When I open my project and afk, CPU usage is near 0%.
Sometimes when I work for a while, and then afk, CPU usage doesn't decrease to 0%, still remains at 5%.
Although it's low, my fan don't think so.
Repro steps
I don't know?
Open a project, randomly delete something and undo?
Or open a HUGE project, and open as many tabs as you can?
Related information
Actually I captured a flame graph. It says
|
let rebuildStore () = |
|
let newStore = |
|
ConcurrentDictionary<'Key, CachedEntity<'Key, 'Value>>(Environment.ProcessorCount, capacity, options.Comparer) |
|
|
|
for entity in evictionQueue do |
|
newStore.TryAdd(entity.Key, entity) |> ignore |
|
|
|
Interlocked.Exchange(&store, newStore) |> ignore |
|
|
|
let processEvictionMessage msg = |
|
match msg with |
|
| EvictionQueueMessage.Add(entity: CachedEntity<_, _>, target) when isNull entity.Node.List -> |
|
evictionQueue.AddLast(entity.Node) |
|
// store has been rebuilt while this message was in the queue. |
|
if store <> target then |
|
store.TryAdd(entity.Key, entity) |> ignore |
|
|
|
// Evict one immediately if necessary. |
|
if evictionQueue.Count > capacity then |
|
let first = nonNull evictionQueue.First |
|
evictionQueue.Remove(first) |
|
|
|
match store.TryRemove(first.Value.Key) with |
|
| true, _ -> |
|
CacheMetrics.Eviction &tags |
|
evicted.Trigger() |
|
| _ -> |
|
CacheMetrics.EvictionFail &tags |
|
evictionFailed.Trigger() |
|
deadKeysCount <- deadKeysCount + 1 |
|
|
|
if deadKeysCount > headroom / 2 then |
|
rebuildStore () |
these two functions consumes 80% of the 5% CPU usage.
I can't investigate further, sorry.