Add potential perf gains in the future
This commit is contained in:
parent
92a66e60dc
commit
099c795cca
1 changed files with 18 additions and 0 deletions
18
src/cache/disk.rs
vendored
18
src/cache/disk.rs
vendored
|
@ -70,6 +70,24 @@ impl DiskCache {
|
|||
db_update_channel_sender: db_tx,
|
||||
});
|
||||
|
||||
// maybe this is better? in theory this sounds like it should be faster
|
||||
// but in practice I haven't seen any perf improvements
|
||||
|
||||
// let ret = Arc::clone(&new_self);
|
||||
// std::thread::spawn(move || {
|
||||
// tokio::runtime::Builder::new_current_thread()
|
||||
// .enable_all()
|
||||
// .build()
|
||||
// .unwrap()
|
||||
// .block_on(db_listener(
|
||||
// Arc::clone(&new_self),
|
||||
// db_rx,
|
||||
// db_pool,
|
||||
// disk_max_size.get() as u64 / 20 * 19,
|
||||
// ))
|
||||
// });
|
||||
// ret
|
||||
|
||||
tokio::spawn(db_listener(
|
||||
Arc::clone(&new_self),
|
||||
db_rx,
|
||||
|
|
Loading…
Reference in a new issue