button of "re-evaluate results" doesn't reflect changes in assembler/ade-xl

Dear monglebest,

monglebest said:

Situation 1: my simulation failed for somereason (disk full and hence the table shows the simulation status is "running") and hence I use the command under netlist/runSimulation to re-run the simulation in linux terminal(after disk is cleaned), but the button in ade-xl/assembler cannot reflect the latest data after the correct simulation is done (the table box sitll shows "running"). How can I have the ade-xl/assembler to refelct the updated data from psfxl files?

I do not believe this it is possible to use the "Re-evaluate results..." evaluation option in Assembler after you re-ran the simulation that did not complete from the Linux terminal. I filed a case on this issue some time ago. The basic issue is that Assembler uses the main psf data directory of a database to access the simulation results located in each individual psf directory for the simulation database. However, when you ran the simulation from the individual netlist directory using its input scs and runSimulation command line, the main psf directory was not updated to reflect that the results for that specific case are complete and valid. A way to use that option for a simulation set that does not complete correctly, is to right click on the database from the History tab and choose the Re-run Unfinished/Error Points option. This will copy over the database and re-run those individual simulations that did not complete correctly and update the main psf directory. In Figure 1, I circled the Re-run Unfinished/Error Points option and also an example of the new database created from the unfinished simulation set Interactive.13. I hope this helps monglebest!

monglebest said:

Situation 2: I have a skill code updated (using "load" command in CIW) for calculator expression and then click the yellow button, want to have the table in ade-xl/assembler to reflect my latest skill code changes, but the evaluation fails or keeps the old skill code results. I have to re-run the simulation in ade-xl/assembler to have the table updated. How can I avoid re-run the simulation to have the data updated correctly in the table?

If you are referring to the example you noted where you manually re-ran the incomplete simulation, you will not be able to evaluate the result for that incomplete simulation as the main database is not aware it is complete. Or, are you referring to the use of new skill code for a completed simulation set? If the latter, perhaps your new skill code is not being applied to the correct test. Note that the outputs in Assembler are all assigned to a particular test. I'm not 100% clear, I am afraid, of the speciifcs of your second situation - sorry!

Shawn

Figure 1

Hi ShawnLogan:

Thanks for your reply.In all situations that I use assembler, I only enable one corner, there is no sweep. I only use assembler's results function to keep many results available at the same time because explorer will overwrite from previous results. so I think the psf directory I access should be the only simulation database.

In situation 1, the reason I don't use "Re-run unfinished/Error Points" is because I sometimes modify the netlist and hence want to simulate the modified netlist (Cadence GUI sometimes have significant delay in response, or just re-generate netlist will takes a few minutes, I don't want to wait). But I prefer the output format assembler gives. Another reason I don't use "Re-run unfinished/Error Points" is because when I have issue with simulation, usually I need to re-start the Cadence to make it clean and work properly. If I don't re-start Cadence, I usually failed to have correct simulation running when "re-run unfinished/Error Points", hence I prefer just start the simulation from the command line, which is more clear/reliable. 

In situation 2, I didn't re-run the simulation. I just re-evaluate the results. But because my skill code is updated, I would expect the assembler results table will reflect such changes. The calculator in assembler can correctly refelct my updated skill code results but the assembler results table cannot. For example, in my old skill code, the result calculated is 43 (positive 43). Now I update my code to make it -43 (minus 43). Calculator can report -43(minus 43) with new skill code loaded, but assembler result table still report 43 (positive 43).

I wonder if there is any commmands in ocean code that can do this job? 

Regards

monglebest

Dear monglebest,

monglebest said:

In all situations that I use assembler, I only enable one corner, there is no sweep. I only use assembler's results function to keep many results available at the same time because explorer will overwrite from previous results. so I think the psf directory I access should be the only simulation database.

Two comments with respect to this comment if I may...

1. If you include the following in your .cdsinit file:

; Increments number of Explorer runs to avoid overwriting

envSetVal("maestro.explorer" "onHistoryNameCollision" 'cyclic "IncrementAsNew")

Explorer will not overwrite the ExplorerRun.0 database but will increment the number of the database. Hence, you will not need to use Assembler to save multiple datasets.

2. In Assembler, even if there is a single corner in your test, there still exists a distinct psf directory that Assembler will access. Hence, without its specific knowledge that the simulation you re-started in the command line completed successfully, it will not update the table of results. You can look in your database with the single corner in the test and you should see the main psf directory.

monglebest said:

In situation 2, I didn't re-run the simulation. I just re-evaluate the results. But because my skill code is updated, I would expect the assembler results table will reflect such changes. The calculator in assembler can correctly refelct my updated skill code results but the assembler results table cannot.

Did you try saving the outputs in a comma-separated file, deleting the outputs and reading in the comma-separated file and then re-evaluating? I don't know if your output expressions contain a reference to your skill code, but if so, make sure the output refers to the proper SKILL function.

Shawn

All that is probably needed here is to use the drop-down arrow next to the Re-evaluate button to select "full" rather than "incremental" evaluation. Normally it only re-evaluates results where the expression had changed.

The reliability issues mention here by monglebest sound a bit odd - I'm not sure why that should be the case and  you would be better working with custmer support to resolve that.

Andrew.

Thanks for Shawn's comment, I will try the explorer increment command definitely and also other method you mentioned.

In MonteCarlo.0 results below, I have my own skill function put in the table, row 1 report 43.4 and row 3 &4 are new function added in skill code. Though I expect after click Re-evaluate button, the row 1 number should be updated to -43 and also row3 &4 should give me a value, it doesn't do so.

Hence, I decide to start a new simulation, called MonteCarlo.1, as you see, the row 1 value has been updated to -43 and row 3 and 4 also have new values. But I should not need to run MentoCarlo.1 if the re-evaluate button in MonteCarlo.0 works.

A quick comment regarding Andrew's reply, I tried "full" evaluation in the situation 2 above, it didn't help update the value. The reliability issue, I guess, is related with ICRP process, which handle simulation job submission to grid server. We use RTDA here and somehow the job will time out with no activity before submitting to grid server sometimes. Once this problem happens, it happens always unless I re-start Cadence. I have case reported to the company CAD here some time ago. pending for their reply.

Regards

Monglebest

I tried "saving the outputs in a comma-separated file, deleting the outputs and reading in the comma-separated file and then re-evaluating?"

My Cadence crashed.

Dear monglebest,

Rats! That certainly was not my intention ! My apologies...I use this technique quite often to, for example, edit or copy new expressions into the set of Assembler outputs and the process has been robust. Unfortunately, it was not for you...sigh...you might restart a new Virtuoso session and try again if you have the patience.

Shawn