Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search

Messages from 37600

Article: 37600
Subject: Re: Certicom challenge and FPGA based modular math
From: "Carl Brannen" <carl.brannen@terabeam.com>
Date: Mon, 17 Dec 2001 11:55:33 +0000 (UTC)
Links: << >>  << T >>  << A >>
Jay, I think that the link you gave is to a Xilinx implementation that would be
suitable for elliptic curves over the finite field F2m, rather than elliptic
curves over the finite field Fp.  I think I mentioned that F2k would be a
somewhat kinder, gentler problem for FPGAs, but they could rip solutions out
for Fp as well.

Here's the Certicom challenge you linked to (i.e. ECCP-109) :
http://www.certicom.com/research/ch42.html

Here's the one that the article would be good for solving (i.e. ECC2-k):
http://www.certicom.com/research/ch4.html

By the way, here's a link to the C code that's required for the ECCP problems.
It looks to me like it would be easily put into an FPGA, and that it wouldn't
take up a lot of bandwidth on the PCI interface if done correctly.  (I'd try to
put all the math algorithms on, and only write control to the PCI, and read
back results on the PCI interface that were "distinctive".

http://www.nd.edu/~cmonico/eccp109/downloads/eccp109-130-3.tar.gz

Carl


-- 
Posted from firewall.terabeam.com [216.137.15.2] 
via Mailgate.ORG Server - http://www.Mailgate.ORG

Article: 37601
Subject: division 64
From: Mardin <chens_w@yahoo.com.cn>
Date: Mon, 17 Dec 2001 05:23:08 -0800
Links: << >>  << T >>  << A >>
i want implement a division  operation in FPGA.Eg.it divide by 64
how do i?

Article: 37602
Subject: Re: SPI interface in VHDL
From: alex_shrab@hotmail.com (AlexS.)
Date: 17 Dec 2001 05:32:54 -0800
Links: << >>  << T >>  << A >>
Hello, Jason.
The SPI-4 IP Cores are available from both ALTERA and Xilinx. In our
project, I use the ALTERA'S one. It could be freely downloaded from
the ALTERA'S site for evaluation.

http://www.altera.com/products/ip/altera/m-alt-posphy4.html
 
It does not include a lot of VHDL code but encrypted netlists. But 
there is a 'USER GUIDE' which gives some information how it is
implemented. It could be useful.
Alex.

Article: 37603
Subject: Re: division 64
From: next <chensw20@hotmail.com>
Date: Mon, 17 Dec 2001 05:34:48 -0800
Links: << >>  << T >>  << A >>
By the way ,how do it implement if it use addition,subtration,multiplication,division ?

Article: 37604
Subject: Re: Multiplying by squaring using Block RAM.
From: Christopher.Saunter@durham.ac.uk (Christopher Saunter)
Date: Mon, 17 Dec 2001 14:22:43 +0000 (UTC)
Links: << >>  << T >>  << A >>
Hi Carl,
	You da man!

I was thinkng about using a BlockRAM to do squaring in a project of mine,
more for the latency issues than device usage in my particular case, and
was left with one 'regular' multiply.  I like your sugestion.  I'll try it
out in the new year and get back to the group with any issues it raises,
or sucesses.

It seems so simple...  I guess I don't need to feel so bad about only
having Virtex hardware to play with, and no hardware multipliers ;-)

As an aside, thanks to the posters who answered my previous article about
modifying blockram contents in a bitstream, I haven't got round to trying
anything yet, but I will do...

Regards,
	Chris Saunter

 Carl Brannen (carl.brannen@terabeam.com) wrote:
: I'm not sure if people are doing this already, but I couldn't find a reference
: on the Xilinx web site.

: Block RAMs make more efficient squaring circuits than they do multipliers.  And
: you can get multipliers out of squarers.

: An explanation for the arithmetic.  Let A and B be the numbers to be
: multiplied.

: Compute C = (A+B) * (A+B) = A**2 + 2AB + B**2
: Compute D = (A-B) * (A-B) = A**2 - 2AB + B**2
: Then C - D = 4AB.

