Josh Baker

1.1K posts

Josh Baker banner
Josh Baker

Josh Baker

@tidwall

https://t.co/6XRpVFBykk | https://t.co/LcGR0jC0y2 | https://t.co/mIsvaahJbj | https://t.co/f38NIqdq9g

Tempe, AZ 🌵 Katılım Şubat 2010
87 Takip Edilen1.9K Takipçiler
Josh Baker
Josh Baker@tidwall·
Tile38 v1.37 released. Various updates and fixes, plus added support for regex filtering on fields and geojson properties. github.com/tidwall/tile38
English
0
2
3
159
John Solly
John Solly@_jsolly·
@tidwall Me too! Going to any of the socials on Monday?
English
1
0
1
43
Josh Baker
Josh Baker@tidwall·
Pogocache 1.2 released The big update is Pogocache now uses mimalloc as the default allocator. This allows for predictable performance across more linux platforms, such as Alpine and musl. github.com/tidwall/pogoca…
English
1
0
9
658
Josh Baker
Josh Baker@tidwall·
@akrylysov As far as I can tell their key space is still one shared hashmap and these recent gains are IO thread optimizations. But interestingly their single-threaded throughput also showed a modest gain in v8.2. ~5-10%.
English
0
0
1
95
Artem Krylysov
Artem Krylysov@akrylysov·
@tidwall The "IO threads" help with IO-bound workloads, but that's not the only type of workload people use Redis for. Does it still have only a single "main" thread responsible for state manipulation, like modifying the hashmap etc?
English
1
0
0
141
Josh Baker
Josh Baker@tidwall·
@kellabyte that's the biggest tradeoff, for every one second you lose about 50 microseconds per 1 million keys. Which I think is negligible on a multicore server. Optionally, autosweeping can be turned off with --autosweep=no.
English
0
0
3
78
Josh Baker
Josh Baker@tidwall·
@kellabyte concurrently locked by a full sweep. So the contention is super low. If you have 10 million keys, that's about 2500 keys per shard. My tests show a shard test with 2500 keys take about 550 microseconds to check for sweeping. So I guess
English
1
0
3
87
Falco Girgis
Falco Girgis@falco_girgis·
rofl a wild @tidwall appears! Dude, not only is my own C Hashset structure based upon your implementation, but I also have it wrapped to a HashSet in C++ which is then wrapped again through an adapter container to a HashMap which--despite the many layers at play--I've profiled and still has faster read/lookup times than C++'s unordered_map in STL AND Qt's QHash... REALLY freaking good perf. Use it on Dreamcast all the time. Great work!
English
1
0
5
399
NewoGame
NewoGame@NewoGame·
What c compiler optimization level should I be using?
English
3
0
2
974
Josh Baker retweetledi
Tile38
Tile38@Tile38DB·
Tile38 1.36 released. Added support for Cloudflare Queues and vector tile rendering (Mapbox vector tiles) github.com/tidwall/tile38…
English
0
3
14
565
Josh Baker retweetledi
Richard Artoul
Richard Artoul@richardartoul·
WarpStream is still hiring for engineers in the US and Canada. Come work on a small team building software that processes many hundreds of gibs/s of traffic. Storage, networking, consensus, algorithms, UI. You'll get to do it all, there are very few opportunities like this!
English
5
16
79
7.9K
Josh Baker
Josh Baker@tidwall·
I just pushed the new docs site for Pogocache. It's not much but it's a start. Took a lot longer to build than I expected. 😅 The best part is now every command has an interactive example. Webassembly to the rescue!! pogocache.com/docs/commands
Josh Baker tweet media
English
0
1
7
2.7K