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
FYI: There's a new book hot off the presses: Splash-2: FPGAs in a Custom Computing Machine, by D. Buell, J. Arnold, and W. Kleinfelder IEEE Computer Society Press, ISBN 0-8186-7413-X This book covers the key aspects of the Splash custom computing project born from the (formerly) Supercomputing Research Center (Bowie, MD). Splash-2 is a large experimental system consisting of (among other things) numerous XC4010s, crossbar networks, and an abundance of SRAM. The book covers the hardware, system software, and applications for the Splash machine. -P. Athanas Virginia TechArticle: 3426
First rule is to get some sleep. For instance, Xilinx position the mode pins in the sequence M1..M0..M2. It could be something really goofy! ft63@dial.pipex.com "Peter" writes: >> Hello, >> >> I have a board (a PC card) with 32 XC3064A (TQ144 package) devices, >> loaded serially (CCLK+DATA) and all 32 simultaneously. Each contains >> an identical circuit, basically a very slow but complex pulse counting >> function. Only a few pins on each FPGA are actually used; the rest are >> unconfigured so they use the internal pullup. <big snip> >> Any help is much appreciated. I have to deliver this by the 29th! >> >> Peter. >>Article: 3427
In article <4o1kfh$nv@serv.hinet.net>, Felix K.C. CHEN <flxchen@diig.dlink.com.tw> wrote: >Dear Friends, > >I might be wrong but I believe that no synthesizer can optimize >finte state machine (in VHDL) today! > My first reaction when I read this post was RTFM! But, I will be generous and assume that you don't have a good synthesis manual available to you.... > >The following statements are typical in a finite state machine >VHDL code (of course we can write it in variation): > >-- state transition process > >if (CLK'event and CLK='1') then > case state is > when S0 => > if (some condition is true) then > state <= S1; > else > state <= S3; > end if; > when S1 => > Your coding style is your first mistake. Most synthesis tools state that case statement are preferable to if else. if-else implies priority. (This generates wasted logic in an FSM.) The second mistake is your method of one-hot encoding. Example: package pk is type state is (S0, S1, S2, S3);--state can take one of these values. attribute ENUM_ENCODING : STRING; attribute ENUM_ENCODING of state : type is "00 10 11 01"; end pk; library work; use work.pk.all ; -- the default library in synopsys is work. entity test is port (X, clock : in bit; Z : out bit); end test; architecture trial of test is attribute STATE_VECTOR : string; signal ST : state; attribute STATE_VECTOR of trial : architecture is "ST"; begin process begin wait until clock' event and clock = '1'; if X='0' then Z <= '0'; else case ST is when S0 => ST <= S1 ; Z <= '0'; when S1 => ST <= S2 ; Z <= '0'; when S2 => ST <= S3; Z <= '0'; when S3 => ST <= S0 ; Z <= '1'; end case; end if; end process; end trial; This is a "normal" state machine and this is one hot: next_state = 8'b0 ; case (1'b1) // synopsys parallel_case full_case state[START]: if (in == 8'h3c) next_state[SA] = 1'b1 ; else next_state[START] = 1'b1 ; state[SB]: if (in == 8'haa) next_state[SE] = 1'b1 ; else begin next_state[SF] = 1'b1 ; state[SC]: next_state[SD] = 1'b1 ; Mos synthesis tools have special compile modes for state machines. It is best to use them. //Celia ClauseArticle: 3428
Dear friends, although I have posted a correction for my earlier post, titled "impossible for Synthesizer to optimize FSM", many of you seem have not read the correction. I am not sure how good Synopsys is with FSM, though I have read its manual about extracting FSM from an entity. With Synopsys you have to specify the identifier of state vector you use so that it can extract the FSM. However, with Exemplar you can direct the synthesizer that how each enumeration type signal should be coded, one-hot, binary, gray and so on. I do not know which is better in result. But I think that for users' convinence, exemplar is better. In article <Ds45yu.B13@bri.hp.com> andrew@bri.hp.com <andrew@bri.hp.com> wrote: > > I have only experienced synthesis with synopsys. These are my thoughts. > > You can have all your state vectors encoded such that only 1 bit is set in any > state, thus: > > constant S0 : std_logic_vector(3 downto 0) := "0001"; > constant S1 : std_logic_vector(3 downto 0) := "0010"; > constant S2 : std_logic_vector(3 downto 0) := "0100"; > constant S3 : std_logic_vector(3 downto 0) := "1000"; > For one-hot encoding for FSM, I believe that VHDL synthesizers could reduce the comparison from a vector to a bit. According to Exemplar's Galileo reference manual, a state constant coded in one-hot is assigned a vector like "----1" or "---1-". It is not assigned "00001" or "00010". Therefore the statement like when state=state_name it implies when state="----1" rather then when state="00001". Since all logic operating on '-' are believed to be eliminated by synthesizer, there is no vector comparison at > With synopsys, it is possible to extract the FSM and then get synopsys to re- synthesys > with suitable directives to indicate that the state-vector should be decoded on a > per-bit basis and not on a per vector basis. This then does give you a true one-hot > encoded state-machine in that only 1 bit is used to decode the state. This methodology > is a pain because requires the FSM to be extracted from the design and resynthised. > I have seen it work but it just strikes me as being inefficient to have to synthesize > twice. Perhaps some others can comment on their experiences of fsm_extract. > I am not sure how good Synopsys is with FSM, though I have read its manual about extracting FSM from an entity. With Synopsys you have to specify the identifier of state vector you use so that it can extract the FSM. However, with Exemplar you can direct the synthesizer that how each enumeration type signal should be coded, one-hot, binary, gray and so on. I do not know which is better in result. But I think that for users' convinence, exemplar is better. Regards, Felix K.C. CHEN -- --------------------------------- Felix, Kuan-chih CHEN (³¯ «a §Ó) Associate Project Manager System Product Division D-Link Co., Hsin-chu, Taiwan Email: flxchen@diig.dlink.com.tw Machines and tools are only as good as the people who use it. ---------------------------------Article: 3429
Felix K.C. CHEN wrote: > [...] > For one-hot encoding for FSM, I believe that VHDL synthesizers > could reduce the comparison from a vector to a bit. According > to Exemplar's Galileo reference manual, a state constant coded > in one-hot is assigned a vector like "----1" or "---1-". It is > not assigned "00001" or "00010". > > Therefore the statement like > when state=state_name it implies when state="----1" rather then > when state="00001". Since all logic operating on '-' are believed > to be eliminated by synthesizer, there is no vector comparison at > [...] Felix, this will not work for synthesis, at least under Synopsys, and simulation results of this code will be different from your intention. A comparison of '0' to '-' (or '1' to '-', resp.) yields FALSE. Hence, if the state vector has the value "00001", and is compared to "----1", the result will be FALSE in simulation, and the particular state will not be recognized. As a rule of thumb, Don't Cares should be used only in assignments, and not in comparisons. If you want to hand-code one hot encoding, you will have to compare single bits of the state vector, like: if state(0) = '1' then ... elsif state(1) = '1' then ... ... I prefer using enumeration types for the state signal. The (Synopsys) FSM Compiler can be applied to encode the states one-hot or "auto" (which means that an optimized binary encoding is computed) and the results of the different encoding styles can be compared, whereas hand-coding requires changes in the VHDL code if you want to move from one-hot to binary or vice versa. Regards, Martin. ------------------------------------------------------------------------ Martin Radetzki e-mail: martin.radetzki@informatik.uni-oldenburg.de Carl v. Ossietzky University D-26111 Oldenburg, GermanyArticle: 3430
david@fpga.demon.co.uk wrote: > The "rough edges" referred to by recent reviewers here apply to any > newly released product. I hope you're right. Although many of my complaints seem to be designed in and, in fact, changes from the arguably better pro-series product. > I imagine that they looked at the first cut version (7.0) rather > than later versions with the rough edges smoothed. The current > release version of the Viewlogic product is, I believe, 7.11. I made my comments based on version 7.10. Based on my experience with Viewlogic, I'd be willing to bet large sums of money that no significant improvement will be made for at least a year.Article: 3431
