ProgrammingMultiCoreSystems.pptVIP

* * * Stupid saying: Right now, threaded software is like a car. You can make it go fast, but then it will probably crash. We want to fix this problem. * * Try to motivate each one of these with a question of why you need to do it. * * * * * * * * * * This slide is also hard to understand. The boxed material especially. I need to mention alias analysis here. Also, change p-m to p-lock or something to avoid confusion. * I also need to mention alias analysis here. * * Say that OCC is a *class* of techniques * I should explain the situations in which STM is worse. Just this benchmark? * Probably should mention soundness here (say it’s in the paper) XFI, OSDI 06 * * Finding a Feasible Lock Order foo bar qux baz functions Constraints m1 p-m’ p-m p-m’ m1 r-m m3 p-m q-m p-m m1 r-m q-m p-m p-m’ Global Lock Order search for infeasible patterns topological sort * Algorithm Summary atomic { … } C program with atomic sections begin acq L1, L2; end lock order cyclic? yes no fail and report potential deadlock C program with acquire statements generate global lock order insert ordered lock acquisitions lock requirements remove redundant acquisitions begin acq L2; end match locks to data * Outline Introduction semantics usage model and benefits Autolocker algorithm computing protections lock ordering acquisition placement Related work Experimental evaluation Conclusion * Comparison Transactional memory with optimistic CC Threads work locally and commit when done If two threads conflict, one rolls back restarts Benefit: no complex static analysis Drawbacks: software versions: lots of copying, can be slow hardware versions: need new hardware both: some operations cannot be rolled back (e.g., fire missile) How does this compare with Autolocker? * Experimental Questions Question 1: What is the performance cost of using Autolocker? How does it compare to other concurrency models? * Concurrent Hash Table Simple microbenchmark Goal: Same performance as hand-optimized low over