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
Dragon wrote: > 1. Count three 'ones' per CLB: > With 2 sum bits you could count at most 3 ones. To count 16 ones you would need > 5 CLBs and have one bit left over. Now you have five 2-bit numbers with one bit > left over. If you add pairs of 2-bit numbers, you need 3 bits per sum. Since we > have 3 pairs, we need 9 more CLBs. Now we have three 3-bit numbers. We can use > 5 CLBs to add them together to get the five-bit result. This solution uses 19 CLBs > with three levels of logic. > > 2. Use all 9 inputs per CLB: > To count 9 ones we need a 4-bit result. In 4 CLBs we count the first 9 ones. In > 3 CLBs we count the remaining 7. Now we have a 4-bit and a 3-bit number. > Use 5 more CLBs to get the five-bit result. This solution uses 12 CLBs and > two levels of logic. > > 3. Somewhere in the middle: > With 2 sum bits you could count at most 3 ones. To count 16 ones you would need > 5 CLBs and have one bit left over. Now you have five 2-bit numbers with one bit > left over. Add three of the pairs of 2-bit numbers into a 4-bit sum. This would take > 4 CLBs. Add the other two pairs and one left over bit into a 3-bit sum. This > would take 3 CLBs. Use 5 more CLBs to get the five-bit result. This solution > uses 17 CLBs and three levels of logic. > > Looks like solution 2 is the best way to go. If speed was not the issue, and you > had access to a clock, I would go with Ray's suggestion of a shift reg and counter. > Shift the bits and tap one stage to the clock enable of a 5-bit binary counter. > That would need only 5 CLBs. > > - Craig Unfortunately solution 2 uses the false assumption that any 9-input function can be mapped to one CLB. You cannot count 9 ones with just 4 CLBs. CatalinArticle: 20351
Uwe Bonnes wrote: > rk <stellare@nospam.erols.com> wrote: > : Hal Murray wrote: > > : For the previous post, I just noted that you can paste notes with Acrobat; obviously > : not with a CD-ROM. The data part of the CD-ROM is about 500 megabytes, so copying the > : whole thing over to the hard disk is not too bad. That does seem to work ok. > : However, the directory and file names are not very descriptive so I couldn't be > : selective in what I loaded and I would rather not load the whole thing. My CD-ROM > : reader isn't that fast so I like to have things on the hard disk, which is pretty > : quick. > > I always wonder who produces those data book CDs. In most cases tons of > senseless DLL are uploaded to your system ( some old TI data books were > great at this) and files and directories have sensles names. > > The recent TI CD has for every PDF file a dirctory of the same name. Argh! > > Bye I guess people who make $ making data book CD's. :-) I do not like when you have to run .exe's and load .dll's and modify this program and that (I believe this one links into the browser). I just want the computer to, er, compute and enough of this stuff and it won't work and one will never find out why. On the other hand, the Xilinx guys did give me a hardcopy databook and I find that I do use that most of the time, then their www site, and finally the CD-ROM. Have a nice day, ---------------------------------------------------------------------- rk The ability to carry out scientific stellar engineering, ltd. observations at a distance is stellare@erols.com.NOSPAM developing so rapidly that I don't Hi-Rel Digital Systems Design see any unique role for man in planetary exploration. -- Gordon MacDonald, National -- Academy of Sciences, 1968Article: 20352
rk <stellare@nospam.erols.com> wrote in message news:3896D039.4C2DF0B6@nospam.erols.com... > Of course, we must mention that > schematics are portable across humanoids. Two humanoids reading the same schematic will > understand the design exactly the same. Two humanoids reading the same VHDL (or Verilog or > C++ or whatever) code can not understand the design exactly the same way since they are > shielded from the design (except if the design is completely structural). ** > Unless my synthesizer is inserting or removing FFs from my design (FPGA Express doesn't seem to), then the only thing synthesis seems to shield me from is the implimentation of the combinatorial logic I specify. If I draw schematics of the same logic using the conventional primitives (and, or, not), the mapper still has to convert that logic into look up tables (in the cases of brand X and brand A), so I'm still shielded from the logic implementation. What's the difference? As an example, my synthesizer tends to remove the redundant logic required to overlap state decodes so that an output doesn't glitch. The solution is to register state machine outputs that can't glitch. Doesn't the mapper do the same thing to the logic in a schematic? Paul ButlerArticle: 20353
