make ade use the spectre data directory when plotting signals

In general you should not have to mess around like this. What normally happens is that after each simulation point is completed, ADE will evaluate any expression with the results from that specific spectre run, and if the results are scalar then the value is returned to ADE and stored in a database. If the result was a waveform, then it is not stored.

Later, if you wish to plot the waveform expressions, it does indeed use the root of the overall results sets so that that waveform is plotted with the context of which parameter values are in effect - so which corner or sweep values are relevant to that point (which is missing if you looked at just the point data).

You didn't explain what the actual problem is - you said "some of my maestro outputs depend on files located in SPECTRE_DD. ade produces an error when I plot the outputs because it uses ADE_DD while the outputs refer to files located in SPECTRE_DDs using relative paths". I'm not sure what is failing, or why, but I think you're trying to subvert the way it works (for good reasons) without giving any details of the actual problem - it's better to understand the actual problem first and maybe that will give you a clear path to resolution.

So, please give the output expression that is failing, and what the error is.

Andrew.

OUTPUTS AND THE ERROR LOG

In the chain of outputs executed I had the two outputs:

‘
return_true’ function returns ‘true’ always

as a side note, the file ‘data.csv’ is used by a third-party software for further processing thereafter.

The simulation has finished. I right click on one of the signals > plot > the first error appearing in the ciw log is:

*Error* ("error" 0 t nil ("*Error* error: basic_string::_S_construct null not valid" nil))







ERROR (ADE-1057): Could not evaluate expression

Name:'save_restore_vs_in'

Expression:'(return_true(psf_dir) && awvSaveToCSV(restore_vs_in strcat(psf_dir "/data.csv")))'

because of the following error(s)

("error" 0 t nil ("*Error* error: basic_string::_S_construct null not valid" nil))

Ensure that the definition of the expression is correct.

FIX

I replaced the built-in awvSaveToCSV with the custom awvSaveToCSV_family:

procedure(

awvSaveToCSV_family(wave path)

if((famIsFamily(wave) == nil)

then

awvSaveToCSV(wave path)

else

famMap('awvSaveToCSV wave path)

)

)

DISCUSSION

with this fix the program works as I need. ‘Plot’ menu item plots waveforms without an error, ‘Plot across corners’ menu item plots a waveform family without an error as well. so I had a bug in one of my commands, the problem was not with ade. The reason why I thought that the problem was in an in-context results directory is because the error disappeared when I amended manually the context in calculator.

in the original post there was a question how to turn existing behavior into desired behavior.

When ade executes outputs during “the first run”, it uses a spectre data directory for each simulation (like …/Interactive.581/1/tran/psf ) (the data directory path is returned by openResults()). If there is a save command, the file will be saved to the spectre data directory. Importantly, during this “mode” outputs are treated as single waveforms (or single scalars).

When ade executes outputs with a user right click, after all simulations have finished, it uses a different, top level data directory (like …/Interactive.581/psf/tran/psf). If there is a save command, the file will be saved to the ade, top level data directory. Importantly, during this “mode” outputs are treated as waveform families (or scalar families), not single waveforms.

Post-simulation effectively ade produces exactly the same results as when it uses spectre data directory (for an individual simulation run), at least in my case. I checked this by manually tweaking in-context results DB in calculator and evaluating an output. The only differences, I noticed, are that 1) the skill working directory is different, 2) waveform and scalar families are created instead of single pieces of data when in the first mode. So, there is no motivation to turn existing behavior into desired one, because effectively the program operates as needed.

Andrew, thank you very much. Your post was very insightful and helped to find a bug. I am sincerely grateful to you. I am satisfied with the way ade works with regards to this matter.

In general, expressions/calculator functions should always make sure they can deal with both single waveform and family data, as then everything works properly - as you've found here. So I'm glad you've resolved it and are happy with the solution!

One thing I should point out is that your expression for psf_dir really shouldn't need the dependency on the previous results; I think it will be OK with just using openResults() to retrieve the results directory. Similarly, the save_restore_in doesn't' need the return_true(psf_dir) because you're already referring to psf_dir in the expression and so that would define the dependency anyway. This is just to clean up your expressions and make them a bit more readable.

Regards,

Andrew

And just for reference, here is the corresponding article from the Cadence Support website: https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O0V000007MkO7UAK 

And just for reference, here is the corresponding article from the Cadence Support website: https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1O0V000007MkO7UAK