In article <31a76353.204485@news.dial.pipex.com> ft63@dial.pipex.com "Peter" writes: "<snip> "If one is using a part which can go into a socket on the *same* PCB "footprint (e.g. DIP, or any of the PLCC devices) then the OTP parts "are fine. One firm (Altera?) even offers to take back the first 100 "programmed parts for a 100% refund! " You can get QFP sockets that solder onto QFP pads. You can also get logic analyzer adapters (i.e. fancy sockets with test points) that solder straight onto QFP pads. There are various sources, but the best-known is Emulation Technology Inc. of Santa Clara. -- David PashleyArticle: 3432
Hello, is there any body there who uses Atmel Serie 6000 FPGAs and maybe even stores the configuration data in AT17C-series memory devices? If you do, you could help me a lot. I have developed a programmer that programs these little memory chips. The programming information is not very clear on the topic of the programmable reset polarity. I would like to have some sample files intended for AT17__ devices and I would like to know, what the reset polarity should be. In the documentation it says, that the reset polarity should be active low if the source file contains the data FF FF FF FF at location 04000h. This makes sence for the AT17C128, but what about the AT17C65 or AT17C256??? I would like to see the output files produced by some design package for these Atmel FPGAs so that I can adapt the programmer software to it. Thanks. Gruss, Paul. FIDO Paul Hoepping@2:2457/315.23 Internet Paul_Hoepping@citybox.fido.de Fax +49-271-3937339Article: 3433
In article <TB.96May28112333@vingmed.vingmed.no>, tb@vingmed.vingmed.no (Torbjoern Bakke) wrote: > >>>>> "P" == Peter <ft63@dial.pipex.com> writes: > > P> The odd thing I have just found is this: the 3030 "control device" is > P> visibly up and running (it is outputting various clocks which go to > P> the other 32 3064s) some 2ms before D/P gets released. This is even > P> though both /INIT and D/P are wire-ored between all 33 devices! One of > P> these clocks is 16MHz, and maybe this is interfering with the config. > P> I will work on this, but how can a FPGA be up & running while D/P is > P> still held low? > > This is odd, and should be looked into. But remember that you can > program the point were D/P changes - Before or After outputs go > active. This is all described in the databook. > > You said that the 3030 was the device driving CCLK. Thus - this > device has to start before the other devices, otherwise they will not > be configured (no CCLK). > > If you are using a serial PROM of some kind, make really sure it > doesn't start clocking out data before the devices are ready to > receive data. If this happens, you could lose the start of the > bitstream, and the devices would not configure properly. > > Hope this can be of help. > > -- > Torbjorn Bakke (tb@vingmed.no) Tel: +47 33 04 21 32 This gets interesting. Here is my analysis: The lead 3064 will go active on the second CCLK tick after it had received the amount of configuration data specified by the length count ( see page 2-29 in the Xilinx data book ). The fact that DONE was still held Low by other devices - that apparently did not receice enough CCLK ticks-, does *NOT* stop the 3030 from configuring and becoming active. ( You can wire DONE permanatly Low, and the chip will still configure properly. That is actually the recommended circuit for applications that don't have to look at DONE, and where a Low pulse on RESET must instigate reconfiguration.) I am certain that the 3030 cannot have active outputs and then spit out more than one rising CCLK edge after that. Check that CCLK stays parked High one tick after the 3030 outputs went active. So the question is: why does DONE mysteriously go High, about 2 ms later ? There is more to this circuit than we have been told. I would suggest counting or timing CCLK, to get at least a rough idea whether it has 22,216 rising edges, or 68,321, and preferrably a few more, as required for the concatenation of a 3030 and a 3064. The software generates the length-count value automatically. Check whether the preamble arrives at all DOUT outputs, and if possible analyze the length count. Has the length count been edited ? Are its bits in the right order? Remember, the value came from a parallel interface, but the chip is only interested in the bit-serial value that appears, MSB first, after the eight ones, followed by 0010 in the header. The next 24 bits are length-count, MSB first. Remember, it does not hurt to have a length-count value that is too large, it just takes longer to configure, but a length-count that is too short prevents configuration. My guess is still that there are double-transitions (reflections) on CCLK. That's why I suggest to lengthen the transition times drastically, e.g with a 50 Ohm series resistor and a 200 pF capacitor to ground, the usual dirty remedy when s signal causes reflections. I hate to make this sound complicated, it really is not. And we will fix this design, if we only get all the information. Peter Alfke. Xilinx Apps.Article: 3434
> If one is using a part which can go into a socket on the *same* PCB > footprint (e.g. DIP, or any of the PLCC devices) then the OTP parts > are fine. One firm (Altera?) even offers to take back the first 100 > programmed parts for a 100% refund! Altera makes some nifty development sockets for their PQFP-packaged devices. they require a bit more clearance around the part than if you just soldered the part to the board, but the socket is footprint-compatible with the PQFP. Works for me. rob rob.pfile@eng.sun.com standard disclaimer; not speaking for Sun, yadda yadda...Article: 3435
Newsgroups: comp.cad.synthesis,comp.lang.vhdl Subject: Re: VHDL synthesis & style questions: From: flxchen@diig.dlink.com.tw (Felix K.C. CHEN) Organization: D-Link Co., Hsin-chu, Taiwan References: <4oia3j$aqb@atheria.europa.com> X-NewsReader: Hi, You do not have to worry (or protect) the wrap around problem if you "add the std_logic_vector" rather than add the integer. The package ieee.std_logic_arith seems support addition on std_logic_vector. If not, please try synopsys's de-facto package std_logic_unsigned.all. The addition on std_logic_vector is very simple. In your case, you can code y <= y + "1"; -- but change the port direction from out to inout Best wishes, Felix K.C. CHEN In article <4oia3j$aqb@atheria.europa.com> Clifford E. Cummings <cliffc@europa.com> wrote: > VHDL synthesis & style questions: > > In VHDL, I could not add to the std_logic_vector output, I had to declare a > separate count register of type integer that I could add to, I had to compare > the integer to 15 because I could not allow the integer to wrap around (run > time syntax error), and I had to use a conversion function to re-assign the > integer to the std_logic_vector output (shown below): > > --------------------------------------------------------------------- > library IEEE; > use IEEE.std_logic_1164.all; > use IEEE.std_logic_arith.all; > > entity count4b is > port (y : out std_logic_vector(3 downto 0); > clk, rstN : in std_logic); > end count4b; > > architecture rtl of count4b is > signal count : integer range 0 to 15; > begin > process (clk, rstN) begin > if (rstN = '0') then > count <= 0; > elsif (clk = '1' and clk'EVENT) then > if (count = 15) then count <= 0; > else count <= count + 1; > end if; > end if; > end process; > > y <= CONV_STD_LOGIC_VECTOR(count, 4); > end rtl; > --------------------------------------------------------------------- > > When I synthesized the two models using Synopsys (Design Optimization>Map > Design>Medium), the Verilog model produced the 4-bit counter with wrap- around > as expected, but the VHDL model produced the same 4-bit counter with logic for > comparing the count to 15. I know about the new IEEE.numeric_std package for > synthesis which would help with the wrap-around problem, but Synopsys does not > seem to support it yet(?). How widely is it supported? > > I suppose I could have declared the output to be of type integer to avoid the > conversion function re-assignment(?) but that brings up a VHDL style question. > Is there a recommended I/O declaration style which makes porting and > instantiating VHDL models easier for other users of foreign models? > > I also ran into a case where it was more convenient to declare an output to be > of mode "buffer" so that I could read the output from within an architecture > that was checking for a full bit. I could not find an index reference in Doug > Perry's book for "buffer", and I only found one example in J. Bhasker's book > on "buffer" and it appears that whatever connects to an instantiated "buffer" > must also be of type "buffer", which again raises the question of I/O > declaration style recommendations? (I seem to recall that the VHDL > participants in John Cooley's "EDA Shootout" ran into problems > (type-mismatches which caused some re-coding delays) when "buffer"-mode was > used. > > Is there a better way to write my VHDL code for this 4-bit counter? > > Regards - Cliff Cummings > > //********************************************************************// > // Cliff Cummings E-mail: cliffc@europa.com // > // Sunburst Design Phone: 503-579-6362 / FAX: 503-579-7631 // > // 15870 SW Breccia Dr., Beaverton, OR 97007 // > // // > // Verilog & Synthesis Training / On-Site Training // > // Verilog, VHDL, Synopsys, LMG, FPGA, Consulting and Contracting // > //********************************************************************// > -- --------------------------------- Felix, Kuan-chih CHEN (³¯ «a §Ó) Associate Project Manager System Product Division D-Link Co., Hsin-chu, Taiwan Email: flxchen@diig.dlink.com.tw Machines and tools are only as good as the people who use it. ---------------------------------Article: 3436
