Re: GSoC: possible IPA-ICF project direction

Project / Subsystem

gcc / gsoc

Date

2026-09-14

Proposer

Rohith Kapelli <kapellirohith@gmail.com>

Source type

public_inbox

Consensus

Proposed

Sentiment

7/10

Technical tradeoffs

  • • Potential for code size reduction through improved ICF.
  • • Risk of incorrect code generation if separate bodies or inlining are handled improperly.
  • • Complexity in managing function visibility and linkage across compilation units.

All attributes

project
gcc
subsystem
gsoc
patch_id
—
discussion_id
51s2p0n4-99ns-30s7-0o93-ss4os0n0r1so@fhfr.qr
source_type
public_inbox
title
Re: GSoC: possible IPA-ICF project direction
headline
GSoC: possible IPA-ICF project direction
tldr
GSoC student explores IPA-ICF with separate function bodies, finding issues with inlining and non-null promises.
proposer
Rohith Kapelli <kapellirohith@gmail.com>
consensus
Proposed
outcome
proposed
sentiment_score
7
technical_tradeoffs
  • • Potential for code size reduction through improved ICF.
  • • Risk of incorrect code generation if separate bodies or inlining are handled improperly.
  • • Complexity in managing function visibility and linkage across compilation units.
series_id
—
series_role
standalone
series_parts
[]
tags
  • • gcc
  • • gsoc
  • • ipa
  • • icf
  • • optimization
bugzilla_url
—
date
2026-09-14T00:00:00.000Z

Re: GSoC: possible IPA-ICF project direction

A GSoC student is investigating a project direction for IPA-ICF (Interprocedural Analysis - Identical Code Folding), focusing on the use of separate function bodies. Initial experiments with test cases for PR119006 and PR123227 reveal issues related to incorrect code generation when function bodies are inlined and problems with non-null promises. The student also observed differences in behavior with GNU extern inline and plans further investigation into IPA-cloning and ICF’s handling of losing bodies.