Labview Multicore support

[GerdW]

Hi Mark-Ren,

on the last LabViewDay in Germany a demonstration vi was shown that compared performance on 1, 2 and 4 cores (and seems to run smoothly...), you can download it here (multicore example, point 3)!

I wrote: "(Overall performance will be better on 8-core as all timed sequences can run in parallel...)" Yes, your code should run faster on an 8-core.

But I meant: In each of your structures you only allow one core. So the pipelined version will run slower than the standard version from (inner sequence) frame 1...

What happens to your code when you run all those processing loops without putting them into timed sequences? All cores can now process all parts of the code, no manual core settings anymore?

[Mark-Ren]

GerdW,

You can test my code on what you want since you have my code.

If you donot assign to different core manually in my code, the overall running speed

will be equal to only that of one core!

By the way, have you found the VI code crashing when you rub it on a 8-core computer? 

Deqing.

------------

Hi Mark-Ren,

on the last LabViewDay in Germany a demonstration vi was shown that compared performance on 1, 2 and 4 cores (and seems to run smoothly...), you can download it here (multicore example, point 3)!

I wrote: "(Overall performance will be better on 8-core as all timed sequences can run in parallel...)" Yes, your code should run faster on an 8-core.

But I meant: In each of your structures you only allow one core. So the pipelined version will run slower than the standard version from (inner sequence) frame 1...

What happens to your code when you run all those processing loops without putting them into timed sequences? All cores can now process all parts of the code, no manual core settings anymore?
 

Best regards,
GerdW

[CarlaU]

Hi Mark-Ren,

Can you please confirm that your  8-core computer fails in all tests listed below:

8 loops on 8 cores - automatically assigned

8 loops on 8 cores - manually assigned one loop to one core

8 loops on 8 cores - automatically assigned, but manually assigned to each core in the right input node

8 loops on 1 core (core 7 (8th))

7 loops on core 0 and 1 loop on core 1

     2 loops per core 0-3 - FAILS after a longer period than the above tests

Thanks in advanced the for the information.

Carla

[Mark-Ren]

Carla,

These are what I can do according to your requirements:

(1) 8 loops on 8 cores,  automatically  assigned--- Crashed !

(2) 8 loops on 8 cores, manually assigned one loop to one core --- Crashed !

(3) 8 loops on 1 core (core 7) -- Crashed !

(4) 7 loops on core 0 and 1 loop on core 1:  no crash

(5) 2 loops per core 0-3 --- no crash.

I cannot do one of your requirement, since I cannot understand the meaning.

Thanks.

Mark_ren.

--------------------

[CarlaU]

Hi Mark-Ren,

This is a known issue that has been brought up to our R&D teams and a Correction Action Request (CAR ID 55660) has been filed.

There are two known workaround for this issue:

• To gain the majority of Timed Loops' functionality in more than 4-core systems, use a Timed Sequence structure around a While Loop.
• To use Timed Loops, disable 4 cores while executing LabVIEW code to prevent the crash. Multicore applications can run more slowly on computers with more cores because of the overhead required to manage data and threads between multiple cores.

If neither of these workarounds are possible for you, then I would suggest calling in/emailing Technical Support and reference this forum post. I hope this helps!

Carla

[Riconquistiamola]

Hello all,

It should be noted this issue is covered extensively in KB 4L6EGCQ0: LabVIEW 8.5.1 Timed Loops Crash on Computers with Four or More Cores. Please contact the Applications Engineering Department with any further questions/comments/concerns.

Cheers.