual basic run on linux? You can always try Wine (the windows emulator) but this is sort of nonsense (apart from the fact that Wine cannot crash the machine). There are many free compilers for unixes, so it is probable that someone made a VB version. >national instruments, do they supply ieee-488 drivers for linux? I know a physicist who drives some experiments with a Linux-PC and IEEE cards. I'll have to ask him about this version. I don't have time to comment upon the others. Some (especially Xilinx tools) do not have a Linux version, which is explanable by market analysis but not technical reasons. The whole point is that if all vendors made Linux versions, everybody would be happier: the vendors who would have an extended market for cheap (good software should be easy to port) and the clients, who could benefit from the power and stability of Linux. Oh well, maybe Bill Gates would not be so happy (niark niark niark). >like i say, "it's the s/w, stupid." [paraphrased from the pres.] who >cares about the os or cpu. it's the apps. Actually, considering the performances and the interface I like, the OS has some importance. I am also not very rich, and free software is much cheaper than commercial software. --Thomas PorninArticle: 9359
One cannot LOC an IPAD that is connected to BUFGLS to any but a corner pad, unless a IBUF is connected into the LOCed IOB. However when placing IBUFs on BUFGLS nets MAP warns of delay problems. So best is to place the pad in the corner. Have FUN!!! Nick Jacob W Janovetz wrote: > Hello, > > I read in the Xilinx documentation that the 4000XL series allows > mapping an internal signal to an available BUFGLS. > (page 4-43 9/96 databook). I have not been able to do this, however. > In Leonardo, I have tried assigning a BUFGLS to the signal and > even tried mapping a component in. However, when it comes time for > the Xilinx tools to place and route, I get the following error: > > -------------------------------------------------- > Processing global clock buffers... > ERROR:baste:263 - The LOC constraint "P175" (a IOB location) is not valid for > IPAD symbol `ix2282' (pad signal=dsptx_wr), which is being mapped to the > following site types: > CLKIOB > -------------------------------------------------- > > Can anyone lend some advice? Thanks! > > Cheers, > Jake > -- > janovetz@uiuc.edu | Once you have flown, you will walk the earth with > University of Illinois | your eyes turned skyward, for there you have been, > | there you long to return. -- da Vinci > PP-ASEL | http://www.ews.uiuc.edu/~janovetz/index.htmlArticle: 9360
William D. Billowitch wrote: > > Maybe its time to take this personal discussion between > the 3 folks off-line and out of the VERILOG, VHDL, Synthesis, FPGA > etc groups its being SPAMMED to... > > This has little to do with the subject of the groups and the cross-posting > is nothing more than SPAM. > > Anyone agree? > > -- > Bill Sorry for starting one of the longest threads these newsgroups have seen in a long time (although admitedly it has been domninated by this rk fellow). Just goes to show that there is a lot of interest in the subject. All of the groups posted to seem to be relevent to the discussion as they all cover parts of the EDA field. I noticed the following in an article at: http://news.com/News/Item/0,4,19783,00.html?latest "Vox populi. The people have spoken," said Aart de Geus, president and CEO of Synopsys, in trumpeting his company's move to NT and Intel. I don't think the people have spoken clamoring for Synopsys tools on NT, it looks more like the $$ have spoken. The article goes on to say that Micro$oft subsidizes - in some cases - the porting of Unix apps to NT. The people are speaking here in this thread. Many of the people on the thread worry about a future dominated by NT. Getting back to the original intent of the thread: How can we go about effectively promoting Linux (or even FreeBSD)as a viable EDA platform? The EDA companies are driven by $$$ (as any business will be - although there are other factors that can motivate business decisions besides money) so how do we at the grassroots come up with the $$$? We've got to learn to speak the business language, though we don't like it, and come up with a way to "market" Linux as an EDA platform. BTW: I noticed that the "Industry Gadfly" column by John Cooley in this week's EETimes was basically asking the question of why move EDA to NT. Perhaps Mr. Cooley could be a spokesman for EDA on Linux. :-) BTW2: FreeBSD is a good, solid OS, but we need to concentrate our efforts in one area. Since the number of Linux users is probably 10X the number of FreeBSD users, we get more clout with Linux. Anyway, this is a lot of fun ;-) jdArticle: 9361
rk:
: >does microsoft office run on linux? need that to communicate these
days.