: This is particularly efficient in Xilinx Spartan2, Virtex, and Virtex2
: architectures because the block RAM is dual port.  That means you can use one
: side for the (A-B)**2 calculation and the other side for the (A+B)**2
: calculation.

: With the Xilinx Spartan2, Virtex or VirtexE, use the RAMB4_16_16.  It has 8
: inputs and 16 outputs in two sections.  Each section can conveniently compute
: the square of an 8-bit number.  Note that the lowest two bits of the two
: squares are going to have to be equal (i.e. C-D = 4AB, so C and D have to match
: two bits), so you don't have to subtract bits 1 and 0 of the two squares.

: If "A" and "B" are both 7-bit, their sum will be no worse than 8-bit, so you
: can compute a 7x7 multiply using only the 8 LUTs for each of "A+B" and "A-B",
: and another 14 LUTs for the result, a total of 30 LUTs (i.e. 15 slices) and one
: block RAM.  Maybe there's a way to get the bit back, and let A and B be 8-bit
: numbers; I haven't looked at it long enough to conclude there isn't.

: The circuit uses about half the LUTs required by the standard algorithm, at an
: expense of one block RAM.

: To put the LUT utilization in perspective, the Xilinx 8x8 multiply takes 39
: slices:
: http://www.xilinx.com/ipcenter/reference_designs/vmult/vmult_v1_4.pdf

: Using RAMB4s alone to implement even a 7x7 multiply would require a huge number
: of them, as multiplies require twice as many address inputs as squares.

: You can iterate on the calculation of the square.  That is, if A is too big to
: square in a single operation, then break A into two parts.  With A broken into
: two parts, say A = AH + AL, you can compute AH**2, AL**2 with block RAM, and
: compute 2*AH*AL by computing the difference between (AH+AL)**2 and (AH-AL)**2.

: Breaking A and B into more than 3 parts may be worth exploring, for certain bit
: sizes.

: Carl


: -- 
: Posted from firewall.terabeam.com [216.137.15.2] 
: via Mailgate.ORG Server - http://www.Mailgate.ORG

Article: 37605
Subject: Re: Certicom challenge and FPGA based modular math
From: Philip Freidin <philip@fliptronics.com>
Date: Mon, 17 Dec 2001 14:38:31 GMT
Links: << >>  << T >>  << A >>
On Sun, 16 Dec 2001 21:09:17 -0800, "Jay Berg" <admin@eCompute.org> wrote:
>Let me see if I can explain this. But given that I'm not a math expert, bear
>with me.
>
>Modulo math (also known as "clock arithmetic") can be thought of as using
>remainders. Imagine the following numbers.
>
>  ....
>
>Since the need is for 128-bit multiplication (128x128=256), the result of
>the multiplication can be 256-bits in size. Following the multiplication,
>the 256-bit result is reduced by the modulus value N. This translates the
>result into a number between 0 and (N-1). With the assumption that N is
>128-bits (or less), the final result of the modulo multiplication will be
>128-bits (or smaller).

So, is the N you want to use a
    variable ( 1 .. (2^128)-1 )
    constant( 1 .. (2^128)-1 )

or easiest of all, the number (2^128)-1


Philip Freidin
Fliptronics

Article: 37606
Subject: How to initialize the block ram of xilinx SpartanII FPGA?(Verilog)
From: "deerlux" <deerlux@hotmail.com>
Date: Mon, 17 Dec 2001 22:43:51 +0800
Links: << >>  << T >>  << A >>
I use the following code to initialize the block RAM of SpartanII FPGA as
the handbook of Xilinx:
module MYMEM(clk,we,addr,din,dout);
input clk,we;
input [7:0]din;
output [7:0]dout;
input [8:0]addr;
wire logic0,logic1;
assign logic0=1'b0;
assign logic1=1'b1;

RAMB4_S8
ram0(.WE(we),.EN(logic1),.RST(logic0),.CLK(clk),.ADDR(addr),.DI(din),.DO(dou
t));
// synopsys translate_off
defparam ram0.INIT_00=
256h'0123456789ABCDEF0123456789ABCDEF0123456789ABCDEF0123456789ABCDEF;
defparam ram0.INIT_01=
256'hFEDCBA9876543210FEDCBA9876543210FEDCBA9876543210FEDCBA9876543210;
// synopsys translate_on
endmodule
But when I implement it using Xilinx Foundation3.1i.It shows INIT_xx
including INIT_00 and INIT_01 not initialized.I use Synopsys FPGA Express to
synthesis.
Who can tell me how to initialize the block RAM?
Thank you!






