projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d9a2b83) | patch
[scudo] Support partial concurrent page release in SizeClassAllocator64
authorChia-hung Duan <chiahungduan@google.com>
Mon, 10 Jul 2023 17:24:35 +0000 (17:24 +0000)
committerChia-hung Duan <chiahungduan@google.com>
Mon, 10 Jul 2023 17:24:37 +0000 (17:24 +0000)
commit57ae8a2a1acb1aa1a5f55c29b1b338a780d649d5
tree96d2b3986a7fc76c9f517c9a71293aa660e75963tree | snapshot
parentd9a2b83dcd93cc5f8c34f6f2a0a75a0f3ca1545ccommit | diff
[scudo] Support partial concurrent page release in SizeClassAllocator64

After extracting memory groups, it's safe to do
1. markFreeBlocks
2. releaseFreeMemoryToOS concurrently with pushBlocks() and
popBatches(). This will improve the throughput of Scudo.

Reviewed By: cferris

Differential Revision: https://reviews.llvm.org/D153608
compiler-rt/lib/scudo/standalone/primary64.h diff | blob | history
Domain: System / Toolchain;