#µs
LightNews LightNews
reactiveashley.cat
Did I just create the fastest cut and fill algorithm?
Case Release build Dev build 400 m over hills 3.91 ms → 72 µs 24 ms → 96 µs 400 m across the valley 3.89 ms → 38 µs 23 ms → 48 µs 1 km over hills 9.80 ms → 188 µs 64 ms → 248 µs 2 km over hills 19.9 ms → 415 µs 325 ms → 530 µs Two ways and their resultant slopes.
October 3, 2026 at 11:52 AM
welltypedwit.ch
that took me forever but this was definitely worth it lol
$ hyperfine -N ./a.out Benchmark 1: ./a.out Time (mean ± σ): 356.1 µs ± 51.0 µs [User: 235.9 µs, System: 27.3 µs] Range (min … max): 264.3 µs … 2623.3 µs 9702 runs
December 4, 2025 at 5:13 PM
honk.foo
hono uses ~10% less memory with ~5% lower latency vs a few days ago, thanks to new gc and other allocation optimizations
./hono_new Latency #1 median 0.417 ms, p90 0.451 ms, p99 0.867 ms App #1 median 0.003 ms, p90 0.003 ms, p99 0.004 ms Throughput (req/s) #1 overall 115719.0, p10 109517.8, p1 82189.2 Memory usage #1 peak 3.58 MB, median 3.56 MB CPU time / req #1 median 8.58 µs, p90 8.61 µs, p99 8.62 µs Binary size #2 731.1 KB Startup #1 7.322 ms ./hono_old Latency #2 median 0.433 ms, p90 0.471 ms, p99 0.898 ms App #1 median 0.003 ms, p90 0.003 ms, p99 0.004 ms Throughput (req/s) #2 overall 111291.0, p10 105300.5, p1 75054.0 Memory usage #2 peak 3.91 MB, median 3.89 MB CPU time / req #2 median 8.90 µs, p90 9.11 µs, p99 9.23 µs Binary size #1 650.4 KB Startup #2 7.486 ms
June 5, 2026 at 1:26 PM
honk.foo
elysia uses >2x less memory with ~15% lower latency and higher throughput vs ~a week ago :)
./elysia_new Latency #1 median 0.377 ms, p90 0.525 ms, p99 0.909 ms Throughput (req/s) #1 overall 118411.4, p10 110169.6, p1 69386.1 Memory usage #1 peak 7.80 MB, median 6.91 MB CPU time / req #1 median 8.26 µs, p90 8.51 µs, p99 8.61 µs Binary size #2 6.07 MB Startup #1 7.630 ms ./elysia_old Latency #2 median 0.443 ms, p90 0.528 ms, p99 0.942 ms Throughput (req/s) #2 overall 104255.5, p10 95517.0, p1 67389.8 Memory usage #2 peak 16.50 MB, median 16.48 MB CPU time / req #2 median 9.46 µs, p90 9.60 µs, p99 9.62 µs Binary size #1 5.80 MB Startup #2 21.27 ms
June 8, 2026 at 2:17 AM
aoikurai.bsky.social
FELIZ LL NAVIDAD!!!!!!!!!🎉🎄🥳🥂🎆
#lovelive #loveliveart #µs #aqours #nijigasaki #hasunosora
December 25, 2025 at 10:16 AM
informaq-de.bsky.social
Qblox und Riverlane haben eine vollständige Fehlerkorrekturschleife in 6.886 µs bei Abstand 3 gemessen und 9.866 µs bei Abstand 7 erreicht – wodurch die für die Skalierbarkeit fehlertoleranter Quantencomputer entscheidende 10-µs-Annahme erfüllt wird.
Qblox und Riverlane demonstrieren Echtzeit-Quantenfehlerkorrektur unter dem 10-Mikrosekunden-Ziel
postquantum.com
October 6, 2026 at 3:12 AM
aki.lethalbit.net
pain
Signal::new/512 time: [4.2052 µs 4.2574 µs 4.3187 µs] change: [+12975% +13144% +13397%] (p = 0.00 < 0.05) Performance has regressed.
April 2, 2026 at 7:25 AM
jamesmunns.com
Neat :)
Screenshot of the following text (Rust Criterion benchmarks): log/postcard/serialize time: [367.94 µs 369.07 µs 370.58 µs] log/postcard-forth/serialize time: [229.93 µs 230.44 µs 231.16 µs] log/postcard/deserialize time: [1.1929 ms 1.1974 ms 1.2026 ms] log/postcard-forth/deserialize time: [1.0045 ms 1.0079 ms 1.0124 ms]
June 2, 2024 at 11:31 AM
hrbrmstr.dev
*nicely done, sir!*
Performance benchmark results in microseconds comparing different data processing methods, highlighting ‘josiah_table’ as significantly faster with a mean of 490.09 µs and median of 471.19 µs, outperforming alternatives like sapply, vapply, and purrr.
December 1, 2024 at 4:08 PM
marcdwyz.bsky.social
I'm still not sure if this is real?! Getting BLAZING speeds utilizing my Actor Pooling system.

