Nikolay Sakharnykh

MAHOUT-1974 CUDA support

11

Initial PR for CUDA bindings support through JCuda

Memory leak in UCXBufferCommunicator

2

`UCXBufferCommunicator` allocates buffers for `count` and `recv_buffer` used in callbacks but never frees them. Although these buffers are small (8 bytes each), we should consider making it clean.

- Currently `UCXCommunicator` uses different communication tag design than `UCXBufferCommunicator`. We should make them in line with each other. - Currently when there's no comm buffer available in the buffer...

[FEA] Add option to report hash collisions

1

**Is your feature request related to a problem? Please describe.** Hash collisions impact performance of hash map insert and probe. It will be useful to find a way to report...

Looks like `buildAtomListGpu()` may be broken when running in L-J mode. Going into `buildAtomListGpu()`, `a_list.n` starts out with the correct value, but when it returns (after copying out the resulting...