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
Each block ram is up to 16 bits wide, at which point is 256 words deep. Is there a reason you can't use 13 block RAMs to get your 200 bit width? Each of course would need an identical address, which depending on your speed requirements might dictate duplicated address and control logic. Another 'trick', which is useful if you only need a single port memory and 128 words is deep enough, is to use both ports to get an effective width of 32 bits. In that case, you tie the upper address bit on one port to '1' and the same address bit on the other port to '0'. The remaining address bits as well as the controls are common, and presto, you've got a single port 32 bitx128 word memory in one block ram. That will get you up to 256 bits wide in an XCV50, without any fancy addressing or double access garbage. I find the COREGen to be more of a hinderance than a help for the BRAMs. It's not like it adds any useful logic, all it does is instantiate the required number of BRAMs. I think it is easier, and frankly, more intuitive to just instantiate the BRAMs directly. Gerhard Griessnig wrote: > > I need to create a RAM in a XILINX-Virtex V300 with the XILINX Foundationtool. > > My problem is that my RAM has a width of 200 bits. > > Can i use the ONBOARD-RAM (only in Virtex-Series? max width is 16?) without a > complex addressing. > > Futhermore i tryed to create a RAM with the XILINX Coregenerator but i got an > error message > > Error L-44/C0 : #0 Error: > D:/programs/xilinx/active/projects/grieda02/ram.vhd line -44 Library > logical name XILINXCORELIB is not mapped to a host directory. (VSS-1071) > (FPGA-dm-hdlc-unknown) > Error L48/C0 : #0 Error: > D:/programs/xilinx/active/projects/grieda02/ram.vhd line 48 No selected > element named C_MEM_SP_BLOCK_V1_0 is defined for this prefix. (VSS-573) > 2 error(s) 0 warning(s) found > > Does anyone has an idea, or does anyone has a vhdlcode ? > > THANKS Gerhard > -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com or http://www.fpga-guru.comArticle: 24976
Hal Murray wrote: > What sort of metastability data is available for antifuze parts? [1] "Metastability Report for FPGAs," Quicklogic Data Book, 1994, pp. 523-526. They show the circuit used, the formulas, and give tables of taw and w. This is for pASIC 1. They also give a graph comparing metastable performance of Quicklogic, Actel, and Xilinx. [2] Actel Data Book, 1992, pp. 5-1 to 5-2. This was how the data was referenced in the Quicklogic report. I remember that in the 1992 book there was a report on Act 1 metastable state performance. Stretching the memory cells, I believe it was on the older versions of this. I think there were two volumes of the book in 1992, one of which was a designer's guide consisting of application notes. I'll dig out the reference if anyone is interested. ================================ > How do ASIC vendors get metastability data? I have data books from Chip Express handy for the QYH500, CX2000, and the CX3000 series. I only found metastability information in the CX2000 book - perhaps the eyes aren't quite working yet. They do supply a nice discussion of the effect and a comprehensive table of values for each of their flip-flops. The text doesn't describe any test circuits or how the values were obtained. Using the search engine on their www site for "metastable," I received 0 hits. ================================ Here's a definition of "fuze" from www.dictionary.com: fuze n : any device by which an explosive charge is ignited [syn: fuse, fusee, fuzee, primer] :-) rkArticle: 24977
As I mentioned earlier, I've run into this too. You can also solve the problem by just setting the TimingChecksOn generic in the unisim clocked logic elements to FALSE when you instantiate the component. I automatically just add the generic in the component declaration and it is done. For example: Architecture mixed_structural_and_RTL of doomsday_machine is -- Component declaration of the "SRL16E(SRL16E_V)" unit -- File name contains "SRL16E" entity: .\src\unisim_VITAL.vhd component SRL16E -- synthesis translate_off generic( TimingChecksOn : BOOLEAN := FALSE); --synthesis translate_on port( D : in std_ulogic; CE : in std_ulogic; CLK : in std_ulogic; A0 : in std_ulogic; A1 : in std_ulogic; A2 : in std_ulogic; A3 : in std_ulogic; Q : out std_ulogic); end component; attribute syn_black_box of SRL16E : component is true; -- Component declaration of the "FDE(FDE_V)" unit -- File name contains "FDE" entity: .