You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With #65 we now use a queue of slabs for selecting free lists. The queue is also how we pick up the bump pointer, this means we are not really benefiting from the bump ptr.
We should have a separate per allocator per size class structure for the slab, we are currently bump allocating from. This would reduce the memory usage and allow us to simplify the data in the meta slab, and would remove some conditionals on the slower paths.
We could just store the pages that have been bump allocated in the meta slab, i.e. just require 4bits.
The text was updated successfully, but these errors were encountered:
With #65 we now use a queue of slabs for selecting free lists. The queue is also how we pick up the bump pointer, this means we are not really benefiting from the bump ptr.
We should have a separate per allocator per size class structure for the slab, we are currently bump allocating from. This would reduce the memory usage and allow us to simplify the data in the meta slab, and would remove some conditionals on the slower paths.
We could just store the pages that have been bump allocated in the meta slab, i.e. just require 4bits.
The text was updated successfully, but these errors were encountered: