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 18000

Article: 18000
Subject: No Subject
From: Mark Harvey <mark.harvey@farsystems.it>
Date: Wed, 22 Sep 1999 11:28:21 +0200
Links: << >>  << T >>  << A >>
Does anyone know how to the time between two consecutive events on a signals
for reporting in a testbench?
(I basically need to check & report a pulse width)


Thanks in advance,

Mark Harvey.



 Sent via Deja.com http://www.deja.com/
 Share what you know. Learn what you don't.
Article: 18001
Subject: Re: No Subject
From: Edwin Naroska <edwin@mira.e-technik.uni-dortmund.de>
Date: Wed, 22 Sep 1999 15:03:25 +0200
Links: << >>  << T >>  << A >>
Hi,

Mark Harvey wrote:

> Does anyone know how to the time between two consecutive events on a signals
> for reporting in a testbench?
> (I basically need to check & report a pulse width)
>

This depends on the HDL language you are using. For VHDL'93 you may try:


   signal sig : ...

   watch: process (sig)
        variable last : time;
        variable first_run : boolean := true;
   begin
        if not first_run then
            report "pulse duration = " & time'image(now - last);
        end if;
        last := now;
        first_run := false;
   end process;


--
Edwin


Article: 18002
Subject: Re: Programming Spartan XL
From: acher@informatik.tu-muenchen.de (Georg Acher)
Date: 22 Sep 1999 15:20:49 GMT
Links: << >>  << T >>  << A >>

In article <37E5DFED.7402@mazsola.iit.uni-miskolc.hu>, Drotos Daniel <drdani@mazsola.iit.uni-miskolc.hu> writes:
|> Hi,
|> 
|> I'm trying to configure a Spartan chip (XCS20XL-PQ208) and it doesn't
|> work. I'm using slave mode, generating 640ns wide CCLK pulses and
|> about 16usec between them. I've checked DOUT, it is OK.
|> 
|> The problem is, that INIT goes low after the first frame which means
|> that frame error is detected. If I turn off the CRC checking when
|> generating the bitstream, then the chip accepts the first frame and
|> pulls INIT low after the second frame.
|> 
|> I've tried to download the configuration via the xchecker cable
|> using JTAG programmer of the Foundation 1.51i. It downloads the
|> bitstream via the boundary scan interface and everything looks
|> OK (DONE goes high for example) but the chip doesn't operate.
|> When I try to verify, it says "to many mismatches".
|> 
|> Does somebody have any idea what can be wrong?

Maybe it's as simple as that:

If you generate the bitstream by hand (ie. not using designmgr but makefile etc),
make sure you use the right partnumber. ppr swallows all combinations of part
numbers, but mostly interpretes them as a non-xl device, thus generating a
bitstream that is too short. For a 10XL in a VQ100, speed grade -4 ONLY the
type s10xlVQ100-4 works. Other types (like XCS10XL, printed on the chip) are
converted to non-XL.

I've searched a long time to find this...
-- 
	Bye
         Georg Acher, acher@in.tum.de         
         http://www.in.tum.de/~acher/
          "Oh no, not again !" The bowl of petunias          
Article: 18003
Subject: Virtex questions
From: husby@fnal.gov (Don Husby)
Date: Wed, 22 Sep 1999 15:57:42 GMT
Links: << >>  << T >>  << A >>
Three simple questions that I couldn't figure out from the online data sheet:

Can the carry chain be routed through both slices of a CLB, or is it more
efficient to run it to the next CLB?

Do both slices of a CLB share a common clock and CE, or are there separate
pins for each slice?

How are CLB outputs routed to the TBUFs?  Do all CLB outputs have access to
both TBUFs or are there some restrictions?



