Change sick/dead processing to use device pointer, not gpu array. Change BFL timing to adjust only when hashing complete (not error/idle etc.).
2012-06-26T14:45:48