Mark Thom
Mark Thom
I created the `abolish_2` branch nearly a year ago to begin scrubbing `Term` from the codebase, then rebased the `pstr_in_heap` branch on top of it.. at some point a discrepancy...
I've done as @notoria suggested and pushed a [rebis-next](https://github.com/mthom/scryer-prolog/tree/rebis-next) branch.
The problem is that `pstr_vec` needs to be expanded separately in parallel to the heap since it's a separate data structure. The act of pushing a cell to the heap...
Yes, but the root nodes only detect live partial string reasons, and then only partially. The marking algorithm works by scanning the heap one cell at a time in either...
I was actually mistaken to attribute the full heap scan of the GC algorithm to the mark phase. it occurs in the compaction/sweep phase. I'm referring to the algorithm detailed...
Yes. Good catch
@triska @haijinSk The issue seems to date to the changes of 025ec377. `render_step` is called twice in both master and rebis-dev, but in rebis-dev, `Chars` is unified to `[H,e,l,l,o,.,.,.]` the...
@Skgland The latest commit is an improvement but there's nothing preventing two threads from taking out pointers to the same offset at once. I'm still unsure of how to handle...
@Skgland Thanks. I hope I haven't disrupted your reading with my latest commit (made just now) but I think it's complete now.
I tried rewriting the commit previous to the latest one by changing the type of `offset_locks` to `Arc` using parking_lot's `ArcRwLockReadGuard` to get around the lifetime issues. Benchmarking `ConcurrentTableImpl` on...