!!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / INDUSTRY GADFLY: "Synopsys Redeemed; Summit Rises" _] [_ by John Cooley, EE Times Columnist Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Roughly 5 years ago, Synopsys tried to sell a tool called FSM Compiler. It was an add-on product to Design Compiler that was supposed to do great things for state machines -- thus supposedly justifying its high price at the time. Those who foolishly bought FSM Compiler found it full of bugs plus it made their state machines far worse! Synopsys customers took great joy in bashing FSM Compiler in the weekly E-mail Synopsys Users Group (ESNUG) discussions. At first, Synopsys desperately tried to fix all the bugs. But, eventually Synopsys backed off of trying to sell FSM Compiler and quietly merged it into Design Compiler as a "free new feature". Synopsys apps engineers and instructors practiced clever yet discrete ways to warn customers to avoid FSM Compiler. Through the years Synopsys quietly fixed the bugs in FSM Compiler -- but it still remained a public embarrassment that Synopsys just wanted to forget. Fast forward to the Great ESDA Shootout at the recent 1996 HP Design SuperCon. Seventeen design teams (including five ESDA vendors) frantically tried to create and synthesize a state machine with the fastest clock-to -clock timing. The big question being asked was: Are ESDA tools up to snuff compared to engineers doing hand coded Verilog or VHDL? Among the ESDA companies, Summit Design did very well in the design entry and ECO stages; for synthesis results, i-Logix and Speed Electronic got second and third place with 2.09 nsec and 2.31 nsec respectively. Hoai Tran of Symbios Logic, a hand coder using Verilog, won with a 1.77 nsec design. All of the other hand coders ranged from 2.46 to 3.35 nsec. Even Bob Painter, who learned basic Verilog, Design Compiler and "vi" from reading a magazine article *during* the contest, got a 2.94 nsec design that easily beat Mentor Graphic's last place 4.25 nsec! Afterwards I was surprised to find out that the first and the third place teams had both used FSM Compiler. This, plus one ESDA vendor's complaint that everyone should have used the *same* Synopsys script to make all non-ESDA things equal, prompted me to personally re-synthesize everything using FSM Compiler. To my surprise, 8 out of 9 hand coded designs synthesized down to 1.73 to 1.78 nsec! (Only newbie Bob Painter's design came in at a slower 2.03 nsec.) Even Mentor's last place 4.25 nsec improved to 2.08 nsec! All but one design got faster by about 50 percent with FSM Compiler! My congrats now go to Summit Design. Summit took third in design entry, first in ECO, and first in synthesis (tying with John Gilbert of Motorola) with a 1.73 nsec design. Gilbert used Verilog and Summit used VHDL. Summit's Verilog also took second place in the synthesis category with 1.76 nsec. My sympathies now go to i-Logix because their contest design data was lost and to Speed Electronic because their 2.30 nsec design now ranks dead last out of the twelve contenders in the synthesis category. I'd like to thank HP Design SuperCon for sponsoring the Great ESDA Shootout; it has reintroduced us cynical engineers to FSM Compiler -- and 50 percent faster state machines! (I'm sure Summit Design didn't mind this new discovery, either!) ------ John Cooley runs the E-mail Synopsys Users Group (ESNUG), is president of the User Society for Electronic Design Automation (USE/DA) & works as a contract ASIC/FPGA designer. He loves hearing from fellow engineers at "jcooley @world.std.com" or (508) 429-4357. [ Copyright 1996 CMP Publications ]Article: 3437
Felix, Kuan-chih CHEN (flxchen@diig.dlink.com.tw) wrote: : Newsgroups: comp.cad.synthesis,comp.lang.vhdl : Subject: Re: VHDL synthesis & style questions: : From: flxchen@diig.dlink.com.tw (Felix K.C. CHEN) : Organization: D-Link Co., Hsin-chu, Taiwan : References: <4oia3j$aqb@atheria.europa.com> : X-NewsReader: : Hi, : You do not have to worry (or protect) the wrap around problem if : you "add the std_logic_vector" rather than add the integer. : The package ieee.std_logic_arith seems support addition on : std_logic_vector. If not, please try synopsys's de-facto : package std_logic_unsigned.all. : The addition on std_logic_vector is very simple. In your case, : you can code : y <= y + "1"; -- but change the port direction from out to inout IMHO, the use of "inout" on a port which is really only an output is a nasty hack. We use "buffer" with very little grief and it reduces the risk of having multiple drives on a signal where we don't want it (these problems are found at analyse/elaborate time and not at run-time). Synopsys "design_compiler" has no problem with ports of mode buffer. There are a few complications in the use of "buffer"s when you wish to mix components with mode "out" with components which have "buffer"s. As long as one is aware of the restrictions then there is little overhead. If you want further clarification, I'd be pleased to help. Andrew library usual; use usual.disclaimers.all;Article: 3438
Rafiki Kim Hofmans (tw38966@vub.ac.be) wrote: : : Hi, : : I'm trying to create a ROM (2x32) with the memgen of Xact 6.0. : After inserting parttype 4010pq160-3 and editing the memorydata the : output states parttype 4005pg156, width 0 and depth0. : : And prosim doesn't recognize the created ROM symbol. : : Anyone experienced with the memgen ? : : Thanks in advance ! : : Kim : We wanted to use Memgen in combination with Mentor DA and had the problem that the automatic designed symbol interface was not created correctly, so we created a schematic out of the xnf file (created by memgen) and then we added the in and outports manually and designed a symbol for this.In the created schematic you can see a HEX-Value on the ROM symbols that represetates the data. For a 32*4 ROM we need about 10 minutes so its shorter than to find out a better way, but if anyone knows please tell me (thread memgen+mentor) -- Manfred Aigner alias maigner@sbox.tu-graz.ac.atArticle: 3439
>My guess is still that there are double-transitions (reflections) on CCLK. >That's why I suggest to lengthen the transition times drastically, e.g >with a 50 Ohm series resistor and a 200 pF capacitor to ground, the usual >dirty remedy when s signal causes reflections. This is probably what it was. I say "probably" because I can make it fail even when the waveform looks clean, on my 250MHz scope. I can make it reliably configure with either of: 1. the 3030's CCLK output driving the 32-off 3064s direct, with 220pF to GND; 2. the 3030's CCLK output driving a 74AC244 buffer, then 470pF to GND, then to the 3064s. I also find that with a very clean but slow edge it does not configure, around 30ns or slower. Any resistor in series makes things worse. The only conclusions I can reach are: that I had transmission line problems; that the edge must be faster than about 30ns; that the devices are sensitive to noise or glitches which are too fast to see on a 250MHz scope. >I hate to make this sound complicated, it really is not. And we will fix >this design, if we only get all the information. Short of posting the schematic and PCB layout, I did describe it as best as I could. But it is working now, so - THANKS TO EVERYONE FOR YOUR HELP. Peter.Article: 3440
I am trying to configure one FLEX81500 using PSS mode. Any one knows what kind of configuration file should I generate from MAXplus2? The "sequential .rbf" file or else? And what about the file format? Feng YunArticle: 3441
