Spectre alter statement very slow

[Joel]

I am trying to simulate process variation with spectre using the alter
statement. I use ocean to load the design, add a model file that
contains a set of alter statements and then run the simulation. This
has been recommended in previous posts.

The problem I have is that there are several process parameters for
each transistor in the design and for large designs there are a large
number of alter statements. A simulation that takes 3 seconds without
the alter statements can take a minute or more with alter statements.

It seems like the alter statement is surprisingly slow. Why does it
take so long? Does anyone else have this problem? Is there a
better/faster way of setting process parameters from ocean? Is it
possible that I am doing something wrong?

 

[fogh]

Joel wrote:

I am trying to simulate process variation with spectre using the alter
statement. I use ocean to load the design, add a model file that
contains a set of alter statements and then run the simulation. This
has been recommended in previous posts.

The problem I have is that there are several process parameters for
each transistor in the design and for large designs there are a large
number of alter statements. A simulation that takes 3 seconds without
the alter statements can take a minute or more with alter statements.

It seems like the alter statement is surprisingly slow. Why does it
take so long? Does anyone else have this problem? Is there a
better/faster way of setting process parameters from ocean? Is it
possible that I am doing something wrong?

Hi Joel,
you can try your luck with
- altergroup,
- a report of the performance problem to cadence support, (first do an
upgrade of MMSIM as necessary),
- a modified version of the model file where you add your own "section"
to define process params with the "parameter" statement.

 

[Joel]

Thank you very much for your response.

I have never used the altergroup statement, but it does not seem to do
what I want it to do. It seems like I would have to rewrite the model
statement when all I want to do is change a few parameters. Is this
correct? I was hoping to avoid that level of complexity.

Adding sections to the model file is certainly a valid option, and I
have explored that to some extent, but it does not provide the
flexibility I was hoping for. It would be nice to be able to set
parameters to values without needing edit the model file each time.

What are other people doing to try to investigate changes in parameters
between a number of Spectre simulations. Is the alter statement the
best method even though it is slow (has anyone else noticed that it is
slow)? Has anyone had success using other methods?