What makes parallel computing one of the more difficult disciplines for programmers to master?
-
Race conditions can lead to data being corrupted in unexpected ways
-
Inadequate memory management can lead to application deadlocks
-
Too many threads or processes can actually decrease performance
-
All of the above
Parallel computing introduces race conditions (data corruption from concurrent access), memory management deadlocks, and performance degradation from too much threading overhead. All these challenges make parallel programming difficult.
Parallel programming is hard for several compounding reasons at once: race conditions can silently corrupt shared data when threads access it without proper synchronization, poor memory management can lead to deadlocks where threads wait on each other forever, and even when things work correctly, spawning too many threads/processes can hurt performance through context-switching and contention overhead. Since all three of these are genuine, distinct challenges, "all of the above" is the most complete answer.