Figured out a small bug within the graybat code which does not occur in a single GPU setup,
thus there is even more need for a testing node containing at least two GPUs for testing (related to #53)
Received results are counted by nReceivedResults
The main loop is finished when all results were received nReceivedResults == samples.size()
All vertices still waiting on request, receive abort message through master.spread(abortMsg)
Figured out a small bug within the graybat code which does not occur in a single GPU setup, thus there is even more need for a testing node containing at least two GPUs for testing (related to #53)
nReceivedResults
nReceivedResults == samples.size()
master.spread(abortMsg)