Martin Maas 1dca72cc48 Implement the lifetime-based allocator. ...

The lifetime-based allocator is intended for handling medium and large allocations. It predicts whether or not the object is short-lived. If so, it puts the object into a separate HugeRegion that is expected to free up quickly. Otherwise, the allocation is ignored, to be handled by the usual allocation path.

The lifetime-based allocator supports a counterfactual mode where objects are not actually stored in a different region but where the allocator tracks the prediction accuracy and memory consumption that such an approach would have.

PiperOrigin-RevId: 412992130
Change-Id: I62e742a9e7b741dfb214503b31a66a01a7e183e7

2021-11-29 15:42:01 -08:00

..

legacy_pageheap.png

Initial commit

2020-02-07 17:55:59 -05:00

lifetimes-counterfactual.png

Implement the lifetime-based allocator.

2021-11-29 15:42:01 -08:00

lifetimes-enabled.png

Implement the lifetime-based allocator.

2021-11-29 15:42:01 -08:00

pagemap.png

Initial commit

2020-02-07 17:55:59 -05:00

per-cpu-cache-internals.png

Initial commit

2020-02-07 17:55:59 -05:00

per-thread-structure.png

Initial commit

2020-02-07 17:55:59 -05:00

spanmap.gif

Initial commit

2020-02-07 17:55:59 -05:00

tcmalloc_internals.png

Initial commit

2020-02-07 17:55:59 -05:00