Does awd cache its data, and if where are they?

Hi,

got into a very interesting problem when I wanted to start the
simulator outside of ADE. Reason is that we start the simulator with
lsf, --Always--. For very small simulations like operating point of a
handful of transistors this is really time consuming so I start a
second shell and start the top level netlist directly in the generated
netlist directory (spectre input.scs). I select the result in ADE with
Results/select... and awd is happy to plot any signals that I select
via the calculator. So far so good.

If I then change the form of a pwl input voltage and rerun the
simulation the same way and then in awd do a File/update results I see
the results of my first simulation. I have to reselect the simulation
result using the Results/select... and then File/update results to get
the modifications.

Now I enter the generated psf directory and deletes all results and
then rerun my simulation on the command line. A File/update results
show me the data from the first simulation. I again have to do a
Results/select... to get my latest simulation results.

But from where did awd get those old simulation results? I thought I
deleted them when I deleted the contents of psf. And I verified that
new data were written into that psf directory. It looks as if awd
moves its data somewhere and keep it, because when I do a
Result/select... on a new dataset, I get a warning in CIW saying that

*WARNING* Data in "/share/.../spectre/schematic" doesn't correspond to
the cellView in window:5

Now, I am not able to find out which window is which, because I have
a couple of cadence windows and awd windows open at a time.

This slowly sounds more like a question to AE, but maybe other people
use lsf and have experienced the same. (It doesn't have anything to do
with lsf, though)

 

svenn> Hi, got into a very interesting problem when I wanted to
svenn> start the simulator outside of ADE. Reason is that we start
svenn> the simulator with lsf, --Always--. For very small
svenn> simulations like operating point of a handful of
svenn> transistors this is really time consuming so I start a
svenn> second shell and start the top level netlist directly in
svenn> the generated netlist directory (spectre input.scs). I
svenn> select the result in ADE with Results/select... and awd is
svenn> happy to plot any signals that I select via the
svenn> calculator. So far so good.

svenn> If I then change the form of a pwl input voltage and rerun
svenn> the simulation the same way and then in awd do a
svenn> File/update results I see the results of my first
svenn> simulation. I have to reselect the simulation result using
svenn> the Results/select... and then File/update results to get
svenn> the modifications.

svenn> Now I enter the generated psf directory and deletes all
svenn> results and then rerun my simulation on the command line. A
svenn> File/update results show me the data from the first
svenn> simulation. I again have to do a Results/select... to get
svenn> my latest simulation results.

svenn> But from where did awd get those old simulation results? I
svenn> thought I deleted them when I deleted the contents of
svenn> psf. And I verified that new data were written into that
svenn> psf directory. It looks as if awd moves its data somewhere
svenn> and keep it, because when I do a Result/select... on a new
svenn> dataset, I get a warning in CIW saying that

svenn> *WARNING* Data in "/share/.../spectre/schematic" doesn't
svenn> correspond to the cellView in window:5

svenn> Now, I am not able to find out which window is which,
svenn> because I have a couple of cadence windows and awd windows
svenn> open at a time.

svenn> This slowly sounds more like a question to AE, but maybe
svenn> other people use lsf and have experienced the same. (It
svenn> doesn't have anything to do with lsf, though) -- Svenn

Hi Svenn

Unix systems don't delete a file. Only the information about it's
existence is removed from the directory listing. The actual inode is
freed only when the last process that opened the file closes the
file (the open count on the inode is zero). So if awd opened the file,
it still has access to it.

Satya