Forums > MaxMSP

how to deal with dim>10000


nSa
February 16, 2010 | 12:35 am

is it correct that matrices can have larger dims but objects (jit.op, jit.expr) can’t handle them?
//this gives me a zero-matrix out

– Pasted Max Patch, click to expand. –


Ad.
February 16, 2010 | 2:22 pm

I’m not sure about what you want to do ! "+10" is not correct.
Replace "in" by "norm" and you’ll see that the output is different than 0.
You can us jit.op @op @val 10, if you want to add 10 to each value in the matrix.
Ad.



nSa
February 16, 2010 | 4:56 pm

take another look. On my machine both expressions(in[0]+10 and @op + @val 10) give a correct (non-zero) output with the 9999 element matrix, not with the 10000 element one. however SOMETIMES the expr version sometimes returns the correct output which makes me think there is actually a bug here…

thanks

`

– Pasted Max Patch, click to expand. –

February 16, 2010 | 5:14 pm

The size was defined in one dimension only. What happens if you specify a matrix of 1 by 20000 instead?

_
johan



nSa
February 16, 2010 | 5:43 pm

explicitly defining dim 10000 1 fixes it. what causes this? it’s not exactly a solution, however, because you can’t rely on objects like jit.op to correctly adapt to inputs of dim>10000.

jit.buffer~ for instance:

– Pasted Max Patch, click to expand. –

February 16, 2010 | 9:56 pm

Unfortunately, I cannot reproduce any of your reported problems with your patches using 5.1.3.

As always, please let us know your Max version, OS, platform, and any other relevant information.

This sounds like a bug with Jitter’s parallel processing and one dimensional objects, which I believe was fixed a while back.



nSa
February 17, 2010 | 8:03 am

The above jit.buffer~ patch has the same behavior on 2 Windows 7 machines, using both 5.1.1 and 5.1.3.

the below patch does work, however, so things are not so dire.

– Pasted Max Patch, click to expand. –

thanks


February 23, 2010 | 8:52 pm

I am able to reproduce on my Win 7 machine, and will look into a fix for the next release. Thanks for the report.


Viewing 8 posts - 1 through 8 (of 8 total)