Oceanscript run is taking more time than ADE run in IC616

Hi,

Can you provide some additional details?  Is the core simulation time greater (i.e. are you comparing the elapsed time reported at the bottom of the spectre.out file) or are you measuring from pressing the run button in ADE vs. executing "ocean -restore oceanScript.ocn" or some similar command?  If it is the latter, there is some overhead associated with starting the OCEAN process, initializing an ADE session within OCEAN, etc. before the simulation actually starts.  Is this for a relatively short simulation where that startup time could be swamping out the simulation time?  How large of a difference in time are you seeing?

Regards,

TOM

Tom,

The other thing I've sometimes seen is that if the analysis is particularly verbose (for a short simulation), the printing of the spectre output into the CIW (or terminal) can slow down the elapsed time noticeably. This can be be controlled somewhat by using the annotate options in each analysis though.

However, I agree entirely - we need some more detail here as to what is slow...

Regards,

Andrew.

Dear Tom/Andrew,

The initial overhead is NOT much.

Both ocean & ADE show same elapsed-time at the end ( 17s in the simple test-simulation).

However, ocean actually takes 2min ( I measured using a stopwatch just for quantification ).

I actually ran a pss & pac simulation. The pss simulation times are same in both cases.

But, as Andrew has mentioned, the PAC is certainly verbose ( 2001 steps (100KHz -step for -100M to +100M sweep) ).

In fact this PAC print on the terminal is very slow and time gets consumed, though cadence reports approximately same time at the end of simulation for both cases.

Also while printing into a file ( by fprint() with %0.8e , it takes almost 38 mins to write above 2001 points, both real part and imaginary part and the frequency ). Paradoxically,this is much much faster in case of IC5141.

Kind Regards,

RFStuff said:

Also while printing into a file ( by fprint() with %0.8e , it takes almost 38 mins to write above 2001 points, both real part and imaginary part and the frequency ). Paradoxically,this is much much faster in case of IC5141.

I don't understand what you mean by this. Writing 2001 points to a file cannot take 38 minutes in SKILL, so I must have misunderstood what you're saying.

Are you running the OCEAN script from the CIW in virtuoso, or are you using the "ocean" executable?

It might be interesting to see how it behaves if you load this code here, and then use:

abMute(run())

instead of just run()

Regards,

Andrew.

/* abMute.il

Author     A.D.Beckett
Group      Custom IC (UK), Cadence Design Systems Ltd.
Language   SKILL
Date       Dec 22, 2011 
Modified   
By         

A macro to mute the output from printf() calls etc.

abMute(printf("test this\n"))

will produce nothing.

Relies on SKILL's dynamic scoping, so will not work in SKILL++
unless you do:

inSkill(abMute(printf("test this\n")))

***************************************************

SCCS Info: @(#) abMute.il 12/22/11.11:03:17 1.1

*/

(defmacro abMute (@rest body)
  `(progn
     (unless (openportp (getqq abMute nullport))
       (putpropqq abMute (outfile "/dev/null") nullport))
     (let ((poport (getqq abMute nullport)))
       ,@body
       )
     )
  )

Dear Andrew,

The SKILL script is disabling the print.

But still the same amount of time is being taken by the Ocean-Script run.

Kind Regards,

Sounds a bit odd in that case. Are you running from "ocean" or "virtuoso" (i.e. do you have a CIW or not)?

Regards,

Andrew.

I am running from  virtuoso.

Dear Andrew,

I have filed a service request regarding this.

But they say, since my OS is CentOS they can't help.

I am scared as I am middle of a design and I can't  go for RHEL-5/6.

I observed following things:-

1:- With APS disabled. ocean runs ok. But with APS enabled only PAC simulation step gets slower.

PSS runs properly

But ocean reports same time elapsed as that of ADE

2:- It takes a longer time for writing into files:

I use " fprintf( myPort1  "%0.8e%s  %0.8e%s  %0.8e \n",Freq_BB,",",BB_GAINr,",",BB_GAINi )".

Is there any other way of writing into files ?

Kind Regards,

I'm rather surprised that they said they couldn't help. What is the case number? I'll check.

This sounds very strange to me. I can't see a good reason why pac should be slower. Also, it's hard to imagine anything faster than fprintf, so something very strange is going on.

Regards,

Andrew.

Dear RFstuff,

> Is  there any other way of writing into files ?

Please refer to the ocnPrint() command in the ocean manual. This command is much faster than fprintf().