Appears in IEEE Micro “Top Picks in Microarchitecture ” Jan./Feb. 2007.ADDING SIMPLE RECORDER HARDWARE TO A MULTICORE CHIP CAN ENABLE DETERMINISTIC REPLAY FOR CYCLIC DEBUGGING.

Hardware vendors are currently transitioning from single-threaded microprocessors to chips that integrate multiple processor cores and threads in what is variously called multicore, chip multiprocessing, and chip multithreading. At the same time, developing and maintaining reliable software continues to challenge software vendors. Effectively debugging the software is critical. Valuable to any debugger, deterministic replay enables a developer to re-execute the (buggy) program and zero in on bugs that faithfully re-appear. Moreover, deterministic replay can also be useful for fault detection/recovery, intrusion detection, and other applications. To effectively use multicores, however, programmers must write and debug multithreaded applications. Unfortunately, neither the software or hardware environments of multicore