1.2.3 (2/7/2012)
* Fixed a bug broke  --max_num_blocks option
* Slightly reduced the total memory allocated to allow future small memory allocation to work 

Thanks to mtisza and Rick (rick@microway.com) for patches.


1.2.2 (8/1/2011)
* Change the "blocks" to "MB" in the printed message to avoid confusion
* In trying to malloc maximum size global memory, the size is decreased by 16 MB per step
  instead of 1 MB to avoid a (possible) bug in cudaMalloc()
* Print out version number

1.2.1 (7/22/2011)
* fixed a message print problem for memory size > 4 GB (M2070/M2090/C2070)