--
Don Husby <husby@fnal.gov>             http://www-ese.fnal.gov/people/husby
Fermi National Accelerator Lab                          Phone: 630-840-3668
Batavia, IL 60510                                         Fax: 630-840-5406
Article: 18004
Subject: Re: Virtex questions
From: Ray Andraka <randraka@ids.net>
Date: Wed, 22 Sep 1999 13:23:45 -0400
Links: << >>  << T >>  << A >>


Don Husby wrote:

> Three simple questions that I couldn't figure out from the online data sheet:
>
> Can the carry chain be routed through both slices of a CLB, or is it more
> efficient to run it to the next CLB?
>

There are two carry chains in the CLB, each runs through a slice and then
connects to the next slice vertically.  The carry chain connection doesn't go
horizontally (to the adjacent slice).  You can get a connection from slice to
slice by going out the YB or XB and in to the carry in on the other slice, but
that connection is slow compared to the dedicated carry chain connections (the
carry chain connections go directly to the slices while the other connections
generally go through the omux and imux.

> Do both slices of a CLB share a common clock and CE, or are there separate
> pins for each slice?
>

Each slice has separate connections for CE, clk and SR input multiplexer in the
CLB.

> How are CLB outputs routed to the TBUFs?  Do all CLB outputs have access to
> both TBUFs or are there some restrictions?
>

Looking at the Omux and Imux in epic, it appears that any of the outputs can get
to any of the TBUFs.  I haven't tried all combinations here, so I'm not sure.

> --
> Don Husby <husby@fnal.gov>             http://www-ese.fnal.gov/people/husby
> Fermi National Accelerator Lab                          Phone: 630-840-3668
> Batavia, IL 60510                                         Fax: 630-840-5406



--
-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/~randraka


Article: 18005
Subject: Reasonable out-of-circuit programming platform desired
From: "John Becich" <johnbecich@csi.com>
Date: Wed, 22 Sep 1999 11:34:11 -0700
Links: << >>  << T >>  << A >>
I have seen the pdf file on how to build an inexpensive in-circuit
programming link to the parallel port of a PC.

Suppose I would like to have an inexpensive way to program chips that is not
an in-circuit method.  Perhaps something that connects to the parallel port
of the same PC that I am using to generate source and object code for these
chips...I burn the chips, and stick them into sockets on my breadboard....Of
course I must be careful about static electricity.

Any suggestions would be appreciated.  Can I get something for a couple
hundred bucks?

I prefer Altera and Lattice chips.

Thanks,
John


Article: 18006
Subject: Reset signal and Altera's FPGAs
From: Farhad Abdolian <farhad.abdolian@tellabs.com>
Date: Wed, 22 Sep 1999 15:23:30 -0400
Links: << >>  << T >>  << A >>
Hi,
I am a new Altera user, have a long background in Xilinx and now trying
to understand the logic behind the Global reset on Altera FPGAs.

I have a lot of code there I reset the signals when the reset_n signal
is '0',  that works very well for simulation and then I thought that
Altera mapper should take care of it too, but it seems that Altera does
not have a global RESET signal and I can not assign a pin to act as one
either! 

Is it true? Because I think it is SICK, how on earth can a user deal
without a global RESET? I know that I can assign my signals to a value
when I declare them, but then I will receive hundreds of warnings when I
synthisize the code (which I absolutely will not do!).

Can anyone help me? Because the only way for me to solve this problem
was to hard code that signal on my top level, and assign it to 
'1' to be able to use my synthesized code!

TIA,
Farhad A.

-- 
 --------------------------------------------
|Farhad Abdolian, Tellabs Operations Inc.    |
|farhad.abdolian@tellabs.com, +1-914-593-4346|
 --------------------------------------------
Article: 18007
Subject: Re: Programming Spartan XL
From: Leon Heller <leon_heller@hotmail.com>
Date: Thu, 23 Sep 1999 07:36:51 GMT
Links: << >>  << T >>  << A >>
In article <37E5DFED.7402@mazsola.iit.uni-miskolc.hu>,
  Drotos Daniel <drdani@mazsola.iit.uni-miskolc.hu> wrote:
> Hi,
>
> I'm trying to configure a Spartan chip (XCS20XL-PQ208) and it doesn't
> work. I'm using slave mode, generating 640ns wide CCLK pulses and
> about 16usec between them. I've checked DOUT, it is OK.
>
> The problem is, that INIT goes low after the first frame which means
> that frame error is detected. If I turn off the CRC checking when
> generating the bitstream, then the chip accepts the first frame and
> pulls INIT low after the second frame.
>
> I've tried to download the configuration via the xchecker cable
> using JTAG programmer of the Foundation 1.51i. It downloads the
> bitstream via the boundary scan interface and everything looks
> OK (DONE goes high for example) but the chip doesn't operate.
> When I try to verify, it says "to many mismatches".
>
> Does somebody have any idea what can be wrong?
>
> Daniel
>

I had the same problem, and found the answer on the Xilinx web site.
Don't use the verify option, and simply check the DONE pin.

