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
"Andreas Kugel" <kugel@ti.uni-mannheim.de> schrieb im Newsbeitrag news:3DC656A3.3080005@ti.uni-mannheim.de... > Hello. > One of our XC95288XL exhibits ecessive heating (untouchable!). It > contains a multiplexor to select an output clock from 3 different input > clocks. Two of the input clocks are attached to a GCK pin while one is > attached to an IO/GTS pin. When selecting the the non-GCK clock source > the parts start to heat-up dramatically. Frequencies are in the range > between 60 and 100MHz. The device contains 180 macrocells in high-perf > mode. Only about 20 macrocells are affected by the input clock > selection, the others are connected to a GCK clock. > > Any hints ? Open inputs? Short cicuits? -- MfG FalkArticle: 49176
"newb" <n/a@n/a.com> schrieb im Newsbeitrag news:ee7a255.-1@WebX.sUN8CHnE... > Hello, > > I was wondering if anyone had any urls, notes, or just general information on using an fpga with cpu core, such as risc, then connect to PC (usb, lpt etc) and have an application send the fpga cpu commands such as "add 3, 5" and the result will be returned to the PC (8). Thanks, newb. Have a look at xapp 213, it contains a plug&play RISC processor, 8 bit, very nice. Xapp 223 has a plug&play RS232 UART. -- MfG FalkArticle: 49177
A lot of people seek the same answer as you and as of yet I have not seen a satisfactory solution. You can do a piece wise (bottom up) synthesis with some complexity in Synplify but the P&R and usually the long pole. This may not be the answer you want to hear but these are the times that you make sure your simulation environment is rock solid so you don't end up getting into the "turn and burn" cycle on that FPGA. Every nite you get another chance to run your P&R so hoipefully that should be enough. Imagine you had an ASIC and it took 3 months for each trial. Regards President, Quadrature Peripherals Altera, Xilinx and Digital Design Consulting email: kayrock66@yahoo.com http://fpga.tripod.com ----------------------------------------------------------------------------- "Peng Cong" <pc_dragon@sohu.com> wrote in message news:<aq5di8$me2$1@mail.cn99.com>... > Hello, every one. > I'm working on a FPGA design project, it includes many module. > It'll take so long a time to run full implement one time when I only > make a little change in one module. > So I wonder how I can do incremental implement. I use Synplify Pro > and ISE 5.1i. > I've read the Xilinx document on how to do Incremental designing, but > not really understand, it seems that I need do it in Synplify Pro GUI mode. > Is it impossible to do it in ISE? And where can I find any example code? > > Thanks for any advenceArticle: 49178
I have had the same problem with FPGA Express and only using attributes for 'Reset' signal forced tool to stop implement sync_reset on 'D' path of FF and to use dedicated sync_reset. > I'm trying to write some VHDL that will infer an FDRE (ff with clock > enable and *sychronous* reset) in a Xilinx Virtex-E or -2 device. > > We've tried several versions of Synplify Pro 7.x, and they all waste a > LUT to implement the synchronous reset function.Article: 49179
Jay, That was very informative and I had a question for you. When you say "a large FPGA design into a small ASIC and 4x speedup", are there any constraints to it OR will all designs of FPGA result into such a speedup and area reduction. I understand why the area reduction and speedup occur. Its just the factor by which they occur that I'm not sure about. What would you say for a design that uses 20,000 LEs in an Altera Apex20K1500E device and about 40,000 bits of internal RAM working @ 40 MHz ? Would such a design speed up 4 times when converted to ASIC ? I understand that the speed up numbers would vary from design to design. But there must be a minimum and maximum possible numbers ? Interested in your comments. Thanks, Prashant kayrock66@yahoo.com (Jay) wrote in message news:<d049f91b.0211011522.4fd5cec2@posting.google.com>... > I've done a few FPGA prototypes of designs that we knew in advance > were going to be ASICs. We used TSMC and IBM. I never heard the > price on the TSMC and I think IBM was charging something like $400k > for chip #1. However, expect foundries to be very agreeable these > days on account of the surplus capacity. The results went fine, > because usually, by the time you've worked out all the bugs to get the > FPGA to work in the lab, you've solved most of your problems. Also > FPGA use sort of forces a certain level of simplicity with repect to > clocking. A BIG FPGA turns into a small ASIC because of the > difference in area efficiency. Also, expect about a 4X speed-up going > to ASIC. And of course, yes you can hand place, super pipeline, > embedded