- Interrupt the thread before waiting on join
- Provide a method for looping threads to exit on demand
- CHECK if start and stop succeed instead of returning an error

- Makes sure each solver accesses a different subset of the data
- Sequential reading of DB for performance
- Prefetches a configurable amount of data to host memory
- Distributes data to solvers in round-robin way for determinism

- Split batches between GPUs, and tree-reduce the gradients
- Detects machine topology (twin-GPU boards, P2P connectivity)
- Inserts a callbak in the solver for minimal code change
- Added gpus flags to caffe.cpp, e.g. '-gpus 0,1' or '-gpus all'
- Deterministic architecture for reproducible runs

Follow expectation of `iter_size` in compounding the effective batch
size by the number of devices.