Source snapshot of github.com/Scottcjn/mlx-power8, also on GitLab with full history. Code only: nothing here runs on Hugging Face infrastructure.

License MLX POWER8 BCOS Certified

MLX Linux Port - POWER8 & x86 SIMD

Port of Apple's MLX to Linux with native SIMD optimizations.

Architectures Supported

Architecture SIMD Vector Width
POWER8/9 VSX 128-bit (4 float, 2 double)
x86_64 AVX 256-bit (8 float, 4 double)
ARM64 NEON (use upstream)

Files

  • simd/vsx_simd.h - POWER8 VSX vectorization
  • simd/avx_simd.h - x86 AVX vectorization
  • simd/type_linux.h - Architecture detection
  • BUILD_POWER8.md - Build instructions

Usage

Copy simd/ contents to mlx/backend/cpu/simd/ in your MLX source tree.

Why?

MLX networking uses standard TCP sockets - already portable! Only the SIMD backend needed porting for each architecture.

Credits

  • Apple MLX team (original framework)
  • Elyan Labs (POWER8/x86 Linux port)

Part of the Elyan Labs Ecosystem

  • BoTTube โ€” AI video platform where 119+ agents create content
  • RustChain โ€” Proof-of-Antiquity blockchain with hardware attestation
  • GitHub
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support