Leon
--
Leon Heller, G1HSM
Tel (Mobile): 079 9098 1221 (Work): 01327 357824
Email: leon_heller@hotmail.com
Web: http://www.geocities.com/SiliconValley/Code/1835


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Article: 18008
Subject: Re: Reset signal and Altera's FPGAs
From: Steve Rencontre <Steve@XXX_REMOVE_XXXrsn-tech.demon.co.uk>
Date: Thu, 23 Sep 1999 10:52:45 +0100
Links: << >>  << T >>  << A >>
Which Altera devices? The devices I've used (MAX7000 and FLEX10K) most
certainly do have global reset!

Go look up 'global' in Max+Plus II help.

On Wed, 22 Sep 1999 15:23:30 -0400, in <37E92CB2.5715C232@tellabs.com>
Farhad Abdolian <farhad.abdolian@tellabs.com> wrote:

>Hi,
>I am a new Altera user, have a long background in Xilinx and now trying
>to understand the logic behind the Global reset on Altera FPGAs.
>
>I have a lot of code there I reset the signals when the reset_n signal
>is '0',  that works very well for simulation and then I thought that
>Altera mapper should take care of it too, but it seems that Altera does
>not have a global RESET signal and I can not assign a pin to act as one
>either! 
>
>Is it true? Because I think it is SICK, how on earth can a user deal
>without a global RESET? I know that I can assign my signals to a value
>when I declare them, but then I will receive hundreds of warnings when I
>synthisize the code (which I absolutely will not do!).
>
>Can anyone help me? Because the only way for me to solve this problem
>was to hard code that signal on my top level, and assign it to 
>'1' to be able to use my synthesized code!
>
>TIA,
>Farhad A.

--
Steve Rencontre, Design Consultant
http://www.rsn-tech.demon.co.uk/  --  remember to despam return address
Article: 18009
Subject: Re: Reset signal and Altera's FPGAs
From: deroberts@my-deja.com
Date: Thu, 23 Sep 1999 10:20:35 GMT
Links: << >>  << T >>  << A >>
In article <37E92CB2.5715C232@tellabs.com>,
  Farhad Abdolian <farhad.abdolian@tellabs.com> wrote:
> Hi,
> I am a new Altera user, have a long background in Xilinx and now
trying
> to understand the logic behind the Global reset on Altera FPGAs.
>
> I have a lot of code there I reset the signals when the reset_n signal
> is '0',  that works very well for simulation and then I thought that
> Altera mapper should take care of it too, but it seems that Altera
does
> not have a global RESET signal and I can not assign a pin to act as
one
> either!
>
> Is it true? Because I think it is SICK, how on earth can a user deal
> without a global RESET? I know that I can assign my signals to a value
> when I declare them, but then I will receive hundreds of warnings when
I
> synthisize the code (which I absolutely will not do!).
>
> Can anyone help me? Because the only way for me to solve this problem
> was to hard code that signal on my top level, and assign it to
> '1' to be able to use my synthesized code!
>
Altera's handling of reset is perhaps not ideal, but is not as bad as
you make out.

If you are using Verilog, then simply pass your global reset signal
through a "global" module i.e.

  input nRESET;
  wire gnRESET;

  global myreset ( .in( nRESET ),   .out( gnRESET )   );

The compiler will spot this and allocate your global reset to one of the
global nets on the device. Depending on the device there are between
roughly 2 and 6 of these (check your Altera data sheet).

