michelez

Several times now I have encountered the situation where I have scaling or position values stored into a single cell matrix that I then send to jit.gen object's second or third inlet. Assuming that the first inlet's matrix is of a much bigger size and knowing that jit.gen automatically adapts matrices coming into the secondary inputs to the size of the first, I am wondering if it's more efficient to transform those values into a max list via jit.iter and send it into jit.gen with a prepended param message.

comment

prepend

jit.iter

jit.gl.mesh

jit.matrix

jit.gen

param

sphere

norm

jit.mo.sin

jit.mo.join

message

jit.gl.handle

toggle

jit.world

Several times now I have encountered the situation where I have scaling or position values stored into a single cell matrix that I then send to jit.gen object's second or third inlet. Assuming that the first inlet's matrix is of a much bigger size and knowing that  jit.gen automatically adapts matrices coming into the secondary inputs to the size of the first, I am wondering if it's more efficient to transform those values into a max list via jit.iter and send it into jit.gen with a prepended param message.
I have attached an example with jit.mo

Several times now I have encountered the situation where I have scaling or position values stored into a single cell matrix that I then send

jit·gen-efficiency-question

I just made a test by making poly~ patches with the objects after the blank 50x50 matrix, 300 instances each including individual instances of the jit.mo stuff.

the left branch of your patch gets around 8-10 fps on my graphics machine.

Probably you could get more efficiency with jit.gl.multiple?

I'd be really interested to know exactly how the bridge between cpu and gpu is handled with gen. kind of amazed by all the cross platform support.

I just made a test by making poly~ patches with the objects after the blank 50x50 matrix, 300 instances each including individual instances of the jit.mo stuff. 

Probably you could get more efficiency with jit.gl.multiple? 

thank you for taking the time to test this! It seems like the difference is not substantial.

Cool. Also, I didn't test @adapt 0 on the gen, didn't think of it at the time.

Not sure, but I think that argument allows swiz to unpack just a 3 plane 1x1 matrix for your coordinates instead of up scaling to match the rest of the shader buffers..

Cool.   Also, I didn't test @adapt 0 on the gen, didn't think of it at the time.

That's a good observation, I wonder if that is the case. If true that would make using the matrix instead of the parameter the best choice for sure!

jit.gen efficiency question