\src\unisim_VITAL.vhd component FDE -- synthesis translate_off generic( TimingChecksOn : BOOLEAN := FALSE); --synthesis translate_on port( Q : out std_ulogic; D : in std_ulogic; C : in std_ulogic; CE : in std_ulogic); end component; attribute syn_black_box of FDE : component is true; --other declarations begin U1:SRL16E port map( D=>t_in(i), CE=> ce, A0=>dqa(0), A1=>dqa(1), A2=>dqa(2), A3=>dqa(3), Q=>ff_d, clk=>clk); U2:FDE port map( Q=>t(i), D=>ff_d, C=>clk, CE=> ce); "K. Orthner" wrote: > > Regarding functinoal simluations that fail, but timing simulations that > pass; there may be another problem. > > (I ran into this a while back, and had to scratch my head for quite a while > to figure it out.) > > Often, the libraries that come with a simulator for a given target part > contain minimum timing delays. An example is the "Unisim" library that > comes with the Aldec VHDL simulator, for Xilinx parts. In many cases, it > assumes a "minimum" delay of 0.01 ns, or something. > > If you happen to instantiate some parts, but infer others, you can run into > problems with your functional simulation. > > In my case, I inferred an IBUFG for my clock inputs, but just left the > other inputs as regular inputs, letting the place-and-route software add > the IBUF/OBUFs automatically. When I ran my simulation, my data would > arrive inside the chip 0.01 ns before my clock signals. > > If I did the post-timing simulation, it was okay, since at that time, all > of the inputs used IBUFs with non-zero delays. > > I solved this by recompiling a copy of the unisim library with all of the > timing parameters set to zero, and then using *that* library when > simulating. > > I hope this helps. > > -Kent -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com or http://www.fpga-guru.comArticle: 24978
In <39A29BC5.949D50D1@pacbell.net>, Neil Nelson <n_nelson@pacbell.net> writes: --snip-- ><!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN"> ><html> > <head> > <title>Anonymous Remailers</title> > </head> --snip-- Neil, Please don't post HTML, particularly irrelevant HTML. Thank you. Frank McKenney, McKenney Associates Richmond, Virginia / (804) 320-4887 E-mail: frank_mckenney@mindspring.comArticle: 24979
My Virtex 400 design that placed and routed in 2 - 2.5 hours with 2.1i (on a 350 mhz PC) is now taking 4 - 5 hours with 3.1i (with service pack 2 installed). The routing delays and timing look no better or no worst than 2.1i. Is anyone else seeing these longer run times?Article: 24980
Jon Kirwan wrote: > > On Wed, 23 Aug 2000 01:25:53 -0400, rickman <spamgoeshere4@yahoo.com> > wrote: > >A Background questionnaire release form, > > Release form for exactly what? For a background check by a private investigator! The investigating company is PROFILES PLUS, INC. I have to authorize all "corporations, companies, credit agencies, financial institutions, educational institutions, persons, law enforcement agencies, former employers and the Military services to release all written and verbal information about me to PROFILES PLUS, INC." I also have to release these sources from liability. The agreement survives for "future reports or updates that may be requested". > >Then they want me to leave behind a urine sample for drug testing. > > As a condition of employment, this has gotten to be quite common. > I've been through that several times, although I keep reminding myself > how much I absolutely love poppy seed strudel at about this time. I > keep wondering if the rumors are true... > > But before you have completed negotiations??? No offer? No pee. > Their only argument is, once again, they'd like to save time on you if > you fail. Frankly, I don't care. They will just have to invest a bit > in me and expect the best. I had to do this once before when I interviewed with a DOD contractor. After talking with them about the avenues of recourse in the event that I failed, I realized that 1) there was no recourse, 2) they did not acknowledge that a drug test can be wrong. It can be wrong for a large number of reasons; poppy seeds (yes that is true), prescription drugs, and many "medicinal" herbs sold in health food stores. 3) to make sure that I could not dispute the result of the test it was given at the interview so that if it was failed they would simply not make an offer. You would never know that you failed the drug test. > >And more importantly, I will not provide such an authorization to use > >this information without an opportunity for me to dispute any erroneous > >information contained in these reports. > > They probably don't even have a policy, other than not continuing > discussions. The two times I asked, I was told that they do NOT > provide another opportunity to be tested and that the results, false > positive or otherwise, are invariably accepted as final. But never > was this asked as part of the interview. This is exactly the problem. You lose all control over the results of the tests. In fact, they don't even promise to keep the results confidential! > >I have heard of people who were fired from their jobs based on incorrect > >credit information from fradulent use of their identities by other > >people. > > Hopefully, this isn't too serious of a problem. I've heard those > stories, too. Probably on some "news" program. I have heard this on several of the "investigative" television programs like 20-20. Some are due to mistakes in the credit report that were hard to have removed. Others were due to another person impersonating the worker and commiting crimes. This resulted in a police file which said that warrents were outstanding. Even after clearing the matter up with the police and getting it in writing, the worker was dismissed from the job. It really ruined their life. -- Rick Collins rick.collins@XYarius.com Ignore the reply address. To email me use the above address with the XY removed. Arius - A Signal Processing Solutions Company Specializing in DSP and FPGA design Arius 4 King Ave Frederick, MD 21701-3110 301-682-7772 Voice 301-682-7666 FAX Internet URL http://www.arius.comArticle: 24981