Alternatively, if you go into the Assign/Global project logic synthesis
settings, then there's an auto global switch; however I have known this
not to allocate a global line if the reset line is used in a non-global
way at all.

Finally, most devices have a dedicated global reset (DEV_CLRn) which can
be enabled from the Assign/Global project device options menu.  The
drawback of using this however is that (so far as I can tell) is that
the compiler won't automatically recognise this from your Verilog code
and so all your explicitly defined reset signals will get synthesised as
well.

I prefer using the global module approach. I can testify that it works
fine and it gives you control of what's going on without having to write
anything too Altera specific in your code.


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Article: 18010
Subject: basic Altera simulation questions
From: "John Becich" <johnbecich@csi.com>
Date: Thu, 23 Sep 1999 08:56:02 -0700
Links: << >>  << T >>  << A >>
I am using the Altera Baseline Max+Plus software, version 9.3, as recently
downloaded from Altera.  I have designed a few state machines, and I would
like to simulate their performance.

I have created an .scf file using the Waveform Editor, and I configured it
so that my waveform simulation would show several input and output signals,
as found in the project .snf file (I used Node menu, Enter Nodes from SNF).
So far, so good.

I notice that the span of time covered during the simulation is 1 us (i.e.,
microsecond).  I would like to change this span to me much longer, perhaps 1
second.  I can't figure out how to do this.

Could anyone please tell me what to do?

**********************

I have read some of the HELP available, and it has told me to find a
"Processing menu in the Compiler."  I haven't been able to find any such
"Processing Menu" anywhere.  Can anyone tell me where to look?  Or is such
menu available only in other software than what I am using?

Thanks,
John Becich


Article: 18011
Subject: Fineline BGAs
From: steves@traclabs.com
Date: Thu, 23 Sep 1999 16:43:56 GMT
Links: << >>  << T >>  << A >>

Hello.  Sorry, this is not entirely FPGA related.  Has anybody used a
0.8mm pitch BGA on their board?  I am using the CS280 package from
Xilinx and have some questions.  What trace/via/pad and clearances
were used on the PCB?  Any links to guidelines on using these devices
would be nice.  My layout guy hasn't found anything on these devices
yet.  Thanks for your time.


Article: 18012
Subject: Re: Vertex Select I/O
From: rseglie@my-deja.com
Date: Thu, 23 Sep 1999 17:50:41 GMT
Links: << >>  << T >>  << A >>
In article <7s8iqa$20ae$1@thoth.cts.com>,
  "Steve Kinkead" <stevek@cts.com> wrote:
> I am using the Select I/O feature in a Xilinx Vertex device to
configure
> my outputs to 3.3v PCI.  In my .ucf constraint file I specify each I/O
> to be PCI using the following:
>
> NET "nFrame" IOSTANDARD = PCI33_3;
>
> But, the place and route software seems to ignore the contraint and
sets
> all the I/O to be LVTTL.
>
> I created the design using Verilog and have not hand-instantiated any
I/O
> cells
> except for the clock buffers in the clock dlls.   A Xilinx app note
says
> that
> if you hand-instantiate any I/O cells you must hand-instantiate all of
them
> due to a limitation with FPGA Express.
>
> Anyone found a work-around?
> --
> ==============================================
> Steve Kinkead                              5951 Seacrest View Rd
> Infinite Logic, LLC                         San Diego, CA 92121
> Hardware Design Consulting         (858) 638-1457
> stevek@cts.com                           (858) 638-0998 : FAX
>
>

I'm using FPGA Express with VHDL and in the "constraints" windows and in
"Ports" you can specify the default I/O type. By default, it's LVT,
12mA, slow but you can change it !
all the best..

