LabVIEW
Why are shift registers are so slow
sima wrote:
> Hi,<br><br>I have a very similar problem. I have 3 arrays connected
> to shift registers, and plotted in 2 XY graphs (2 data arrays and 1
> time array). Initially, I was using a Build Array block to generate
> the arrays, but noticed that as time (hours) went by and the arrays
> got bigger, each loop iteration was taking longer and longer. I need
> the loop to run at a particular rate, but it was slowing down over
> time.
>
> I followed the advice of this post (determine the size of the array
> beforehand, initialize it, and then use Replace Array Element instead),
> but I had the same problem. My 1-D arrays can grow up to 200,000 rows,
> and I have 3 of them. Perhaps graphing all this data in an XY graph is
> the problem? Is there another way around it?
Shift register add very little overhead to your VI, in fact except in a
few rare cases in earlier versions of LabVIEW it is almost always the
fastest way if you need handling large data repetitevely.
As you already suspected the plotting of such large arrays is the real
problem. What can you do?
Do you really need to plot all that data? Maybe a shorter history would
be enough? If that is not an option then put another loop inside that
loop (with also a shift register) which acquires several data elements
(10 or more depending on your input speed) repetitively before leaving
that loop and plotting the new resulting array on the front panel. This
will reduce overhead greatly.
Rolf Kalbermatter
> Hi,<br><br>I have a very similar problem. I have 3 arrays connected
> to shift registers, and plotted in 2 XY graphs (2 data arrays and 1
> time array). Initially, I was using a Build Array block to generate
> the arrays, but noticed that as time (hours) went by and the arrays
> got bigger, each loop iteration was taking longer and longer. I need
> the loop to run at a particular rate, but it was slowing down over
> time.
>
> I followed the advice of this post (determine the size of the array
> beforehand, initialize it, and then use Replace Array Element instead),
> but I had the same problem. My 1-D arrays can grow up to 200,000 rows,
> and I have 3 of them. Perhaps graphing all this data in an XY graph is
> the problem? Is there another way around it?
Shift register add very little overhead to your VI, in fact except in a
few rare cases in earlier versions of LabVIEW it is almost always the
fastest way if you need handling large data repetitevely.
As you already suspected the plotting of such large arrays is the real
problem. What can you do?
Do you really need to plot all that data? Maybe a shorter history would
be enough? If that is not an option then put another loop inside that
loop (with also a shift register) which acquires several data elements
(10 or more depending on your input speed) repetitively before leaving
that loop and plotting the new resulting array on the front panel. This
will reduce overhead greatly.
Rolf Kalbermatter
@antonella86 wrote:
Hi, I have the same issue. I need to acquire a 108000x8 data matrix. I initialize an array that size, enter a while loop with a shift register and replace one row at each iteration with the data that I am gathering online.
I need an update rate of 1000 Hz so loop with a wait 1 ms)
Everything works fine with smaller array (i.e. 20000), but loop slows down when I request a larger file.
Any idea?
Yes, we need to see the VI. You talk about "online" and "request file", so what does each loop iteration actually do? Is the data coming from a network location? How exactly is it read?
Also be careful with terminology. The term "matrix" is a special datatype. I assume you have a plain 2D array.
