Case statements in Design Variables and AMS UNL

Dear Christian,

Without seeing the full input.scs, this is difficult for me to assess what the syntax error might mean.  The error codes, however, suggest it may be due to a missing comma in your case statement due to the ternary operator and if your design variable "conf"  is a local design variable. Have you seen the following Cadence on-line Support article Christian?

https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1Od0000000nXT2EAM&pageName=ArticleContent

This article provides a couple of workarounds that may. or may not(!), be helpful to you...

I hope this helps!

Shawn

Dear Christian,

Without seeing the full input.scs, this is difficult for me to assess what the syntax error might mean.  The error codes, however, suggest it may be due to a missing comma in your case statement due to the ternary operator and if your design variable "conf"  is a local design variable. Have you seen the following Cadence on-line Support article Christian?

https://support.cadence.com/apex/ArticleAttachmentPortal?id=a1Od0000000nXT2EAM&pageName=ArticleContent

This article provides a couple of workarounds that may. or may not(!), be helpful to you...

I hope this helps!

Shawn

Actually, the case statement is a SKILL function, and so should be evaluated before it even produces the netlist (case would not be understood by spectre). That suggests there is some kind of evaluation error - normally what happens is that ADE trues to evaluate it as SKILL, and if that fails, it passes it on to spectre with the hope that spectre can handle it (which would fail in this case).

I initially thought that the missing spaces between the close and subsequent open parentheses would be an issue, but SKILL actually handles that without problem.

I've not had a chance to test it - rather busy this week. I would advise customer support to follow this up... (you could always check with the latest hotfix first, given that there have been a few fixes related to how dependent expressions are handled due to something breaking recently due to trying to handle parameters locally within the simulator where possible - it's possible that this is confusing matters). The fix for that was in ISR15 (one after the version you're using). You could try doing:

setenv AXL_NETLIST_WITH_VALUE 1

or

export AXL_NETLIST_WITH_VALUE=1

(depending on the shell you're using) before starting virtuoso (in the version you have; in the later version this shouldn't be necessary).

Note, this is a wild guess based on other issue with dependent expressions - I've not tested it. My theory is that the check for expressions that should be handled by the simulator is mistakenly passing the expression onto the simulator rather than evaluating it in SKILL, rather than it being due to an evaluation error.

Andrew.

Hi Andrew -

This morning I woke with the same hunch as you, and tested running more sweeps (in LSCS mode) - low and behold - it parses the netlist. I guess it then properly evaluates the SKILL code before parsing it to the input.scs? I still can't netlist using RMB -> netlist -> create..., only through starting a simulation does it pass.

I tried to add setShellEnvVar("AXL_NETLIST_WITH_VALUE=1") to my cdsinit (per this suggestion), with no luck - it still doesn't evaluiate the SKILL code before parsing it to netlisting.

Is there someway to force evaluation of SKILL before netlisting?

BR,
Christian

----

EDIT: Per Shawns suggestion, I added the SKILL code to global variables, instead of local - now it works with a single run.

Hi Shawn,

I've tried with several different syntax around the SKILL - it seems to be fine as shown, however, it is something to the evaluation of the SKILL.

Thank you for the suggestion - I have tried to set the SKILL coded variables all to global variables, and it works! At least when pressing simulation, and not when netlisting. So that seemed to do the trick for now.

Now I have a different issue with  ERROR(ASSEMBLER-2709) but I should have fixed that error with a env variable - I will contact cadence support about that issue.

BR,

Christian

Christian,

I suggest you follow this up with customer support - I think it should work as local (test) variables too. Hopefully it's resolved by the latest hotfix anyway, but if not, it would be good to ensure this is fixed.

Andrew

I will do that, thank you.