Rémi SEGLIE
Company : CELOGIC
www : http://www.celogic.com/



Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Article: 18013
Subject: Re: Vertex Select I/O
From: vlogvin@my-deja.com
Date: Thu, 23 Sep 1999 18:55:40 GMT
Links: << >>  << T >>  << A >>
In article <7s8iqa$20ae$1@thoth.cts.com>,
  "Steve Kinkead" <stevek@cts.com> wrote:
> I am using the Select I/O feature in a Xilinx Vertex device to
configure
> my outputs to 3.3v PCI.  In my .ucf constraint file I specify each I/O
> to be PCI using the following:
>
> NET "nFrame" IOSTANDARD = PCI33_3;
>
> But, the place and route software seems to ignore the contraint and
sets
> all the I/O to be LVTTL.
>
> I created the design using Verilog and have not hand-instantiated any
I/O
> cells
> except for the clock buffers in the clock dlls.   A Xilinx app note
says
> that
> if you hand-instantiate any I/O cells you must hand-instantiate all of
them
> due to a limitation with FPGA Express.
>
> Anyone found a work-around?
> --
> ==============================================
> Steve Kinkead                              5951 Seacrest View Rd
> Infinite Logic, LLC                         San Diego, CA 92121
> Hardware Design Consulting         (858) 638-1457
> stevek@cts.com                           (858) 638-0998 : FAX
>
>
It is a problem!
I work on VHDL with Xilinx Express under Windows.
I had same problem.
I have connected library Unisim.
library unisim;
use unisim.vcomponent.all;
It for simulation in ModelSim
In library Virtex there are components PCI buf
Further, in the text VHDL I used these components.
component OBUF_PCI33_3
.....
end component;
It has helped.
Good luck
 =============================
Logvin Victor.
Rossia.
vlogvin@mydeja.com&#137;


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Article: 18014
Subject: Re: basic Altera simulation questions
From: bob elkind <eteam@aracnet.com>
Date: Thu, 23 Sep 1999 13:25:09 -0700
Links: << >>  << T >>  << A >>
Open .SCF file (waveform editor)

Menu bar: File/End_time

enter value.

Voila!
   (BTW, I agree that if you don't know where to look, it's almost
      impossible to find!)

-- Bob Elkind

John Becich wrote:
> 
> I am using the Altera Baseline Max+Plus software, version 9.3, as recently
> downloaded from Altera.  I have designed a few state machines, and I would
> like to simulate their performance.
> 
> I have created an .scf file using the Waveform Editor, and I configured it
> so that my waveform simulation would show several input and output signals,
> as found in the project .snf file (I used Node menu, Enter Nodes from SNF).
> So far, so good.
> 
> I notice that the span of time covered during the simulation is 1 us (i.e.,
> microsecond).  I would like to change this span to me much longer, perhaps 1
> second.  I can't figure out how to do this.
> 
> Could anyone please tell me what to do?
> 
> **********************
> 
> I have read some of the HELP available, and it has told me to find a
> "Processing menu in the Compiler."  I haven't been able to find any such
> "Processing Menu" anywhere.  Can anyone tell me where to look?  Or is such
> menu available only in other software than what I am using?
> 
> Thanks,
> John Becich
Article: 18015
Subject: Modelsim,synplify,Leonardo,MAX+plus, you name it!!
From: flexlm666@my-deja.com
Date: Thu, 23 Sep 1999 21:33:24 GMT
Links: << >>  << T >>  << A >>
Can you or your company afford paying 10000$ at least for these
programs ?!!? If you can then ignore this posting, otherwise I can
offer an alternative at 1/100 th of the original price. I offer perfect
solutions for small enterprizes and I would like to emphasize on
perfect.

Just send me an e-mail to freelicenses@hotmail.com or flexlm666@my-
deja.com


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.
Article: 18016
Subject: Synopsys inside Foundation 2.1i does not infer fast-adder
From: Steve Martindell <s-martindell@_nospamm_ti.com>
Date: Thu, 23 Sep 1999 19:18:32 -0500
Links: << >>  << T >>  << A >>
I'm trying to map some existing VHDL code(describing a datapath with a
32bit adder)
into the Virtex FPGA using Foundation2.1i on the PC.  Previously this
data path was
mapped to a standard ASIC library using DC(Design Compiler) on a Unix
work station.
When I set the clock constraint to >50MHz on the Unix version of DC, it
inferred a "fast-
carry" adder(used porpagate/generate terms) and the design meets timing.

