Learning Forum

[AndyJP]

I am trying monitoring simulation performance, but HFSS does not report when simulation was started.

So I tried to output the simulation start time using embedded VBA scripting engine.

But it appears, message output is also not described in the scripting manual.

 

So How can I put the simulation start time in the message log from a script directly or indirectry?

[rtk]

Hello,

If the aim is to get the report on when simulation have started, you can go for analyzing the Profile Data. (Result >> Solution Data >> Profile).

This will give in detail simulation flow.

 

Regards,

[AndyJP]

AH, thank you very much!

 

And is there any way to reflect it in the message log? (because that is what logs were invented for)

... or putting a user string in the message log, like printf() or disp() ?

[AndyJP]

Why would I need it? Because I am scripting. And every step is separate calculation. But the script execution definitely has a start and finish time.

Did you try figuring out the start time from hundreds of variations in out of order collection? Almost impossible. Logs instead are stored in chronological order, where the start time of first calculation is actually a start time.

[mbeliaev]

you can use all functions of IronPython.

 

Please use logging module, writing to file or write messages to the AEDT message window.

 

[AndyJP]

I am not familiar with Python. Isn't there a way to print text to the message window from VBA script?

 

I tried debug.write, it gives just "text parsing error", debug.print does not work.

 

It looks like VBA implementation is crippled in EDT2020R. Really, why ANSYS stopped maintaining scripting engine? It does not even allow simple array declaration, other than using array(). Types are also seem limited to default

[mbeliaev]

VBA is previous century

 

please switch to Python. It has a wide community and it would be easy to use

[AndyJP]

I understand your point.

I would even switch to AMD from Intel, and to RedHat from Windows, and even install 4 Tb of Optane memory one day.

But since it takes efforts and time, and I need it urgently...

Is there a way of putting a text string from a script to the message log? I can't believe there is no, because it is not so new, and in 30 years there should have been many requests and orders for such feature... at least debug.write produces some effect, just not that was expected.

[mbeliaev]

my personal opinion, is better to stay with Intel

 

 

https://ansyshelp.ansys.com/Views/Secured/Electronics/v195/home.htm#../Subsystems/Maxwell/Subsystems/Maxwell%20Scripting/Content/AddMessage.htm%3FTocPath%3DMaxwell%7CMaxwell%2520Scripting%7CDesktop%2520Object%2520Script%2520Commands%7C_____1

[AndyJP]

Ah, I got it. I need a an HFSS-F1 token to access that page. You have to mention such things.

 

Thank you, I have not seen it in the PDF scripting manual.

Will try it today

[AndyJP]

 BTW, why Intel?

Before Zen, AMD's FPU had better precision and error-shaping, but was slower.

Now, Zen2 has comparable, or better FPU speed than Intel. AVX is still not 512bit (2x256 bit fusing), but in Zen3 it should become multiple of 512x AVX blocks per core, with flexible simultaneous access; bye-bye buldozzer scheduler problems.

Do you have information that it is still inferior to Intel?

[mbeliaev]

It is also in PDF scripting help:

[AndyJP]

Thank you very much! I was using an old PDF manual. The new one has much better scripting part, despite worse general HFSS usage sections. I just had to access it from machine with HFSS installed, or I could not access the page.

It did exactly what I wanted: a message followed by a timestamp.

 

About CPU's, I am not speaking about promoton, but in the past Ansoft and Ansys were releasing benchmarks. So I am curious about current state of things, since AMD made a big leap with Zen2, and will make bigger one with Zen3 in 2021. Intel's progress seems too slow, they didn't change general FPU design for decades, just adding more features above the old bones.

[mbeliaev]

I am not sure that we have any public available benchmarks.

 

Any case this is a separate topic and current thread I suppose can be closed