Article: 37607
Subject: Re: division 64
From: Peter Alfke <palfke@earthlink.net>
Date: Mon, 17 Dec 2001 15:28:08 GMT
Links: << >>  << T >>  << A >>


Mardin wrote:

> i want implement a division  operation in FPGA.Eg.it divide by 64
> how do i?

Vry simple:
Just shift the binary number six positions beyond the LSB.

Peter Alfke, Xilinx Applications



Article: 37608
Subject: Re: division 64
From: "Pallek, Andrew [CAR:CN34:EXCH]" <apallek@americasm01.nt.com>
Date: Mon, 17 Dec 2001 10:31:33 -0500
Links: << >>  << T >>  << A >>
If you just want to devide by 64, shift right by 6 places.  The modulo is what was shifted
out.

To do the others, I recomend reading HDL Chip Design, by Douglas J. Smith.
Chapter 9 has a good number of examples on how to implement various
arethmetic functions, and pros and cons of each.

next wrote:

> By the way ,how do it implement if it use addition,subtration,multiplication,division ?


Article: 37609
Subject: Re: Certicom challenge and FPGA based modular math
From: "Jay Berg" <admin@eCompute.org>
Date: Mon, 17 Dec 2001 07:57:12 -0800
Links: << >>  << T >>  << A >>
The current SW does not lend itself well to parallel processing at the level
we're talking. The results of the first series of operations provide the
results used to achieve the next series. Block operations would be quite
difficult to achieve without redoing the entire SW application.

I'll start work on thinking how to redo the method into a process that would
allow working multiple results in a parallel fashion.

        Jay


"Steven Derrien" <sderrien@irisa.fr> wrote in message
news:3C1DBD59.ECD96E70@irisa.fr...
>
> Hello,
>
> On a typical PCI FGPA bord, it is likely that your performance is limited
by the
> PCI bandwidth rather than by the FPGA processing power. Assuming N fixed,
you
> need 3*128 bits (96 bytes, 2Wr 1Rd) I/O  per iteration.
>
> If you use PCI MMAP IOs, you will hardly get more than 15MBytes/sec
between the
> host and the board. This poses a bound on your achievable peformance
(15/96*10^6
> Mul operation per second  ~ 150 000 Mul/sec), which will be less than what
you
> get by software.
>
> If your alogorithm has no data dependecies between the different
multiplication
> results (which I doubt), you could use blocked I/O (or DMA) operations,
and
> maybe reach 60-80Mbytes/sec, but even then, you would not get more than 1
> million multiplication per second.
>
> The only solution would be to implement a larger part of the algorithm
(like a
> whole loop nest) on the FGPA board, which is much more difficult (unelss
your
> algorithm is very regular, and requires little control) but this generally
> reduces the amount of I/O operations on the PCI bus.
>
> Steven




Article: 37610
Subject: Re: Certicom challenge and FPGA based modular math
From: Christopher.Saunter@durham.ac.uk (Christopher Saunter)
Date: Mon, 17 Dec 2001 16:02:42 +0000 (UTC)
Links: << >>  << T >>  << A >>
I recently came across a 'Pilchard' - an FPGA prototype system built on an
SDRAM module, which when used in a Linux host allows very fast PC -
cryptoprocessor etc. comms.  This project was headed by Prof. Philip Leong
at the Chinese University of Hong Kong.  

They used a Virtex 300 chip, in QFP format, which I believe is pin
compatible with larger chips (although if you used a Virtex 800 on there,
you might want to run a seperate power line to the card...)