Maybe someone asked this question before. but i don't understand why the designers of the Virtex family did not include the possibility to download from an EPROM without an additional PLD. The XC4000 family had this feature and I liked it very much. Does Xilinx has to much low density pld and want to ship with every virtex one of these. But I think a lot of Virtex users will take an Altera for this job. Or is there a more lucid explanation ? Kai -- ----- Dipl. Ing. Kai Troester ------------------------------------- IMMS - Institut fuer Mikroelektronik- und Mechatronik-Systeme gGmbH Langewiesener Strasse 22, 98693 Ilmenau, Germany Tel: +49(3677)6783-42 Fax: +49(3677)6783-38 mailto:kai.troester@imms.de http://www.imms.de/~troester -------------------------------------------------------------------Article: 20354
Its a mix of html, pdf and ppt files, which would be fine except they are all named wcd#####.pdf/html/ppt and are separated into folders called wcd#####. Makes it kind of hard to find anything without going through their program. Hal Murray wrote: > > HELLO XILINX, do you hear all this. S**tcan the stuff you need to install and > > give us plain old PDFs like you used to do. The latest Applinx disk is not very > > useful for much anything but a high tech frisbee. > > Is most of the info on the CD normal pdf files? Does the new stuff > do anything useful? If so, it should be possible to have both modes. > > Run the installer if you want that mode. Bypass it and browse > directly if you don't want it "helping" you. > > -- > These are my opinions, not necessarily my employers. -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 20355
Its a mix of html, pdf and ppt files, which would be fine except they are all named wcd#####.pdf/html/ppt and are separated into folders called wcd#####. Makes it kind of hard to find anything without going through their program. Hal Murray wrote: > > HELLO XILINX, do you hear all this. S**tcan the stuff you need to install and > > give us plain old PDFs like you used to do. The latest Applinx disk is not very > > useful for much anything but a high tech frisbee. > > Is most of the info on the CD normal pdf files? Does the new stuff > do anything useful? If so, it should be possible to have both modes. > > Run the installer if you want that mode. Bypass it and browse > directly if you don't want it "helping" you. > > -- > These are my opinions, not necessarily my employers. -- -Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email randraka@ids.net http://users.ids.net/~randrakaArticle: 20356
I have a clock that I want to bring in on a regular I/O pin. I have it assigned to a non-BUFGP pin in the .ucf file...but prior to using that file, Synplicity forces the use of a BUFGP... I tried this: input I_S_LF_SCK /* synthesis xc_padtype = "IBUF" */ ; And that didn't stop it from using the BUFGP. When I run the Xilinx tools, it tells me I can't use the regular I/O pin with a BUFGP, and I need to get Synplicity to NOT assign this input to a BUFGP. Any suggestions? The design is in Verilog, and I don't want to use the 'black box' instantiations if I can avoid it.Article: 20357
I found it: /* synthesis syn_noclockbuf = 1 */ seems to work fine... Austin Franklin <austin@dar88kroom.com> wrote in article <01bf7185$382cc370$207079c0@drt1>... > I have a clock that I want to bring in on a regular I/O pin. I have it > assigned to a non-BUFGP pin in the .ucf file...but prior to using that > file, Synplicity forces the use of a BUFGP... > > I tried this: > > input I_S_LF_SCK /* synthesis xc_padtype = "IBUF" */ ; > > And that didn't stop it from using the BUFGP. When I run the Xilinx tools, > it tells me I can't use the regular I/O pin with a BUFGP, and I need to get > Synplicity to NOT assign this input to a BUFGP. > > Any suggestions? The design is in Verilog, and I don't want to use the > 'black box' instantiations if I can avoid it. > > >Article: 20358
On Fri, 04 Feb 2000 19:42:13 GMT, in comp.lang.vhdl Mikhail Matusov <matusov@ANNTIsquarepegSPPAMM.ca> wrote in article <pYFm4.131$2a6.154350@news.magma.ca>: > Mike Treseler <tres@tc.fluke.com> wrote in message > news:389B1405.8529A3F8@tc.fluke.com... >> >> I define some boolean constants and jump around them. > Can you tell me please how exactly you are jumping? I am kinda > missing a syntax book at the moment so I am not sure what kind of > "goto" statement I could use... Generate statement as it was pointed > out in another post requires code to be syntactically correct and > what is worse I believe it works only inside of processes. If it is > not generate then what is it? if ... generate works OUTSIDE of processes. if ... then works INSIDE of processes. Neither is a preprocessor construct, so all code, excluded or not, must be at least statically semantically correct. You can also use any number of preprocessors. See the FAQ, located at http://vhdl.org/comp.lang.vhdl, for more info. Paul -- Paul Menchini | mench@mench.com |"Outside of a dog, a book is Cadence Design Systems | www.orcad.com | probably man's best friend, and P.O. Box 71767 | 919-479-1670[v] | inside of a dog, it's too dark Durham, NC 27722-1767 | 919-479-1671[f] | to read." --Groucho MarxArticle: 20359
