This directory contains a test to check the speed
at which a GPU thread can access a B-field map
compared to the CPU. A 70k element table is allocated
and filled to represent a B-field map. Then a set
of checksums is generated for each of the 8 elements
in the Bfield_Map_t objects. High resolution timers
are used to time how long it takes on both the 
CPU and GPU to access the map Naccesses times.
The checksums of a few of the members
are printed from both so they can be compared.

Note that this tests 1 GPU thread vs. 1 CPU thread.
The value of 1M accesses is chosen such that the
overhead of launching and sync-ing a GPU thread seems
small compared to time doing the task. Increasing
this to 10M accesses does not affect the ratio of
speeds. This ratio is about 52:1 in favor of the
CPU on qcd10i1 which has a GTX285.

This was originally written in Jan. 2011 and then
modified in Oct. 2011 to include device
synchronization calls so as to properly time the
GPU.