LLVM Weekly - #532, March 11th 2024

Welcome to the five hundred and thirty-second 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 http://llvmweekly.org and pass it on to anyone else you think may be interested. Please send any tips or feedback to [email protected], @llvmweekly or @asbradbury on Twitter, or @[email protected] or @[email protected].

News and articles from around the web and events

LLVM 18.1.0 (and the quick follow-up 18.1.1 fixing a version numbering issue) was released. Congratulations and thank you to everyone involved. Note that this is the first release using the new versioning scheme.

The next LLVM Bay Area meetup will take place on Monday 18th March.

MaskRay blogged about a compact relocation format for ELF.

According to the LLVM calendar in the coming week there will be the following (note, the US entered daylight savings time this weekend):

• Office hours with the following hosts: Aaron Ballman, Alexey Bader, Alina Sbirlea, Kristof Beyls, Johannes Doerfert.
• Online sync-ups on the following topics: pointer authentication, SPIR_V, new contributors, OpenMP, Flang, BOLT, libc, MLIR, RISC-V.
• For more details see the LLVM calendar, getting involved documentation on online sync ups and office hours.

On the forums

• Jeremy Morse provided another update on work to eliminate debug intrinsics.

• Nikita Popov summarised recent improvements to the IR parser, including the removal of the requirements to have declarations for intrinsics and for unnamed values to be consecutive, and an -allow-incomplete-ir option for opt.

• Michael Buch posted an LLDB RFC on enabling more reliable completion of record types.

• Minutes from the February LLVM Foundation board meeting have now been posted.

• The 63rd MLIR News is now available.

• Nikita Popov proposed adding nowrap flags to the trunc IR instruction, motivated for use cases like induction variable widening. The responses so far are very positive.

LLVM commits

• The frontend performance tips documentation was expanded with tips about loads/stores and aggregates. ff66e9b.

• update_test_checks now keeps meta variables (TMBnn) stable by default, reducing the noise in diffs. 3846019.

• A range attribute was introduced for parameter and return values. 4028267.

• STRICT_BF16_TO_FP and STRICT_FP_TO_BF16 SelectionDAG nodes were introduced. 8300f30.

• Vector reduction instructions were added to the SPIR-V backend. 540d255.

• AMDGPU gained a -lower-buffer-fat-pointers pass which rewrites away address space 7. 6540f16.

• Support was added for parsing non-instruction debug info from textual IR. 464d9d9, 01e5d46.

• JITLink gained support for Intel VTune. 00f4121.

Clang commits

• The C++23 [[assume]] attribute was implemented. 2b5f68a.

• It’s now possible to use a shorthand for GitHub issue links in the Clang release notes. 6e36ceb.

• -Wmissing-designated-field-initializers was added. 7df43cc.

• __builtin_cpu_supports was documented. dcd08da.

Other project commits

• LLVM’s libc GPU documentation pages were overhauled. 0cbbcf1.

• Support for additional Linux kernel sections were added to BOT. 0262979, 143afb4, f51ade2, ccf0c8d.

• The MLIR inline pass was refactored ahead of adding cost model hooks. 2542d34.