Apologies for any intrusion but I am looking for ASIC Designers to work on the very latest in digital broadcast projects in the UK. Please mail me for more info. You should be eligible to work in the UK. See www.vertex-solutions.co.uk for other exciting opportunities Sent via Deja.com http://www.deja.com/ Before you buy.Article: 20360
Hi We are 2 engineers sharing 1 FPGA Express license. It should be all right excepted that the project manager won't launch if it doesn't find a valid license, so only 1 can work at a time. I thought the floating license was only used with Synopsys, not the whole program. Now that my design is synthesized, I can't improve the p&r if the other one uses Foundation at the same time? Nicolas MATRINGE DotCom S.A. Conception electronique 16 rue du Moulin des Bruyeres Tel 00 33 1 46 67 51 11 92400 COURBEVOIE Fax 00 33 1 46 67 51 01 FRANCEArticle: 20361
I would recommend Xilinx Foundation because it includes Synopsys FPGA Express, which I think most would agree is a better HDL synthesizer than Altera's... You also get a state machine entry program, which might make it easier for beginners... Alas, neither come with a good VHDL simulator... I use ModelSim, so I can't comment on the simulator in Foundation... I know you can use the Aldec VHDL simulator right in the Foundation design flow, but that's extra money. I don't think you can put a VHDL simulator seamlessly in the MAX-PLUS II design flow, but I'm not sure on this... Altera's graphical waveform simulator is very easy to use for small designs. MAX-PLUS II is quite mature and easy to use, but Xilinx has improved in this regard in the last few years. In my limited use of the schematic editors, they seem to be a wash. Foundation is a better platform for VHDL, though. Of course, you should see what sort of deal you can cut with the vendors. Cheap (or free) is good... Good Luck, JJ Shawki Areibi wrote in message <3891C2B3.9FB4614E@uoguelph.ca>... >We are introducing a new course on Digital Design at the School of >Engineering >at the University of Guelph. We would like to choose between the Altera >MAXII-Plus and the Xilinx Foundation Series Software. >Anyone with the experience of using the two systems? Which is better to >use >by the students (i.e simpler to use e.t.c) >Thanks, >Shawki Areibi > >-- >Shawki Areibi >Assistant Professor >School of Engineering >University of Guelph >Guelph, Ont, Canada N1G 2W1 >Tel: (519) 824-4120 >Fax: (519) 836-0227 > >Article: 20362
Nicolas Matringe wrote: > > Hi > We are 2 engineers sharing 1 FPGA Express license. It should be all > right excepted that the project manager won't launch if it doesn't find > a valid license, so only 1 can work at a time. > I thought the floating license was only used with Synopsys, not the > whole program. Now that my design is synthesized, I can't improve the > p&r if the other one uses Foundation at the same time? Looks like I had a problem and Synopsys had not released the license after ending. Everything works fine now. Sorry. Nicolas MATRINGE DotCom S.A. Conception electronique 16 rue du Moulin des Bruyeres Tel 00 33 1 46 67 51 11 92400 COURBEVOIE Fax 00 33 1 46 67 51 01 FRANCEArticle: 20363
Paul Butler wrote: > Unless my synthesizer is inserting or removing FFs from my design (FPGA > Express doesn't seem to), then the only thing synthesis seems to shield me > from is the implimentation of the combinatorial logic I specify. If I draw > schematics of the same logic using the conventional primitives (and, or, > not), the mapper still has to convert that logic into look up tables (in the > cases of brand X and brand A), so I'm still shielded from the logic > implementation. What's the difference? It is possible to make a schematic using only native logic blocks. In this case the schematic will be very close to the primitive netlist. However, if the schematic designer uses higher level schematic elements like shift registers or large muxes, your point is well taken. The schematic is no longer directly related to the netlist. A schematic of TTL functions is very easy to read for some humanoids. Others find collections of algorithmic processes easier. Mappers and synthesizers are getting smarter with each release. Since you can get 1000 gates for less than $.40 in large devices, I think designers should use whatever description lets them work fastest. > As an example, my synthesizer tends to remove the redundant logic required > to overlap state decodes so that an output doesn't glitch. The solution is > to register state machine outputs that can't glitch. Doesn't the mapper do > the same thing to the logic in a schematic? There may be arcane ways to shut off logic reduction. But I agree with you, that synchronizing inputs and outputs is both easier and more reliable. -Mike TreselerArticle: 20364
