Dev Ojha
Dev Ojha
Its fairly hard to pin with certainty, but in some of our benchmarks that aren't doing much beyond creating many iterators, theres a ton of time going into mstart(), and...
Looked into this some more, it definitely appears like our benchmark delay (50x) is coming from here, and this probably isnt from unclosed iterators. In the benchmark, there are 859926...
Oh yikes lol. I thought it was already used throughout the codebase, otherwise would've just re-imported a BTree :( This can be added in a non-breaking way
agreed/understood, working on it here!
No perms to re-open this. Can this remain open though, I am very much hoping to find someone with availability to work on this, its very much needed. This problem...
I was investigating with traces, and basically ~all of our time is spent in deadtime with these chan.recv' and mutex unlocks aligning, and processes starting and stopping, between here and...
Its not state breaking. This purely an in memory representation. I don't think this is where the big wins are though, as it only impacts cacheKV store.
Yes, but the CacheKV store is not where any bottleneck is coming from right now.
I like the commit history option, though I have the same concern regarding fixing / updating changelog data. (Though I may not be the right person to opine on this,...
Good point, I misread the file proposal and thought it said PR number (not issue number). I didn't like the idea of having to figure out the PR number /...