Dan Kuechle wrote: > My Virtex 400 design that placed and routed in 2 - 2.5 hours with 2.1i > (on a 350 mhz PC) is now taking 4 - 5 hours with 3.1i (with service pack > 2 installed). The routing delays and timing look no better or no worst > than 2.1i. > Is anyone else seeing these longer run times? I've seen multiple problems with 3.1i that I'm still trying to sort out. Some of the problems are: - FF's not being put into IOB's when F2.1i successfully put them there (Spartan-II). This was not fixed with SP2. - More logic resources being taken up (separate than what was covered in one of their "Answers" Database). - PAR times 2 to 4 times longer. The first one is amusing, since it is very obvious but somehow slipped through their QA. Unfortunately that's where the amusement ends. I have a project that is in limbo because of that bug. I've been trying to create a work-around, but it is difficult given the much longer PAR times (1 hour vs 15 minutes). David Kessner davidk@free-ip.comArticle: 24982
Jon Kirwan wrote: > >Now I have been giving a suggesting/question of which the response is > >as yet absent: why do we not post the company names with the behavior > >we have a problem with? > You want to start? I have not yet indicated that there is a company whose behavior I have a problem with, unlike others in this thread. Why are they so determined not to disclose company names that they surely must know? I would not, of course, suggest signing anything without reading it, but how many of us read the fine legalistic print, though we perhaps should, on every agreement and rather depend on a common understanding with a reputable person or company. Because we not giving company names and, as we are apparently concerned with the details of these agreements, the specific passages of the agreements about which we are concerned, it tends to reduce the credibility that there are any such companies or problem agreements. We maybe concerning ourselves with some extremely rare or imaginary conditions. Rick Collins wrote: > A completed application, > > A "consumer report" (otherwise known as a credit check) authorization, > > A Background questionnaire release form, > > Then they want me to leave behind a urine sample for drug testing. > > I can see the potential need for all of these things even if I don't > agree that they should be used for employment screening. But I strongly > disagree with the requirement to authorize them prior to an interview. > And more importantly, I will not provide such an authorization to use > this information without an opportunity for me to dispute any erroneous > information contained in these reports. In fact I am not given any idea > of how the information will be evaluated. Finally the company retains > the right to perform further checks in the future (if hired) for > "deciding whether to continue your employment", and " when making other > employment related decisions directly affecting you". > > I have heard of people who were fired from their jobs based on incorrect > credit information from fradulent use of their identities by other > people. > > I am sure that this will get responses both pro and con to the reasons > that a company may feel the need to "protect" themselves against harm > from an employee. But these agreements go far beyond "protecting" a > company and can very easily be used against an employee to permit > dimissal without recourse or even explanation. I disagree. These requests do not create a continuing liability on the employee as against the case of a long term NDA that may be expected to continue beyond employment. There will likely be, as indicated, subsequent requests for credit reports and urine samples, but these are not particularly onerous. But primarily, each requirement or screen placed on an applicant reduces the effective number of people that may be accepted with the result that the supply of labor is reduced (in what has been said is already a job seeker's market) such that some reward would flow to that reduced supply. Or more generally, if you want a higher quality product, you are going to pay more. Whether or not this company is overbearing, confused, or just doing what would be expected will be better determined once we have more information about the company. And though I would not be combative about these things if you decide to go to the interview, it seems quite appropriate and expected as a normal process of interviewing the company to ask them why they have these requirements on applicants that other companies to not. We seem to be making an assumption that all companies should have the same requirements in these indicated categories for all employees. This is obviously not the case. These fears that some terrible thing may happen because we essentially become less private for some purposes I suggest, as above, needs to be given a concrete, objective basis. I am sure that all manner of crazy, illegal, or unethical actions have been taken by some companies and/or individuals at some time, but what is their effective relevance in the current case? (And this is another reason why we need specific names and actions so that we can deal with that apparently smaller proportion that is a problem instead of globally characterizing a large number for which there is no problem.) A meteorite may strike any one of us dead today, but we do not worry about it. Regards, Neil NelsonArticle: 24983
