Steve Blackburn

Eric, I paid for Dmitry to develop it. He completed a baseline port. My funds for that project are exhausted. I hope the community will take the port further. --Steve...

The issue with (2) is that it is not sound w.r.t. the "valid object" semantics. Some random memory may return a false positive.

Separate point: The API should include symmetric functions to set and clear the vob: `set_vob()`, `clear_vob()`. In a non-optimized allocation implementation, the binding would call `set_vob()` at allocation time. In...

A few comments about the abstractions... - The structures discussed here are _not_ visible at the language level. The abstraction is a single object. - For the above reason, we...

I don't understand the motivation for `has_buffer` and `handle_buffer`. I think that the contract should be that responsibility lies entirely with the VM to identify any buffer when it scans...

Lots of interesting thoughts above. We're doing a full re-write of [MMTk](https://github.com/JikesRVM/JikesRVM/tree/master/MMTk) in Rust. Although our implementation is not yet public (I hope to have it for you soon), you...

@Techcable I'm keen to hear of your MMTk experience. What language were you porting? Any thoughts on what you found difficult? We have a lively community with a lot of...

This should be generalized to allow N single-threaded workpacket instances of the same type. The simple case (above) has N=1, however we can make N > 1 to match the...

There's a detailed overview of the issues in [this](http://users.cecs.anu.edu.au/~steveb/pubs/papers/zero-oopsla-2011.pdf) paper.

To be clear, the Java MMTk does not support work stealing. But yes, we absolutely want support for work stealing. https://conf.researchr.org/details/ismm-2019/ismm-2019-papers/8/Scaling-Up-Parallel-GC-Work-Stealing-in-Many-Core-Environments