Yi Lin

Results 120 issues of Yi Lin

For testing, we include a `DummyVM` in the mmtk-core repo. We should rename it to `MockVM` to properly reflect its usage. And we should allow proper mocking for testing.

C-enhancement
A-test
F-project

C-missing
A-utils

`ImmixAllocator` uses the same fast path allocation as `BumpAllocator`. Also a VM's compiler can generate immix fastpath in the same way as bump pointer allocator (except that the field offsets...

F-question
C-enhancement
A-allocator

We need a general solution to allow extra header words for allocators. There are different scenarios we need this: * We may globally add some extra header words for different...

C-enhancement
A-allocator

Some GC metadata (such as work packets, remember set) are allocated in Rust, and those are not counted in the heap size. This is incorrect. We need to know how...

C-bug
A-heap
P-normal

Currently we allocate and initialize `SFTMap` during `MMTK::new()`. This takes roughly 100ms and can make the VM start up time noticeably slower. Ideally we could directly allocate and initialize a...

C-bug
A-heap

This PR closes https://github.com/mmtk/mmtk-core/issues/520. This PR 1. adds a proper `max_non_los_default_alloc_bytes` for mark compact 2. adds a `prepare()` call for mark compact before the compact transitive closure 3. adds some...

PR-testing

MMTk provides API `flush_mutator()` and `destroy_mutator()`. We would also need to have flush/destroy for each allocator, and call to each allocator for `flush/destroy_mutator()`.

C-missing
A-allocator

Each policy provides their `trace_object()` method and that should be used for a GC transitive closure. However, if a plan needs to do multiple traces, do we expect `trace_object()` be...

C-bug
A-policy
P-normal