David Kessner wrote: > > Dan Kuechle wrote: > > My Virtex 400 design that placed and routed in 2 - 2.5 hours with 2.1i > > (on a 350 mhz PC) is now taking 4 - 5 hours with 3.1i (with service pack > > 2 installed). The routing delays and timing look no better or no worst > > than 2.1i. > > Is anyone else seeing these longer run times? > > I've seen multiple problems with 3.1i that I'm still trying to sort > out. Some of the problems are: > > - FF's not being put into IOB's when F2.1i successfully put > them there (Spartan-II). This was not fixed with SP2. > - More logic resources being taken up (separate than what > was covered in one of their "Answers" Database). > - PAR times 2 to 4 times longer. > > The first one is amusing, since it is very obvious but somehow > slipped through their QA. Unfortunately that's where the amusement > ends. I have a project that is in limbo because of that bug. I've > been trying to create a work-around, but it is difficult given the > much longer PAR times (1 hour vs 15 minutes). > > David Kessner > davidk@free-ip.com I am not trying to be dense, but why can't you go back to the 2.1 tools if this is holding up your project? -- Rick Collins rick.collins@XYarius.com Ignore the reply address. To email me use the above address with the XY removed. Arius - A Signal Processing Solutions Company Specializing in DSP and FPGA design Arius 4 King Ave Frederick, MD 21701-3110 301-682-7772 Voice 301-682-7666 FAX Internet URL http://www.arius.comArticle: 24984
Elftmann wrote: > Actel set out on the same mission to characterize metastability > characteristics in our newer devices, but ran into the same problem as the > Xilinx team did. I'll check back with Product Engineers next week and see > where it sits on the priority list these days. As a result of past > discussions in this group I continue to push on Product Engineering to make > metastability characterization standard operating procedure. Yes, please continue to PUSH. This should be part of the product specification. If an exact number can not be found, than a bounded limit should be specified so that designers can do a bounded analysis. Going through a lot of companies' data books and application notes, some are quite good, others do a little, others do none. Up to date information is critical. As some of the manufacturer's produce products for the high-reliability applications, that information is really required. ---------------------------------------------------------------------- rk History will remember the twentieth stellar engineering, ltd. century for two technological stellare@erols.com.NOSPAM developments: atomic energy and Hi-Rel Digital Systems Design space flight. -- Neil Armstrong, 1994Article: 24985
Peter Alfke wrote: > Metastability is caused by the (limited) gain bandwidth product of the > master latch in the flip-flop. As such it has nothing to do with the > programming technology. Xilinx ( and Altera :-) ) flip-flops are > hard-implemented, optimized structures that contain no programming > elements in the critical path. > I suppose the same is true of most ASICs. The original Actel devices > composed their latches out of multiplexers, and we always pointed out > their (theoretically) poor metastable performance. I assume that Actel > now also has "hard-coded" flip-flops. (?) As Dan replied, they do have "hard-coded" flip-flops although one can make "soft-coded" flip-flops in any of the other familes using logic resources. Now, my question: Is "hard-coded" the standard term? I have been using "hard-wired" and "routed" to describe flip-flops in programmables. Comments? Thanks, rkArticle: 24986
Jim Granville wrote > Does that mean these chips are getting worse / degraded to the point > you are then able to measure metastable events again ;-) > ( I know it's not what you meant, but it's {almost} what you said ) That's obviously not what I meant, and not what I said. :-( Are the summer doldrums getting to us? Nothing meaningful to say, so we start posting about line breaks, create deliberate silly misunderstandings and have a hundred posts about NDAs during job interviews. And Moore vs Mealy ad infinitum... Must mean that there are no more serious subjects to talk about. PeterArticle: 24987
rk wrote: > Elftmann wrote: > > > This should be part of the product > specification. If an exact number can not be found, than a bounded limit > should be specified so that designers can do a bounded analysis. > As long as we understand: There is no limit, and there will not be any limit, ever. It is a statistical thing, and we can only promise an MTBF. I have published Xilinx metastability test circuits and results in every data book since 1989. And we will, again, for the newer parts (Virtex). For the reasons that I described, they got a little long in the tooth in recent years. Peter AlfkeArticle: 24988