W A N T E D ASIC/FPGA Engineers Papillon Research Corporation (formerly Chrysalis Research Corp.) Due to continued growth, Papillon Research in Concord, MA is seeking to expand its hardware engineering group. We are looking for experienced Digital Hardware Engineers to participate in leading edge ASIC, FPGA, and board-level designs using the latest in HDL synthesis and simulation tools. If you desire to work in a casual, startup atmosphere and have practical experience in these areas, please contact us (PRINCIPALS ONLY): Papillon Research Corporation 52 Domino Drive Concord, MA 01742 jobs@chrysal.com 508-371-9175 (fax) Chrysalis Research Corp. is not related to Chrysalis Symbolic Design.Article: 3442
In article <Ds7AIE.2ED@world.std.com>, John Cooley <jcooley@world.std.com> writes > >Among the ESDA companies, Summit Design did very well in the design entry >and ECO stages; for synthesis results, i-Logix and Speed Electronic got >second and third place with 2.09 nsec and 2.31 nsec respectively. Hoai Tran >of Symbios Logic, a hand coder using Verilog, won with a 1.77 nsec design. >All of the other hand coders ranged from 2.46 to 3.35 nsec. Even Bob >Painter, who learned basic Verilog, Design Compiler and "vi" from reading a >magazine article *during* the contest, got a 2.94 nsec design that easily >beat Mentor Graphic's last place 4.25 nsec! > >Afterwards I was surprised to find out that the first and the third place >teams had both used FSM Compiler. This, plus one ESDA vendor's complaint >that everyone should have used the *same* Synopsys script to make all >non-ESDA things equal, prompted me to personally re-synthesize everything >using FSM Compiler. To my surprise, 8 out of 9 hand coded designs >synthesized down to 1.73 to 1.78 nsec! (Only newbie Bob Painter's design >came in at a slower 2.03 nsec.) Even Mentor's last place 4.25 nsec improved >to 2.08 nsec! All but one design got faster by about 50 percent with >FSM Compiler! > >My congrats now go to Summit Design. Summit took third in design entry, >first in ECO, and first in synthesis (tying with John Gilbert of Motorola) >with a 1.73 nsec design. Gilbert used Verilog and Summit used VHDL. >Summit's Verilog also took second place in the synthesis category with >1.76 nsec. My sympathies now go to i-Logix because their contest design >data was lost and to Speed Electronic because their 2.30 nsec design now >ranks dead last out of the twelve contenders in the synthesis category. >I'd like to thank HP Design SuperCon for sponsoring the Great ESDA Shootout; >it has reintroduced us cynical engineers to FSM Compiler -- and 50 percent >faster state machines! (I'm sure Summit Design didn't mind this new >discovery, either!) > >------ IMHO the admission that contestants were not using the same synthesis tools makes the results highly suspect. Also you left out the changes in area! which is of course relevant. I'm not entirly sure what this exercise has proven maybe you could post a summary. ------------------------------------------------ Steven Bird VIZEF Limited Tel: 44 (0)1628 481571 Fax: 44 (0)1628 483902 ------------------------------------------------Article: 3443
landmarktrvl <landmarktrvl> wrote: >Dear Reader, >Are you running out of destinations for your VACATION LIST? >Are you interested in knowing more about history, culture and traditions? I will assume that you are new to the internet -- what you have done is known as spamming, and is most unwelcome. Do it if you want to non technical newsgroups (where this kind of thing is more popular), but this kind of posting to a technical newsgroup will just cause flames. I suggest you find out the meaning of the word "spam", and also "mail bomb" before doing this again -- we can fight back -- and the simple reaction is that because of the rudeness of this kind of advertising knowone ever reads what you are offering. -- Gavin Melville, gavin@cypher.co.nzArticle: 3444
[ Enclosed is a review of last year's (1995) DAC in San Francisco. If you bump into me at DAC next week in Las Vegas, tell me what you think is hot and what's not, so I can include it in this year's review! :^) - John ] !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / The Third Annual ESNUG/DAC Awards: _] [_ "Sleep Deprivation, MacGyver & DAC '95" - or - "One Engineer's Review of DAC'95 in San Francisco, CA, June 12-16, 1995" by John Cooley Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion." [ Check out http://techweb.cmp.com/eet/docs/eetff.html (WWW EE Times) for a photo of the DAC freebies & the awards they received! - John ] It's amazing what the mildly hallucingenic effect five days worth of sleep deprivation can do to the engineering mind. I remember in college psychology there was some question as to whether humans actually dream in color or not. By my fourth day of getting only 50% of my normal amount of sleep at DAC, not only did the demos start breathing and exhibiting 3-D properties, I could now personally attest that we do dream (or at least sleep walk) in color. It was then I started pondering the big questions... Why do engineers count by going "0, 1, 2..." instead of "1, 2, 3..."? Is it possible to build a single pin chip with power, ground, input and output all time multiplexed? Would "MacGyver", "Batman", "James Bond", and "Dr.Who" suddenly all become glamorless technerds if they owned up to having an engineering background? Instead of a circuit built on the assuption that every question can be answered with a "yes" or a "no" (a binary system), what would it look like if we based it on "yes", "no", or "no answer" (a trinary system)? If engineers are so smart, why are they in the middle class? What if there were no hypothetical questions? Then I hit upon the one question that I could answer: "What were the ESNUG User DAC awards this year?" BIGGEST SURPRIZE AT DAC: Silerity. At first, I thought Alain Hanover (CEO of ViewLogic) bought Silerity just as part of the deal to get Will Herman back on board at ViewLogic. But after looking at the Silerity PathBlazer I was surprized that there was real value there. What I saw was a datapath compiler that, through brute force, went through all the datapath part and layout permutations to produce an optimal final design. Cool! Runner up: Mentor buying Exemplar. (Rumor was MGC might buy IST. IST is focused primarily in VHDL and is still at the start-up stage. Exemplar is well known in Verilog and VHDL in both the PC and UNIX worlds. Good choice Exemplar.) WEIRDEST PRE-DAC IMAGE: Before DAC, ArcSys sent a postcard to customers with pictures of six people in a "Name The Great Thinkers" contest. The odd thing was that one "great thinker" appeared to be Lizzie Bordon, a famous ax murderer from Massachusetts who killed her father and stepmother. Another appeared to be Machiavelli, author of a book advocating gaining power via cunning and ruthlessness with no regard to morality (and, incidentally, was reputed to be the favorite bedside reader of Hitler, Stalin, and Mussolini!) MOST TROUBLED HEIR APPARENT: SpeedSim. With high speed Chronologics falling into disarray, Verilog vendors like Intergraph, Frontline, and Simucad have been salivating at taking second place in the Verilog simulator market. It appears that SpeedSim's super fast cycle based Verilog racehorse is the heir apparent, but it's under a dark cloud. Synopsys has been heavily hinting that they're working on a cycle based simulator of their own, but won't say if it is Verilog, VHDL, or both. BIGGEST CINDERELLA AT DAC: The IC place and route companies. With all the engineering conferences chanting the mantra of "deeeeep suuuub miiiicroooon", the usually ignored IC place and route companies like ArcSys, Cooper & Chyan, ISS plus parts of Mentor and Cadence have suddenly come out of the closet telling mainstream engineers: "Look at me! I'm hot technology!" The grand pappy of all IC place and route, the once public domain TimberWolf, even made its commercial debut at this year's DAC. FIRST TO HACK DACNET AWARD: Unknown. An unidentified PCB EDA vendor managed to hack DACnet's internal mailing list, get a partial listing of DACnet account names and send junk e-mail every day promoting their wares. (I can't get confirmation as to exactly which company did this, because most people ignored the e-mail infomercial after the first day.) Despite this, a lot of people enjoyed having the ability to check their secure home e-mail via DACnet. Afterwards, the DACnet people forwarded the leftover e-mail in our conference acounts to our home accounts. (A class act!) BIGGEST LIE AT DAC: The NTT "Consortium." Angry attendees reported that through Harmonix, Nippon Telegraph & Telephone (NTT), Japan's