Xilinx 4013 80% utilized but won't route

Xilinx 4013 80% utilized but won't route

Post by Eric Ryher » Tue, 14 May 1996 04:00:00



Hi,

Anyone have experience with Exemplar and Xilinx 4013s where it
was only 80% utlilized but wouldn't route?
In synopsys language, I think I need to increase the "porocity"
of the design. I have other designs that are 90% and route, but
for some reason this one is real bad.

The design in nearly completely random logic. only 42 DFFs.
I've tried all of the settings in Xilinx PPR, various seeds, max
place & route effort... I even tried PPR on some of
the less than optimal netlists from Exemplar. But still
won't route.

Any options that might help???

We'll use the 4020 if we have too, but they're hard to get...

The design is in VHDL. IOs are locked (don't get on my case here!!!
unlocking the IOs only goes from about 200 to 100 unroutes).

I tried synthesizing with another synthesizer but it came up with 105%
utilization...

--

VAutomation Inc.           Synthesizable HDL Cores
20 Trafalgar Square        http://www.vautomation.com
Suite 443 Nashua NH 03063  (603) 882-2282 FAX:882-1587

 
 
 

Xilinx 4013 80% utilized but won't route

Post by Pete » Tue, 14 May 1996 04:00:00


Sounds like you have used up all your interconnects.

I have had register-intensive designs (XC3064) which were 98% utilised
and routed fine (with old 1991 software) and I have had designs which
were about 70% utilised which only their latest place/route s/w would
route, and only just about at that.

I would imagine VHDL generates lots of random logic, which is not very
good for register-rich FPGAs like Xilinx.

Peter.

 
 
 

Xilinx 4013 80% utilized but won't route

Post by Patrick McCab » Wed, 15 May 1996 04:00:00



> Anyone have experience with Exemplar and Xilinx 4013s where it
> was only 80% utlilized but wouldn't route?
> In synopsys language, I think I need to increase the "porocity"
> of the design. I have other designs that are 90% and route, but
> for some reason this one is real bad.

> The design in nearly completely random logic. only 42 DFFs.
> I've tried all of the settings in Xilinx PPR, various seeds, max
> place & route effort... I even tried PPR on some of
> the less than optimal netlists from Exemplar. But still
> won't route.

> Any options that might help???

> We'll use the 4020 if we have too, but they're hard to get...

> The design is in VHDL. IOs are locked (don't get on my case here!!!
> unlocking the IOs only goes from about 200 to 100 unroutes).

> I tried synthesizing with another synthesizer but it came up with 105%
> utilization...

Don't use the "ungroup" switch or command - that causes a very dense
circuit which is nearly impossible to route. Try to use hierarchy, and
maintain it throughout the compile. Maybe try structuring and
multiple-output minimization to share logic elements (you get less
logic). Or, alternatively, try the flatten feature, and see what that
buys you.

I was having similar problems with a 5215 design. Now that I'm routing,
though, I can't seem to get acceptable timing. I found that the timing
is incredibly placement sensitive, and PPR is not very good at
placement, even with a high level of effort.

You may be forced to use the floorplanner to come up with a more
routable placement, and use the resulting LCA as a guidefile. This is
not a pretty option, due to possible iterations of this step due to
design changes, though.

A major problem that I see is that to do a synthesis followed by a place
and route takes about 6 hours for my design. That is too low a frequency
to try playing with different options. I currently have a Xilinx AE
helping me out to try to improve my timing. You may want to see if you
can get similar help.

Pat
________________________________________________________________________
 Patrick McCabe                     | AT&T Paradyne

 Ph: 813/530-8776 FAX: 813/532-5244 | Largo, FL  34649-2826
________________________________________________________________________
"Just because something doesn't do what you planned it to do doesn't
 mean it's useless."
-- Thomas Edison.
________________________________________________________________________

 
 
 

1. RFI:Info on 80's GE CALMA displays

We still have one or two systems here at IDT. We use them as a backup
for old designs. (Sometimes we need to massage the database before
translation.) I could probably dig up the manuals if you do not get the
info you are looking for.

There were several companies that serviced them. The biggest were G.E.
Computer Services, ASJ, and Ostari. I could also dig up the phone
numbers of at least ASJ and Ostari if I looked real hard. Let me know if
you  need them.

Just curious, WHY? are you looking to use display monitors that are no
longer being built, thus hard to find and hard to maintain? As I
remember, they were ALWAYS burning out parts. In fact, when we had 5
Calma systems on site (at another company) we made sure we had at least
one complete monitor available to use as a backup.

===================================================
|Chuck Phillips                                   |                          
|Integrated Device Technology, Inc                |
|2670 Seeley Ave                                  |
|San Jose, CA 95134                               |
|(408)944-2052                                    |
|(408)944-2195 (FAX)                              |
|uunet.uu.net!idtg!cwp                            |
|International Cadence User Group - Co-Chair      |
|Bay Area Cadence User People, BACUP, Chair       |
===================================================

2. subnet-zero

3. Atari ST as CAD machine back in the 80's

4. Book FS: Object Success (by Bertrand Meyer)

5. If you've seen this....it finally has an SPR# 80)

6. AMY....msjet35.dll problem....still here

7. Cherche listing des sociétés de l'est de la france qui utilise Pro/ENG

8. Core memory

9. PowerScope's view of Modeler files utilizing parts!!!

10. Install\Config R11c2 (dos) on Win95-r1b 486DX2-80 w\24mb

11. HELP!! 80 X 87 ERROR

12. Edutech Acad 12 and PowerMac 8100/80 SLOW!