[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAELLoAhC3emyLF26DNNb0m-D+eUJ2z9=G6r1Px+oYwOztzvZ=g@mail.gmail.com>
Date: Tue, 20 May 2025 00:57:21 +0530
From: Vikram R <vikram@...lexai.com>
To: linux-kernel@...r.kernel.org
Cc: linux-newbie@...r.kernel.org, kernelnewbies@...nelnewbies.org
Subject: Re: [ANNOUNCE] Call for Contributors: ASIOS – AI‑Native OS (Kernel Work)
ASIOS – Envisioned to Host and Sustain Cognition & Synthetic Intelligence
The world’s first AI-native operating system.
ASIOS is a new class of OS built on Ubuntu 24.04 LTS, designed from
first principles for AI workloads. We’re creating the foundational
infrastructure layer to support ever-more sophisticated
intelligence—from today’s deep-learning models to tomorrow’s advanced
cognitive and quantum-ready AI.
Building on our May 3 thread, we’ve landed a resilient config-overlay
and build script for ASIOS on both x86_64 and aarch64. Key artifacts:
asios-config-overlay.sh: idempotent Kconfig overlay that tolerates
missing symbols
build-asios-kernel.sh: automated clone to config to compile workflow
test-asios-config.sh: verifies .config flags post build
Initial perf on Ubuntu 24.04 HWE 6.11:
x86_64: mbw=7.3 GB/s, hackbench=53 ms, fio=1.6 GB/s
arm64: mbw=27.4 GB/s, hackbench=4.8 ms, fio=3.2 GB/s
GitHub (scripts and tests): https://github.com/asi-os/asios-core/
Discord: https://discord.gg/rWuU7cWU4E
Areas for review and test:
1. Overlay logic: skipping versus failing on absent Kconfig symbols
2. Build-matrix logic: host-only versus cross-build
3. Test completeness: missing flags
4. Scripting style and robustness
Any pointers or suggestions are highly appreciated before we push to a
formal RFC series.
—
Vikram R
Founder & CEO, KarLex AI, Inc.
https://asios.ai
On Sat, May 3, 2025 at 12:22 PM Vikram R <vikram@...lexai.com> wrote:
>
> ASIOS is a new AI‑native OS (Ubuntu 24.04 LTS base) bringing
> deterministic scheduling, NUMA‑GPU optimizations, zero‑copy I/O, and
> eBPF telemetry into the Linux kernel.
>
> Key directions—
> — Deterministic CPU scheduling for reproducible AI runs
> — NUMA‑aware memory placement tuned for GPU DMA
> — Zero‑copy GPU I/O via GPUDirect RDMA/Storage
> — eBPF‑based telemetry hooks
>
> **Call for contributors:** scheduler/MM, GPU/accelerator integration,
> eBPF instrumentation.
>
> Project links—
> GitHub: <https://github.com/asi-os>
> Discord: <https://discord.gg/rWuU7cWU4E>
> Roadmap: <https://github.com/asi-os/asios-docs/blob/main/docs/ARCHITECTURE.md>
>
> Please reply on‑list with questions or join us on Discord to dive into
> the design.
>
> Vikram Karlex R
> KarLex AI, Inc. | <https://asios.ai> | <https://karlex.ai>
Powered by blists - more mailing lists