version of AT&T, tried to create the image that there was a new "consortium" promoting their NTT proprietary C-like SFL language as an industry standard in reconfigurable computers. What angered the researchers at the meeting was that they were indirectly told that the company with the money (NTT) would control the "consortium." Also, all technical issues where handled in a hand waving fashion. One researcher said: "Harmonix/NTT never mentioned the problems of a reconfigurable logic operating system (the biggest sticking point in the concept today) nor described how they were going to solve the dynamic place and route problems, time-based partitioning, nor anything else. These people were going to decide the industry standard for reconfigurable computing?!?" WHAT ENGINEERS TALKED ABOUT: This was the year for the fruity niche tool. Other than GLOBEtrotter's licence management and admin tools, there was no one hot EDA "toy" that everyone raved. What people saw was heavily influenced by what they wanted to see. The power junkies were seen shooting up around the Epic Design booth on their 'Mill products, Mentor for the Lsim Power Analyst, Systems Science for their PowerSim and PowerFault, plus Synopsys for their rumored power tools. The engineering Survivalists saw the world shortage of foundry capacity as a bad sign and sought out the foundry-independent ASIC solutions offered by SiArch, Mentor, Compass, Cascade, Sagantec, Aspec, and Meta-Software. The I-don't-want-learn-another-damned-tool ASIC designers gravitated to the tools made for them. Synplicity provided a simple, one button FPGA synthesis tool that took all of 30 seconds to learn and run. Interconnectrix offered a pricey PCB place and route tool that took in constraints in EE terms (instead of geometric rules of thumb) to build PCB's that passed timing and signal integrity specs. HLD Systems offered a general floorplanner that hooks rather nicely into the Synopsys environment eliminating the need to learn the 25 or so ASIC specific floorplanners. Intellitech offered a general purpose foundry/synthesis independent BSDL file generator. Engineer members of the "Glowing Path Pure RTL Level Design" cult sought out Synopsys HDL Advisor to help them write and analyze their RTL source code for synthesis *before* synthesis, InterHDL for VeriLint (a killer Verilog syntax checker) plus their V-to-V (a two way Verilog/VHDL translator) and Cadence's newly announced RTL level floorplanner called SiliconQuest. Cultists also noted that Leda pushed a universal VHDL source encrypter while Chronologics pushed a universal Verilog model encrypter. Pugilistic engineers said: "Looks like AT&T Design Automation is going to get into some interesting fist fights with Chrysalis and Abstract Hardware in the formal verification market, and LogicVision in the BIST market." A few EDA vendors themselves happily noticed that although Microsoft had a booth at DAC last year, they didn't this year. (They figured that a 2 billion dollar industry was chump change as far as Bill Gates was concerned.) MOST CURIOUS NEW COMPANY AWARD: Savantage. These guys created a tool that looks at a design as a bunch of raw dies and juggles partitioning through the packaging, PCB, heat exchanger, connector, bonding, assembly, backplane and enclosure issues with an eye on costs. A manufacturing engineer's dream. BEST USE OF TERRAIN FOR A DAC PARTY: Synopsys took 400+ customers to Fisherman's Warf in trolley cars to jump on a boat ride to Alcatraz. (After having crashed three of Synopsys's DAC parties over the years, I was personally kind of uneasy when, this time, they sent me an invitation to visit a prison...) Anyway, we were given a great guided tour of the island prison and then fed a yummy dinner on a evening cruise in San Francisco bay. Even anticipating that the customers might get cold on the boat, Synopsys gave us white sweaters tastefully embossed with a subdued "SYNOPSYS" in the fabric! (The only down side was after dinner when the ferry was maneuvering underneath the Golden Gate Bridge with a full moon and a fantastic view of San Francisco at night. I really didn't want to be next to three engineers from Data General plus two Synopsys R&D types. I wanted my girlfriend!) Runner Up: Quickturn took customers for the best night view of the city, the 52nd floor of the Bank of America Building. The jazz band was excellent and having their own home brew of HDL ICE beer was a hoot! BEST CONGRESSIONAL JUNKETEER: Yatin Trivedi. Yatin, in a move that would embarrass even the most corrupt congressman, managed to not only get himself on the Design Acceleration sponsored harbor cruise, he got his wife, his kid, his business partner and his business parter's wife on board! WORST PLANNED DAC PARTY: The Official DAC "Cruise The Mediterranean" party. Like a fool I thought we'd have a boat ride, but it just turned out to be one HUGE room with food, beer and music. I did like seeing all the EDA vendors and customers there, but thought it odd that at a conference which is 95% male, they had a band playing slow dance music. Then I remembered: "Oh, yea! That's right. We're in San Francisco!" PANEL WITH THE MOST POLITICAL INTRIGUE: For a few months the Users Society for Electronic Design Automation (USE/DA) has been surveying EDA customers about proposed changes in the way EDA tools are sold. When USE/DA tried to have a lunch panel on Monday at DAC to discuss the results with the CEOs of the Big Four EDA companies, the DAC Committee nixed the idea for fear of stealing from Ron Collett's panel on Tuesday. They wouldn't list the USE/DA panel in any schedule and forbid the use of signs directing engineers to where the USE/DA lunch panel was. Plus, they asked the CEOs not to attend! At the last minute, Collett put an engineer from IBM to represent foundries plus one of his unknown clients from Siemens to represent user views on his Tuesday panel. (That is, no USE/DA people who have been working on these issues were invited.) Despite the intrigue, the Monday USE/DA panel drew CEOs Wally Rhines (of Mentor) and Aart De Geus (of Synopsys) along with 200 people as an audience. Interestingly, we found some real working data from polling the audience; customers prefered a model of having all tools always available to use as needed and to pay later on a per use basis (as opposed to the PC shrink wrap sales model.) WORST YET MOST ENTERTAINING DAC PANEL: Collett's five EDA CEOs DAC panel. The panel consisted of Collett baiting each of the CEOs with embarrassing issues they had dealt with throughout the year. At any moment I expected the CEOs to pull out whiffle ball bats and start hitting each other. (Most attendees voiced dual reactions of thinking the panel was a complete waste of time, yet also very entertaining.) Richard Goering did a great write-up of it on page 29 of the June 19th issue of EE Times. One fun quote left out, though, was Alain Hanover (CEO of ViewLogic) on his rebellious Chronologic: "Our founding forefathers guaranteed us the right to life, liberty, the pursuit of happiness and the right to sue each other." PANEL WITH THE MOST TECHNICAL PROBLEMS: The EE Times Forum. Although this won best panel last year with its hand held electronic voter boxes that engineers used to provide immediate feedback to the panelest's statements, this year the voting boxes went haywire. The first indication was in the supposed demographics of the audience. Polling indicated that they were about 50% Synopsys, 50% Mentor, 50% Cadence, 50% Viewlogic and 58% Intergraph users. (This is virtually impossible.) The second tip off was that no matter what the panelists said they all got a 2.8 to 2.9 score. (Scoring went from 1 for "lame" to 5 for "fantastic!".) "After seeing Penny Herscher of Synopsys get a 2.9 for a great answer, I knew the voting boxes were broken," said Cadence Marketing bigwig Tony Zingale, "I was tempted to announce all Cadence software was now free." (I'm not sure whether Tony was trying to get positive or negative responses from customers with this offer.) Keeping in the spirit of technical difficulties on the EE Times DAC Forum, attendees were given inadvertently defective pins that, when you pressed hard while writing, the spring loaded guts would unexpectedly explode out the back of the pin! (They make for great office joke pins!) EDA SEGMENT WITH THE MOST SPYING: ESDA companies. My first impression of Escalade was Michiel Ligthart whining: "Half of my demos today were given to Synopsys employees." Yet, the very next day, I laughed finding Michiel watching the demo of his competitor, Summit Design. (I'm sure he found out that Summit was adding more pure Verilog features plus beefing up its test