I have nothing to do with this proejct, other than I saw it and thaught
'that's sweet'... ;-)

Try a google search for: pilchard fpga
There is a .pdf availible but I can't find the address right now... (It's
linked to from slashdot.................)

Best of luck with the project!

Cheers
	Chris Saunter
 Jay Berg (admin@eCompute.org) wrote:
: After making the mistake of getting involved in the current ECCp109
: distributed computing project (see URL below), I'm now casting around to
: determine if there's a possibility of finding a PCI board with an FPGA
: co-processor capable of handling a small set of modular math functions.

< big snip>

Article: 37611
Subject: Re: Certicom challenge and FPGA based modular math
From: "Jay Berg" <admin@eCompute.org>
Date: Mon, 17 Dec 2001 08:17:09 -0800
Links: << >>  << T >>  << A >>
        <see below for comments>


"Carl Brannen" <carl.brannen@terabeam.com> wrote in message
news:382a3bda7feccf0799b4a71c935b2c57.51709@mygate.mailgate.org...
> Jay, I think that the link you gave is to a Xilinx implementation that
would be
> suitable for elliptic curves over the finite field F2m, rather than
elliptic
> curves over the finite field Fp.  I think I mentioned that F2k would be a
> somewhat kinder, gentler problem for FPGAs, but they could rip solutions
out
> for Fp as well.
>
> Here's the Certicom challenge you linked to (i.e. ECCP-109) :
> http://www.certicom.com/research/ch42.html
>
> Here's the one that the article would be good for solving (i.e. ECC2-k):
> http://www.certicom.com/research/ch4.html
>
> By the way, here's a link to the C code that's required for the ECCP
problems.
> It looks to me like it would be easily put into an FPGA, and that it
wouldn't
> take up a lot of bandwidth on the PCI interface if done correctly.  (I'd
try to
> put all the math algorithms on, and only write control to the PCI, and
read
> back results on the PCI interface that were "distinctive".
>
> http://www.nd.edu/~cmonico/eccp109/downloads/eccp109-130-3.tar.gz
>
> Carl


Yes, the last URL (for the SW download) you provided is for the ECCp109
challenge that I'm speaking of.

There are a total of three paths through the math to achieve results. Each
is very slightly different, but share many characteristics. As you can see,
the only steps used are a series of add, subtract, and multiply. With each
of the math operations being modulo of the same N value. Also note that all
functions use a parameter list in the format of:
        function (result, inputX, inputX)
And last, each of the three paths result in a single result value which is
checked to for being one of the target values.


Path 1 - Total of input parameters needed: 5
        PY
        PX
        op_list[i].y
        op_list[i].x
        needInverting [i<<2]

    submod_p109 (lambda, PY, op_list[i].y);
    mulmod_p109 (lambda, lambda, &needInverting [i << 2]);
    addmod_p109 (temp_ul, op_list[i].x, PX);
    mulmod_p109 (temp2_ul, lambda, lambda);
    submod_p109 (tempx, temp2_ul, temp_ul);
    submod_p109 (temp_ul, op_list[i].x,  tempx);
    mulmod_p109 (temp_ul, lambda, temp_ul);
    submod_p109 (res_list[i].y, temp_ul, op_list[i].y);


Path 2 - Total of input parameters needed: 5
        QY
        QX
        op_list[i].y
        op_list[i].x
        needInverting [i<<2]

    submod_p109 (lambda, QY, op_list[i].y);
    mulmod_p109 (lambda, lambda, &needInverting [i << 2]);
    addmod_p109 (temp_ul, op_list[i].x, QX);
    mulmod_p109 (temp2_ul, lambda, lambda);
    submod_p109 (tempx, temp2_ul, temp_ul);
    submod_p109 (temp_ul, op_list[i].x,  tempx);
    mulmod_p109 (temp_ul, lambda, temp_ul);
    submod_p109 (res_list[i].y, temp_ul, op_list[i].y);