thomas:
: There is StarOffice, that understand the Microsoft "standards" and is
free
: for personnal use. Works pretty well. Really looks like Office.
rk:
ok, but does it understand for reading and writing *all* of the latest file
formats, not just the standards to be 'office compatible.' one can argue
whether we should be using office, pagemaker, tex, latex, or whatever.
but, micro$oft has one, and to communicate effectively with others and do
joint writing, it *is* the standard. also, i can take the output of my cae
tools ('office compatible') and drop them into the microsoft tools for
communication or publishing.
rk:
: >does aol run on linux? very popular in america.
thomas:
: Dunno. I know several people connected at home with a modem and a
Linux-only
: machine, so this might be possible.
rk:
part of the discussion, clipped, had to do with why people would use one os
over the other. a lot of people want to use aol. to get on aol, the most
popular service, you need to run aol code. very popular amongst cae
professionals that i know that travel a lot, so they can communicate
anywhere in the country for email and transmitting cae files. you can
always find a local number, plug into the telephone network, and you're
off.
rk:
: >visual basic run on linux?
thomas:
: You can always try Wine (the windows emulator) but this is sort of
nonsense
: (apart from the fact that Wine cannot crash the machine). There are many
: free compilers for unixes, so it is probable that someone made a VB
version.
rk:
this is important as there is a lot of investment in visual basic code,
amongst engineers. not me (delphi fan), but i know a lot of people who do
it. very important application.
rk:
: >national instruments, do they supply ieee-488 drivers for linux?
thomas:
: I know a physicist who drives some experiments with a Linux-PC and IEEE
: cards. I'll have to ask him about this version.
rk:
the national cards, at least in america, are most probably the most
popular, they are fairly 'standard,' and give us code portability from dos
to windows running on itty bitty 286 machines to pentium ii's, and from
desktops to rackmounts to laptops and even have versions of their drivers
which allow one to communicate with the target device over the net. they
put a lot of effort into maintaining their api to the user doesn't have to
redo their code. that represents a substantial investment in software for
them and for us users. and yes, machines that i use for cae i also use for
laboratory use.
thomas:
: I don't have time to comment upon the others. Some (especially Xilinx
tools)
: do not have a Linux version, which is explanable by market analysis but
: not technical reasons.
rk:
what succeeds in the market is not determined solely by technical
considerations, which is the point of this whole argument. we can discuss
os/2 vs. win 3.1. microchannel vs. isa vs. eisa. c vs. pascal. etc. etc.
etc.
thomas:
: The whole point is that if all vendors made Linux
: versions, everybody would be happier: the vendors who would have an
: extended market for cheap (good software should be easy to port) and
: the clients, who could benefit from the power and stability of Linux.
rk:
that's a hell of a big IF. but there's a lot of legacy software out there.
and i'm still running dos apps that i wrote years and years ago. people
sometimes bug me to upgrade them into 32-bit programs for '95/nt. i say
no, runs fine, want to *use* those programs and put s/w development effort
into new applications. sometimes better doesn't win.
thomas:
: Oh well, maybe Bill Gates would not be so happy (niark niark niark).
rk:
don't underestimate that <fill in choice word here>. i'm sure he'd find a
way to make more billions off of it, selling more upgrades, convertors,
bleah, bleah, bleah. he'd use it to avoid the justice department and any
attempt to break his company up for being a monopoly. look what he's doing
to netscape. first he got them to give away their product for free. now
he has them giving away their source code for free. what do they give away
next to attempt to stay in business <good taste prevents me from filling
that in here>.
rk:
: >like i say, "it's the s/w, stupid." [paraphrased from the pres.] who
: >cares about the os or cpu. it's the apps.
thomas:
: Actually, considering the performances and the interface I like,
: the OS has some importance. I am also not very rich, and free software
: is much cheaper than commercial software.
rk:
performance and interface is secondary. like i said, you can have the best
os, the nicest interface <one where the truly lazy don't have to move their
hands over to the arrow keys, a windows complaint i have recently heard>
and have a worthless machine.
if the applications aren't there, no work gets done. go ask ibm about
os/2. ask apollo about their os? well, you can ask some apollo users,
still running the machines, keeping them for their applications, locked in.
