Show Buffer Allocations on Feedback Nodes

[chris.b]

JackDunaway (mechelecengr) wrote:

What I REALLY wanted to hear is that Feedback Nodes became magical in LV2009, that are infinitely fast and consume zero memory. Until NI is able to implement that execution priority [...] we will need to depend on the Show Buffer Allocations for FN code validation.

---

While Feedback Nodes aren't quite magical, they actually are pretty smart...

 

In the first case, LabVIEW is able to (magically?) reuse the buffer from the initializer terminal.  This is exactly what would happen with a standard shift register.

In the second case, we see an alloc at the increment primitive, as this output value may or may not be stored in the feedback node, based on the Enable input.  We must keep a copy of the last "Enabled" FN value, as this may be needed for the next iteration. 

[chris.b]

I forgot to note, these screenshots were in LV 2009.

As far as the CAR goes, we can definitely look into this, but here's my theory:

The dialog is correct; the "missing" alloc in 2009 may simply be there result of internal optimizations.