CUB 1.5.2

Summary

CUB 1.5.2 enhances cub::CachingDeviceAllocator and improves scan performance for SM5x (Maxwell).

Enhancements

• Improved medium-size scan performance on SM5x (Maxwell).
• Refactored cub::CachingDeviceAllocator:
  • Now spends less time locked.
  • Uses C++11's std::mutex when available.
  • Failure to allocate a block from the runtime will retry once after freeing cached allocations.
  • Now respects max-bin, fixing an issue where blocks in excess of max-bin were still being retained in the free cache.

Bug fixes:

• Fix for generic-type reduce-by-key cub::WarpScan for SM3x and newer GPUs.