How to parallellize a loop with arrays?

[Yamaeda]

Hi,

Trying to improve the performance of a program i want to parallellize a loop. This loop uses a vi which returns an array, and i assume this can be of different lengths each iteration. The current solution is a classic Concatenate arrays via shift register disallowing parallellization, but there should be some workaround, right?

/Y

[GerdW · Accepted solution]

Hi Yamaeda,

whenever you need to concat elements in an array in a loop you need a shift register, either by BuildArray or ReplaceArraySubset (with a predefined array)...

One possible workaround: When the arrays from your subVI differ in length you should put them in a cluster. So you can use autoindexing to build an array of cluster of array - with the disadvantage of needing another loop to concatenate the subArrays later on...

Is it worth it?

[Yamaeda]

Yeah, that's a good solution, i'll give it a try! If it's worth it? Well the sub-vi does some analysis which should take some time, this is more of a check for possible routes to speed it up. 🙂

/Y

[altenbach]

---

@Yamaeda wrote:

The current solution is a classic Concatenate arrays via shift register disallowing parallellization, but there should be some workaround, right?

---

If you use the concatenating tunnel, the second loop can be parallelized just fine. 😄

 

[GerdW]

Hi,

ok, that seems to be a good reason to finally upgrade from LV2011...

When the tunnel supports parallelization you could fulfill the task with just one loop (and without the bundle/unbundle step in between)!

[Yamaeda]

Good point Old creek. However i'm sitting with 2010 right now. It sure improves parallellizability in general though!

/Y