When I set the same clock constraint on the same VHDL code in
Foundation2.1i
(targetting a Virtex XVC400) it doesn't meet timing,  running somewhat
less than
50MHz.  Note: the version of Synopsys Design Compiler running inside
Foundation2.1i
is called "FPGA Express" I believe,  and may not be the same DC that
runs on Unix.

When I reviewed the long path in the static timing analysis, it showed
the long path
"rippling" through 32 CLB's(each CLB =2bit adder + carryin).  This
version of DC
did not infer the "fast-carry" adder required to meet timing.  This is
not a case of
improper clock constraints because the timing analyzer reports that it
missed the
50MHz target and only runs ~30MHz!

Note: I do have  the following in the VHDL code:
                               library IEEE;
                               use IEEE.std_logic_1164.all;
                               use IEEE.std_logic_signed.all;

When Xilinx talks about fast carry generation logic apparently they mean
generate
a "ripple-carry" fast!?

I've discussed this w/ with an engineer at Xilinx Technical Support. He
ran a testcase
where he described a 32-bit adder between two registers. He used the
same code
that I am using for the add function:

SUM <= A(31 downto 0) + B(31 downto 0);

with the same clock constraint on Foundation2.1i and got the same
results: one
big ripple-carry adder which did not meet timing.  To get around this
problem
I am coding the 32-bit propagate/generate fast-carry adder by hand.
Anyone else seen this problem?

         Steve Martindell
             real email address is:
                 s-martindell@ti.com



Article: 18017
Subject: virtex clock questions
From: rk <stellare@NOSPAM.erols.com>
Date: Thu, 23 Sep 1999 21:00:40 -0400
Links: << >>  << T >>  << A >>
hi,

a few questions on the virtex clock system (+):

1. let's assume that i get the fastest speed grade part.

2. let's assume i get the xcv1000.  i need the biggest one for it's
height.  actual utilization will be very low.  let's say i use about 4
columns, max.

3. let's assume that i use a linearly regulated power supply and there
is very little ripple on it.  nothing else is on the supply.

4. can i run a low skew clock at 250 MHz?  i think that speed will work
for my application(*).  what speed is achievable?

5. what is the power in mwatt/MHz for that clock?

6. what is the maximum flop-flop skew for the low skew clock?

7. anyone have a feel for how much jitter would be on the clock?  see 3,
above.

8. ball park estimate for part cost?  i am w/out clue.  (yeah, i can try
the www, but getting into the distributor www site was way slow this
evening).

thanks very much in advance,

a bit urgent,

rk

(+) yeah, i can read the whole data sheet, app notes, etc., etc.  but
need some estimates quick for day job.  and am under the cannon to get
some stuff out NOW! (more than just this and it's gonna be another
loooooooooooooong night :-)

(*) the application is right now a proposal and feasibility study i'm
doing.  so, no sales critter follow-ups, please.  if we do it, we can
figure out where to buy them.

Article: 18018
Subject: Re: Synplfy 5.21 and 5.08a
From: Andrew Dauman <andrew@synplicity.com>
Date: Thu, 23 Sep 1999 18:51:58 -0700
Links: << >>  << T >>  << A >>
Hello Ozgur,

Stuart is a distributor of the Exemplar product, a direct competitor of
ours.
I thought it might be helpful if I corrected his statement.
Synplify 5.2.1 is an official release of Synplify, as was version 5.1.5a.
Synplify 5.2.1 adds support for the new Apex20K architecture from Altera.
It is identical to 5.1.5a for other vendors. Therefore, I expect you will
experience
the same behavior for Spartan in either 5.1.5a or 5.2.1. It is perfectly ok
to use 5.2.1
for Spartan.

We do not know of any problems with inferencing or creation of counters in
our software
since the 5.0.8a release. If you have a test case that demonstrates the
reduced
counter detection, we will gladly take a look at it. You can send it to  us
at support@synplicity.com
and we will do our best to understand the difference.

Best regards,
Andrew Dauman
Director of Corporate Applications Engineering
Synplicity, Inc.