In article <J9Qj4.8623$uI1.146653@typhoon2.kc.rr.com>, "Matt Billenstein" <mbillens@one.net> wrote: > We're actually getting into the layout now and I'm not so scared now that > I've found the "correct" dimensions and a suggested escape routing note for > the FG parts... it turns out the solder balls on the FPGA are substantially > larger than the foot print that is suggested goes on the board... thus, I > get a ton of more room between pads to route stuff and place vias. I'll > forward the appnote to anyone who'd like to see; I don't think it can be > found on the Xilinx website. > > m > > Matt Billenstein > http://w3.one.net/~mbillens/ > mbillens@one.net > I would love to see this appnote. Please forward the appnote (or a link to it) to me. Thanks! -- Greg Neff VP Engineering *Microsym* Computers Inc. greg@guesswhichwordgoeshere.com Sent via Deja.com http://www.deja.com/ Before you buy.Article: 20365
I have an application where I would like to be able to remotely re-program a FLASH configuration PROM (XC18V00). The PROM and FPGA reside on an expansion card inside a PC. Basically I would like to be able to download a new FPGA configuration file to the Host over the network, re-program the Flash PROM, and then re-configure the FPGA with the new code. The FPGA will be connected to an embedded processor which will handle the task of interfacing to the Host via the PCI bus and passing the configuration bits to the FPGA, which will then have the task of re-programming the Flash, I'm guessing through the JTAG port. At that point the embedded processor will force the FPGA to reconfigure. Does anyone have any experience or ideas on how to do this? We basically want a "simple" way to download feature changes and/or bug fixes to the FPGA without having to remove the board from the PC. Thanks in advance for any help you can provide. -Paul P.S. I have read a little about "Xilinx Online" and think that this method is more complicated than what we need.Article: 20366
<mench@mench.com> wrote in message news:87ms3h$nns$1@mench.mench.com... > On Fri, 04 Feb 2000 19:42:13 GMT, in comp.lang.vhdl Mikhail Matusov > if ... generate works OUTSIDE of processes. I am probably doing something dumb but I can't get this thing through the syntax check. Here is the idea (not the actual code as I don't want to post several pages) of what I am doing: library IEEE; use IEEE.std_logic_1164.all; entity XXX is port ( ); end XXX; architecture xxx_rtl of XXX is constant enableP2 : BOOLEAN := TRUE; begin P1: process (...) begin ... end process; if enableP2=TRUE generate P2: process(...) begin ... end process; end generate; end xxx_rtl; Any suggestions why I am getting syntax error on if enableP2 line? I tried constants, signals, variables of different types but it seems to be something else. If I use simple "if" it passes. I am using latest rev of the Xilinx Foundation tools. Thanks, Mikhail MatusovArticle: 20367
Mikhail Matusov wrote: > if enableP2=TRUE generate > > P2: process(...) > begin > ... > end process; > > end generate; > > end xxx_rtl; > > Any suggestions why I am getting syntax error on if enableP2 line? I tried > constants, signals, variables of different types but it seems to be > something else. If I use simple "if" it passes. I am using latest rev of the > Xilinx Foundation tools. Try if enableP2 generate -Mike TreselerArticle: 20368
Hey guys, I recently found a website where I was able to determine my market value based on my experience, industry, geography, and skills. It is an easy way to get information about salary and market trends. My individual market value was e-mailed back to me and I was better able to make decisions about my career based on this valuable information. I just thought I'd pass this along since it is so important to be aware of your current market value. Check it out.....http://www.gdsweb.com/survey.asp Mark Enwright Firmware Engineer Real-Time Solutions 716.425.8009 --Article: 20369