frank_mckenney@mindspring.com wrote: > In <39A29BC5.949D50D1@pacbell.net>, Neil Nelson <n_nelson@pacbell.net> writes: > --snip-- > ><!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN"> > ><html> > > <head> > > <title>Anonymous Remailers</title> > > </head> > --snip-- > > Neil, > > Please don't post HTML, particularly irrelevant HTML. > > Thank you. > > Frank McKenney, McKenney Associates > Richmond, Virginia / (804) 320-4887 > E-mail: frank_mckenney@mindspring.com I can avoid the HTML. As to relevancy, I would say it is very relevant, the reasons of which were given in that post. Would you care to give an argument as to why it is not relevant? Regards, Neil NelsonArticle: 24989
"Paul E. Bennett" wrote: > In article <39A375B6.1BC937B9@pacbell.net> > n_nelson@pacbell.net "Neil Nelson" writes: > > > I think I have said that NDAs for the purpose of an interview are of > > little material relevance one way or the other such that it should not > > be a critical factor in interviewing with a company. I.e., if I > > understood that the NDA was for the purpose of the interview I would > > sign it. > > Many legally binding documents have a much longer term associated than > you realise. My legal guides always tell me to read the document and > understand its implications before I sign it. If the terms do not seem > fair to both parties then do not sign the document until it has been > corrected to reflect such fairness. > > We as logicians (I would hope all of us here are anyway) we can often > be better placed than legal counsel at spotting when the logic of an > agreement is broken. As I stated before, I live just 3 miles from one > establishment where the visitors register contains a (very fair) NDA. > I have no problem signing theirs when visiting them (I still read it > every time before signing though). I have, however, massively carved > other NDA's about when the emphasis was too much in favour of the other > party. So that is the test. Is the document fair to both parties. I have not said: do not read the document before signing it. But I am wondering specifically what passages should we be concerned with to a degree that would cause us to not sign it? I understand that an NDA for the purposes of an interview to be that if the company divulges during the interview company specific information of a nature that would damage the company if made widely known or known to another company, that this information should not be made widely known or known to another company. Perhaps what we are calling an NDA consists, in our different interpretations, of a variety of things that might be better labeled differently. Does anyone have a document with a specific passage that they hold objectionable and can post? Regards, Neil NelsonArticle: 24990
I think Xilinx needs to fix the Unisims library. I can't think of any reason why it should contain default timing information, given that it's meant for functional sims. IIRC, the problem didn't exist before DLLs were added to the library. Before DLLs, you could set a timing resolution of nanoseconds in your sim, and all the sub-nanosecond default delays in the library were magically truncated to zero, which is required for a functional sim. DLL functional sims require a picosecond timing resolution, and suddenly all those 10ps defaults in the library become real requirements, which you can't meet in a unit-delay simulation. So, if you haven't got a DLL in your design, the simple fix is to set your sim to nanosecond resolution. If you have got a DLL, then my own preference is to turn off timing checks on the simulator command line or GUI - for ModelSim it's 'vsim +notimingchecks', for example. EvanArticle: 24991
On Mon, 21 Aug 2000 17:09:14 -0700, Philip Freidin <philip@fliptronics.com> wrote: > >Send a request to > xdl_support@xilinx.com > >On Mon, 21 Aug 2000 21:19:14 GMT, William Chow <choww@eecg.utoronto.ca> wrote: >>I'm looking for documentation on the Xilinx xdl file format. I've tried >>looking for it off Xilinx's support page. Can't get to it. Can someone >>help me out? I seem to remember this resulting in a response that it wasn't supported. You can get the docs by installing the userware option if you haven't already done so, and looking for $XILINX/userware/doc/xdl.html. EvanArticle: 24992