multiplier, etc your FPGA to get a faster design, but I'm > speaking in general for random logic writen by your average ASIC > designer, not spending all the time to get so deep into the > implimentation details. > > President, Quadrature Peripherals > Altera, Xilinx and Digital Design Consulting > email: kayrock66@yahoo.com > http://fpga.tripod.com > ----------------------------------------------------------------------------- > > > > "alla" <alng23@hotmail.com> wrote in message news:<aps669$cdd$1@tilde.itg.ti.com>... > > Just want see anyone here has any experience of converting a Xilinx FPGA > > design into an ASIC implementation. If so, which vendor did you use? What's > > the cost? Are you happy with the result? We are using the Virtex series and > > considering this option. ThanksArticle: 49180
Noel, > We bang on about QoR, size, area performance...why don't we use > schematic capture and layout, this will give us serious efficiency? > Probably because multi-million gate designs, or even multi-hundred > thousand gate designs would take an eternity to complete and the > product would be defunct before it got anywhere near the market. That's simply not true. The Alpha CPUs were designed using schematic capture, as well as many other multi-million gate designs. It IS entirely possible to design multi-million gate designs with schematics equally as fast and obviously far more efficiently with schematics. IF you know how to use schematics. You don't draw things by the gate, you use a library of heriarchical symbols, and it's basically drag and drop. AustinArticle: 49181
That GTS signal is a high capacitance net because its routed to every single I/O cell regardless of whether or not its used. You probably should pick another I/O. Also, why operate in high perf mode if the design is only 100MHz or less? I'm assuming that you know that the static power dissipation doubles for each macrocell you select for that mode. Regards President, Quadrature Peripherals Altera, Xilinx and Digital Design Consulting email: kayrock66@yahoo.com http://fpga.tripod.com ----------------------------------------------------------------------------- Andreas Kugel <kugel@ti.uni-mannheim.de> wrote in message news:<3DC656A3.3080005@ti.uni-mannheim.de>... > Hello. > One of our XC95288XL exhibits ecessive heating (untouchable!). It > contains a multiplexor to select an output clock from 3 different input > clocks. Two of the input clocks are attached to a GCK pin while one is > attached to an IO/GTS pin. When selecting the the non-GCK clock source > the parts start to heat-up dramatically. Frequencies are in the range > between 60 and 100MHz. The device contains 180 macrocells in high-perf > mode. Only about 20 macrocells are affected by the input clock > selection, the others are connected to a GCK clock. > > Any hints ? > > Thanks, > Andreas > > P.S. One idea is that the IO/GTS input drives the GTS net with the clock > even if the GTS signal is not used ???Article: 49182
Tomas Lopez wrote: > Has anyone had a problem with the latest Leonardo version (2002d) and the "virtex2_multipliers" variable? Even though I set it to FALSE, Leonardo insists on using one of the embedded multipliers. I haven't had a problem directly related to your question but Leonardo on the whole is horribly buggy. We've had to try different versions just to try and minimise the number of synthesis bugs we've been finding and we currently have 1 or 2 that we have to manually edit after synthesis! Move to Synplify if you can. Shareef.Article: 49183
Andreas, I would radically reduce the clock frequency ( down by at least a factor ten), and then check the temperature. If ( and I do not believe it) it's dynamic power, the temperature should be much lower. If, as I suspect, it's internal or external contention, the power stays the same... Peter Alfke =========================== Andreas Kugel wrote: > Hello. > One of our XC95288XL exhibits ecessive heating (untouchable!). It > contains a multiplexor to select an output clock from 3 different input > clocks. Two of the input clocks are attached to a GCK pin while one is > attached to an IO/GTS pin. When selecting the the non-GCK clock source > the parts start to heat-up dramatically. Frequencies are in the range > between 60 and 100MHz. The device contains 180 macrocells in high-perf > mode. Only about 20 macrocells are affected by the input clock > selection, the others are connected to a GCK clock. > > Any hints ? > > Thanks, > Andreas > > P.S. One idea is that the IO/GTS input drives the GTS net with the clock > even if the GTS signal is not used ???Article: 49184
