Re: [PATCH 2/2] malloc: introduce ifuncs for malloc functions
Project / Subsystem
glibc / malloc
Date
2026-04-17
Proposer
DJ Delorie <[email protected]>
Source type
public_inbox
Consensus
Proposed
Sentiment
6/10
Technical tradeoffs
- • Introduces ifuncs for potential performance benefits via target-specific implementations.
- • Renaming core functions requires careful consideration of naming conventions.
- • Targets without ifunc support will alias external functions to their respective *_core symbols.
All attributes
- project
- glibc
- subsystem
- malloc
- patch_id
- —
- discussion_id
- [email protected]
- source_type
- public_inbox
- title
- Re: [PATCH 2/2] malloc: introduce ifuncs for malloc functions
- headline
- malloc: Introduce ifuncs for malloc functions
- tldr
- A patch introduces ifuncs for malloc functions, renaming core implementations with a `_core` suffix. Discussion questions the suffix choice.
- stakes
- May impact performance and ABI by introducing ifuncs for malloc functions, potentially allowing for target-specific optimizations.
- proposer
- DJ Delorie <[email protected]>
- consensus
- Proposed
- outcome
- proposed
- sentiment_score
- 6
- sentiment_rationale
- The discussion raises a valid API naming question, indicating a neutral but engaged review process.
- technical_tradeoffs
-
- • Introduces ifuncs for potential performance benefits via target-specific implementations.
- • Renaming core functions requires careful consideration of naming conventions.
- • Targets without ifunc support will alias external functions to their respective *_core symbols.
- series_id
- glibc:malloc: introduce ifuncs for malloc functions
- series_role
- reply
- series_parts
- []
- tags
-
- • malloc
- • ifunc
- • performance
- • ABI
- bugzilla_url
- —
- date
- 2026-04-17T00:00:00.000Z
Re: [PATCH 2/2] malloc: introduce ifuncs for malloc functions
This patch introduces ifuncs and resolvers for malloc-related functions. Core implementations are renamed with the _core suffix, and a generic resolver is provided. The patch includes an AArch64-specific resolver as an example. A question is raised about the choice of the _core suffix vs. the _generic suffix.