#Microbenchmarks
benchfmt.rs : un runner de benchmarks Rust qui apporte à vos microbenchmarks le workflow de go test -bench et de benchstat, pour comparer vos versions en A/B.
Il suit les mêmes flags que Go, sans dépendance, et mesure les allocations heap (B/op, allocs/op) […]

[Original post on mastodon.social]
November 9, 2026 at 12:20 PM
benchfmt.rs : un runner de benchmarks Rust qui apporte à vos microbenchmarks le workflow de go test -bench et de benchstat, pour comparer vos versions en A/B.
Il suit les mêmes flags que Go, sans dépendance, et mesure les allocations heap (B/op, allocs/op). ⬇️
news.humancoders.com/t/rust…
November 9, 2026 at 12:20 PM
Benchfmt.rs: Bringing Go's benchmarking and benchstat workflow to Rust

https://www.bwplotka.dev/2026/rust-microbenchmarks-benchstat/
November 4, 2026 at 7:10 PM
harness used to matter so much. Remember the difference using GPT 5.2 in opencode, codex and others? It was visible even beyond microbenchmarks.

Now models have the entire moat and in fact everyone is trying to get OUT of the way with things like code mode
November 5, 2026 at 10:57 PM
#Linux 7.4 flushes idle workqueues 10-100x faster on big machines, but unbound queues get 15-20% slower when all are busy. Author's microbenchmarks on a 192-CPU box. The commit says where it loses.
github.com/torvalds/lin...
October 25, 2026 at 12:00 AM
Set the regression threshold per stack, not per team. Go microbenchmarks wander more than criterion medians, so the percentage that means something in Rust is noise in Go. Agree the number where merges get decided, and gate each repo against its own baseline branch. #golang #rustlang #benchmarking
October 20, 2026 at 11:00 AM
Understanding NVIDIA Grace-Blackwell Unified Memory Through Placement-Controlled Microbenchmarks

#CUDA #Performance #NVIDIA #Blackwell #Memory

hgpu.org?p=31407
Understanding NVIDIA Grace-Blackwell Unified Memory Through Placement-Controlled Microbenchmarks
NVIDIA Grace–Blackwell systems make CPU–GPU memory increasingly unified at the programming level, but unified addressability does not make the memory system uniform. A CUDA managed allo…
hgpu.org
October 18, 2026 at 9:21 PM
Nathan Graddon, Aaron Jarmusch, Sunita Chandrasekaran: Understanding NVIDIA Grace-Blackwell Unified Memory Through Placement-Controlled Microbenchmarks https://arxiv.org/abs/2610.20354 https://arxiv.org/pdf/2610.20354 https://arxiv.org/html/2610.20354
October 15, 2026 at 6:40 AM
I’ve been building Willow, a native programming language in Rust + Cranelift, mostly solo.

Latest local microbenchmarks:
• yield_switch: 2.33× faster than Go
• idle_spawn: 1.74× faster
• parked task: 765 B vs Go 2,741 B

Still experimental. Repo ↓
github.com/lechatthecat...

#GarbageCollector
October 14, 2026 at 3:19 PM
Worth a read: @mrale.ph writes about microbenchmarks in the age of AI.

Also: @gmicloud.bsky.social raised a $223M Series B. And FormEnergy.com is hiring a Manager, Site Reliability Engineering in Berkeley.

theconsensus.dev
October 7, 2026 at 12:33 PM
JDK 28 FFM allocations will be much faster: Arena.ofConfined() can reuse native-memory pools without source changes while preserving lifetime, confinement, and zero-initialization guarantees. Early microbenchmarks show up to 18x improvements!
inside.java/2026/10/05/c...
#java #FFM
Making Arena.ofConfined() Even Cheaper in JDK 28 - Inside.java
Pooled confined arenas provide a significant performance improvement for short-lived native allocations.
inside.java
October 5, 2026 at 6:30 PM
Z.ai stood up GLM-5.3-Flash serving on 100k+ Chinese accelerators with an infra agent doing much of the work. Throughput tripled in under two weeks. The reusable part is dense feedback: correctness tests, traces and microbenchmarks, not one aggregate number.

