cd /news/ai-chips/amd-upstreams-avx10-v2-aux-support-i… · home topics ai-chips article
[ARTICLE · art-138991] src=phoronix.com ↗ pub= topic=ai-chips verified=true sentiment=↑ positive

AMD Upstreams AVX10_V2_AUX Support into LLVM Clang 24

AMD's initial AVX10_V2_AUX ISA support was merged into LLVM 24 Git on Monday, spanning 15,000 lines of new code across 20 patches and exposed in the Clang compiler via the new -mavx10v2aux flag. AVX10_V2_AUX is part of the AI Compute Extensions (ACE) developed by Intel and AMD through the x86 Ecosystem Advisory Group, adding instructions to convert standard data types to compressed formats such as FP4 and FP6, additional OCP microscaling formats, and new rounding modes to accelerate AI workloads. LLVM/Clang 24.1 stable is expected around next March.

read2 min views1 publishedSep 24, 2026
AMD Upstreams AVX10_V2_AUX Support into LLVM Clang 24
Image: Phoronix (auto-discovered)

In addition to

AVX10_V2_AUX adds dedicated data processing and format conversion for helping accelerate AI workloads. This adds onto AVX10 with instructions for converting standard data types to compressed data formats like FP4 and FP6, additional OCP microscaling formats, new rounding modes, and more. AVX10_V2_AUX is part of the AI Compute Extensions (ACE) worked on by both Intel and AMD as part of the x86 Ecosystem Advisory Group.

Engineers from both Intel and AMD have worked on various bits of software around the

Merged to LLVM 24 Git on Monday was the initial AVX10_V2_AUX ISA support and is being exposed with the Clang compiler via the new

LLVM/Clang 24.1 stable should be out around next March so we'll see what more work from the Intel/AMD x86 Ecosystem Advisory Group makes it into this next open-source compiler release.

AMD working on AVX10_V2_AUX support for the GCC compiler, this summer they also posted AVX10_V2_AUX support for the LLVM/Clang compiler. As of yesterday that enablement was merged for LLVM 24 for that important step toward AI Compute Extensions (ACE). AVX10_V2_AUX adds dedicated data processing and format conversion for helping accelerate AI workloads. This adds onto AVX10 with instructions for converting standard data types to compressed data formats like FP4 and FP6, additional OCP microscaling formats, new rounding modes, and more. AVX10_V2_AUX is part of the AI Compute Extensions (ACE) worked on by both Intel and AMD as part of the x86 Ecosystem Advisory Group.

Engineers from both Intel and AMD have worked on various bits of software around the

AI Compute Extensionswith AMD engineers seemingly responsible for the AVX10_V2_AUX work. Merged to LLVM 24 Git on Monday was the initial AVX10_V2_AUX ISA support and is being exposed with the Clang compiler via the new

-mavx10v2aux compiler flag. The AVX10_V2_AUX implementation spans 15k lines of new code over 20 patches. Seethe LLVM pull requestfor those curious about any other details of this implementation. LLVM/Clang 24.1 stable should be out around next March so we'll see what more work from the Intel/AMD x86 Ecosystem Advisory Group makes it into this next open-source compiler release.

── more in #ai-chips 4 stories · sorted by recency
── more on @amd 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/amd-upstreams-avx10-…] indexed:0 read:2min 2026-09-24 ·