[PD] reading and handling of multichannel data request

[Mathieu Bouchard]

On Tue, 18 Apr 2006, IOhannes m zmoelnig wrote:

> > Btw, in [matrix~], where's the inlet for modifying the matrix's values? is 
> > it leftmost or rightmost? what is it with [mtx_*~] ?
> in order to be consistent:
> [matrix~]'s "matrix" inlet is on the right-side (it is not rightmost,
> since that is the fade-time, but it is 2nd rightmost): the object does I^*A
> [mtx_*~]'s "matrix" inlet is the left-most inlet: the object does A*I^

ah ok. In GridFlow, [#inner] also uses left to mean left and right to mean
right as in the ordinary matrix product. However I wouldn't be able to
make a swapped version mean anything because I'm often not multiplying
vector*matrix, so e.g. replacing the current image*matrix product by a
transposed matrix*image product would require transposing each image twice
(instead of zero times) because the vectors are pixels and the channels of
the pixels are the last dimension. ([#inner] has to be consistent with the
standard definition of tensorial contraction)

> > When there's no standard it's normal to be non-standard.
> but half of my institute nudged me because of [matrix~]'s behaviour, so
> we finally decided to change it in order to satisfy them....

aaah... peer-to-peer pressure... how many kilopascals was it like?

 _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801 - http://artengine.ca/matju
| Freelance Digital Arts Engineer, Montréal QC Canada