Path 3 - Total of input parameters needed: 4
        A
        op_list[i].y
        op_list[i].x
        needInverting [i<<2]

    mulmod_p109 (temp_ul, op_list[i].x, op_list[i].x);
    addmod_p109 (temp2_ul, temp_ul, temp_ul);
    addmod_p109 (temp2_ul, temp2_ul, temp_ul);
    addmod_p109 (lambda, temp2_ul, A);
    mulmod_p109 (lambda, lambda, &needInverting [i << 2]);
    mulmod_p109 (temp_ul, lambda, lambda);
    submod_p109 (temp_ul, temp_ul, op_list[i].x);
    submod_p109 (tempx, temp_ul, op_list[i].x);
    submod_p109 (temp_ul, op_list[i].x, tempx);
    mulmod_p109 (temp_ul, lambda, temp_ul);
    submod_p109 (res_list[i].y, temp_ul, op_list[i].y);


If it was possible to put all three paths into firmware, it would be easy
enough for the SW to preload the correct parameters and trigger the correct
firmware operation. Is this in line with what you were thinking?




Article: 37612
Subject: Re: division 64
From: "Wolfgang Loewer" <wolfgang.loewer@elca.de>
Date: Mon, 17 Dec 2001 17:28:25 +0100
Links: << >>  << T >>  << A >>
Dividing by a power of two is easy and can be accomplished by just shifting
in the direction of the LSB.
In case you want to implement more sophisticated divisions check out the LPM
(Library of parameteized Modules) in the Altera tools MAX+plus II and
Quartus II. It contains a free divide Megafunction called LPM_DIVIDE as well
as VHDL and Verilog simulation models that go with it.

- Wolfgang
http://www.elca.de

"Mardin" <chens_w@yahoo.com.cn> schrieb im Newsbeitrag
news:ee73c01.-1@WebX.sUN8CHnE...
> i want implement a division  operation in FPGA.Eg.it divide by 64
> how do i?



Article: 37613
Subject: Re: Certicom challenge and FPGA based modular math
From: "Jay Berg" <admin@eCompute.org>
Date: Mon, 17 Dec 2001 08:34:45 -0800
Links: << >>  << T >>  << A >>
The N (modulo) value for the ECCp109 challenge is:
    00001BD579792B380B5B521E6D9FB599

As you can see, it does not utilize the full 128-bit domain. As I said
elsewhere in the thread, N does not change during operation. But when the
current challenge is completed, a new challenge would be launched and the N
factor would need to become a new value.

But if we are to move the entire equation into firmware rather than discrete
math operations, the firmware solution becomes unique to this challenge. As
a result, a new design would be needed for later challenges.

Be aware that the current challenge was started in April and is currently
estimated to be about 14% completed. It is believed that 58 million DPs
(Distinguished Points) will need to be found to achieve a solution. As of
this morning, there have been 8.6 million DPs found.

To give an idea of the complexity of this challenge, I have one 450p2 that
has been running the client for the last month or so. It has a total of
459,756,170,368 iterations performed (max of 190,000 iterations per second)
and has found a total of 829 DPs. Each iteration consists of approximately
8-11 math operations on 128-bit numbers.

            Jay Berg
            jberg@eCompute.org



"Philip Freidin" <philip@fliptronics.com> wrote in message
news:i01s1u445chatt8sjimupf54m04j8gmp4n@4ax.com...
> On Sun, 16 Dec 2001 21:09:17 -0800, "Jay Berg" <admin@eCompute.org> wrote:
> >Let me see if I can explain this. But given that I'm not a math expert,
bear
> >with me.
> >
> >Modulo math (also known as "clock arithmetic") can be thought of as using
> >remainders. Imagine the following numbers.
> >
> >  ....
> >
> >Since the need is for 128-bit multiplication (128x128=256), the result of
> >the multiplication can be 256-bits in size. Following the multiplication,
> >the 256-bit result is reduced by the modulus value N. This translates the
> >result into a number between 0 and (N-1). With the assumption that N is
> >128-bits (or less), the final result of the modulo multiplication will be
> >128-bits (or smaller).
>
> So, is the N you want to use a
>     variable ( 1 .. (2^128)-1 )
>     constant( 1 .. (2^128)-1 )
>
> or easiest of all, the number (2^128)-1
>
>
> Philip Freidin
> Fliptronics