This is a multi-part message in MIME format. --------------98F33DBA5A2ABD0EB25F5EC3 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Andreas, It seems strange that it only heats up when you have the GTS input selected. How is the mux implemented - are you sure that the Synthesis tool hasn't implemented the mux as tristates? If you are seeing this - I think you may be causing contention. Dave Andreas Kugel wrote: > Hello. > One of our XC95288XL exhibits ecessive heating (untouchable!). It > contains a multiplexor to select an output clock from 3 different input > clocks. Two of the input clocks are attached to a GCK pin while one is > attached to an IO/GTS pin. When selecting the the non-GCK clock source > the parts start to heat-up dramatically. Frequencies are in the range > between 60 and 100MHz. The device contains 180 macrocells in high-perf > mode. Only about 20 macrocells are affected by the input clock > selection, the others are connected to a GCK clock. > > Any hints ? > > Thanks, > Andreas > > P.S. One idea is that the IO/GTS input drives the GTS net with the clock > even if the GTS signal is not used ???Article: 49185
ssy wrote: > I use APEX20k400E, and I use the lpm_ram_dp in quartus II to construct > my register file, all input signal is registered , so read and write > are all sync > > so if I read and write the same address of this memory at the same > clock edge, can I get the just writen value? You get an unspecified value. You have to add logic to arbitrate this. For a FIFO, this is the "empty" state. You can't pop more than you push. -- Mike TreselerArticle: 49186
"Peng Cong" <pc_dragon@sohu.com> wrote in message news:<aq5di8$me2$1@mail.cn99.com>... > Hello, every one. > I'm working on a FPGA design project, it includes many module. > It'll take so long a time to run full implement one time when I only > make a little change in one module. > So I wonder how I can do incremental implement. I use Synplify Pro > and ISE 5.1i. > I've read the Xilinx document on how to do Incremental designing, but > not really understand, it seems that I need do it in Synplify Pro GUI mode. > Is it impossible to do it in ISE? And where can I find any example code? > > Thanks for any advence I think synplify 7.2 has a feature for what you want. It's called Multipoint synthesis. http://www.synplicity.com/literature/pdf/synpro_multi_xilinx.pdf There is a tutorial in this apps note.Article: 49187
My design is using too many slices for the device I am targetting (XCV2000), although my LUT and slice flip flop count is under 100%. Any suggestions for logic and / or slice reduction? Number of slices: 5,246 out of 5,120 Number of slice flip flops: 6,175 out of 10,240 60% Number of 4 input LUTs: 8,516 out of 10,240 83%Article: 49188
Amy, What is your design entry tool? Austin "Amy Mitby" <amyks@sgi.com> wrote in message news:2d2a8f5d.0211041056.edc3a80@posting.google.com... > My design is using too many slices for the device > I am targetting (XCV2000), although my LUT and slice > flip flop count is under 100%. Any suggestions for logic > and / or slice reduction? > > Number of slices: 5,246 out of 5,120 > Number of slice flip flops: 6,175 out of 10,240 60% > Number of 4 input LUTs: 8,516 out of 10,240 83%Article: 49189
One of the most cited reasons for using C is the claimed simulation speed, write in C & the simulations is bound to be vastly faster than the same behavioural code in Verilog/VHDL. Somthing like an FFT in C (classic rad 2) will run orders faster than same in behavioral HDL. I was of the same opinion too, but eventually the C model still had to be rewritten in Verilog to synthesize it. I specifically used an RTL C style of coding in instances so the Verilog RTL was almost a 1 to 1 map. Problem was the C code looked bloody ugly, hierarchy was very difficult to manage, & was manually levelised and ended up being a very poor mans Verilog. Worse still, all the revisions were done to the Verilog & not all the way back to the C as it was often quite difficult to express in poor mans C what is easy in HDL. At the end of the road I even had the C env auto magically spit out the corresponding Verilog using lots of ugly macros that did both simulation & equiv HDL code output. This was maybe ok in the gate level world, lots of companies did this too! I got real tired of doing that flow. The cleverer the tool got to be more like Verilog, the slower the sims became. In the end I junked it & wrote a direct RTL subset of Verilog to C compiler. NOW I can model in RTL Verilog AND I have the speed of C in simulation. There is a penalty, the backend doesn't optimize as well as it could, but it is only a few times slower than raw C. Expect to fix that one day, gee might even add x86 raw output since passing the C code to VC6 tends to really burden it (100k lines in a fn() will kill most C compilers). I call it Vpp (like the old Cpp from Bell Labs) or sometimes V2C. One day might put it out under open source (BSD/MIT) if there is enough interest. Other than that, I have to agree with Ray & others, there is some room for HandelC, Occam, StreamsC, CSP etc for Reconfigurable Computing with a FPGA board to speed up a PC, but to create an actual ASIC or FPGA product, I am much more skeptical? I would expect to see a performance penalty of at least a decade or so but the more the C app looks like a DSP flow, the better the result will close in to hand RTL. I have seen the Celoxica demos & I was quite impressed with the Venus de Milo show at the Xilinx VirtexPro fair last spring. Now I would like to draw your attention to CSP or Occam. 