multicore delay time

[ssara]

hi guys..

i am using the multi-core toolkit in labVIEW, as a simple example of matrix multiplication i got a speedup reach to 2.7 (2 thread vs 1 thread), but there is no more speedup when i used 3 or 4 threads, the delay times seem to be equaled. and the CPU usage is not fully 100%.

 

 I wounder if i should separate the matrix in to # of logical processors or what???

the second question is : can i reach the 8 logical processors in LabVIEW or only the 4 cores and the O.S distribute the threads?  

I am using an Intel i7-7700HQ

[tyk007]

I haven't use this toolkit, but have you read through this : http://www.ni.com/white-paper/14113/en/

[ssara]

---

@tyk007 wrote:

I haven't use this toolkit, but have you read through this : http://www.ni.com/white-paper/14113/en/

---

yes , i read this paper and i built my vi as its example 

[altenbach]

What happens if you operate on datasets with more than 3 rows?

[ssara]

---

@altenbach wrote:

What happens if you operate on datasets with more than 3 rows?

---

I have vertices of 3D points with 3 values x y z 

[altenbach]

---

@ssara wrote:
I have vertices of 3D points with 3 values x y z 

---

I see that, but that was not the question. Maybe the parallel processes operate on rows and you are limited by that. Hard to tell without doing some research. Easiest would be to generate some fake data with more rows and see if things are different. You could easily do that. 😄

[ssara]

---

@altenbach wrote:

Maybe the parallel processes operate on rows and you are limited by that. Hard to tell without doing some research. Easiest would be to generate some fake data with more rows and see if things are different. You could easily do that. 😄

---

u r right, i multiplied two dummy square matrices of size 1000, the result seem to be satisfied, but how can i add a fake rows without affect the output result?! the result values of matrix  mul is important and i dont want to change it. 

are there any other way to do that?