Benchmarks of speedup of using CUDA UVA-access with permanent scan.

Source code requires: CUDA >= 4.0 and (C++11 or boost 1.53)

Recomended: CUDA 5.5 and C++11 in MSVS 2012 with Nsight VS 3.1

Comparison latencies for memory access of small blocks 64 bytes - 256 KBytes:

• standard MPP-approach with using DMA-controller and launching kernel-function for each block of data (good for block_size >= 1 MB)
• permanent scan of CPU-pinned-memory for the new data by using Unified Virtual Addressing and launching kernel-function once at all time (good for block_size < 1 MB)

Result of benchmarks on GeForce GTX460 SE CC2.1 see in: result.txt