TSX branch/locking tests

Hi…

Does anyone have any OpenMP tests which stress branching/locks? I’m porting the runtime to another platform and curious about both

  1. If the locking we’re doing on the other platform is correct
  2. The performance difference compared to Intel

If anyone (Intel/IBM/friends) has any specific tests floating around they can share that would be great.

Thanks

Not that I have any tests that I could give you :-(, but it might help to clarify for others…

Are you looking for tests for locking at the OpenMP level (omp_{set,unset,test}_lock etc), or are you re-implementing some of the locking used inside the runtime itself?

Since you mention TSX, I get the feeling you mean that you’re changing the internal locks, but it’s hard to tell from the question.

p.s. If you are interested in enabling speculative locking at the OpenMP level, you should look at our IWOMP paper, and the implementation of those interfaces in the recently published openmprtl.org code that will be coming here soon…

It is our intention to propose that interface to the language committee for standardization, so if it meets your users’ needs it would be good to know that (and it could save you having to invent something else!)

– Jim

James Cownie james.h.cownie@intel.com
SSG/DPD/TCAR (Technical Computing, Analyzers and Runtimes)

Tel: +44 117 9071438