tool.) I'm also sure that the Cadence and Synopsys employees watching the i-Logix demo noted that i-Logix had added a Motiff style GUI, graphical simulation that runs with vaguely defined descriptions, and how to generate different styles of Verilog/VHDL/C code from one behavioral description (useful for HW/SW codesign.) Runner up: the IC place and route companies. MOST ENTERTAINING FLOOR SHOW: Compass Design hired a professional comedian who told submicron jokes that were so corny they were good. Two were: "If a clock tree falls alone in the woods, does it make a glitch?" and "When I was young I got into some trouble. My father tied me to VSS. I was grounded for life." Runner Up: Quickturn did a skit where an engineer was going to leap off a building to his death because of design verification problems. Even when the predictable Quickturn Man came to save the day, the customers in the audience kept yelling: "Jump! Jump! Jump!" MOST CONTENT-FREE FLOOR SHOW: Either Hewlett-Packard workstations or HP Developers or HP EEsof (tells you how memorable the sales pitch was) managed to continuously draw noticeably large crowds for their talks. Quietly asking some of the crowd why this was such a draw, I got: "Shhh! We just want the T-shirt!" (Later, I asked engineers what they remembered from the HP presentation; no one could remember one quote or even what the talk was on!) "WHERE'S THE BEEF?" AWARD: Cadence. They didn't have a customer party and even though their exhibit booth was big, their demo suite was two cubicals in size! Quite a few people were surprized how Cadence appeared to be backing out of DAC, the various EDA committees like CFI, EDAC, VIUF, and OVI, and the EDA industry in general. (I guess you don't need this presence to sell design services instead of design products.) MOST IMPROVED EDA COMPANY: Mentor Graphics. Watch out Cadence, Synopsys and Viewlogic! Quite a few engineers noticed that Mentor Graphics now appeared to be serious about mainstream EDA tools by offering a "real" VHDL simulator plus adding full Verilog support for their simulation, synthesis and FPGA synthesis offerings. Adding Verilog allows Mentor to sell to customers who would have normally not given them the time of day. (Also, Mentor won the Best Demo Suite Area award with their open tables, free food & soft drinks.) MOST DISAPPOINTING DAC FREEBIE: ViewLogic's T-shirt. ViewLogic used to be so hip and with the times with their DAC freebies. The year of the L.A. riots, ViewLogic gave out baseball bats. The year that the U.S. sponsored the World Cup in soccer, ViewLogic gave out soccer balls. This year, ViewLogic joined the ranks of the mundane in giving out bland, white T-shirts. Blah. BACK TO THE "REAL" WORLD: Checking my answering machine the morning after DAC, I found out that my girlfriend had phoned the management of EE Times, because she hadn't heard from me in a week and was worried. The front page news in the San Francisco Chronicle that day was the mayor "blessing" the new public toilets on Market street. In the weekend section, the "25th Annual Lesbian, Gay, Bisexual and Transgender Pride" parade was scheduled on Father's Day. I thought to myself: "Gosh, it's great to be back in the real world!" and then happily daydreamed of what the freebies would be like at next year's DAC in the notoriously decadent city of Las Vegas... - John Cooley part-time EDA industry gadfly full-time contract ASIC/FPGA designer P.S. If you thought this review was on-the-money or out-to-lunch, please tell me. I love getting frank, honest feedback from fellow engineers. P.P.S. In replying, *please* don't copy back this entire article; a 14,400 baud modem attached to a 386 on a sheep farm can handle only so much! :^) =========================================================================== Trapped trying to figure out a Synopsys bug? Want to hear how 3567 other users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, _] [_ Verilog, VHDL and numerous Design Methodologies. Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion."Article: 3445
Hello, all! Some days ago we decided to use Xilinx FPGA in our PCMCIA design and found that our Cadence - Xilinx library is very old. Particularly we needs the latest files with packages description ($XACT/data/partlist.xct, $XACT/data/speeds.xct $XACT/data/*.pkg and may be some else). Who knows were I can get this information or send me this files through e-mail sol@elvis.msk.su? Help, please. Alex.Article: 3446
Yun Feng (yunfeng@cse.ucsc.edu) wrote: : I am trying to configure one FLEX81500 using PSS mode. Any one : knows what kind of configuration file should I generate from : MAXplus2? The "sequential .rbf" file or else? And what about the : file format? : Feng Yun There isn't a 'PSS' mode. You probably mean 'passive serial', 'PS'. (There is however a PPS mode, a parallel scheme). What file you need depends on whether you are using Bitblaster or a download cable from the PC Programmer. It's all in App Note 33. Try www.altera.com. -- Best Regards, Steve.Article: 3447
Hi, I'm a new Xilinx user. I work with Xact 6.0.0 and OrCAD 386+. Are there other Xilinx - OrCAD users? Thank you.Article: 3448
LAST MINUTE DAC NOTES: - The Users Society for Electronic Design Automation (USE/DA) is going to have its Birds of a Feather meeting on Wednesday, June 5th, from 6:00 to 7:30 PM in the Hilton in Ballroom "D". Come join us to install the new president, to swap industry dirt and to set our agenda for next year. I hope to see ya'll there! - Here's what we got for questions to put to the industry bigwigs at this year's BIG DAC CEO panel on Tuesday. As the outgoing president of USE/DA, I'll be moderating it. We'll have: Mike Bosworth, CEO of OrCAD Joe Costello, CEO of Cadence Aart De Geus, CEO of Synopsys Richard Goering, EE Times Technology Editor Alain Hanover, CEO of ViewLogic Wally Rhines, CEO of Mentor Graphics Gary Smith, EDA Analyst from Dataquest for 90 minutes of user-driven questioning! Feel free to slip me any more last minute questions you may have by leaving them for me at the Las Vegas Hilton front desk. See you at DAC! - John Cooley USE/DA President -------------------------------------------------------------------------- To: All CEOs I don't have a question, but a thank you note. "By making your tools difficult to use, poorly integrated and full of bugs, you have provide jobs for thousands of people like me to support CAD tools within companies that buy them. Thank you very much." Ameesh Oza Teradyne To: Aart De Geus Your in-Sync program has much higher entry requirements than Cadence Connections, Mentor OpenDoor, or Viewlogic PowerPartner. Why is that? Despite the positive image in the press, there still seems to be a raging EDA Cold War. I've been trying to get into inSync for 12 months and I just get stalled. (I been passed to three different people, they'd lose the required customer references, claim the customers I reference are unreachable yet when I pick up the phone the customer answers and/or claimed to be on the road. They raised this stalling to an artform.) Speed Electronic and Summit Design, my competitors, got into inSync very quickly. Why can't Mentor? - Daniel Payne Mentor Graphics To: Mike Bosworth Was PC pricing a "cargo cult" belief for OrCad?; price it cheaply to sell millions of copies in a market of only ~300,000 engineers? - Sean Murphy Leader-Murphy, Inc. To: All When will EDA providers finally realize PC platform support (Windows NT, in particular) actually provides *more value* to users than UNIX support? The typical shrink-wrapped PC software price structure need *not* apply. The trick is how the EDA providers spin the value to customers in terms of hardware price-performance, network support, integration with other office tools, etc. etc. etc. - Elliott Berger Pixel Magic To: All I'm working on a mid-range sized design of around 200K gates, and my ASIC vendor recommends 512MB of RAM; I also require several GB of disk for storage. The cost of the underlying computer is not very significant at this point. I don't think it will long before 200K will be considered a relatively small design, so this isn't a problem likely to go away soon. Windows is, to my mind, a plainly inferior development environment when compared to the power and features of a Unix environment, so a really strong incentive would be needed to get me to consider moving to a PC. For the reason's given above I don't really see cost as being a good reason, the Window's GUI has got to be the worst of the readily available interfaces, and if anyone wants to they can run PC applications on workstations via Wabi or emulators. Where is the driving force for the change coming from ? - Gavin Brebner