Yuri S retweeted
For io_uring nerds...
uringscope: eBPF tracer and analyzer for io_uring. Reconstructs per-request latency, io-wq punts, and batching from kernel tracepoints, then names the problem and suggests a fix.
github.com/rch0wdhury/uri…
uringscope: eBPF tracer and analyzer for io_uring. Reconstructs per-request latency, io-wq punts, and batching from kernel tracepoints, then names the problem and suggests a fix.
github.com/rch0wdhury/uri…
3
30
228
9,002
Hmm wayland ecosystem in Rust (pure Linux + wayland) for UI is really good, you can get very efficient especially if most of time your UI is static. You can go very far even using pure CPU based engine.
Randomly creating my own simple bar in Rust
Randomly creating my own simple bar in Rust
3
0
0
49
I do also learn things about pipewire protocols i guess while developing this. Contrary to people believes, the protocols is quite simple although theres definitely some weird parts (8 byte padding for bools etc). Though it's mostly control plane rather than data plane itself.
0
0
1
2
damage tracking need to be very fine grain also if you want to use CPU, a bit diff to what GPU. GPU might probably ok (in term of latency not resource usage) for drawing more than CPU. GPU you can batch things up and impact less until your vertex counts is hitting some threshold
1
0
0
3
but then again I doubt for simple bar, the number of vertices are that big :)), I simple use CPU just because I knew that my simple bar are mostly idle and I just want simple shell that doesn't do animations and such (and can do precompute things ahead of time).
0
0
0
5
The layout is static though (I'm not using taffy for layout engine :') so there are some manual calculation happening, it's being helped by AI for this one for full layout calculation :)) )
But I guess even if I use taffy the total cpu usage won't be much different
But I guess even if I use taffy the total cpu usage won't be much different
0
0
0
`All tech debt: gone. I solved it.`
I'm not sure about this one, when one find enough or integrated well enough, there always going to be a debt in Product, Business, Operations and Technical... And most of time, it will be converted into Tech Debt ...
I'm not sure about this one, when one find enough or integrated well enough, there always going to be a debt in Product, Business, Operations and Technical... And most of time, it will be converted into Tech Debt ...
Oh boy, many such cases across every dev team.
0
0
0
29
chawan.net i think this app for tui browser is amazing not sure its being rarely talks about hmm. For quick browsing if you're so tired of fancy browser interface you might want to take a look. Not sure about safetyness and sandboxing hmm..
0
0
0
177
Seriously is there any thing that we can NOTIFY/NUDGE to our govt related to REAL LIFE? A lot of folks got an ISPA ...
1
0
0
46
wayland-rs is definitely one of the good crates that interface things Linux in good way, they did it right for modelling callback behavior into Dispatch, compare to pipewire-native-rs, it enable folks to integrate it on whatever event loop they have ..
0
0
0
36
Hmm, i think somebody need to package arkts pipeline to something like deno or bun like cli. Most of typescript ecosystem already in strict mode most of the time ... the gain will be quite insurmountable ...
0
0
1
36
Yuri S retweeted
New on the blog: part 1 of 3 on how Apache Iggy became a cluster. Why Viewstamped Replication Revisited over Raft, one VSR group per partition, how we swapped the whole server, and what it costs, with benchmarks.
iggy.apache.org/blogs/2026/09/…
#iggy #asf #vsr #cluster #rust #streaming
iggy.apache.org/blogs/2026/09/…
#iggy #asf #vsr #cluster #rust #streaming
0
8
17
534
I swear to god, dcn 3.5 last kernel codes was so incomplete even compare to dcn version before and after. Seriously, we do have a lot of device on dcn 3.5 right? This should be right since I think steam dev* use this amd zen 5 variant?
3
0
0
52