kiper220/mimalloc
1
0
Fork 0
mirror of https://github.com/microsoft/mimalloc.git synced 2025-05-06 23:39:31 +03:00
Code Issues Projects Releases Packages Wiki Activity Actions

add comment on wait strategy in region allocation

Browse source
This commit is contained in:
daan 2019-08-11 08:18:51 -07:00
parent f5902a7a82
commit 6435887657
1 changed files with 7 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
src/memory.c
View file

@ -139,7 +139,13 @@ static bool mi_region_commit_blocks(mem_region_t* region, size_t idx, size_t bit
start = ALLOCATING; // try to start allocating start = ALLOCATING; // try to start allocating
} }
else if (start == ALLOCATING) { else if (start == ALLOCATING) {
mi_atomic_yield(); // another thead is already allocating.. wait it out // another thead is already allocating.. wait it out
// note: the wait here is not great (but should not happen often). Another
// strategy might be to just allocate another region in parallel. This tends
// to be bad for benchmarks though as these often start many threads at the
// same time leading to the allocation of too many regions. (Still, this might
// be the most performant and it's ok on 64-bit virtual memory with over-commit.)
mi_atomic_yield();
continue; continue;
} }
} while( start == ALLOCATING && !mi_atomic_compare_exchange_ptr(&region->start, ALLOCATING, NULL) ); } while( start == ALLOCATING && !mi_atomic_compare_exchange_ptr(&region->start, ALLOCATING, NULL) );