How to create a multi-threaded VI?

[LyLee]

I am working on a VI that uploads/downloads files from multiple FTP servers. Since FTP transmission is slow, I would like to open multiple FTP session at the same time, so as to maximize the network usage.  In other environments, it seems easy: I just create several worker threads, each establishes one session with a FTP server. However, how can I do this in LabVIEW?  Because the number of threads should be adjustable, I can not just put multiple FTP nodes on the diagram.  If I put them in the for loop, then those threads will be executed sequentially.  
Is there anyway in Labview to programmatically create worker threads?

[mishklyar]

Have you considered multiple instances of the FTP connection vi?

Just for fun I wired two VI's that launch multiple instances of random number generation.

I suppose that instead of random number it can be FTP connection

[LyLee]

can you post the diagram graph or save it in 8.0 format? Can't read it...

[tbd]

Hi LyLee,

      I think you want to make a reentrant VI for handling a single thread, then launch as many copies as you need (using VIs server.)

Cheers.

[Pnt]

this is mishklyar 's vis.. if you want to take a look..

Although i argee with tbd

[mishklyar]

Hi tbd,Pnt what is the difference between vi's that I attached and VI server? Probably I miss something

...attach here vi's for 8.0 version

[rolfk]

---

@mishklyar wrote:

Hi tbd,Pnt what is the difference between vi's that I attached and VI server? Probably I miss something

...attach here vi's for 8.0 version

---

The difference is that you use templates and the suggestion was to use VIs which have been set to be reentrant. Templates spawn a completely independant VI for each load operation which would be in terms of comparison to C more like a new task each time. Reentrant VIs just create a new data space for each instance but let thv e rest remain unspawned. This means you do not really have front panels and such for the reentrant instances but for a background thread that is perfectly fine once you debugged it properly.

Reentrant VIs instantiation is less costly in terms of memory usage and also time needed for the instantiation in comparison to templates.

Rolf Kalbermatter

Message Edited by rolfk on 05-04-2008 03:58 PM

[mishklyar]

Thanks rolfk ,very useful answer, (unfortunately I can't vote yet   ).

Where one can learn such a details?

[rolfk]

---

@mishklyar wrote:

Thanks rolfk ,very useful answer, (unfortunately I can't vote yet   ).

Where one can learn such a details?

---

Start as NI Application Engineer and work 5 years in there, then go to an alliance member and work another 12 years there, in all that time participate in various discussion groups such as Info-LabVIEW, NI forums, LAVA and keep being curious and inquisitive all this time and you should be about there

And this is only half way meant as joke.

Rolf Kalbermatter

[tbd]

Hi Rolf,

      Thanks for the nice description, but maybe a slight correction or qualification is in order?

---

rolfk wrote:

The difference is that you use templates and the suggestion was to use VIs which have been set to be reentrant. Templates spawn a completely independant VI for each load operation which would be in terms of comparison to C more like a new task each time. Reentrant VIs just create a new data space for each instance but let thv e rest remain unspawned. This means you do not really have front panels and such for the reentrant instances...

Message Edited by rolfk on 05-04-2008 03:58 PM

---

      In LV  7.1 this was true, however, as of LabVIEW 8.2, reentrant VIs can have their own FP.  I don't know about VITs - haven't used them much.

Cheers.