Article: 37614
Subject: Re: SPI interface in VHDL
From: arichter@atmel-du.de (Alexander Richter)
Date: 17 Dec 2001 09:13:57 -0800
Links: << >>  << T >>  << A >>
Hi Jason,

look at http://www.disi.unige.it/person/AnconaD/Architettura/vhdl_man/spi_ex.htm

Regards
Alex

"Jason Berringer" <jberringer@trace-logic.com> wrote in message news:<S5bT7.2519$NC5.476993@news20.bellglobal.com>...
> Hello again
> 
> I'm curious to know if anyone out there knows where there are some examples
> of an SPI interface coded in VHDL. Just curious as I have to code one in the
> near future and I always like to compare the various approaches taken by
> others.
> 
> Thanks
> 
> Jason

Article: 37615
Subject: Xilinx ChipScope - experiences ?
From: Rick Filipkiewicz <rick@algor.co.uk>
Date: Mon, 17 Dec 2001 17:43:07 +0000
Links: << >>  << T >>  << A >>

As usual I'm in the position of trying to shut the stable door when the
horse is already 2 counties away and accelerating fast but ...

Has anyone on C.A.F used the ChipScope ILA stuff ?

Does it work as advertised ?

Had sucesses/failures ?

Does it take up a lot of space per embedded analyser ?

In short where does it lie in the spectrum

[essential ... helpful ... difficult to use ... waste of time & gates] ?


Article: 37616
Subject: research on fast carry chains for FPGA
From: saliard30@caramail.com (supaman)
Date: 17 Dec 2001 09:50:04 -0800
Links: << >>  << T >>  << A >>
hello,
i'm new in this domain of computer science, so i don't know where to
look for site or article on the specific sibject, that is :
fast carry chains for FPGA.

i've searched on citeseer.nj.nec.com/cs and found nothing interesting.

so, do you have any idea where i can find resources about this
subject?

supaman.

ps : i'm mainly interested in article, i have already 2 :
"high-performance carry chains for FPGA's" by S. Hauck, M. M. Hosler,
and T. W. Fry.
"FPGA adders : performance evaluation and optimal design" by Shanzhen
Xing, and William W.H. Yu.

Article: 37617
Subject: Re: Dual-port ram templates
From: David Dye <davidd@xilinx.com>
Date: Mon, 17 Dec 2001 12:24:11 -0700
Links: << >>  << T >>  << A >>
Ray and Adarsh,

XST can infer dual port RAM.  Supported templates are documented in the XST User
Guide within the Xilinx online documenation.  Not every possible variant is
supported; please let us know if the configuration you wish to use is not listed.

FPGA Express does not infer any RAM.

thanks,
david.

David Dye
Xilinx Boulder


Ray Andraka wrote:

> XST won't infer a dual port RAM, and I don't believe express will either, which
> is why you are getting FF's.   Instantiate the RAMB4_S16_S16 directly from the
> unisim library.  Any generics have to be made invisible to the synthesizer
> using translate_off/on pragmas (and you'll have to put matching attributes in
> to pass the parameters to the netlist).
>
> It might help for you to post the errors along with snippets of your code.
>
> adarsh wrote:
>
> > we are having a similar problem.
> > we need a  Dual Port Ram for our design and were trying to instantiate one
> > of the Block Rams available on the Virtex - E device using Verilog.
> > Initially we just declared the required memory as an array of
> > registers.something like
> >
> > reg [15:0]  memory [255:0] ;
> > The synthesizers (XST, FPGAexpress) are not inferring this as a RAM but as
> > FFs
> >
> > Then we tried with the Xilinx Language template  RAMB4_S16_S16, this gave a
> > synthesis error.
> > Last we tried with CorGen that gives an error when we hit the Generate
> > button.
> >
> > Any suggestions ?
> >
> > adarsh kumar jain,
> >
> > Ray Andraka wrote:
> >
> > > I think he was looking for a low cost or free tool that would infer
> > > one.  If your tools do not support RAM inference, then you can always
> > > instantiate the RAM primitive.  I usually just instantiate the primitive
> > > because it gives me more portability between tools and more flexibility
> > > in describing what I want (and you don't have to rely on the tool for
> > > doing the right thing, especially in regards to a dual ported memory).
> > >
>
> --
> --Ray Andraka, P.E.
> President, the Andraka Consulting Group, Inc.
> 401/884-7930     Fax 401/884-7950
> email ray@andraka.com
> http://www.andraka.com
>
>  "They that give up essential liberty to obtain a little
>   temporary safety deserve neither liberty nor safety."
>                                           -Benjamin Franklin, 1759


