Change O'Megas strategy to compute large colorful processes

[Bijan Chokoufe Nejad]

Trying to compute 2 gl -> 7 gl scattering with O'Mega on a machine with 64 GB RAM and an Intel Xeon CPU E5-2440 0 @ 2.40GHz doesn't seem possible.

After more than 24 hours, O'Mega still hasn't stopped growing in memory and hasn't produced any output yet. It has grown to over 56 GB and I kill it now before the inevitable happens.

This is in contrast to e+ e- -> 12 A in QED, where O'Mega needs less than one GB and finishes in minutes.

Possible steps toward higher multiplicities in QCD are

• Avoid the construction of the color factor table completely with an option -no-color-factor-table. It takes A LOT of memory as it grows like (n-1)!^2 (note the square) in n gluon amplitudes and as such faster than the amplitude itself. For 2 gl -> 6 gl scattering only the symmetric part of this table represents 93 % of the bytecode. This information has to be computed on-the-fly as storing it becomes impossible.
• Use a different, maybe continuous color representation that allows to pass color as an external quantum number to the constructed amplitude
• Try to optimize the outweeding algorithm? Maybe avoid trying to construct everything at once and reduce the expressions at intermediate steps?
• Weyl spinors and analytical selection rules (#40) will probably also help a lot

[Juergen Reuter]

Ticket #428 offers a possible solutions to (parts of) this problem.

[Juergen Reuter]

Milestone renamed

[Juergen Reuter]

Transferred to Gitlab Issue 331.