Avg. Time '1.62756 ms' | TEST 'Unreal Spawn BP'
Avg. Time '40.45800 µs' | TEST 'Actor Pool Spawn BP'

#UnrealEngine5 #UnrealEngine #gamedev
November 30, 2024 at 1:46 AM
connorshea.bsky.social
Love to benchmark
Table titled "Results: main vs 3.0.0" comparing benchmark timings between version 3.0.0 and main across three fixtures (Radix at 2.5 KB, kitchen-sink at 716 KB, and cal.com at 1.4 MB). Columns are Benchmark, Fixture, 3.0.0, main, Change, and Speedup. read_span: Radix 1.62 µs → 173 ns (-89.5%, 9.4×); kitchen-sink 457 µs → 10.5 µs (-97.7%, 43.8×); cal.com 900 µs → 21.0 µs (-97.6%, 42.8×). render (color): Radix 11.3 µs → 1.35 µs (-88.1%, 8.4×); kitchen-sink 1.40 ms → 11.6 µs (-99.2%, 120×); cal.com 2.79 ms → 22.5 µs (-99.2%, 124×). render (mono): Radix 10.1 µs → 819 ns (-91.9%, 12.4×); kitchen-sink 1.37 ms → 11.0 µs (-99.2%, 124×); cal.com 2.90 ms → 21.9 µs (-99.2%, 132×). render_multi_label: Radix 22.9 µs → 4.06 µs (-82.0%, 5.6×); kitchen-sink 3.60 ms → 13.2 µs (-99.6%, 273×); cal.com 7.09 ms → 24.3 µs (-99.7%, 292×). Caption below the table: for read_span, throughput went from ~1.5 GiB/s to ~63 GiB/s on the large files.
July 23, 2026 at 3:47 AM
ponceto91.bsky.social
permettant d'unifier l'accès à plusieurs fournisseurs avec une API compatible OpenAI ⬇️

github.com/maximhq/bifr...
GitHub - maximhq/bifrost: Fastest LLM gateway (50x faster than LiteLLM) with adaptive load balancer, cluster mode, guardrails, 1000+ models support & <100 µs overhead at 5k RPS.
Fastest LLM gateway (50x faster than LiteLLM) with adaptive load balancer, cluster mode, guardrails, 1000+ models support & <100 µs overhead at 5k RPS. - maximhq/bifrost
github.com
January 25, 2026 at 9:02 AM
jeffgeerling.com
Taking my Pico GPS mini rack clock from µs to ns with PPS (and an LCD lens upgrade!): www.youtube.com/watch?v=w_EJ...
Taking my Pico clock from microseconds to nanoseconds
YouTube video by Level 2 Jeff
www.youtube.com
January 16, 2026 at 4:02 PM
advantage87.bsky.social
6.886 µs นั่นเป็นเวลาที่เร็วมาก
October 6, 2026 at 5:55 AM
eikopf.com
nothing about TPOT is good news claude
#1169 (preferred affinity off t 5. TPOT error-log spam - µs-lev 6. Good new - this morning's #
August 13, 2026 at 8:30 AM
welltypedwit.ch
and part 2 ^^
(this was actually pretty straightforward: instead of 3 separate arrays for rows i could just use three pointers into the same large array that i shift along.)
~/Code/aoc2025/day4 (main)$ hyperfine -N ./a.out Benchmark 1: ./a.out Time (mean ± σ): 505.6 µs ± 83.5 µs [User: 333.5 µs, System: 65.1 µs] Range (min … max): 403.5 µs … 3410.3 µs 4718 runs
December 4, 2025 at 7:11 PM
informaq.bsky.social
Qblox and Riverlane measured a complete error-correction loop in 6.886 µs at distance 3, reaching 9.866 µs at distance 7—meeting the 10 µs assumption critical for fault-tolerant quantum computing scalability.

