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
Oftentimes, I'm not interested in alloc costs when micro-benching a piece of code with iai-callgrind. In that case, memory footprint is quite small and dealloc isn't needed for the duration of benchmark.
So, I ended up replacing GLOBAL_ALLOCATOR with super minimal allocator. I find the results a lot stable and reproducible.
Thus, I think providing such one as opt-in from iai-callgrind could be useful as well. Note there's some caveats for it and some proper doc will be needed, of course.
The text was updated successfully, but these errors were encountered:
Oftentimes, I'm not interested in alloc costs when micro-benching a piece of code with
iai-callgrind
. In that case, memory footprint is quite small and dealloc isn't needed for the duration of benchmark.So, I ended up replacing
GLOBAL_ALLOCATOR
with super minimal allocator. I find the results a lot stable and reproducible.Thus, I think providing such one as opt-in from
iai-callgrind
could be useful as well. Note there's some caveats for it and some proper doc will be needed, of course.The text was updated successfully, but these errors were encountered: