Xilinx ISE 11.x lossage

I installed the full-up Xilinx ISE 11.x tools on a spare machine so I
could give it a test-drive. (We have a site license.)

I checked out a fresh copy of the trunk of a working, shipping design
that was created with ISE 10.1.3. I used 10.1's rather-nifty "Source
Control | Export" feature to boil the binary ISE project file down to
a source-control-friendly tcl script, and as such the only parts of
the project that are in the repository are the VHDL sources, the UCF
and the three tcl scripts created by the Export. With 10.1, this is
sufficient to completely recreate the ISE project file in all of its
bloated glory.

Then I launched ISE 11. I clicked on the "Project | Source Control |
Import ..." menu item, hoping to import the project. Instead of
happiness, I got a dialog box saying that this feature was shitcanned,
and that I should import the project in the previous version to create
the .ISE project file, and then open that project file in 11.x. The
dialog also said something about using the shell to expand the project
file, but I couldn't figure out how to do this.

So, Xilinx: thanks a whole lot for taking one of the few actually
useful features of the Project Navigator and throwing it -- and your
customers who were foolish enough to rely on it -- under the bus.
There's really no reason for this feature to be removed, except that
you wanted to prove how much you hate your customers (as if moving to
FlexLM wasn't enough).

I opened a WebCase, which I suspect will be ignored like every other
WebCase.

-a

On Jul 23, 12:36=A0am, Andy Peters <goo...@latke.net> wrote:
> I installed the full-up Xilinx ISE 11.x tools on a spare machine so I
> could give it a test-drive. (We have a site license.)

...

> I opened a WebCase, which I suspect will be ignored like every other
> WebCase.
>
> -a

Well, Andy, I hate to say that, but it is your own fault.
Perhaps you are a new user ...

Using anything but only the fundamental features of the
tools is asking for trouble. Even when just using the
fundamental tools like "map <option>" in a perl script
you will see that they tend to rename and change the way
some switches work from release to release.

This is a good example of a closed industry that locks out
innovation and growth though patents. It is impossible to
design a new and fresh FPGA architecture and tool chain
without stepping on one or another patent. That's why
everybody whines about how bad the tools are how bad and
expensive the the FPGAS and support are, and why feature x,
y and z are not there ...

Xilinx just like all the other players has to make a trade
off as to how much money to invest in to tools upgrades and
developments and improving FPGAs. Being a large company with
many levels of management, they tend to be less efficient
than they could be and innovation is at it's minimal level.

You want better tools and better FPGAS ? Let you wallet
speak. Chose FPGAs from vendors that meet all your needs,
that respect you even as a small end-user and meet you
budget. I bet you'll be back at X & A !

The moral of it all ? Think positive and you will live
longer :*)

Cheers,
rudi

On Jul 24, 1:17=A0am, luudee <rudolf.usselm...@gmail.com> wrote:
> On Jul 23, 12:36=A0am, Andy Peters <goo...@latke.net> wrote:
>
> > I installed the full-up Xilinx ISE 11.x tools on a spare machine so I
> > could give it a test-drive. (We have a site license.)
>
> ...
>
> > I opened a WebCase, which I suspect will be ignored like every other
> > WebCase.
>
> > -a
>
> Well, Andy, I hate to say that, but it is your own fault.
> Perhaps you are a new user ...

No, actually, I've being doing this for a long time -- remember XACT?

> Using anything but only the fundamental features of the
> tools is asking for trouble. Even when just using the
> fundamental tools like "map <option>" in a perl script
> you will see that they tend to rename and change the way
> some switches work from release to release.

I realize that things change all the time, which is why I like to
minimize my dependency on the tools. But unfortunately, that's not
always possible. And this little archiving feature was one of Xilinx'
better ideas, which is probably why they shitcanned it.

Andy Peters wrote:

> No, actually, I've being doing this for a long time -- remember XACT?

Yes. That was a good demo of the halting problem.

> I realize that things change all the time, which is why I like to
> minimize my dependency on the tools. But unfortunately, that's not
> always possible.

I need synthesis for STA and for the .bit file,
but I have started using svn tags for archiving the files.
Not exactly easy to use, but it does work and is unlikely
to vanish without a trace.

         -- Mike Treseler

On Jul 24, 10:57=A0am, Mike Treseler <mtrese...@gmail.com> wrote:
> Andy Peters wrote:
> > No, actually, I've being doing this for a long time -- remember XACT?
>
> Yes. That was a good demo of the halting problem.
>
> > I realize that things change all the time, which is why I like to
> > minimize my dependency on the tools. But unfortunately, that's not
> > always possible.
>
> I need synthesis for STA and for the .bit file,
> but I have started using svn tags for archiving the files.
> Not exactly easy to use, but it does work and is unlikely
> to vanish without a trace.

The project is in an svn repo. The great thing about this now-deleted
feature was that it boiled the ISE project file down to a couple of
tcl scripts which are scc-friendly. The only other things in the repo
are the sources and a .ucf.

So I'd check out the project, then open ISE, use the Project | Source
Control | Import feature and it would reconstruct the ISE project
file.

When the design is final, I tag it by svn add-ing the bitfile and
the .mcs and creating a tag from the working copy. Then I revert the
add because my working copy is still the trunk.

I suppose I could go back to scripts and Makefiles, but EDK adds
another whole level of bullshit to the process and that makes it
difficult to use without the GUI.

-a

On Jul 24, 3:35=A0pm, Andy Peters <goo...@latke.net> wrote:
> I suppose I could go back to scripts and Makefiles, but EDK adds
> another whole level of bullshit to the process and that makes it
> difficult to use without the GUI.

if you're using edk you should *definitely* use scripts and
makefiles.  edk uses make for a backend after all.  sometimes i use
platform studio to make project modifications (like adding chipscope),
since it does some error checking, but all builds are with makefiles
from a cygwin shell.

perhaps it is different if you use ise to drive xps, but i use xps to
drive ise.

On Fri, 24 Jul 2009 15:35:37 -0700 (PDT)
Andy Peters <google@latke.net> wrote:

> On Jul 24, 10:57=A0am, Mike Treseler <mtrese...@gmail.com> wrote:
> > Andy Peters wrote:
> > > No, actually, I've being doing this for a long time -- remember
> > > XACT?
> >
> > Yes. That was a good demo of the halting problem.
> >
> > > I realize that things change all the time, which is why I like to
> > > minimize my dependency on the tools. But unfortunately, that's not
> > > always possible.
> >
> > I need synthesis for STA and for the .bit file,
> > but I have started using svn tags for archiving the files.
> > Not exactly easy to use, but it does work and is unlikely
> > to vanish without a trace.
>=20
> The project is in an svn repo. The great thing about this now-deleted
> feature was that it boiled the ISE project file down to a couple of
> tcl scripts which are scc-friendly. The only other things in the repo
> are the sources and a .ucf.
>=20
> So I'd check out the project, then open ISE, use the Project | Source
> Control | Import feature and it would reconstruct the ISE project
> file.
>=20
> When the design is final, I tag it by svn add-ing the bitfile and
> the .mcs and creating a tag from the working copy. Then I revert the
> add because my working copy is still the trunk.
>=20
> I suppose I could go back to scripts and Makefiles, but EDK adds
> another whole level of bullshit to the process and that makes it
> difficult to use without the GUI.
>=20
> -a

Unless I'm mistaken, the new ISE project files are already SVN
friendly text files.  Only took a decade to get there.

--=20
Rob Gaddi, Highland Technology
Email address is currently out of order

"Andy Peters" <google@latke.net> wrote in message 
news:89bb1088-36f1-435c-8107-284367a3ac23@13g2000prl.googlegroups.com...
The project is in an svn repo. The great thing about this now-deleted
feature was that it boiled the ISE project file down to a couple of
tcl scripts which are scc-friendly. The only other things in the repo
are the sources and a .ucf.

=========
The <project>.xise file serves the need, without exporting or importing.

Now if only I can get it to stop polluting the project root with all those 
generated files.

MikeWhy wrote:

> The <project>.xise file serves the need, without exporting or importing.

Looks like xml text file with
all the files names and synthesis settings.
This may be worth getting the license server working.

 -- Mike Treseler

Andy Peters <google@latke.net> writes:

> I suppose I could go back to scripts and Makefiles, but EDK adds
> another whole level of bullshit to the process and that makes it
> difficult to use without the GUI.

I've integrated XPS acceptably (not perfectly :) into my script-driven
flow.  As EDK is makefile controlled, I have: 

xbash -c "cd some_xps_dir && make -f system.make netlist program" 

at the start of my synth/PAR script.

Cheers,
Martin

-- 
martin.j.thompson@trw.com 
TRW Conekt - Consultancy in Engineering, Knowledge and Technology
http://www.conekt.net/electronics.html