LWN reports that Linus Torvalds has posted his first large-language-model-assisted kernel patch series, after telling the Maintainers Summit that he had used an LLM to work on a longstanding problem. The ten-patch set speeds up folio lookup by noticing when folios are hot, delaying their release until after an RCU period and then reading the folio data itself inside the RCU-protected region without taking a refcount on the fast path where the user copy cannot fault. The result is an order-of-magnitude speedup on a read()-heavy benchmark, but review has already identified a race condition, so further work is needed. The submission split commenters: one accused Torvalds of ‘setting a bad example’ by using AI while pushing changes that make contributing harder, arguing newcomers will copy him; others replied that the criticism misrepresented his carefully argued summit position and that an experienced maintainer’s use of a tool says nothing about how newcomers would use it.
Linus Torvalds's first LLM-assisted patch series
LWN reports that Linus Torvalds has posted his first large-language-model-assisted kernel patch series, after telling the Maintainers Summit that he had used an LLM to work on a longstanding problem. The ten-patch set speeds up folio lookup by noticing when folios are hot, delaying their release until after an RCU period and then reading the folio data itself inside the RCU-protected region without taking a refcount on the fast path where the user copy cannot fault. The result is an order-of-magnitude speedup on a read()-heavy benchmark, but review has already identified a race condition, so further work is needed. The submission split commenters: one accused Torvalds of 'setting a bad example' by using AI while pushing changes that make contributing harder, arguing newcomers will copy him; others replied that the criticism misrepresented his carefully argued summit position and that an experienced maintainer's use of a tool says nothing about how newcomers would use it.
Source: Lwn