Article: 9362rk wrote: > > rick kwan: > : However, even in small networks, if you are or have a systems > : wiz around, automation on a UNIX/Linux network comes very > : cheaply. > > rk: > ah, the wiz, gotta have one for unix, that i know (and i hear the same > about linux). me, not a unix wiz, key tapper, know how to yell for help! > but for pc/win, any idiot can setup and maintain a decent, functional > networked environment. don't even need a network, can just string the > 'puters together with some cable, t's and 50 ohm resistors. that's what i > like. see earlier note (prob. in previous post) about controlling one's > destiny. important when it's you *ss on the line. and i just don't have > the time to become unix wiz. friends i know says it takes 1 year of pain > to get there. I knew I was exceedingly brief. (I am guilty of enough clutter elsewhere already.) I said "wiz" because I couldn't say "sys admin" which nowadays has too little knowledge, or "guru" which implies too much knowledge. Perhaps the best desription is a UNIX tools and shell script person. Most UNIX CAD software engineers qualify. A year for non-computer scientists might be true. I would expect a few months for folks who were CS majors with or (like me) came off the mainframes, neither group having previous UNIX experience. Pain ?? The first few weeks for me were rather painful, but it was incredibly intriguing, which got me thru. (You learn compiler theory and regular expressions in school; but here, you have real tools to build them!) --Rick KwanArticle: 9363
This thread is still going strong ??!! (I know... I'm probably doing my fair share to enflame things.) I was apparantly unclear about something, so I'll just elaborate on that. (Unfortunately, "elaborate" might be putting it lightly.) Peter wrote: > > >I am primarily a UNIX developer; so I have had to rely on NT experts to > >get me past some wicked problems. They made an interesting observation: > >if you develop your application within the strict confines of the > >"Microsoft path," you should have no problem; but if you try to do > >something really innovative that they haven't thought of, you are going > >to get into lots of trouble (which apparantly was why my problems were > >so wicked :-). > > Not sure what this means in specifics. All apps end up calling the > API, one way or another. Maybe some version of unix does more rigorous > checking on the entry parameters than NT, so a dodgy app is less > likely to crash under it. This "Microsoft path" is a bit hard to explain because no one has ever had to explain it to me. But I've heard roughly the same thing from independent sources in the Windows world. (It seems that no one ever questions this until they've had a myriad of problems, so that by the time such a phrase is used the context forces it to make perfect sense.) In effect, you need to develop software the way Microsoft wants you to develop it. If you don't, you risk major schedule delays and even the experts won't be able to dig you out. For example, if you try to write an event-driven program, you really ought to allocate a window. It doesn't matter that a window is immaterial to the problem at hand; you need one in order to handle asynchronous I/O events. They allow you to make the window invisible, but the kernel must allocate that resource for you. If you choose the traditional UNIX select() mechanism, it might work, depending on what you're doing. In UNIX, select() works on any I/O file descriptor; in WIN32, it only works on network sockets. To read the console or handle local files, you need to find other mechanisms; but now you have to write code to handle three classes of mechanism instead of a unified model for all. Strangely, the only people who seem to have a problem with this are UNIX developers who insist on concept unification whenever possible, and breaking things into orthogonal sets otherwise. But if you're developing this way on a Microsoft platform, you're probably wasting your time. Microsoft seems to provide a myriad of software development kits (SDKs) to help get apps to market fast. If your app falls into the domain of an SDK, then it will get you to market sooner and shield you from the messy guts you would otherwise have to slave through. (UNIX developers typically slave through messy guts, but the orgonality of features really helps.) If the SDKs don't help, a lot of stuff can be done with Visual Basic. (...so hard-core Windows developers tell me.) If you really have to write C or C++ code, there is Visual C++ with Microsoft Foundation Classes (MFC). If MFC can't help you, then you'd better study WIN32 really hard. When you get to DLLs, jettison your notion of global or shared variables in the C language definition because it won't apply anymore; you will need to migrate to the DLL notions of "import" and "export". (There's a lot more development issues at this level, but I'm about to get tossed off the podium...) And by this time, you finally develop the notion that there is a "Microsoft path" which you ought to be walking... By the way, WIN32 doesn't define the APIs of NT. It is a "personality" which calls NT APIs, which are mostly unpublished, and hence virtually no one outside Microsoft uses. In principle, this allows Windows 95 apps to run on NT. I haven't tried to port 95<->NT, so I can't speak to this. ....whew... --Rick KwanArticle: 9364