Ozgur Kayalar wrote:

> Hi
> I use Synplify version 5.21.
> Version 5.21 cannot find counters as the previous version v5.08a does.
> Ver.5.21 finds 1 counter while v5.08a finds 5 counters in my same
> design.
> This is valid for the target Xilinx Spartan 30 device, and I wonder if
> v521 cause Design Manager not to use carry logic for the 'missing'
> counters.
> Maybe it is better to use older version of Synplify in such a condition
> Does anyone observed such a situation?Thank you for your time,
> Ozgur

Article: 18019
Subject: Re: Synplfy 5.21 and 5.08a
From: rk <stellare@NOSPAM.erols.com>
Date: Thu, 23 Sep 1999 21:59:19 -0400
Links: << >>  << T >>  << A >>
er, what was exactly the problem in the earlier release?

<rk gets nervous>

thanks,

rk

Andrew Dauman wrote:

> Hello Ozgur,
>
> Stuart is a distributor of the Exemplar product, a direct competitor of
> ours.
> I thought it might be helpful if I corrected his statement.
> Synplify 5.2.1 is an official release of Synplify, as was version 5.1.5a.
> Synplify 5.2.1 adds support for the new Apex20K architecture from Altera.
> It is identical to 5.1.5a for other vendors. Therefore, I expect you will
> experience
> the same behavior for Spartan in either 5.1.5a or 5.2.1. It is perfectly ok
> to use 5.2.1
> for Spartan.
>
> We do not know of any problems with inferencing or creation of counters in
> our software
> since the 5.0.8a release. If you have a test case that demonstrates the
> reduced
> counter detection, we will gladly take a look at it. You can send it to  us
> at support@synplicity.com
> and we will do our best to understand the difference.
>
> Best regards,
> Andrew Dauman
> Director of Corporate Applications Engineering
> Synplicity, Inc.
>
> Ozgur Kayalar wrote:
>
> > Hi
> > I use Synplify version 5.21.
> > Version 5.21 cannot find counters as the previous version v5.08a does.
> > Ver.5.21 finds 1 counter while v5.08a finds 5 counters in my same
> > design.
> > This is valid for the target Xilinx Spartan 30 device, and I wonder if
> > v521 cause Design Manager not to use carry logic for the 'missing'
> > counters.
> > Maybe it is better to use older version of Synplify in such a condition
> > Does anyone observed such a situation?Thank you for your time,
> > Ozgur



Article: 18020
Subject: Re: Synopsys inside Foundation 2.1i does not infer fast-adder
From: tim <timcat@flash.com>
Date: Fri, 24 Sep 1999 04:24:34 GMT
Links: << >>  << T >>  << A >>
Steve,
    You need to include the basic design ware libraries. DW01 and DW02.
These should come with the basic Synopsys package. Hope
this helps.
change the c to a k and com to net to reach me. I try to avoid spam

Tim

Steve Martindell wrote:

> I'm trying to map some existing VHDL code(describing a datapath with a
> 32bit adder)
> into the Virtex FPGA using Foundation2.1i on the PC.  Previously this
> data path was
> mapped to a standard ASIC library using DC(Design Compiler) on a Unix
> work station.
> When I set the clock constraint to >50MHz on the Unix version of DC, it
> inferred a "fast-
> carry" adder(used porpagate/generate terms) and the design meets timing.
>
> When I set the same clock constraint on the same VHDL code in
> Foundation2.1i
> (targetting a Virtex XVC400) it doesn't meet timing,  running somewhat
> less than
> 50MHz.  Note: the version of Synopsys Design Compiler running inside
> Foundation2.1i
> is called "FPGA Express" I believe,  and may not be the same DC that
> runs on Unix.
>
> When I reviewed the long path in the static timing analysis, it showed
> the long path
> "rippling" through 32 CLB's(each CLB =2bit adder + carryin).  This
> version of DC
> did not infer the "fast-carry" adder required to meet timing.  This is
> not a case of
> improper clock constraints because the timing analyzer reports that it
> missed the
> 50MHz target and only runs ~30MHz!
>
> Note: I do have  the following in the VHDL code:
>                                library IEEE;
>                                use IEEE.std_logic_1164.all;
>                                use IEEE.std_logic_signed.all;
>
> When Xilinx talks about fast carry generation logic apparently they mean
> generate
> a "ripple-carry" fast!?
>
> I've discussed this w/ with an engineer at Xilinx Technical Support. He
> ran a testcase
> where he described a 32-bit adder between two registers. He used the
> same code
> that I am using for the add function:
>
> SUM <= A(31 downto 0) + B(31 downto 0);
>
> with the same clock constraint on Foundation2.1i and got the same
> results: one
> big ripple-carry adder which did not meet timing.  To get around this
> problem
> I am coding the 32-bit propagate/generate fast-carry adder by hand.
> Anyone else seen this problem?
>
>          Steve Martindell
>              real email address is:
>                  s-martindell@ti.com