Article: 37618
Subject: Problems Installing Foundation 4.1 Under '98SE
From: lowry@htc.honeywell.com (Dave Lowry)
Date: 17 Dec 2001 19:32:11 GMT
Links: << >>  << T >>  << A >>
Has anbody done this?  The installer complains about MFC42.DLL and hiccups but
continues.  When in Project Manager everything synthesis-related is grayed-
out.  I've got my LM_LICENSE_FILE var pointed to \xilinx\data\license.loadngo.

Any help most appreciated.  Thanks.

-Dave

Not speaking for my employer, etc.

Article: 37619
Subject: Re: Configuring Xilinx FPGA through parallel port
From: Nick Macias <nmacias@cellmatrix.com>
Date: Mon, 17 Dec 2001 13:45:16 -0700
Links: << >>  << T >>  << A >>
Steven,

I too use the parallel port for both FPGA confiuration and
post-configuration control. I've been using a switchbox (the kind for
sharing two printers on one port) and that works great. If that's an
option for you, it might be easier than debugging the PCI port.

Occasionally, things get in a mode where I get the DONE FAILED TO ASSERT
message repeatedly, but cycling power on the BurchEd board seems to fix
it.

	Nick

Steven Derrien wrote:
> 
> Hello,
> 
> This might be slighlty off-topic, but I guess several people in this NG
> had to face this kind of problem.
> We are using a BurchEd Board, with a parallel port download cable,
> however, because we need to communicate with the board once it is
> configured, we use two parallel ports, the one on the motherboard (for
> communication in EPP mode) and another one connected on a PCI parallel
> port extension board (using netmos 9705 chip) for configuration*.
> 
> The PCI // port, does not work properly when it comes to configure the
> FPGA board (I managed to make it work for a week or so, but now for a
> mysterious reasons, the FPGA DONE signal does not behave correctly). BTW
> configuration with the motherboard // port works fine.
> 
> The general PCI // port behavior is correct (checked by feeding-back
> CTRL signal on STATUS), so I really don't understand where this problem
> is coming from. Has anybody faced the same kind of problems ?
> 
> * We have no choice since the PCI board does not seem to allow anything
> else than SPP
> 
> Thank you for your help,
> 
> Steven

Article: 37620
Subject: Re: newbie Xilinx Foundation ISE4.1 questions
From: Andy Peters <andy@exponentmedia.nospam.com>
Date: Mon, 17 Dec 2001 22:58:29 GMT
Links: << >>  << T >>  << A >>
sdfjsd wrote:
> 
> Last year, I used Xilinx Foundation Express 3.3i, to develop
> for a Virtex300 part.  I recently went to Xilinx's hoomepage,
> and found that the 'Foundation ISE' has replaced the older
> Foundation (non-ISE.)
> 
> Does this mean :
> 
>   1)  goodbye old Windows 16-bit legacy code
>      (3.3i would crash on average, every 4-6 compiles, and
>       sometimes take down my NT4 workstation)

What the heck were you doing with your computer that it would cause it
to crash?  I ran NT 4 SP6 with the Xilinx service packs, and it never
crashed the computer.

--a

Article: 37621
Subject: Re: division 64
From: Andy Peters <andy@exponentmedia.nospam.com>
Date: Mon, 17 Dec 2001 23:02:53 GMT
Links: << >>  << T >>  << A >>
Peter Alfke wrote:
> 
> Mardin wrote:
> 
> > i want implement a division  operation in FPGA.Eg.it divide by 64
> > how do i?
> 
> Vry simple:
> Just shift the binary number six positions beyond the LSB.