40 yrs ago Tony Hoare was a real live EE & did design work in RTL (Resister Transister Logic hehehe). he later moved into SW (at Oxford U) & later became world famous for his work in Par programming. In v1 of his work, I believe (& so does he now) that he & others got it all wrong, hence this is why the SW world has such a hard time with Par programming since most SW people work with spinlocks, P & V, deadlocks, mutexes, semaphores etc. These have nothing to do with HW abstraction so they don't help make that Par code any use to HW design. It is also why Java & even C# with threads are Thread unsafe & why that kind of Par programming is extremely hard when HW guys find Par so easy but Seq much harder. In v2 of his work only a few years later (1968 IIRC), he must have sobered up & remembered he was once a HW guy. CSP & esp Occam is a HDL (or PDL) using processes & messages to describe logic. Channels in Occam are essentially equiv to wires connecting the outputs of one process to the inputs of another. Now the Transputer executed Occam programs in a similar way to how a HDL simulator works. It included a round robin scheduler to effect the par execution of all the par expressions, & shuffling the outputs back to inputs (eventually). At Inmos, Occam was even used to write a simulator & to describe the T800 FPU. So I though I might also mention I am working on a modern Transputer called T2 for FPGA using Vpp etc and lots of work still to do. Perhaps next yr more news. Now with a Transputer in FPGA, you really could could array them in the bigger FPGAs and benefit from the near linear increase in performance since using lots of Transputers was alot like partitioning a big HW block into smaller blocks with the interconnect timeshared over the channels/links. I have now come to the conclusion that Transputing is the soft side of RC & FPGAs are the hard side, both complimentary and somewhat interchangeable. Now if you put a Transputer or 2 (or more) in FPGA, it is not only the ideal embedded cpu, but it makes it easy to divy up the project into HW (FPGA) & SW (T/Occam/.. code). my 2cArticle: 49190
Jan wrote: > > If there *are* other 16-bit C-programmable soft processor cores > under (say) 400 LUTs out there, please let us know. > When last compiled to an XC4005E a couple of years back, my homebrew 16/32 bit RISC weighed in at 335 LUTs in 16 bit datapath mode, including onboard RAM/ROM/IO ports: http://groups.yahoo.com/group/fpga-cpu/message/118 The 16 bit datapath hasn't been tested very much lately, as XST has many bugs handling constant range checking inside conditional generates, and won't compile the VHDL source in 16 bit mode without manual edits. An alpha version under LGPL should be ready by year's end (as unsupported free code with no warranty, support, etc.) BrianArticle: 49191
john jakson wrote: > <snip good stuff > > I have now come to the conclusion that Transputing is the soft side of > RC & FPGAs are the hard side, both complimentary and somewhat > interchangeable. Now if you put a Transputer or 2 (or more) in FPGA, > it is not only the ideal embedded cpu, but it makes it easy to divy up > the project into HW (FPGA) & SW (T/Occam/.. code). > > my 2c Interesting post. What's the status of Transputer tools ? Don't STm still have some flavour of the Transputer targeting the set-top-box market ? - jgArticle: 49192
Hi! I'm new to FPGAs and i don't know in which language should i begin to design( vhdl, verilog, c). I currently have Webpack 4.1. regardsArticle: 49193
I think Synplify has come up with something called multi-point synthesis. They say its used for incremental synthesis. Try searching their docs. "Peng Cong" <pc_dragon@sohu.com> wrote in message news:<aq5di8$me2$1@mail.cn99.com>... > Hello, every one. > I'm working on a FPGA design project, it includes many module. > It'll take so long a time to run full implement one time when I only > make a little change in one module. > So I wonder how I can do incremental implement. I use Synplify Pro > and ISE 5.1i. > I've read the Xilinx document on how to do Incremental designing, but > not really understand, it seems that I need do it in Synplify Pro GUI mode. > Is it impossible to do it in ISE? And where can I find any example code? > > Thanks for any advenceArticle: 49194
