Take the following survey on API usage to help JDK planning. (Yes, we're not supposed to use it directly. Caveat Utor. But many folks do, including myself.)
Yes, I do, to implement a better serialization framework. I also think of relying on it to access some off heap memory
Hi Thomas, make sure to take the survey to make your position known. Personally (and I'm not speaking for Oracle in this context), I'd like to see unsafe remain accessible in its current form so that sophisticated users can continue to make good use of it.
Yes, I do use it for performance when serializing objects. What I understand from my conversations with fellow developers is that if anybody actually uses Unsafe it is for serialization.
Yes I've used it before when working with OpenGL native interop to manage a large number of floats
FWIW, on the JRuby project we have been using a "mock" Unsafe to allow compiling in newer features. We then choose whether to use those features based on JVM version at runtime, rather than anchoring our codebase to the lowest common denominator.
I hope there is a chance to see the surveys results afterwards. It would be awesomely interesting to see what and how others are using sun.misc.Unsafe.
I agree - it'd be interesting to see the span and relative frequency of the use cases. I'll try to prod the owner of the survey to make the results known.
Malthusian Locks appears in EuroSys 2017. An extended version is in
arxiv.Abstract:Applications running in modern multithreaded...
Pointers: Experiences with HTM-Based Reference Counting in C++by Maria Carpen-Amarie , Dave Dice, Gaël Thomas and Pascal...
The following is inspired by some discussions with John Rose on how we might
implement atomic access to value type "tuples" for subsequent...