News for hackers. Stuff that matters.
Re:Rich Hickey: STMs vs Locks [2008] by lucian1900 on Wednesday November 30, 2011 @10:55AM (#3295400)
You put your interfering bits (critical sections) in transactions. Starting with a read and ending with a write. If you were to retry right after a read (but before a write), you would get incorrect results. And read/write pairs can get complex.

Rich is right about correctness. With an STM, it's comparatively easy to get correct behaviour from your program, which is more important than performance.

There are no comments on this comment yet.