v5.3.3

CUDA v5.3.3

Diff since v5.3.2

Merged pull requests:

• Rework context handling (#2346) (@maleadt)
• fix kernel launch logic (#2353) (@xaellison)

Closed issues:

• Excessive allocations when running on multiple threads (#1429)
• Fix and test multigpu support (#2218)
• Bitonic sort exceeds launch resources (#2331)