In article <01bd4973$5a8d0560$aa85accf@homepc>, rk <stellare@erols.comNOSPAM> wrote: >rk: >: >don't know why people think you can't do good work with non unix/linux >: >systems Perhaps some people can't understand how you can get any work done at all on Win95 (and to a lesser extend with NT). Win95 and NT lock you into a graphical paradigm for tools that is difficult to automate. Let's say I'd like to run several simulations over night and keep track of potential failures in each. This is easy in Unix/Linux. Even if a tool has a GUI, it always has a way to execute from the command line - this makes it very easy to put together a perl script to automate the overnight simulation runs. In the Win world, most apps don't have a way to execute from the command line - so you've got to stay around all night and push the buttons if you want to batch up several runs. >rk: >does microsoft office run on linux? need that to communicate these days. Who needs it? There are other ways to 'communicate'. On Linux there is StarOffice (can read and write M$Word doc format), Applix (can also read and write M$Word format as well as others), WordPerfect, LyX (a GUI frontend for LaTeX - great for equations, MSWord can't touch it). >does aol run on linux? very popular in america. AOL is a joke. Who needs it, get a real ISP ;-) >visual basic run on linux? Again, who needs it? Especially with the existence of the following on Linux: Perl (and Perl/TK) TCL/Tk Python (and many different graphical extentions) Eiffle Scheme/Tk Java ... others too numerous to mention... Given the availability of these languages, who in their right mind would us VizBasic >national instruments, do they supply ieee-488 drivers for linux? I donno... but I'll bet that someone has made a driver for Linux somewhere (just may not be National Inst) >labview run on linux? >matlab run on linux? I believe the answer is yes here (but not too sure). If not there are several free equivilents that are available (Octave comes to mind). Mathmatica runs on Linux. >orcad run on linux? Now to address the EDA apps together: No, most of them don't run under Linux and that's the point of this thread. Many of us in the design community would like to see some of these apps on Linux. The fact that they are not here now does not point to some technical deficiency in the Linux OS. It merely points to the fact that a large, well financed company with monopolistic tendencies has been able to convice the EDA companies that its OS will ultimately rule the universe. The OSs from Redmond are not in the position they are in because of any technical merit, but because of the marketing muscle (and strong arm tactics) of Micro$oft. >viewlogic run on linux? >actel designer run on linux? >xilinx tools run on linux (pete)? >altera tools run on linux? >orca tools run on linux (stu)? >synplicity run on linux? >examplar run on linux? i think this is a yes, do they still support? Answer was yes here, not too sure now. >aldec run on linux? >does it network directly with win 95/nt machines with drag and drop? >delphi run on linux? See the languages discussion above. >sigma plot run on linux? gnuplot (and its X frontend). Several plotting programs are available (most free). >statecad run on linux? >mentor run on linux? they did move *to* unix and c. >cadence run on linux? >pads-pcb run on linux? >ms-dos apps run on linux? (still running lots of apps we wrote - and even > link them into '95 programs). Yep. I hear MS-DOG apps run quite well under the DOS emulator (I forget the name of it now). >eudora run on linux? There are equivilents. >modeltech run on linux? Addressed above. (actually, I've heard reports that ModelT runs under the WINE windows emulator under Linux). Now you're back to non EDA stuff.... >quicken run on linux? No, but there is the free Check Book Ballancer (CBB) that is quicken compatible. Also, there is the GnoMoney (yeah, interesting name ;) project that seeks to produce a suite of financial tools for Linux (all free under GPL). The point here is that Linux users are resourceful. If they have to, they'll build it themselves. >can i play the cd-rom interactive disks for the kids on linux? What's this got to do with getting your design done? >can i read the cd-rom databases from manufacturers on linux? >print shop deluxe run on linux? What's this got to do with getting your design done? >turbo tax run on