Paul Butler wrote: > Unless my synthesizer is inserting or removing FFs from my design (FPGA > Express doesn't seem to), then the only thing synthesis seems to shield me > from is the implimentation of the combinatorial logic I specify. If I draw > schematics of the same logic using the conventional primitives (and, or, > not), the mapper still has to convert that logic into look up tables (in the > cases of brand X and brand A), so I'm still shielded from the logic > implementation. What's the difference? First I would like to point out that FPGA Express is perfectly capable of removing FFs. I tried to do a one hot FSM and found that I only got 15 FFs in a 16 state design. The IDLE state was being coded as no FFs set. This was very disturbing to me since it meant that my FSM logic would change depending on how I used the outputs! If I used the IDLE output, it would remove one of the other FFs. This made it very hard to simulate since I had to constantly redefine my vectors. The other really big problem I had was that the synthesizer was capable of vastly enlarging my logic by nearly an order of magitude. Again in the one hot FSM, I found it very difficult to find the exact coding style that would prevent the synthesizer from decoding all 15 FSM variables (FF outputs) as the inputs to all 15 variables. Kinda defeated the purpose of one hot encoding. In general I always had to play with the code to get a good implementation of any logic. The mapper never blows up my logic in that way. > As an example, my synthesizer tends to remove the redundant logic required > to overlap state decodes so that an output doesn't glitch. The solution is > to register state machine outputs that can't glitch. Doesn't the mapper do > the same thing to the logic in a schematic? If this is all the synthesizer does, you are lucky. My project changed synthesizers after I found that the one I started with was for all purposes non-functional. That meant I had to adopt a new coding style to get a good implementation. Then Xilinx changed synthesizers on me with an upgrade. That meant I had to change coding styles again. Each time required that I reverse engineer the logic produced and use trial and error to get good results. I will go back to and HDL someday, but it won't be on any design where I need to optimize for speed or size. It will have to be a project where I need to meet a tight deadline with no constraints on chip size or recurring costs. -- Rick Collins rick.collins@XYarius.com remove the XY to email me. 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: 20370
Austin Franklin wrote: > I have a clock that I want to bring in on a regular I/O pin. I have it > assigned to a non-BUFGP pin in the .ucf file...but prior to using that > file, Synplicity forces the use of a BUFGP... > > I tried this: > > input I_S_LF_SCK /* synthesis xc_padtype = "IBUF" */ ; Try: input I_S_LF_SCK /* synthesis syn_noclockbuf=1 */ ; > > > And that didn't stop it from using the BUFGP. When I run the Xilinx tools, > it tells me I can't use the regular I/O pin with a BUFGP, and I need to get > Synplicity to NOT assign this input to a BUFGP. > > Any suggestions? The design is in Verilog, and I don't want to use the > 'black box' instantiations if I can avoid it.Article: 20371
Rickman wrote: > Paul Butler wrote: > > Unless my synthesizer is inserting or removing FFs from my design (FPGA > > Express doesn't seem to), then the only thing synthesis seems to shield me > > from is the implimentation of the combinatorial logic I specify. If I draw > > schematics of the same logic using the conventional primitives (and, or, > > not), the mapper still has to convert that logic into look up tables (in the > > cases of brand X and brand A), so I'm still shielded from the logic > > implementation. What's the difference? Of course, it depends on the manufacturer that you're using. There are other brands. When I use schematics, it is often at the hard macro level and I know exactly what I am getting. All I need to do is look at what I drew on the schematic. Some of the newer back end software has some optimizations in it. If I don't want a circuit element to disappear since it is not "logically needed," I can attach a preserve attribute and it stays just the way I put it down. For many of the designs that I do with schematics, I would say that I am not shielded from the logic implementation at all. ================ > First I would like to point out that FPGA Express is perfectly capable > of removing FFs. I tried to do a one hot FSM and found that I only got > 15 FFs in a 16 state design. The IDLE state was being coded as no FFs > set. This was very disturbing to me since it meant that my FSM logic > would change depending on how I used the outputs! If I used the IDLE > output, it would remove one of the other FFs. This made it very hard to > simulate since I had to constantly redefine my vectors. > > The other really big problem I had was that the synthesizer was capable > of vastly enlarging my logic by nearly an order of magitude. Again in > the one hot FSM, I found it very difficult to find the exact coding > style that would prevent the synthesizer from