08-23-2012 02:05 AM - edited 08-23-2012 02:09 AM
@P Anand wrote:[..]I know there are Pros and Cons between the Shift register and the feedback node but where is given feedback node is correct compared to shift register. [..]
You do know, that during the transforms (DFIR i think it is in this case), ALL feedback nodes are replaced by shift registers?
Norbert
EDIT: Altenbach pointed out correctly, that this transform was, i think, modified with LV 2010. So on older versions of LV, FN and SR do have different performance.
08-23-2012 02:12 AM - edited 08-23-2012 02:14 AM
@P Anand wrote:
This is what I get when I run your code still you stick on to your point?. ( LV2011 version)
Can you post your benchmark VI?
Also please give the system parameters (RAM, OS, CPU and so on).
I see no actual reason for such a huge difference between the performance since, as written before, there should not be any performance difference if both mechanism are used in a comparable way.
So either you have a flaw in how you benchmark the timing or you are using both in a different way!
Norbert
EDIT: And again, as Altenbach pointed out correctly, you DID DISABLE debugging for you benchmark, no?
08-23-2012 07:38 AM
I feel sorry for the OP since this thread has morphed into lively exchange regarding shift registers and feedback nodes.
08-23-2012 07:45 AM
The VI was posted here:http://forums.ni.com/t5/LabVIEW/how-to-do-every-600-seconds-to-execute-For-loop/m-p/2132502#M690963
They are looking for a difference between an expanded shift register and 3 feedback nodes. With debugging off, I get fairly comparable results between the two methods (FBN actually slightly faster).
08-23-2012 08:21 AM
@Norbert_B wrote:
@P Anand wrote:
This is what I get when I run your code still you stick on to your point?. ( LV2011 version)
Can you post your benchmark VI?
Also please give the system parameters (RAM, OS, CPU and so on).
I see no actual reason for such a huge difference between the performance since, as written before, there should not be any performance difference if both mechanism are used in a comparable way.
So either you have a flaw in how you benchmark the timing or you are using both in a different way!
Norbert
EDIT: And again, as Altenbach pointed out correctly, you DID DISABLE debugging for you benchmark, no?
That is not my benchmark vi and crossrulz has given the link. When I run the code in my machine (2011) I got SR outperformed FN ( I am not sure debugging enabled or not). I support neither SR nor FN just trying to know what drives FN more than SR in performance.
08-23-2012 09:11 AM
Hi 多多,
Here is my VI attached. Sorry, i dont read the message before.
Good Luck.
About SR vs FBN
At LabVIEW 2009, debugging off, I got SR~9.6sec and FBN ~8.1sec.
Best Regards,
08-23-2012 09:35 AM
08-23-2012 09:41 AM - edited 08-23-2012 09:41 AM
Here is a much simpler VI that will do the exact same thing. I even added a stop button to illustrate a point I made earlier.
08-23-2012 10:10 AM
Hi crossrulz,
Yeah your VI is very simple. Now understand why you call my VI a Rube Goldberg Machine . I was seeing only the images with for loops. And because my english is unskilled, i didnt understand when you explained me.
Probably my mind wanted to use FBN.
Best Regards,
08-23-2012 10:41 AM - edited 08-23-2012 10:44 AM
@Luis_AM3C wrote:
Probably my mind wanted to use FBN.
I know, find a new shiney toy and you want to use it for everything. We've all been there. And we all then use them in ways they were not really meant to be used. Don't be afraid of FBNs, but remember that other constructs may work better. It really depends on the situation.
Just for fun, here's a couple more ways it can be done, one of which is still using a feedback node, just for you. I would still with the shift register version because I have ran into cases where an event will determine the timeout. So it is cleaner to use a shift register.