4 results found
This article explores how a 10-year-old Intel Xeon E5-2620 v4 server with 128 GB DDR3 RAM and no GPU can run a modern LLM like Gemma 4 26B-A4B at reading speed. It highlights that LLM inference is often memory-bound and showcases deep optimization techniques using `ik_llama.cpp`, including speculative decoding, CPU-aware MoE routing, advanced memory management, and specialized attention kernels. The success demonstrates that granular software control can unlock significant performance on older, abundant-RAM hardware.
Rust in the Linux Kernel: A Beacon Against AI-Driven Bugs Quick Verdict Greg Kroah-Hartman's assertion that "Rust will save Linux from AI" is a compelling argument rooted in practical security needs. While not a magic
An AWS engineer reported a dramatic 50% performance drop for PostgreSQL on the upcoming Linux 7.0 kernel, caused by changes to kernel preemption modes. While a revert was proposed, kernel developers suggest PostgreSQL should adapt using Restartable Sequences (RSEQ). This could mean significant performance issues for databases on Linux 7.0 until PostgreSQL is updated.
This article details the journey of debugging mysterious system freezes caused by eBPF programs in the Linux kernel. We uncovered an issue where an NMI-driven eBPF sampling program would self-deadlock by attempting to acquire a spinlock already held by another eBPF program on the same CPU, leading to 250ms kernel timeouts. The analysis highlights the complexities of spinlocks, NMIs, and cache coherence in kernel development.
