-
Type: Improvement
-
Resolution: Won't Do
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Performance
-
StorEng - Refinement Pipeline
This ticket includes a proof of concept for a replacement for certain scratch allocations. These allocations would be initially backed by stack memory, so would be fast for allocation and free.
Part of the reason for this would be to identify potential places where WT could be sped up. For example, we know that bounded cursors would like to be more performant. They do use scratch allocation, so maybe this would help?
The second reason is that it might offer more opportunities or "change the way we program" if we knew that scratch allocation was very cheap in the typical case.
- is related to
-
WT-13621 8.20% decrease in Read count in ubuntu2004-perf-tests-arm64/ perf-test-short-key-btree/ short-key-btree.wtperf - c50a427, 2024-10-02
- Closed