linux? big seller this time of year, gotta pay the man. Gotta admit, this is the only reason I keep OS/2 around - so I can run tax software under its Win3.1. But again.... What's this got to do with getting your design done? > >these are mostly pretty common apps. some for general, some for eda, some >for engineering. don't know answers, perhaps you can feel us in. > Well, I'm not that kind of guy. ;-) >don't think it's re-installing the os. not at all. even linux guy (fan) >at work says hardly any s/w. He doesn't get around much then, does he? I run only Linux at home (except, my wife's laptop runs OS/2 so we can do taxes) and find no lack of apps to run. And more are being produced every day. > >like i say, "it's the s/w, stupid." [paraphrased from the pres.] who >cares about the os or cpu. it's the apps. > Yeah, and its the stupid software too ;-) But seriously, the OS does have an impact in how you get your work done. If the GUI gets in the way (as it does with WinXX) it can be a hinderance. philArticle: 9365
Also note that Lucent Technologies for their Orca Product line has as series of crossbar switch macros available. 8x8, 8x16, and many others.. check out their web site at www.lucent.com/fpga TonyArticle: 9366
In article <6drvss$gbr$1@user1.teleport.com>, Phil Ptkwt Kristin <ptkwt@user1.teleport.com> wrote: >Yep. I hear MS-DOG apps run quite well under the DOS emulator (I forget >the name of it now). It is called dosemu. Last time I checked, it could run smoothly Warcraft II (this 50 MB ftp is slow ? no problem, let's play a bit while waiting). If it can run that sort of games (which are rather big and require a very precise emulation) then it should have no problem in running some work applications. It could even run the os/2 version of win 3.1 (not quite well but it could). So some design tool that is more cpu-intensive than OS-intensive should perform well under dosemu. --Thomas PorninArticle: 9367
In a previous article "rk" <stellare@erols.com.NOSPAM> writes: ; :rk: ;: :solution: sit in front of the machine. : ; :wen: ;: Sorry, I can't do that. The machine I sometimes need to use would have :: so many processors, disks, and memory and powersupply that it would take ;: a forklift to carry. I also routinely use manage jobs that runs on a :: farm of workstations and I need to do it remotely and service them at ;: odd hours of the day. : ;rk: :for an application like that, at the high end, you need the high end ;horsepower. that's not a good spot for NT. but that has nothing to do :with suitability of nt for eda, as peter pointed out. it only has to do ;with the suitability of nt for a particular class of job. don't think :there's any disagreement there. and i've bought some big machines for big ;jobs (and spent big $). But that is the direction of EDA future. Larger chips more simulation and analysis, and more difficult place and route. I already have jobs that takes 24 hours per random seed on the fastest pc available. :wen: ;: The problem is not with NT, the problem is with how Microsoft is having :: an effect on how EDA programs were written. There is absolutely no ;reason :: why EDA vendors couldn't compile their program on NT but link them with ;: X-window libraries so that the NT machine would be useful as a remote :: compute engine within a network of unix workstations. NT still has a ;: long way to go to match Unix for networking and user interface. : ;rk: :yup, microsoft monopoly, which is not good [see my earlier post on ;competition and open standards]. and, i believe, microsoft does care about :you; well about your $, anyways. for some networking features, the ;microsoft stuff is pretty good (and win 3.11 was horrible); on others it's :a pain and not up to unix or another of other os'es. their user interface ;is not bad, very mac-like, and it's easy to write your own programs with :nice, easy to use and understand humanoid interfaces. having programmed ;both, i'd go with the win '95/nt interface. Humanoid interface? You should know by now that engineers are more machine than human. :-)Article: 9368
We would like to know where we could get 100 MHz A/D's?Article: 9369
Phil Ptkwt Kristin wrote: > Win95 and NT lock you into a graphical paradigm for tools that is > difficult to automate. Let's say I'd like to run several simulations over > night and keep track of potential failures in each. Bah! NT runs make just fine. Even Altera's amazingly integrated FPGA environment is happy being called from a command line. Same for Modeltec. I imagine that if I want something heavier than I can build in a makefile, I can drop to perl. So far no need, and no desire, to learn perl... Tools that won't let me run them from a command line are broken, no matter what OS they (don't) run under. (I note that I live in a VHDL -> FPGA environment. I'm aware it's a limited case) Anyway, that's not the point. Surely it's the case that people prepared to pay $$$ for an OS (Solaris / NT / Whatever) are more likely to pay $$$ for tools than people who won't. I know which market I'd head for if I had a limited development budget... As for basing my choice of synthesis / simulation tools on which platform rund Word / VB / whatever, heck, that's what the sub-$K PC / portable was put on this earth for. Don't want to be polluting my 'real' machine with cycle-hungry dancing paperclips now, do I? Of cource, I could roll out the panacea that is java, and we can all run the same tools. After all, cycles are cheap *8-) Steve.Article: 9370