Article: 18021
Subject: Altera's MaxplusII: incremental compilation
From: Prashant Arora <arora@ics.uci.edu>
Date: Thu, 23 Sep 1999 22:58:38 -0700
Links: << >>  << T >>  << A >>

Hi,

We have synthesized a design for Altera's Flex10k family fpga  
(EPF10K100ABC600-1) starting  from VHDL. We are using Synopsys for
synthesis and Maxplus II ver 9.25 for place and route. This returns pin
assignments for the IO ports in our design. Our FPGA is located on a PCB. 

We fixed the pin assignments and hence the PCB routing using one of the
synthesis runs that met our clock requirements (50Mhz). Subsequently, we
have discovered a bug in our design that needs a small change to
fix (adding a flip-flop). Now since the PCB routing is fixed, we have to
provide Maxplus with pin constraints (otherwise it produces a totally
different pin assignment). This causes Maxplus to give us a best case 
clock of 33MHz and the p&r takes more than 10 hrs even though the change
to the design is minimal. 

1) Is there a way to retain most of the placed and routed design from a
previous run and add a small 'patch' to the design in Maxplus without
affecting the existent placement and routing?

2) Is there a way in which the p&r algorithm can be made to provide the
same pin assignment as we got for the previous run without constraining it
with pin assignments?

Thanks,
Prashant




Article: 18022
Subject: only test
From: Also-Antal Csaba <antalcs@mail.matav.hu>
Date: Fri, 24 Sep 1999 08:43:31 +0200
Links: << >>  << T >>  << A >>
 
Article: 18023
Subject: Re: F 1.5
From: "user" <tsaimc01@ms35.hinet.net>
Date: Fri, 24 Sep 1999 15:46:11 +0800
Links: << >>  << T >>  << A >>
the book has send it.
isbn-0-13-977-398-3
thank you


Adam Biniszkiewcz <A.Biniszkiewicz@pz.zgora.pl> wrote in message
news:7q3gtr$o5j$1@main.amu.edu.pl...
> Hi,
> I want to buy Xilinx Foundation, version 1.5 with FPGA Express.
> Can you provide me with pricing, if someone is interested in?
> Regards - Adam
> A.Biniszkiewicz@pz.zgora.pl
>
>
>


Article: 18024
Subject: Re: virtex clock questions
From: allan.herriman.hates.spam@fujitsu.com.au (Allan Herriman)
Date: Fri, 24 Sep 1999 08:47:42 GMT
Links: << >>  << T >>  << A >>
On Thu, 23 Sep 1999 21:00:40 -0400, rk <stellare@NOSPAM.erols.com>
wrote:

[snip]
>3. let's assume that i use a linearly regulated power supply and there
>is very little ripple on it.  nothing else is on the supply.

>7. anyone have a feel for how much jitter would be on the clock?  see 3,
>above.

I have data sheets for both Virtex and Apex (Altera) parts.
Interestingly, the jitter spec for both is the same (200ps assuming
50ps input jitter), even though the Xilinx part uses a DLL and the
Altera part uses a PLL.

I think that the DLL should give better jitter performance.  (Or
perhaps Xilinx made some cost tradeoff.)
Certainly, the jitter transfer functions should be different.  (PLL
should be lowpass; DLL should be flat.)

Has anyone observed this on a real part?

Allan.


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