How to duplicate input matrix M number of times?
Need some more help from the smart jitter people...
Let's say I have an input matrix that is a single dimension of length N, and a dial control that lets the user chooser a number, M, between 0 and lets say 5. M is the number of duplicates*.
What is the best way to duplicate this matrix M times (creating discrete copies, i.e., not references to the original but independent copies) and concatenating each duplicate to a final output matrix that is of length N*(M+1) - i.e. if M is 0, then the output is just the original matrix, if M is 1, then the output is the original matrix concated with a copy... etc...
Here is a simple block diagram:
I have been struggling with this all day and not getting very far...
I was trying to find an elegant solution - but it feels like I am going to have to use some type of scripting object to add/delete the duplicate matrices and handle all of the dynamic connections. I was really hoping there would be a more elegant solution - but I am not good enough at jitter to figure that out.
Also, I can't figure out how to use jit.glue to glue together a set of matrices. I keep getting mismatch dim errors - all of my input matrices are 1 dimensional and it seems jit.glue really wants to use a 2 dimensional matrix. How do I tell jit.glue that my matrix only has 1 dimension? Uhg - pulling my hair out with this one.
Here is my playground patch.
Anyone have any suggestions?
Update: fixed connection issue with duplicate matrices
i would glue once into the largest size and then use submatrix to reference a specific dimension:
if you specify "5 1" as dims for you matrices, jit.glue accepts them.
Then you can set the maximum number of rows in jit.glue and dynamically change it using the dial:
if you specify "5 1" as dims for you matrices, jit.glue accepts them.
Then you can set the maximum number of rows in jit.glue and dynamically change it using the dial:
Beautiful! Thank you guys very much!
I did not know that specifying the dim as "5 1" is different than just specifying the dim as "5"... interesting... but your example is almost exactly what I was looking for. I can stop hating jit.glue now!
The only issue I see now (which could be a negligible one - will need to do some testing), is that say I set the number of duplicates to 0 - all of the duplicate matrices are still running through their op objects - taking up needless CPU cycles - they just aren't being concatenated by jit.glue.
In my real world patch, the size of the input matrix is actually something like 1000 / (M+1) 3d points (3 planes), i.e., the total number of points in the final matrix is always the same. So when M is 0, the original matrix is going to contain 1000 or so 3d points - and my ops are actually much more mathematically complex than just + val. But so far, CPU usage is staying under control given the nice multithreaded nature of MOPs.
I'll have to look into if this is a performance issue or not - maybe I can use routes to "mute" or not send output matrices to the jit.op objects - reducing the CPU load.
Thanks everyone!