on my comment on 10K100 v 4062 As I said, your mileage may vary, and one architecture is certainly better than the other for specific applications. Strangely enough though, I saw a lot of the video signal processing community in the UK using big Altera parts, primarily because they were considered cheap. Of course, many users would hand craft pipelined systems to get the speed when needed, so at 9, 13.5, and 27MHz, that's probably indicative of some of the "issues" StuartArticle: 9371
On Fri, 06 Mar 1998 22:51:23 GMT, s_clubb@die.spammer.netcomuk.co.uk (Stuart Clubb) wrote: >Buffer_sig BUFGLS clk sorry, that should have been BUFG_INT, I think. >pad IBUF clk StuartArticle: 9372
I reckon 100+ pricing is the best comparative guide. "Volume" pricing is hugely dependent on a) who the customer is and b) whether the vendor has specific strategic ambitions with regard to that customer. And none of this happens when purchasing via distribution - you have to be a "direct" account. Hence one often sees totally silly prices in the 10k+ *direct* business. I know someone who was buying XC3030 (10k+) for about $7 - a few years ago when the 100+ price was about $30. This wasn't a big firm (T/O about $5M) but they were using a lot of chips from Xilinx. Peter. Return address is invalid to help stop junk mail. E-mail replies to zX80@digiYserve.com but remove the X and the Y.Article: 9373
In comp.arch.fpga rk <rich.katz@gsfc.nasa.gov.NOSPAM> wrote: : peter: : : : : :There are lots of stupid things in NT but none of them make it in : any : : : : ;way unsuitable for EDA. This conversation has evolved(?) beyond the scope of any of the forums that it's posted to. Please move to a more appropriate venue. I would suggest some advocacy groups.... which I never read. :) MarkArticle: 9374
hi rick, now this *is* getting a bit off topic (we better duck the net police) so just a quick comment or two. rk Rick Kwan <rkwan@ix.netcom.com> wrote in article <35017F2E.2BEB9420@ix.netcom.com>... <snip lots of good s/w stuff> : Strangely, the only people who seem to have a problem with this : are UNIX developers who insist on concept unification whenever : possible, and breaking things into orthogonal sets otherwise. : : But if you're developing this way on a Microsoft platform, : you're probably wasting your time. Microsoft seems to provide : a myriad of software development kits (SDKs) to help get apps : to market fast. If your app falls into the domain of an SDK, : then it will get you to market sooner and shield you from : the messy guts you would otherwise have to slave through. : (UNIX developers typically slave through messy guts, but the : orgonality of features really helps.) : : If the SDKs don't help, a lot of stuff can be done with : Visual Basic. (...so hard-core Windows developers tell me.) rk: <insert plug here for delphi. nice environment, easy to develop for, shields you from most of the windows stuff and rarely need to deal outside of the delphi world. and not bad to link into dll's for things like hp-ib cards. fast code, fast compiles, good amount of oop for most jobs. lots of 3rd party stuff available, too. also, i believe visual basic is now compilable, for better performance, although i haven't used that yet> rick kwan: : By the way, WIN32 doesn't define the APIs of NT. It is : a "personality" which calls NT APIs, which are mostly : unpublished, and hence virtually no one outside Microsoft uses. : In principle, this allows Windows 95 apps to run on NT. : I haven't tried to port 95<->NT, so I can't speak to this. rk: well, we've (day job) been writing a lot of '95 code and we did get one nt machine, to make sure the code runs on both, anticipating a possible migration to nt (and the point of this thread! - so we do get to escape the net police with a technicality). we had to change one "standard" driver that i/f's us directly to the hardware. other than that, no difference, except the code ran faster on the nt/pentium pro 200 than the 95/pentium 166. -------------------------------------------------------------- rk "there's nothing like real data to screw up a great theory" - me (modified from original, slightly more colorful version) --------------------------------------------------------------
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