Even simpler: select the bits you want.  F'rinstance,

	signal foo : unsigned (15 downto 0);
	signal result : unsigned (9 downto 0);

	result <= foo (15 downto 6);

You could probably use an alias to do the same thing...

OK, so I'll admit that it's not bleedingly obvious that you want to
divide by 64, and I should've written the example in Verilog, but...


-------a

Article: 37622
Subject: Download byteblast circuit with byteblasterMV mode(Maxplus II baseline) :
From: xmchen_eda@yahoo.com (shawn chen)
Date: 17 Dec 2001 15:31:03 -0800
Links: << >>  << T >>  << A >>
Hi,

    I am debugging an Altera FPGA board with old "byteblaster"
interface on board ( Just a 74HC244 ). Now I have only
Maxplus II baseLine software which support only "byteblasterMV".
    The circuit of "byteblasterMV" added some pull up resistors,
and also other slight difference.

Problem : When program with JTAG chain through "byteblasterMV" mode,
       MaxplusII can't detect the device on board (An EPC2L20 ).


Question:
    1. Should I change the circuit on board to "byteblasterMV"? 
(I mean cut some wire, add some resistors).
    2. Or maybe there is some patch or driver can be installed
into Maxplus II baseline, make it support old byteblaster mode.

    Thank you very much.

Shawn

Article: 37623
Subject: Re: division 64
From: next <chensw20@hotmail.com>
Date: Mon, 17 Dec 2001 16:35:15 -0800
Links: << >>  << T >>  << A >>
Thanks!
how should i  get the remainder ?

Article: 37624
Subject: Re: division 64
From: Philip Freidin <philip@fliptronics.com>
Date: Tue, 18 Dec 2001 01:10:30 GMT
Links: << >>  << T >>  << A >>
On Mon, 17 Dec 2001 16:35:15 -0800, next <chensw20@hotmail.com> wrote:

>Thanks!
>how should i  get the remainder ?

It's the six bits you were going to throw away.

Philip Freidin
Fliptronics



Site Home   Archive Home   FAQ Home   How to search the Archive   How to Navigate the Archive   
Compare FPGA features and resources   

Threads starting:
1994JulAugSepOctNovDec1994
1995JanFebMarAprMayJunJulAugSepOctNovDec1995
1996JanFebMarAprMayJunJulAugSepOctNovDec1996
1997JanFebMarAprMayJunJulAugSepOctNovDec1997
1998JanFebMarAprMayJunJulAugSepOctNovDec1998
1999JanFebMarAprMayJunJulAugSepOctNovDec1999
2000JanFebMarAprMayJunJulAugSepOctNovDec2000
2001JanFebMarAprMayJunJulAugSepOctNovDec2001
2002JanFebMarAprMayJunJulAugSepOctNovDec2002
2003JanFebMarAprMayJunJulAugSepOctNovDec2003
2004JanFebMarAprMayJunJulAugSepOctNovDec2004
2005JanFebMarAprMayJunJulAugSepOctNovDec2005
2006JanFebMarAprMayJunJulAugSepOctNovDec2006
2007JanFebMarAprMayJunJulAugSepOctNovDec2007
2008JanFebMarAprMayJunJulAugSepOctNovDec2008
2009JanFebMarAprMayJunJulAugSepOctNovDec2009
2010JanFebMarAprMayJunJulAugSepOctNovDec2010
2011JanFebMarAprMayJunJulAugSepOctNovDec2011
2012JanFebMarAprMayJunJulAugSepOctNovDec2012
2013JanFebMarAprMayJunJulAugSepOctNovDec2013
2014JanFebMarAprMayJunJulAugSepOctNovDec2014
2015JanFebMarAprMayJunJulAugSepOctNovDec2015
2016JanFebMarAprMayJunJulAugSepOctNovDec2016
2017JanFebMarApr2017

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

Custom Search