Tellabs To: Joe Costello and Wally Rhines In what design process areas do they see the on-site EDA consulting business grow? In those areas where there are such high demand for EDA consulting, is the on-site EDA consulting business induced by the complexity of the EDA company's own tools? Margaret Lee Tandem Computers To: Alain Hanover The Viewlogic IC PowerTeam was formed to tightly integrate with back-end IC tools from HLD Systems, Casacade, SVR, EPIC, and ISS/Avant! What happened to the IC PowerTeam concept? I don't hear about it anymore. - Daniel Payne Mentor Graphics To: Gary Smith Firms like Dataquest, Collett International, and EDA Today all pretty much make their money doing marketing research and selling specialized reports to the commercial EDA vendors. As an EDA analyst often quoted in the industry press how can you be objective about, say, Mentor Graphics, if Dataquest also has a $200,000 market research contract with Mentor Graphics? - John Cooley Holliston Poor Farm To: All 1. CAE CEOs talk about wanting to deliver value by helping customers shorten design cycles. They also talk about guaranteeing results as part of their consulting contracts. Yet customer design cycles are lengthened when CAE companies slip software delivery schedules or ship buggy code. (For example, I have not received Cadence's Q4 '95 maintenance upgrade as of 5/29/96.) Do the CEOs see this as a problem? What are they doing to solve it? Are they willing to guarantee delivery schedules in maintenance contracts? 2. Here's an audience question. Given a choice between buying either: A. An integrated tool set from one vendor, or B. High performance point tools with clean interfaces for user integration from many vendors which would you choose? 3. When CAE companies merge, they often cite synergy as a motivation. How often is synergy actually achieved? How often do mergers simply create customer expectations for synergy and integration, which are not technically feasible? - Mike Murray Acuson To: Richard Goering 1.) How does the impact of a story effect whether or not it goes into EE Times? That is, many times small start-ups have serious product or staffing goofs that, if publically known, will kill the company. How do you handle these situations? 2.) Have you ever known an advertiser in EE Times or other industry trade publications try to use these ad dollars to influence what's reported in that publication? How is this handled? - John Cooley Holliston Poor Farm To: All EDA Software patents. Most EDA software patents are held by IBM or by large Japanese companies -- is this a potential problem for the industry? Is patent protection a good idea for EDA software? If not, what's the alternative? Benchmarking. What do you think in general about independent benchmarks, and why Viewlogic pulled out of the DA Solutions HDL simulation benchmark? - Richard Goering EE Times To: All I have been involved in the EDA industry for almost 17 years and somehow we seem to keep solving the same problems. One of these problems is data transfer between different EDA vendors. I believe the solution is for all users of EDA/CAD/CAM products to demand one standard file format from their vendors. As a user since the most widely accepted standard is EDIF 2.0.0 and EDIF 4.0.0 is coming out in July that is the format we need to rally behind. The vendors I have talked to have agreed that they will supply EDIF 4.0.0. interfaces it is just a matter of priorities as to when it will be implemented. Is this the truth or so much hot air? cathy_olsen@nt.com To: All Being from a PC company, I am extremely interested in EDA applications on PC hardware. There have been several panels recently about "PC versus UNIX"; These panels tended to gravitate toward the cost issues related to the software. What are the technical advantages/disadvantages of PC hardware & OS versus traditional workstation hardware & OS you see as an EDA CEO?? Ben Buzonas Compaq Computer Corporation To: All When will ALL the EDA tools talk to each other and work on standard file formats? I see that some companies are trying to adapt to std file formats but on the other hand, we spend many man years solving those problems with TONs of inhouse scripts/wrappers/translators. -Hiren Majmudar Intel Corporation To: Alain Hanover Viewlogic - isn't a vital user group essential to developing a innovative products? - Sean Murphy Leader-Murphy, Inc. To: All Q: How do the EDA vendors plan to resolve the business conflict of supporting EDA standards needed to fuel the electronics and EDA industry growth curve, while at the same time being concerned that offering their technology for standardization may not give them the competitive advantage? Q: Given the results of the USE/DA Business Practices user survey, what specific plans or commitments are the CEO's willing to make to the users to address their users' concerns and needs? - Steve Schulz Texas Instruments To: All 1. When I ask vendors what they are doing to ensure interoperability between their tools and others' (sometimes even tools within the same company), the most common reply I get is, "Our tools are really the most interoperable. It is *their* tools that aren't". Can anyone care to explain what that means? 2. The consulting pie seems to have given a significant boost to some companies' revenues. What has it done to their long-term profitability? Is "value-based" selling a myth or reality in the consulting world? Do companies really expect to be profitable in consulting on "value based" propositions during hard times? If so, don't the tool-less consulting companies have an obvious edge over the conflict-of-interest'ing companies? Or, are the vendors giving their tools away for *free*? -Shankar Hemmady Gurutech To: All One firm advertises a "make like" utility for managing design files. That utility doesn't even monitor the date stamp of those files; the result is a nightmare. Another case adds insult to injury when a tech support person tells you that they will forward your "feature request" after you have just phoned in a system crashing bug. After rushing a bug-filled product out to market, exorbitant fees are charged for "maintenance". Is predatory marketing practices, deceitful advertising, and poor product quality going to be the EDA industry's continued future? - Bill McGinley Chisholm To: Joe Costello Last year Cadence had something like 7 to 9 major product announcements. This year I can think of only one. Cadence has also pulled out of EuroDAC this year. Is Cadence getting out of the EDA business? - John Cooley Holliston Poor Farm To: Wally Rhines, Joe Costello, Aart De Geus When are Mentor, Cadence and Synopsys going to deliver Windows products? - Abbie Kendall OrCAD To: All Q1: EDA companies are making it easier to create more and more complex designs. How are your companies' products going to make it possible to fully debug these complex designs before the heat death of the universe? Q2: Datapath designs are really easy to design, yet control heavy designs such as bus interfaces are notoriously difficult to design. Why are you automating the easy Datapath designs, and not doing much to help those of us with complex control heavy designs? [ Anon @ Intel ] To: All I am evaluating whether use VHDL or Verilog HDL for high level design in a situation where there is no prior baggage to carry. What are the advantages/disadvantages comparing the two? - Larry Sheu AMD To: All In a developing countries like India, EDA vendors enjoy selling to and then abandoning customers. With Cadence local support has been hopeless, since we bought design tools in DEC 1994. Synopsys support in India is not that bad and we are able to take out help from hotline. The main problem has been their O/S compatibility and recent S/w releases. Viewlogic (thro' its local representative ) always cribbs that PRO series of tools it sold to you thro' some other vendor is a less powerful tool and you should use OFFICE version of tools (obviously buy from us !). We do have Viewsynthesis on PC but because of some technical problem it still is unoperational. Why are developing countries EDA dumping grounds? - TANAY KRISHNA C-DOT, India To: Aart De Geus What's up with Synopsys and CCT and Synopsys and IBM and Synopsys and Cadence? Aren't these eventually mutually exclusive P&R solutions? Also, you've touted Behavioral Compiler for two years now, yet, from the fact that I very rarely see anything about it from customers on the Internet newsgroups tells me you just don't have many buyers. It behavioral synthesis too early for its time, a flop, or what? - John Cooley Holliston Poor Farm To: All When, and how, will we have the interoperability issue solved. I.e. when will