#QuantumErrorCorrection #FaultTolerance #News
Qblox and Riverlane Demonstrate Real-Time Quantum Error Correction Below 10-Microsecond Target
iq.fp2.dev
October 6, 2026 at 3:13 AM
aoikurai.bsky.social
Hoy es el cumple de mi primera best girl de Love Live!!!! Mi dulce y preciosisisima Hanayo!!
FELIZ CUMPLEAÑOS KAYOCHIN!!!!!!
#小泉花陽生誕祭2026
#小泉花陽誕生祭2026
#loveliveart #lovelive #µs #hanayokoizumi
January 16, 2026 at 4:40 PM
danciruli.cloud
We knew that latency would have to improve by at least an order of magnitude (for things like DB calls, you want to measure your latency in 𝜇s, not ms).
February 19, 2026 at 12:23 AM
informaq-pt.bsky.social
Qblox e Riverlane mediram um ciclo de correção de erro completo em 6,886 µs na distância 3, atingindo 9,866 µs na distância 7—atendendo à suposição de 10 µs crítica para a escalabilidade da computação quântica tolerante a falhas.

#CorreçãoErrosQuânticos #TolerânciaFalhas #Notícias
Qblox e Riverlane Demonstram Correção de Erros Quânticos em Tempo Real Abaixo da Meta de 10 Microssegundos
postquantum.com
October 6, 2026 at 3:16 AM
informaq-es.bsky.social
Qblox y Riverlane midieron un bucle completo de corrección de errores en 6.886 µs a distancia 3, alcanzando 9.866 µs a distancia 7—cumpliendo la asunción de 10 µs crítica para la escalabilidad de la computación cuántica tolerante a fallos.

#CorreccionCuantica #ToleranciaAFallos #Noticias
Qblox y Riverlane Demuestran Corrección de Errores Cuánticos en Tiempo Real Por Debajo del Objetivo de 10 Microsegundos
postquantum.com
October 6, 2026 at 3:12 AM
lernerpython.com
What's the fastest way to retrieve the first 2 items from a #Python #Pandas series?

s = pd.Series([10, 20, 30], index=list('abc'))

s.loc[['a','b']] # 108 µs
s.loc['a':'b'] # 22.7 µs
s.iloc[[0, 1]] # 26.7 µs
s.iloc[:2] # 18 µs
s.head(2) # 22.3 µs
June 15, 2026 at 3:30 PM
xkeeper.net
it looks like someone has at least done the mystery gift protocol; the flipper won't show anything on the ir scope because the game is waiting for an acknowledgement before doing anything: projectpokemon.org/home/forums/...

wondering how deep this hole goes
Mystery Gift: Reverse Engineering of IR Protocol
The Mystery Gift IR Protocol HIGH means: IR LED is turned on LOW means: IR LED is turned off There are two message types: 1. Hello Message HIGH for 126 µs LOW for 537 µs HIGH for 126 µs 2. Data Messag...
projectpokemon.org
January 14, 2025 at 1:57 AM
informaq-fr.bsky.social
Qblox et Riverlane ont mesuré une boucle de correction d'erreur complète en 6.886 µs à distance 3, atteignant 9.866 µs à distance 7—satisfaisant l'hypothèse de 10 µs essentielle pour la scalabilité du calcul quantique tolérant aux pannes.

#CorrectionErreurQuantique #ToléranceAuxPannes #Actualités
Qblox et Riverlane Démontrent une Correction d'Erreur Quantique en Temps Réel en Dessous de l'Objectif de 10 Microsecondes
postquantum.com
October 6, 2026 at 3:12 AM
lauriewired.bsky.social
Shuffling memory around in host ram is trivial with standard memcpy.


What if we treated GPU memory the same way?



GDRCopy creates a user space mapping of GPU VRAM, allowing direct CPU manipulation.

With tiny writes, we can hit sub-µs timing. Here’s why that’s interesting.
April 16, 2025 at 10:36 PM