Hello Peng, This is a common complaint for people doing large FPGA designs. I work for a new EDA company called Hierarchical Design Inc. We are coming to market with an FPGA floorplanner to solve this specific issue as well as a host of others. I'd be happy to share some more details about our product with you. Feel free to contact me at anytime. Regards, Brian Jackson Vice President, Business Development Hierarchical Design Inc. www.hierdesign.com 408-982-8277 "Peng Cong" <pc_dragon@sohu.com> wrote in message news:<aq5di8$me2$1@mail.cn99.com>... > Hello, every one. > I'm working on a FPGA design project, it includes many module. > It'll take so long a time to run full implement one time when I only > make a little change in one module. > So I wonder how I can do incremental implement. I use Synplify Pro > and ISE 5.1i. > I've read the Xilinx document on how to do Incremental designing, but > not really understand, it seems that I need do it in Synplify Pro GUI mode. > Is it impossible to do it in ISE? And where can I find any example code? > > Thanks for any advenceArticle: 49195
Amy, Just wanted to make sure you're aware that you can change the pack factor in map to allow the merging of unrelated logic into your slices. It's design dependent and can hurt timing, but I'd guess that you could probably use it to slim your slice utilization down below 100%. For information check out the DSR Guide at: http://support.xilinx.com/support/sw_manuals/xilinx5/index.htm and reference the section on MAP. Regards, Tim Jaynes AAE Amy Mitby wrote: > My design is using too many slices for the device > I am targetting (XCV2000), although my LUT and slice > flip flop count is under 100%. Any suggestions for logic > and / or slice reduction? > > Number of slices: 5,246 out of 5,120 > Number of slice flip flops: 6,175 out of 10,240 60% > Number of 4 input LUTs: 8,516 out of 10,240 83%Article: 49196
My name is Brian Jackson and I am the business development manager for Hierarchical Design Inc. We are an EDA company bringing the next generation heirarchical physical floorplanner to market for Xilinx VirtexII devices. Our product allows you to easily partition and floorplan your FPGA design. It allows incremental updates from synthesis and will perform block based PAR using Xilinx ISE tools. The results can then be assembled and modified inside of the floorplanner. There are quite a few more features that I won't bother to get into here. We are currently looking to sign up a small FPGA design consultant as a technology development partner to beat on it within a "real" design environment. This way we can work out any kinks before making a big splash in the market. Preferrably, this would be someone in the Silicon Valley area. We would compensate with either licenses or with a prearranged consulting fee. If you are interested and do complex FPGA designs in VirtexII, please don't hestitate to contact me. Regards, Brian Jackson Vice President, Business Development Hierarchical Design Inc. www.hierdesign.com brian@hierdesign.com 408-982-8277Article: 49197
Hi, I've done several experimental designs with multiple CPUs without any problems. One hint, though, in case you use custom instructions: each custom instruction is defined and synthesized as its own design entity (whether that is .v, .vhd, .edf, etc). If you have two CPUs and each has a custom instruction, there can be some tool confusion if both CPUs use a custom instruction of the same name (as all SOPC Builder HDL code is generated in a common project directory). You can easily safeguard this by ensuring each processor's custom instructions (if any) have unique names inside the SOPC Builder GUI. Have fun with this - I'd be happy to hear about the progress of your project! Regards, Jesse Kempa satchit_h@hotmail.com (satchit) wrote in message news:<ddf018f6.0210300811.32d31548@posting.google.com>... > Hi, > > has anybody tried a 2-nios parallel design using SOPC builder? I am > trying to build one which can run separate pieces of code > concurrently. Any comments would be helpful. thanks, > > regards, > SatchitArticle: 49198
"Andreas Kugel" <kugel@ti.uni-mannheim.de> ha scritto nel messaggio news:3DC656A3.3080005@ti.uni-mannheim.de... > One of our XC95288XL exhibits ecessive heating > (untouchable!). What do you mean with "untouchable"? It could be 50-60 C, which is completely safe. -- LorenzoArticle: 49199
Does anyone know of an available VME Master core or reference design? Scott Munroe smunroe@phillipsaerospace.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