Test of fine-grained all-to-all communication, comparing aggregation using a 2D
TRAM grid topology to direct sending. Sends are done concurrently across PEs,
with the order of destinations randomized. Sends of more than 32 B are performed
in rounds at each sender, where in each round a single 32 B unit is sent to each
destination PE. When using TRAM, individual sends are aggregated into larger
buffers before being sent.

Usage:

ataTest dataSizeMin(default = 32) dataSizeMax(default = 16384) aggregationBufferSizeInBytes(default = 16384)
