Tag
一篇70分钟阅读量的深度技术长文,从第一性原理推导投机解码(speculative decoding),构建涵盖EAGLE-3、多token预测(MTP)以及块扩散drafters(DFlash、DSpark)的统一框架,并讨论drafter训练中最小化KL散度的问题。
An in-depth look at the evolution of the Hetzner Cloud network stack, tracing how its networking architecture has been built and changed over time.
The article provides a hands-on exploration of BSD disklabels on NetBSD 11, detailing their format, documentation sources, and structure through hex editing and C programming examples.
This article details the process of GPU memory writes, tracing the STG.E instruction through various hardware components like the load/store unit, coalescer, and L1 cache on an RTX 4090.
Valkey uses different encodings like listpack and hashtable for its data types to balance memory and performance; understanding these internal structures can lead to significant memory savings.
The article explains why making the entire stack out of guard pages is problematic, as it could lead to unbounded memory allocation and system hangs. Instead, fixed guard pages are preferred for bounded fault handling.
A detailed technical article explaining the Linux kernel scheduler, including task_struct, scheduling classes, context switching, and the EEVDF algorithm.
This article explains how rqlite, a distributed SQLite database, takes control of SQLite's Write-Ahead Log to enable efficient snapshotting for Raft consensus, avoiding full database copies by using the WAL as incremental state.