rickman wrote: > I am not trying to be dense, but why can't you go back to the 2.1 tools > if this is holding up your project? If Xilinx doesn't resolve this issue within the next day or two then I will return to 2.1i. As will many other Xilinx users, I'm sure. Take a look at the following code: entity test is port (reset :in std_logic; clk :in std_logic; din :in std_logic_vector (7 downto 0); dout :out std_logic_vector (7 downto 0) ); end test; architecture arch_test of test is signal a :std_logic_vector (din'range); signal b :std_logic_vector (din'range); signal c :std_logic_vector (din'range); begin process (reset, clk) begin if reset='1' then a <= (others=>'0'); b <= (others=>'0'); c <= (others=>'0'); elsif clk'event and clk='1' then a <= din; b <= a; c <= b; end if; end process; dout <= c; end arch_test; Essentially there are three 8-bit wide registers. The tools should place two of these registers in the IOB's. When compiling for a 5 volt Spartan it does. But when compiling for Spartan-II or Virtex it will not use the IOFF's. I can't speak for everyone, but I can say that _ALL_ of my designs do something similar to the above code so _ALL_ my my designs would have problems with 3.1i. It would be a safe guess that at least a significant percentage of the FPGA designers out there would also have problems. It's <insert your favorite verb> that this type of bug wasn't found by Xilinx's own internal test suite and/or QA department. David Kessner davidk@free-ip.comArticle: 24993
Peter Alfke wrote: > rk wrote: > > > Elftmann wrote: > > > > > This should be part of the product > > specification. If an exact number can not be found, than a bounded limit > > should be specified so that designers can do a bounded analysis. > > > > As long as we understand: > There is no limit, and there will not be any limit, ever. > It is a statistical thing, and we can only promise an MTBF. > > I have published Xilinx metastability test circuits and results in every data > book since 1989. > And we will, again, for the newer parts (Virtex). > For the reasons that I described, they got a little long in the tooth in recent > years. Oops, perhaps I didn't write quite clearly enough. If we can get numbers that specify at least how good the devices will be, then that would be beneficial. The limits I specify are on the parameters to plug into the equations for MTBF, which some manufacturers provide for engineer's use in calculating MTBF for their circuit. So, even if the calculated MTBF is wrong, as long as it is wrong in the conservation direction, I would be happy. With measurements that I make at day job, even when getting a null response, because of the statistical nature of the measurements, a bound is provided. For example, for SEU measurements, which apply equally to Xilinx and Actel parts, for example, a test is frequently run with a certain fluence, measured in particals/sq. cm. One can get a perfect measurement if one exposes the part to a VERY LARGE number of particles per sq. cm. But how much is enough? 10^6 p/cm^2? Many have used that as a standard. How large is your flip-flops? Pretty small, I'm sure, particularly at the lower feature sizes. Is 10^7 p/cm^2 enough? Much more than that and the measurements become not that practical to run. While there are many flip-flops and that helps with the measurements, in some cases one can't count on that, like a reset state flip-flop, TAP controller flip-flops, etc. In any event, in those types of measurements, it is standard, when a null result is observed, it is assumed that a single event could have occurred and that is used for the calculation of the estimated bound, with a note attached (or by use of a special symbol) that no errors were detected. That's sort of what I was looking for here. Then we can do bounded calculations for a particular sitiuation and then hopefully say: "We got plenty of margin and we knot that Peter's or Dan's number were conservative on top of that." That works better than saying, Peter and Dan say their flip-flops are GREAT! <in the tone of Tony the Tiger> and we won't have any problem. Yes, metastable state resolution time is a statistical thing and no one has shown a circuit to bound that, that has been accepted by the general community. There, that should prevent us from re-opening that can of worms! :-) Thanks! rkArticle: 24994
Ben Franchuk wrote: > > Peter Alfke wrote: > > > Are the summer doldrums getting to us? Nothing meaningful to say, so we > > start posting about line breaks, create deliberate silly misunderstandings > > and have a hundred posts about NDAs during job interviews. And Moore vs > > Mealy ad infinitum... > > Must mean that there are no more serious subjects to talk about. > > Yes - we have to get back to "Use only Xlinix Parts" > "No open source of FPGA's" "No Free software tools" "My latest RISC cpu" > "My latest Forth cpu" "Why suppliers can't supply products" "Who has the > fastest chip (Not counting routing delays)". You've got a Forth cpu??? Tell us all about it! ;) -- Rick Collins rick.collins@XYarius.com Ignore the reply address. To email me use the above address with the XY removed. Arius - A Signal Processing Solutions Company Specializing in DSP and FPGA design Arius 4 King Ave Frederick, MD 21701-3110 301-682-7772 Voice 301-682-7666 FAX Internet URL http://www.arius.comArticle: 24995
"Jimmy Roberts" <j_robby@hotmail.com> writes: > Hi Folks, > > What are the advantages and disadvantages of using Mealy model of FSMs vs > Moore model, on FPGA. > In Mealy model, the outputs depends on the states only, whereas in Moore > model, the outputs depend on the states and the actual inputs. I presume > Mealy model is better (less combinatorial logic). Do you agree? Any other > thoughts. Since in a Mealy state machine the output vector depends both on the state and the inputs, usually you need less states to achieve your goal compared to a Moore machine. The price that you pay is going through the additional combinational logic from the inputs to the outputs, which may add delay to the critical timing path. -ArrigoArticle: 24996
