@aryd reports that the combined module MatchProcessor suffers truncation, leading to a small loss in tracking efficiency. This is caused by the large number of Projections arriving at the MatchProcessor input for layers 3 & 4. Of the four phi regions A,B,C,D, the problem is worst in the central ones B & C.
The proposed solution is that the MP for layers 3 & 4 should be doubled up, with half processing seeds formed from layers 1 & 2 and half processing seeds formed from layers 5 & 6.
This is low priority.
@aryd reports that the combined module MatchProcessor suffers truncation, leading to a small loss in tracking efficiency. This is caused by the large number of Projections arriving at the MatchProcessor input for layers 3 & 4. Of the four phi regions A,B,C,D, the problem is worst in the central ones B & C. The proposed solution is that the MP for layers 3 & 4 should be doubled up, with half processing seeds formed from layers 1 & 2 and half processing seeds formed from layers 5 & 6. This is low priority.