mirror of
https://github.com/facebook/sapling.git
synced 2024-10-07 23:38:50 +03:00
df93ba8548
Summary: Eden performance can degrade when something (a status operation in particular) queues a ton of work on the hg importer queue. In that case, the 12 RocksDB threads are all blocked waiting to insert more work into the importer queue. Thus, even RocksDB queries that would hit cache cannot make progress until the importer queue clears up. In the long term, we need a more comprehensive approach to managing our concurrency and boundedness of scheduled work. In the short term, though, just make the importer queue unbounded. Reviewed By: bolinfest, strager Differential Revision: D10368779 fbshipit-source-id: 9200002338bea091fb353b2a96ac543df98701c2 |
||
---|---|---|
.. | ||
config | ||
fuse | ||
inodes | ||
journal | ||
model | ||
rocksdb | ||
service | ||
sqlite | ||
store | ||
takeover | ||
testharness | ||
utils | ||
CMakeLists.txt |