Kunal Sareen

Improve register allocation for Virgil

7

Issue for tracking implementation of a graph-coloring based register allocation for Virgil with both spilling and splitting of live ranges. The splitting heuristic implemented should be cheap to compute and...

Introduce Bounded and Fixed nurseries

10

Introduce Bounded and Fixed nurseries This commit adds Bounded and Fixed nursery types and changes how the nursery size is set. A Bounded nursery has a lower bound of 2...

We should add a doc comment which explains what the different statistic fields are (both standard and work-packet ones) and their units.

Cleanup `PageResource` and `{FreeList,Monotone}PageResource`

1

While I was debugging heap accounting for the generational GC performance issue, I found it really hard to read and understand the `PageResource` trait and its implementations: `{FreeList,Monotone}PageResource`. It uses...

I think the SemiSpace `get_available_pages()` [1] should be dividing the heap into two like in the JikesRVM mmtk [2]. Though it seems like `get_available_pages()` is only used by the generational...

Odd results for mutator retired instructions for `biojava`

1

During performance evaluation, I've noticed that `biojava` has a very odd result for the number of retired instructions for the mutator. Every plan in the current MMTk (so `GenImmix`, `Immix`,...

Generational plans are slower than their non-generational counterparts

7

During some performance evaluation, I have noticed that our generational GC plans (`GenCopy` and `GenImmix`) are slower than their non-generational counterparts (`SemiSpace` and `Immix`). `Immix` has the best GC time...

Number of (non-precise) stress GCs is sensitive to number of mutator threads

4

It seems like for the same stress factor, if we change the number of mutator threads, the number of GCs occurring due to non-precise stress (i.e. `MMTK_PRECISE_STRESS=false`) is wildly different....

Add extreme_assertion check for edges in ProcessEdges

1

This adds an assertion which checks that edges we get in`ProcessEdgesWork` are where we expect them to be. It is guarded by `extreme_assertions`. The assertion check is simple: it takes...

We should check that the edges we get when creating a new `ProcessEdgesWork` are where we expect them to be. This check is plan-dependent since each plan has a different...