LLVM Weekly - #653, July 6th 2026

Welcome to the six hundred and fifty-third issue of LLVM Weekly, a weekly newsletter (published every Monday) covering developments in LLVM, Clang, and related projects. LLVM Weekly is brought to you by Alex Bradbury. Subscribe to future issues at https://llvmweekly.org and pass it on to anyone else you think may be interested. Please send any tips or feedback via email: asb@asbradbury.org, or Mastodon: @llvmweekly@fosstodon.org / @asb@fosstodon.org, or Bluesky: @llvmweekly.org / @asbradbury.org.

News and articles from around the web and events

The new issue of Communications of the ACM has a piece from Vikram S Adve and Chris Lattner on the LLVM compiler infrastructure, the impact it’s had, and the effect of the federal research funding it received.

According to the LLVM Calendar in the coming week there will be the following:

• Office hours with the following hosts: Kristof Beyls, Johannes Doerfert.
• Online sync-ups on the following topics: Flang, MLIR C/C+ frontend working group, modules, libc++, lifetime safety, OpenMP for flang, LLVM libc, HLSL, memory safety, qualification, reflection.
• For more details see the LLVM calendar, getting involved documentation on online sync ups and office hours.

On the forums

• Michael Kruse wrote up a detailed analysis of the impact of Flang’s heavy template use on compile times and memory usage and additionally gives figures on some proposed patches to improve the situation. One initial series of patches reduces compilation wall time by 35% on Clang.

• LLVM 23.x will branch next week, on Tuesday 14th July.

• Satyajanga kicked off an RFC discussion on address spaces support in LLDB.

• Jonas Devlieghere proposes a “health check” mode for LLDB, inspired by brew doctor and Neoim’s :checkhealth.

• Kim J Smith started a libc++ RFC thread on adding __is_complete_here and __is_complete_or_unbounded_here functions to __type_traits.

• Sidhartha Kumar proposes enabling -fstack-clash-protection by default for Android targets.

• I somehow missed it when first posted, but Joseph Huber proposed a concurrency sanitizer for AMD GPU targets and has now followed up with some benchmarks and PRs.

LLVM commits

• The semantics of the dereferenceable attribute were refined and usage within LLVM modified to match. 5969d64.

• LLVM’s coding standards doc gained guidance on constants and minimising startup overhead. 2d1dd6c.

• LLVM intrinsics can now be annotated with required subtarget feature expressions. 93da31b.

• A scheduling model was added for the X86 C86-4G-M8 processor. 77de767.

• The AARch64 C1-Ultra scheduling model was expanded with support for SME instructions. 2427da5.

• The SLPVectorizer gained initial support for masked stores. 42e9751.

• lit process output is now truncated to 10KiB by default. 524b97a.

• GlobalISel’s LegacyLegalizerInfo was removed. 076af82.

Clang commits

• Clang now emits alignment information for indirect arguments. 947dc92.

• Texture2DArray for HLSL was implemented. 829440b.

• Various Clang docs were migrated to Markdown. 609b651.

• The workflow was documented for PGO for HIP device code and also source-based code coverage. 397eecf, fb83b9b.

Other project commits

• An alternative representation of std::vector was added under libcxx’s unstable ABI. bd33880.

• Dexter’s structured script model was documented. b43a0db.

• Implementation of enumeration types in Flang continued. 5c02f75.

• The AST for a regex implementation in LLVM’s libc was implemented. ca80a06.

• A bugreporting LLDB plugin kind was introduced along with a GitHub bug reporter. 8cf09c5, 1556da0.