I was playing with a rather simple GEM patch and discovered to my surprise that it doesn't seem to matter where in the chain of events one puts a geo (pls see attached).

I am trying to get a better handle on the sequence of matrix transformations here.

Any clarification would be very much appreciated.

d

my bad here's is the patch

http://www.pdpatchrepo.info/hurleur/rotating_color_square_R.pd

you have multiple connections "fanning out" from one outlet. this means the execution order is undefined (read Pd's html manual for more detailed explanation). solution for this subproblem is [trigger] aka [t] - check the help patch.

Pd's depth-first chain of processing coupled with Gem's "not particularly dataflow" way of working (it modifies OpenGL's internal state directly at each object instead of passing the actual state along the patch cords) it leads to weirdness.

However, there is a solution: use both [trigger] for ordering, and [separator] to save/restore the OpenGL state (see also [pix_separator]).

something like:

[gemhead]
|
[trigger a a]
| \
[separator] [separator]
| |
|
or even maybe just use separate [gemhead] objects, if you have a simple scene?