LabvIEW 2013 vs 2015 - will my built exe consume less CPU if I upgrade?

[josborne]

Hi all,

I have a few applications that are built into executables using LV2013.  The applications are running fine, buyt they are resource hogs.  They can sometimes consume so much CPU that I ask my customers to not run any other applications when using my EXE.  I know there are lots of tweaks I can do to improve the performance of my code.  I have done many of those tweaks, and I am somewhat satisfied with the result.  But would upgrading to LV2015 allow my resultant EXEs to perform even better (use less CPU?)?

I hear that LV2014 and LV2015 focus a lot of performance and stability.  But does NI have any metrics for the performance enhancements?

I am very underwhelmed by the last two releases of LabVIEW (2014 and 2015), so the only reason I would upgrade is if my EXEs would perform "faster".  Any idea if this is true?

I suppose I could download a demo and do some benchmark tests.  But that would probably take me 2-3 days seriously.  

[ToeCutter]

Well there are optimisations in 2015, but if it will imrove the performance of your application is hard to say. Depends on how you coded it. If you have spinning loops polling for input without a wait in them, it probably won't improve things.

[altenbach]

There is no way to tell without trying, but my guess is that if your excessive CPU useage is due to inefficient programming it probably won't make a difference. It would only matter if the high CPU is due to a specific bug. We would need significalty more information.

Do you know why the program uses so much CPU? (Just a lot of work to do? Fast loops recalculating the same resut from the same inputs as fast as the loop allows? Excessive data dependencies preventing parallelization? Tight polling loops without wait? etc.)

[Hooovahh]

Yeah I'm with altenbach.  It doesn't hurt to try, but inefficient code will continue to be inefficient code in 2015, if that is your problem.  If you do make the plunge please post back here your results, the community would be interested in real word improvements.  NI can say 9X and 4X all they want and while I believe we are generally moving to a better performing LabVIEW, I'd like to see some real test results.

[josborne]

Thanks, everyone.  I think I'll just try it out.  I will report my findings here.

I agree that my code probably probably has some inefficiencies.  Its been a while since I looked at it.  But I believe the major CPU drain was related to graphics and panel updates.  If I recall, I used property nodes to do some rather frequent and complex graph updates (updating and scrolling a graph).  I'll get back to you soon.