Mesh

Mesh: Compacting Memory Management for C/C++

Mesh is a drop in replacement for malloc(3) that can transparently recover from memory fragmentation without any changes to application code.

Mesh is described in detail in a paper (PDF) that appeared at PLDI 2019.

Or watch this talk by Bobby Powers at Strange Loop: Mesh runs on Linux and macOS. Windows is a work in progress.

Mesh uses bazel as a build system, but wraps it in a Makefile, and has no runtime dependencies other than libc:

$ git clone https://github.com/plasma-umass/mesh
$ cd mesh
$ make; sudo make install
# example: run git with mesh as its allocator:
$ LD_PRELOAD=libmesh.so git status

Please open an issue if you have questions (or issues)!

But will it blend?

If you run a program linked against mesh (or with Mesh LD_PRELOADed), setting the variable MALLOCSTATS=1 will instruct mesh to print a summary at exit:

…

Not all workloads experience fragmentation, so its possible that Mesh will have a small 'Meshed MB (total)' number!

Implementation Overview

Mesh is built on Heap Layers, an infrastructure for building high performance memory allocators in C++ (see the paper for details.)

The entry point of the library is libmesh.cc. This file is where malloc, free and the instantiations of the Heap used for allocating program memory lives.

DEFINITIONS

• Page: The smallest block of memory managed by the operating system, 4Kb on most architectures. Memory given to the allocator by the operating system is always in multiples of the page size, and aligned to the page size.
• Span: A contiguous run of 1 or more pages. It is often larger than the page size to account for large allocations and amortize the cost of heap metadata.
• Arena: A contiguous range of virtual address space we allocate out of. All allocations returned by malloc(3) reside within the arena.
• GlobalHeap: The global heap carves out the Arena into Spans and performs meshing.
• MiniHeap: Metadata for a Span -- at any time a live Span has a single MiniHeap owner. For small objects, MiniHeaps have a bitmap to track whether an allocation is live or freed.
• ThreadLocalHeap: A collections of MiniHeaps and a ShuffleVector so that most allocations and free(3)s can be fast and lock-free.
• ShuffleVector: A novel data structure that enables randomized allocation with bump-pointer-like speed.