topic Re: time spent in loop in LabVIEW

time spent in loop

xradian — Fri, 09 Mar 2007 20:29:12 GMT

Hi,
I wanted to know if there is a way to measure how much time (millisecond) is spent in a loop. I have a for loop with 200 iterations (attached as jpg) and would like to know how long the VI stays in the loop.

Thanks

Re: time spent in loop

Beaton — Fri, 09 Mar 2007 20:43:48 GMT

Here is a quick and crude way to do it

Re: time spent in loop

BrianInTheForest — Fri, 09 Mar 2007 20:49:50 GMT

Here is a method I use. I use two tick count(ms) functions one and subtract the difference.

Brian

Re: time spent in loop

AndrewAlford — Fri, 09 Mar 2007 20:50:58 GMT

You can place the elapsed time express vi in the loop, make sure to wire the Reset and Auto Resest Inputs to false.

Message Edited by AndrewAlford on 03-09-2007 01:51 PM

Re: time spent in loop

xradian — Fri, 09 Mar 2007 23:00:08 GMT

thanks for all the replies. I used the last method suggested as it was the easiest.

Re: time spent in loop

BrianInTheForest — Sun, 11 Mar 2007 21:41:43 GMT

I would not add anything into my loop that might add to the total process time. The easiest method to implement might not be the best for performance.

Brian

Message Edited by brianlachlan on 03-11-2007 05:42 PM

Re: time spent in loop

altenbach — Sun, 11 Mar 2007 22:08:43 GMT

Brian,

For me it is still too expensive to take the tick count at every iteration, even though you only need the last value. If you take the second tick count only after the loop has finished, the VI shown is yet another 8x faster.

(Note that I cranked it up to 1000000 iterations).. Tick "inside": 64ms, as shown: 8ms. 🙂

Yes, sometimes a sequence structrure us useful! 😉

(I also wired the array across the sequence to make sure the indicator update happens after the second tick probe.)

Message Edited by altenbach on 03-11-2007 03:11 PM

Re: time spent in loop

BrianInTheForest — Sun, 11 Mar 2007 22:36:17 GMT

Christian,

Thanks for the advice, would not have thought the tick count added so much.

Brian

Re: time spent in loop

altenbach — Mon, 12 Mar 2007 08:33:08 GMT

Well, it's actually not adding not that much per iteration (56ms/1000000 iterations, or about 56ns). It will only make a real difference for very fast loops. 🙂

I agree that the express VI is a drag, though. 😮

Re: time spent in loop

xradian — Mon, 12 Mar 2007 16:37:30 GMT

when i use the tickcount method it takes a longer time for my loop!?
When I used the time elapsed method 10 iterations took 0.25 sec but with the tickcount method I'm getting 1392ms.

Re: time spent in loop

altenbach — Mon, 12 Mar 2007 17:00:58 GMT

I think your loop time is entirely determined by the hardware interaction, so the the timing is insignificantly fast in both cases.

However you current timing code is heavily flawed because it is impossible to tell at what time the first tick count is taken (No data dependency!). Most likely, the first tick count executes BEFORE (or in parallel to) the "OPEN1" and "OPEN2" nodes, so you are actually including their times in the measurnements. Remember, LabVIEW does not execute left-to-right. Any node that does not need to wait for inputs is ready to execute immediately.

You need to place the first tick count inside a sequence frame too to ensure it executes right before the FOR loop. For exact timings use a 3-frame sequence as in the image.

Make sure that everything that should not be timed is before or after the sequence. Also make sure there is no code that can run parallel to the sequence.

Message Edited by altenbach on 03-12-2007 10:01 AM