View Single Post
  #3 (permalink)  
Old 04-25-2006, 03:39 AM
Gerhard Hoffmann
Guest
 
Posts: n/a
Default Re: ISE 8.1 Sub module Synthesis

On Mon, 24 Apr 2006 17:29:34 +0200, Zara <[email protected]> wrote:


>I don't know if it will work in verilgo, but I soppes it will. It
>works on VHDL. Right click on source node , "Select as Top Module". Do
>remember to reselect original Top module when done.


I did this, too (vhdl), but then the "real top module" user constraint file was
changed to empty. Gave not so funny results with the router deciding arbitrarily
which pin is wich. I had connected new hardware to the eval board and
assumed that I had killed the Virtex4 :-(
Luckily, it was too late already to order a new one.
It would help if one could see in the log files what the reason was for a
signal to end up at a certain pin.

Dependency checking seems to be messed up in 8.1 . That could be a
reason for the above, too.
Today, I got warnings about incompatible components and instantiatons.
I removed some instantiations completely ( commenting out large
blocks is _such_ a fun in VHDL..) but still they appeared in the error list.
It looks like ISE has made own copies of my source files.
I don't use the internal editor but gvim, so maybe ISE does not see every
file update?
Using Project -> CleanupProjectFiles reduced the number of error messages
by 80%, but not completely.

And, because I'm in the mood: The ISE gui is almost dead after kicking
off a synthesis. Sometimes, the routing has run to completion before
the gui wakes up again. The first run after booting seems not to be
affected by this. I use the windows task manager to keep track of
what's going on (shows top CPU eating processes).

When I double click on "ConfigureDevice" to kick off synthesis & everything,
and do 5 times edit-compile-load-measure, then I end up with 5 incarnations of
IMPACT, as if one wasn't enough, and each one asking me what to do and
if it should update the project file on exit, as if anything had changed.

ISE knows enough about my project so that one doubleclick should be enough
to go from source to logic analyzer. We had that already in XC3020 times.
Impact should not ask ME if it needs a mask file for verify.

I changed from 7.1 to 8.1 because 7.1 crashed may computer abt. 6 times
a day. Those crashes are history with 8.1, but all in all, things have
gone from bad to worse.
Retreating to 6.3 like others here do is probably no option for a virtex4.
Perhaps I should convert everything to makefiles.

regards, Gerhard

environment: Athlon64-4000+, 2 GB RAM, striped raptors :-), XP,
Modelsim PE, ISE 8.1.3, ML402, platform cable USB, vim
Reply With Quote