David, When I looked at 3.1i for virtex under implementation, option, optimize and map, in program options, there was a choice pack I/O registers into IOB's for Inputs, outputs, both or none. The default on my set-up is none. Have you tried this, or are you saying it does not work? Bob In article <39A3FCDA.D7749279@free-ip.com>, davidk@free-ip.com wrote: > rickman wrote: > > I am not trying to be dense, but why can't you go back to the 2.1 tools > > if this is holding up your project? > > If Xilinx doesn't resolve this issue within the next day or two > then I will return to 2.1i. As will many other Xilinx users, I'm > sure. > > Take a look at the following code: > > entity test is > port (reset :in std_logic; > clk :in std_logic; > din :in std_logic_vector (7 downto 0); > dout :out std_logic_vector (7 downto 0) > ); > end test; > > architecture arch_test of test is > signal a :std_logic_vector (din'range); > signal b :std_logic_vector (din'range); > signal c :std_logic_vector (din'range); > begin > process (reset, clk) > begin > if reset='1' then > a <= (others=>'0'); > b <= (others=>'0'); > c <= (others=>'0'); > elsif clk'event and clk='1' then > a <= din; > b <= a; > c <= b; > end if; > end process; > > dout <= c; > end arch_test; > > Essentially there are three 8-bit wide registers. The tools should > place two of these registers in the IOB's. When compiling for a > 5 volt Spartan it does. But when compiling for Spartan-II or Virtex > it will not use the IOFF's. > > I can't speak for everyone, but I can say that _ALL_ of my designs > do something similar to the above code so _ALL_ my my designs would > have problems with 3.1i. It would be a safe guess that at least a > significant percentage of the FPGA designers out there would also > have problems. It's <insert your favorite verb> that this type of > bug wasn't found by Xilinx's own internal test suite and/or QA > department. > > David Kessner > davidk@free-ip.com > Sent via Deja.com http://www.deja.com/ Before you buy.Article: 24997
On Wed, 23 Aug 2000 08:10:35 -0700, Neil Nelson <n_nelson@pacbell.net> wrote: >These requests do not create a continuing liability on the >employee as against the case of a long term NDA that may be expected to >continue beyond employment. Didn't rickman say the following? >>Finally the company retains >>the right to perform further checks in the future (if hired) for >>"deciding whether to continue your employment", and " when making other >>employment related decisions directly affecting you". You might dispute rickman's interpretation, but I believe you have no basis for doing so from his post. Certainly, you have absolutely no evidence from rickman's post that would allow you to properly say, "these requests do not create a continuing liability..." I'd consider any document that suggests to rickman that they can use his permission at a later time, as rickman suggests it would do, as being a "continuing liability." If not "liability", then at least it should be a concern of his. We can argue the semantics of this word and that, but the essence is simply that there is no good reason for a company to ask anyone to sign such a document, particularly if it seems to allow the other side to use its power from time to time, at their discretion. Further, such things may make sense as a condition of employment, but can't possibly be justifiable as a condition of further talks. >There will likely be, as indicated, >subsequent requests for credit reports and urine samples, but these are not >particularly onerous. But primarily, each requirement or screen placed on >an applicant reduces the effective number of people that may be accepted >with the result that the supply of labor is reduced (in what has been said >is already a job seeker's market) such that some reward would flow to that >reduced supply. Hehe. There's a point in there. The set of applicants willing to sign idiotic arrangements that give away all manner of self-control and place their trust and faith in people they little know must indeed be a select group! I do believe that their just rewards will be forthcoming, too. hehe. But the essence of your point is simply that anything a company does that reduces the set, is a good thing for those willing to cooperate? That's the basis for your argument here, that this is an okay thing? I love it when people argue as you do. >Or more generally, if you want a higher quality product, >you are going to pay more. Hehe. Somehow, I don't consider this a measure of a "higher quality product," but rather a measure of business idiocy. Probably a good way to get people who are "technical hot house plants," gawking around without a clue about what is reasonable and what isn't. If you call that "higher quality," fine. Good luck to you. >Whether or not this company is overbearing, >confused, or just doing what would be expected will be better determined >once