Initial AVX-512 support#201

Closed

Shnatsel wants to merge 25 commits into

linebender:mainfrom

Shnatsel:avx512

Commits on Feb 21, 2026

Add initial AVX-512 level; delegates to AVX2 for everything for now
Shnatsel
committed
Implement native 512-bit vector support for AVX-512

claude
authored and
Shnatsel
committed
Use native AVX-512 intrinsics for unary and binary ops

claude
authored and
Shnatsel
committed
Use native AVX-512 intrinsics for select, shift, reinterpret, and widen/narrow

claude
authored and
Shnatsel
committed
Use native AVX-512 intrinsics for slide operations

claude
authored and
Shnatsel
committed
Use native AVX-512 intrinsics for comparison operations

claude
authored and
Shnatsel
committed
Remove force_generic_op override for x86

claude
authored and
Shnatsel
committed
Bump MSRV to 1.89 to get access to AVX-512 intrinsics
Shnatsel
committed
Fix clippy cast_possible_wrap warning

claude
authored and
Shnatsel
committed
Expand test coverage to include AVX-512
Shnatsel
committed
Fix AVX-512 bugs in shift, precise conversion, and zip operations

claude
authored and
Shnatsel
committed

Commits on Feb 22, 2026