decoding all 15 FSM > variables (FF outputs) as the inputs to all 15 variables. Kinda defeated > the purpose of one hot encoding. In general I always had to play with > the code to get a good implementation of any logic. The mapper never > blows up my logic in that way. If you read an HDL file, as you point out, you don't know how the logic is mapped. Your experience with the "pretty much" one hot logic is similar to what I've seen from several synthesizers. Of course, by just looking at HDL code, how can you tell if the state assignment is one-hot, binary, gray, or whatever unless there are directives in it? Frequently they are not there, you just get a piece of HDL code to look at. I've seen logic synthesizers do "logic replication" unless you get it shut off. As I've seen someone else here post previously, the back end software can do so also. Another case that I've seen, when analyzing a failed design, is when a piece of CAE code took a flip-flop and an inverter and converted it into two flip-flops, one with a complemented output. The two flip-flops wound up in an "inconsistent state" and since the outputs were controlling some high current switches in a motor, well, BOOM! While originally mis-diagnosed as a failed FPGA - how can a flip-flop output and it's inversion be the same value? - it turned out the designer was not aware of how the CAE software helped him out. Lastly, if it is important to the application, one must be able to ensure that there are no lockup sets of states in the state machine. For many of the applications that I do, that is important. Many times the output of the synthesizer is dependent on manufacturer, version, input (as I saw, same as Rick, dependent on what output of the FSM that you use), phase of the moon, etc. Some applications it's a don't care. For many hi-rel systems, people do care. Have a good evening, rkArticle: 20372
Dear all If we found these errors in XC4020E, how can I do? I have already install .BSDL file of XC4020E. ERROR:basut - Boundary-scan chain test failed at bit position '1' on instance 'mult_c(Device1)'. Check that the cable, system and device JTAG TAP connections are correct, that the target system power supply is set to the correct level, that the system grounds are connected and that the parts are properly decoupled. ERROR:basut - Boundary scan chain has been improperly specified. Please check your configuration and re-enter the boundary-scan chain information. Boundary-scan chain validated unsuccessfully. ERROR:basut - : The boundary-scan chain has not been declared correctly. Verify the syntax and correctness of the device BSDL files, correct the files, reset the cable and retry this command. Please suggest me. Best Regard WannaratArticle: 20373
Hi, I would like to obtain more info about EDIF especification. Thanks. -- Jose (a_k_y@hotmail.com)Article: 20374
Hi Everybody. I want to design a FPGA (virtex E) with the following features: I've got 2 data sources DataA and DataB, they are synchronous and they toggle at 50 MHz. I want to process them with the same component. Therefore, i use a DLL to generate a 100 MHz Clock with the 50 MHz data Clock. Now, i have to switch between DataA and DataB to generate the input data of the component. Therefore, I use the output CLK90 (or CLK270) of the DLL. DLL_Inst: CLKDLL port map( CLKIN => Clk50MHz, -- Input Clk CLKFB => ClkSimple, -- 50 MHz Clk RST => Rst, CLK0 => CLK0, -- 50 MHz Clk CLK90 => CLK90, CLK180 => CLK180, CLK270 => CLK270, CLK2X => CLK2X, -- 100 MHz Clk CLKDV => CLKDV, Locked => Locked ); BufGSimple: BUFG port map ( -- 50 MHz Clk I => CLK0, O => CLKSimple ); BufGDouble: BUFG port map ( -- 100 MHz Clk I => CLK2X, O => CLKDouble ); GenClkSimpleR : process (ClkDouble, Rst) begin -- process GenClkSimpleR if Rst = '1' then ClkSimpleR <= '0'; elsif ClkDouble'event and ClkDouble = '1' then ClkSimpleR <= CLK270; end if; end process GenClkSimpleR; SwitchData : process (ClkDouble, Rst) begin -- process SwitchData if Rst = '1' then DataCompIn <= (others => '0'); elsif ClkDouble'event and ClkDouble = '1' then if ClkSimpleR = '1' then DataCompIn <= DataA; else DataCompIn <= DataB; end if; end if; end process SwitchData Component_Inst : Comp port map( Clk => ClkDouble, DataIn => DataCompIn , DataOut => DataCompOut ); Now, my problem is I want to put a timing constraint on the net CLK270 and i don't know how to do this. This constraint can't neither be FF to FF nor Pad to FF. Does anybody know how to specify it ? (I Use Foundation) Second question : what about the skew between ClkSimple and ClkDouble has anybody got any problem with this ? J-P GOGLIO GETRIS S.A. 13 Chemin des Prés 38240 Meylan Tel : (33) 4 76 18 52 10 Fax : (33) 4 76 18 52 01 E-mail : goglio@getris.com
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