Add heuristics to speed up recognition of forbidden processes

[ohl]

e.g.

• e+ e- -> gl gl gl
• flavor changing processes in flavor conserving models

see also the (closed) ticket #178.

[ohl]

added x-ref to #178.

[ohl]

There is hope: one can simply use the uncolored model for testing whether a given flavor combination is allowed. This should waste only a little time and will weed out all forbidden flavor combinations befor attempting all color combinations.

NB: this can not filter out the forbidden ghost amplitudes in multi gluon amplitudes or in e+ e- -> gl gl gl ... for allowed Hgg and Haa vertices. But these are pathological cases and the former could be handled specially.

I have a working version with disabled caches, but I still need to find a way to avoid that the two models step on each other's cache.

This can go back to v2.0

[ohl]

A working proof of principle implementation can be found in

branches/ohl/omega-development/process-filter

as of r1832. We need to implement multiple cache files though - the 2->2 CATPISS tests are too slow otherwise.

[ohl]

As of r1850 in

branches/ohl/omega-development/process-filter/

the time for

e+ e- -> u:ubar:d:dbar:gl u:ubar:d:dbar:gl u:ubar:d:dbar:gl u:ubar:d:dbar:gl u:ubar:d:dbar:gl

has gone down from 45min to 9min on my laptop.

I will merge this to the trunk.

[ohl]

Merged as of r1852.

Closing the ticket.

[Juergen Reuter]

Milestone v2.0.0final deleted