be able to freely substitute one tool by the other, even from different vendors without spending so much time on converting formats, libraries, scripts, etc. As it seems today, it looks like some vendors prefer that to be the old fashion way, so if you locked into their system, you will not be able to try other systems. Sometimes it is even worse. For Example, Cadence was pushing hard for Verilog to be a standard, which finally came true, thru OVI. But Even Verilog XL is "not" OVI compliant, i.e. it has additional features that are not part of the OVI standard. So, even if we have a standard, it is not a real one (Not talking about edif, ...) - Erez Naory IES Electronics Agencies Ltd. To: Joe Costello Will EDA Vendors (who are becoming merchants of on-site EDA consulting) decide that their tools can give them a proprietary advantage over other designers begin to keep their best tools in-house? (This naturally implies that the tools they sell are out-house tools.) -Martin Miller Charles Stark Draper Laboratory To: Richard Goering Why are upgrades to EDA tools more a product of marketing departments, with enhancements that reflect trendy GUIs and zippy ICONs? What kind of statistics can a EDA manager take that give useful detailed information about design times and might aid in comparing button pressing and Icon grabbing versus quick key pressing? The Design community is very diverse but selecting and upgrading EDA tools is left more to touch and feel than to real measures for performance. Hardware graphics vendors will quote screen redraw rates or zoom rates but when have EDA vendors given us real numbers to compare? P.C. wholesalers are frequently evaluated in magazines on how fast hot line calls are answered and problems solved. Where are these numbers for the big EDA companies? - Marcus Larwill larwill@fnal.gov To: Gary Smith and All Since commercial EDA tools are inadequate, large internal EDA groups exist to meet corporate needs. As a result, commercial tools are of relatively lower value to the dominant part of the semiconductor market. As examples, DRC and schematic capture are essentially solved problems so they are viewed as commodities. Do you see the semiconductor companies seeing their specialized tool abilities as a key strategic advantage that they will not give up? - Doug Walker Texas A&M University To: All 1.) All EDA tools seem to be GUI-happy, with 413 levels of sub-menus, organized in some seemingly arbitrary fashion, obviously driven by some marketing guys that have never made a living at designing chips (i.e. good demos, but poor productivity). What are you doing to measure productivity of designers using your tools, and then increasing designer productivity based on your research? 2.) We've all wasted a bunch of time with Verilog versus VHDL religious wars - but really neither language is adequate and most designers don't give a shit and understand that "the smaller the difference the larger the war" (e.g. Pepsi vs. Coke) - otherwise it would be blatently obvious which one is better. What are you doing to discourage the language religion and to allow the desiger to concentrate on his design and raising the likelyhood of success instead of letting the poor guy wallow in the implementation and language details? - Victor J. Duvanenko Truevision To: All Migrating a toolset to a PC-based platform seems like it could be very difficult if the target operating system is DOS/Windows. If the code was originally written to work with Unix and X-Windows, it seems like porting to DOS/Windows would require alot of work. Have you given any thought to targetting low-cost PC-based Unix operating systems, especially Linux? Linux is a solution that many people are using to supplement Sun and HP workstations using lower cost PCs. - Tom Mayo General Electric To: Wally Rhines and All When are the EDA vending companies going to support PC's running some kind of UNIX like Linux or SCO or something else?? That way we could get the best of two worlds; cheap machines and a *real* operating system. Linux is now so mature that you can get an officially POSIX certified distribution (set of CD's). Digital is officially backing Linux! (When asked by Mentor what OS we'd like supported I answered Linux and they just said 'What's that?' (what a nightmare!)) - Michael Dantzer-Sorensen Nokia Mobile Phones To: All List the 3 things, in order, you have done that your customers feel have improved your software quality the greatest. Given unlimited resources, what one thing would you start doing NEW to improve software quality even more? I ask these questions because ECAD software is expensive, capabilities are over-sold, and quality is lacking. - John Swan Motorola To: All 1.) Could Car sales people do just as good a job at selling EDA tools as the current staff? Will the EDA industry support a return policy like CompUSA's or Software Etc? 2.) Why is Customer support so far down on the list of importance within a company and yet they pay lip service to their importance? (Internal negative attitudes to the ability of CAD/Customer Support Engineers abounds.) 3.) Why does the EDA industry believe in infinite growth patterns and infinite customer pockets? - [ Anon @ an EDA vender ] To: All When will the quality of user documentation be improved? Almost all user documents I have seen, and attempted to use, from the major EDA players are basically reformatted reference documents. There are very few examples to supplement syntax statements. The examples that exist frequently do not match the shipped code. Common problems are not listed. Time spent improving user documentation should not be considered time lost; every constructive person-hour spent improving documentation will equal at least that much time saved in hotline/AE hours. - Erik Kusko IBM To: All Why is it, that we could spend mega dollars for CAD-TOOLS, but we still have to spend mega-labor on libraries? Millions of Dollars of labor and Millions of hours of labor is wasted due a not uniformly supported library. - Robert Schultz M/A-COM To: Joe Costello You claim your customers have over 10 million user written lines of SKILL, yet TCL/TK seems to be a dominate command language and GUI builder for both PC and UNIX environments. What's happening with your extension language, SKILL? Is TCL/TK replacing it, or what? - Daniel Payne Mentor Graphics To: All While "deep-submicron" (0.35um) processing technology is available, tools that enable designers to take advantage of the technology are, in my opinion, not mature. (One indication of this is that the Cyrix's 6x86 implemented in 0.65um can still out-perform Intel's 0.35um-Pentium. Note: these are full-custom designs, not even ASICs.) Concerning this Deep Submicron rush: Do we need new tools & methodologies? If yes, what key characteristics do they have? Have these approaches been tested? What would be the indicator that tools/methodologies advocated by commercial EDA vendors are mature enough for production use? - Steven Leung Tandem Computers =========================================================================== Trapped trying to figure out a Synopsys bug? Want to hear how 4258 other users dealt with it ? Then join the E-Mail Synopsys Users Group (ESNUG)! !!! "It's not a BUG, jcooley@world.std.com /o o\ / it's a FEATURE!" (508) 429-4357 ( > ) \ - / - John Cooley, EDA & ASIC Design Consultant in Synopsys, _] [_ Verilog, VHDL and numerous Design Methodologies. Holliston Poor Farm, P.O. Box 6222, Holliston, MA 01746-6222 Legal Disclaimer: "As always, anything said here is only opinion."Article: 3449
husby@fnal.gov (Don Husby) wrote: > > david@fpga.demon.co.uk wrote: > > The "rough edges" referred to by recent reviewers here apply to any > > newly released product. > > I hope you're right. Although many of my complaints seem to be designed > in and, in fact, changes from the arguably better pro-series product. > > > I imagine that they looked at the first cut version (7.0) rather > > than later versions with the rough edges smoothed. The current > > release version of the Viewlogic product is, I believe, 7.11. > > I made my comments based on version 7.10. > Based on my experience with Viewlogic, I'd be willing to bet large > sums of money that no significant improvement will be made for at least > a year. > > The loss of the command line interface is a serious setback. I never realized how much I used it until I tried the new SW. I went back to my workview plus, partly because of shortcomings with 7.1 and partly because of win95 problems with some of my other tools. Hopefully some of this will be fixed by the time I am really ready to migrate to Win 95.
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