No current major DCC app out there that I know has a thread safe SDK for concurrency, maybe for some specific parts, but not for the actual application and data objects outside such domains, not even the bare fundamentals.
They were all conceived in the late 90s and revisited somewhat in the early 2k (if you're lucky). Even just reading a lot of the structures can get you in trouble because of stupid iterator magic or non contiguous containers being compacted making it possible to race practically anything in there. Concurrent writes and STL friendly containers are completely alien to most architectures, and modern (2005-8 modern) things like a live thread pool with an exposed lock and extensions to containers for concurrency (concurrent vect etc.) you can completely forget about. You stand a somewhat remote chance to have a go at some of it with Maya if you use Intel's compiler and TBB (VC will not keep the thread pool live, you pay through the nose all the time), but it's more of an access venue, forget about any of the SDK playing nice with it. Occasionally even just having non concurrent multithreading will become super dodgy as more black magic goes on under the hood and you find memory between the more expensive passes got shuffled somehow, unexpectedly, and without any apparent reason. Some times they will work, but you are practically always rolling a dice and it might run fine for months out of coincidence until baffling bugs crop up because you did something a mile away that interferes tangentially with said black magic. It's a sad state of affairs in general. If you want some less than trivial threading going you'd better start right away from bubbling it up entirely with flat out copying at the ends. Designing accordingly is a pain in the arse. /rant On Thu, Sep 26, 2013 at 3:38 AM, Matt Lind <[email protected]> wrote: > For the record, the Softimage SDK is not thread safe. Only the ICE SDK > can support multithreading.**** > > ** ** > > ** ** > > Matt**** > > >
-------------------------- To unsubscribe: mail [email protected] with subject "unsubscribe" and reply to the confirmation email.