z.ai/blog/glm-bui...
Z.ai - Advanced AI Chatbot & Agent powered by GLM-5.3-Flash
Meet Z.ai, the AI assistant powered by GLM-5.3-Flash. Build websites, write code, handle long-horizon tasks, and get instant answers. Fast, smart, and reliable.
Z.ai
September 26, 2026 at 8:00 AM
@tendstofortytwo big caveat tho: "on microbenchmarks that don't take too long and aren't memory bound". Won't quite be 3x when the phone sized memory bus is the bottleneck and it's running hot and throttling lol
September 10, 2026 at 9:01 PM
🐧 **LIKWID – performance monitoring and benchmarking suite**

LIKWID is a command-line performance monitoring and benchmarking suite for CPU topology, counters, affinity, energy and microbenchmarks. The post LIKWID – performance monitoring and benchmarking su...

📰 Source: LinuxLinks
🔗 Link […]
Original post on igeek.gamer-geek-news.com
igeek.gamer-geek-news.com
September 6, 2026 at 6:34 PM
Why #microbenchmarks require statistical analysis. The screenshots are results from my repo
github.com/chrisarg/ben...
of alternative computational for the popcounts under different optimization.
(-O0, -O3, -O3 -flto). Is it the algorithm, the compiler or noise the driver? #hpc
August 24, 2026 at 6:02 PM
Why #microbenchmarks require statistical analysis. The screenshots are results from my repo
https://github.com/chrisarg/bench_popcount
of alternative computational for the popcounts under different optimization.
(-O0, -O3, -O3 -flto). Is it the algorithm, the compiler or noise the driver? #hpc
August 24, 2026 at 6:04 PM
There should be no question that benchmarking is important to know if your system is capable of delivering expected performance—and that no single benchmark can answer that for a range of apps. A set of component-specific microbenchmarks plus diverse apps can, and also makes a good regression test!
November 26, 2024 at 1:41 PM
Reminds me of the typical design patterns seen on viral microbenchmarks 😂
Linear Sort xkcd.com/3026
December 24, 2024 at 10:18 AM
"Benchmarks are not numerology. Their results are not a divine revelation. Benchmarks are experiments. Their results are meaningless without interpretation and validation."

Microbenchmarks are experiments: mrale.ph/blog/2024/11... by @mrale.ph
September 29, 2025 at 6:32 PM
For services that need low latency, benchmark candidate post-quantum key operations under realistic concurrency rather than relying on single-threaded microbenchmarks that miss tail effects.
August 17, 2026 at 11:11 AM
ATLAHS: An Application-centric Network Simulator Toolchain for AI, HPC, and Distributed Storage
Network simulators play a crucial role in evaluating the performance of large-scale systems. However, existing simulators rely heavily on synthetic microbenchmarks or narrowly focus on specific domains, limiting their ability to provide comprehensive performance insights. In this work, we introduce ATLAHS, a flexible, extensible, and open-source toolchain designed to trace real-world applications and accurately simulate their workloads. ATLAHS leverages the GOAL format to model communication and computation patterns in AI, HPC, and distributed storage applications. It supports multiple network simulation backends and handles multi-job and multi-tenant scenarios. Through extensive validation, we demonstrate that ATLAHS achieves high accuracy in simulating realistic workloads (consistently less than 5% error), while significantly outperforming AstraSim, the current state-of-the-art AI systems simulator, in terms of simulation runtime and trace size efficiency. We further illustrate ATLAHS's utility via detailed case studies, highlighting the impact of congestion control algorithms on the performance of distributed storage systems, as well as the influence of job-placement strategies on application runtimes.
arxiv.org
May 15, 2025 at 4:22 AM