adexl corner simulations

Dear Paulinho,

If I understand your question correctly, I believe you can. I have used the method many, many times. From the list of simulation results in the History panel, select the results you wish to complete the simulation for cases that are not simulated. If you right click on the specific results, as I show on Figure 1, there is an entry "Re-run Unfinished/error points". Choose this option and Assembler/Explorer/ADEXL will determine those corners that did not complete, duplicate the completed corners and re-run the simulation corners that did not complete. It will re-name the database with a suffix "...rerun.XX".

Does this help or, perhaps, did I misunderstand your question Paulinho?

Shawn

Figure 1

Figure 1 follows:

thanks Shawn. Thts exactly wht i had asked. In fact I had tried the same shown above, when the simulations were still running. tht option to rerun appears only after that job is finished. So couldnt figure it out earlier.

Dear Paulinho,

paulinho said:

Thts exactly wht i had asked. In fact I had tried the same shown above, when the simulations were still running. tht option to rerun appears only after that job is finished. So couldnt figure it out earlier.

Well, I'm glad I managed to understand your question anyway!

I believe it did not work while the simulation set was still running as I assume Assembler/Explorer/ADEXL has not completed writing its results (or pointers to the individual result corners) to its main psf directory. I believe it may use the main psf directory to determine which specific corner cases did not complete successfully. Hence, intuitively, your observation makes sense to me in that the simulations must end before the option to re-run the incomplete corners or corners with simulation errors is enabled.

In any case, thank you, Paulinhp, for letting us know and I am happy to read the option helped out!

Shawn

I do have another related query, so posting in the same thread. As shown above, I see an option to combine 2 result sets using 'Merge & Load to active". Contrary to this, do I have an option to split a result set into 2 or more sets ?

paulinho said:

Contrary to this, do I have an option to split a result set into 2 or more sets ?

No. I struggle to see how this would work (what would the criteria be to split it up, for example? I can see this would get very complicated, and I can't quite see why this would be useful).

Andrew

Hi Andrew,

Say I ran a simulation for corners C1 to C8. Later on I realised that I need to simulate for C9,C10 as well. I run it seperately and then will try to merge both the result sets using the option shown above ('Merge and load to active'). Is that the utility of tht feature ('Merge')? 

I run a sim for corners C1-C30 and analyse the min max values of the simulated results. Later on I realise I want to analyse, not on the whole set of corners, but on C1-C10, C11-C20, C21-C30 seperately. Is there an option to do that, which seems contrary to the 'merge' option ?

Or am i misunderstanding the 'merge' option ?

paulinho said:

Say I ran a simulation for corners C1 to C8. Later on I realised that I need to simulate for C9,C10 as well. I run it seperately and then will try to merge both the result sets using the option shown above ('Merge and load to active'). Is that the utility of tht feature ('Merge')?

Yes. It could be that other ways of extending the data (such as additional sweep points) might be the reason for wanting to do a merge. This also commonly ties in with the use of reference histories.

paulinho said:

I run a sim for corners C1-C30 and analyse the min max values of the simulated results. Later on I realise I want to analyse, not on the whole set of corners, but on C1-C10, C11-C20, C21-C30 seperately. Is there an option to do that, which seems contrary to the 'merge' option ?

I'm not entirely sure why you would need to split the corners up to do that analysis, but this isn't possible - there's no "split" - and if there was, it would potentially need some complex way of describing how you did the split (I'm not sure I recall anyone asking for this so far - but I didn't check today).

Andrew

paulinho said:

Say I ran a simulation for corners C1 to C8. Later on I realised that I need to simulate for C9,C10 as well. I run it seperately and then will try to merge both the result sets using the option shown above ('Merge and load to active'). Is that the utility of tht feature ('Merge')?

Yes. It could be that other ways of extending the data (such as additional sweep points) might be the reason for wanting to do a merge. This also commonly ties in with the use of reference histories.

paulinho said:

I run a sim for corners C1-C30 and analyse the min max values of the simulated results. Later on I realise I want to analyse, not on the whole set of corners, but on C1-C10, C11-C20, C21-C30 seperately. Is there an option to do that, which seems contrary to the 'merge' option ?

I'm not entirely sure why you would need to split the corners up to do that analysis, but this isn't possible - there's no "split" - and if there was, it would potentially need some complex way of describing how you did the split (I'm not sure I recall anyone asking for this so far - but I didn't check today).

Andrew

Dear Paulinho,

You might consider using an ocean script to post-process your simulation results. This is what I do. In this fashion, you can easily combine results from different simulation sets (as I think this is what you want to do by adding a number of corner cases) or do an analysis of a subset of existing corners. In this fashion, you do not have to manipulate in any way the Assembler/Explorer/ADEXL database, but simply point your script to the appropriate set or subset you want to study.

I hope my comment makes some sense to you. Let me know if I need to clarify it!

Shawn