we have more information about the company. More information is always helpful. Tossing out a softball like this doesn't make the rest you said sound any better to me, though. >And though I would >not be combative about these things if you decide to go to the interview, it >seems quite appropriate and expected as a normal process of interviewing >the company to ask them why they have these requirements on applicants >that other companies to not. > >We seem to be making an assumption that all companies should have the >same requirements in these indicated categories for all employees. This is >obviously not the case. I don't think rickman suggested that a company shouldn't make requests that other companies haven't, on that basis alone. The fact that some companies may have some requirements that other companies do not and that such requirements may in fact be reasonable, something we can all agree on, has absolutely no bearing on whether this particular issue is reasonable or not. This kind of arguing from you doesn't deal with the requests rickman mentioned at all. It just says that if some are okay, maybe this one is okay too. In effect, you throw away everything rickman said in his post by way of information and then argue from a position of complete ignorance of his points. Such statements from you could just as well be applied to any request from any company at any time. It would explain any action or none of them, equally well. And it's exactly zero, in terms of making your point, I believe. >These fears that some terrible thing may happen because we essentially >become less private for some purposes I suggest, as above, needs to be >given a concrete, objective basis. Is this the kind of vapid arguments I hear from folks who say, "Why worry about your rights? If you aren't guilty, you have nothing to worry about?" I can't believe may be tredding down this path! >I am sure that all manner of crazy, illegal, >or unethical actions have been taken by some companies and/or individuals >at some time, but what is their effective relevance in the current case? You seem to be saying that if you don't have evidence of specific wrong-doing on the part of the company you hope to intervew at, that you should just sign anything they ask you to? Do you really think so? >(And this is another reason why we need specific names and actions so that >we can deal with that apparently smaller proportion that is a problem >instead of globally characterizing a large number for which there is no problem.) Who exactly argued a global characterization, what is that characterization you refer to and where is it? So far as I can tell, we've been discussing specific requests and actions, not companies. >A meteorite may strike any one of us dead today, but we do not worry >about it. Is this one of your finer points? JonArticle: 24998
On Wed, 23 Aug 2000 09:21:41 -0700, Neil Nelson <n_nelson@pacbell.net> wrote: >I have not said: do not read the document before signing it. But I am >wondering specifically what passages should we be concerned with to >a degree that would cause us to not sign it? Rickman's original post says: >>One of the paragraphs was a brief, but broadly worded NDA. Words >>to the effect that I would not divulge any information that I obtained >>while at this facility. I thought that's what we were discussing. JonArticle: 24999
Surprisingly, I don't recall having seen one point throughout this entire thread of 22 (now 23) messages: The big problem with true Mealy machines is when you cascade them. An input that affects outputs which then become inputs to a second machine, affecting its outputs, to a third, and so on is a pain to manage from a timing perspective. Good tools will find the long paths, but then it's too late and the designer needs to redesign those Mealy machines or do lots of path analysis. When the outputs are registered, or at least are combinational from registered state, the longest paths are from one machine to the next, and you don't have to worry about outputs which eventually propagate through two or more machines. I think this is a common problem when engineers say "I synthesized my modules separately and the slowest one ran at 92MHz but then when I hook them together my rate drops to 55MHz." The tools are smart enough to recognize the multi-machine paths. By the way, Rickman (since I know you read all of these posts), I hope you appreciate that I formatted this email to fit within your viewer's constraints. I know I wasn't the poster to whom you were referring, but I understand your issue. Indeed, I had the same problem with Alliance 2.1i: I have a number of pulldowns on IOs and the translater's output log (I use Design Manager) concatenated all of the messages into one line: Added a PULLDOWN primitive on net PLB_SCLKAdded a PULLDOWN primitive on net PLA_SCLKAdded a PULLDOWN primitive on net PLA_D[13]Added <snipped about 40 more of these...> I must apologize. That previous line wasn't wrapped! :^) What newsreader do you use? OE5 wraps lines for me automatically. But it has other problems, including the OS it comes with. Marc
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