power8 开源项目推荐（28 个）

noloader /SHA-Intrinsics

SHA-1, SHA-256 and SHA-512 compression functions using Intel, ARMv8 and Power8 SHA intrinsics

Scottcjn /ram-coffers

LLM infrastructure cost reduction via NUMA-aware weight banking: 147 t/s (8.8x stock llama.cpp) on refurbished enterprise POWER8. Self-hosted inference, no cloud APIs. Part of the Proof of Physical AI stack.

Scottcjn /llama-cpp-power8

llama.cpp optimizations for IBM POWER8: vec_perm non-bijunctive collapse, PSE hardware entropy, DCBT prefetch. Sovereign inference. Part of the Proof of Physical AI stack.

open-power /skiboot

OPAL boot and runtime firmware for POWER

Scottcjn /claude-code-power8

The first POWER8/ppc64le port of Claude Code CLI

Scottcjn /amd-rocm-power8-patches

ROCm stack for AMD GPUs on IBM POWER8 - First ppc64le build

Scottcjn /nvidia-power8-patches

World first: Modern NVIDIA driver for IBM POWER8 via OCuLink - RTX 5070/4090/3060/V100 support

Scottcjn /cuda-power8-patches

Run NVIDIA CUDA 10.2, 11.8, and 12.0 on IBM POWER8 systems with V100 GPUs

Scottcjn /ltx-video-power8

LTX-Video 13B inference on IBM POWER8 S824 - First 13B video diffusion model on PowerPC

Scottcjn /mlx-power8

Apple MLX ported to POWER8 Linux with VSX SIMD optimizations

Scottcjn /power8-fpga-tools

FPGA programming tools ported to IBM POWER8 (ppc64le) - xc3sprog, xvcd for Platform Cable USB II

Scottcjn /hailort-ppc64le

HailoRT SDK port for IBM POWER8/POWER9 ppc64le - Hailo-8 AI accelerator on PowerPC

Scottcjn /power8-projects

POWER8 Projects - Ubuntu 22.04 build, PSE LLM, Darwin cross-compile

Scottcjn /oculink-gpu-bypass

GPU bypass for IBM POWER8/POWER9 and PowerPC Mac - internal PCIe rescan + OCuLink support

Scottcjn /vulkan-matmul

Vulkan + ROCm GPU matmul offload server for POWER8 with AMD Navi 12 OCuLink

Scottcjn /xrt-ppc64le

Xilinx Runtime (XRT) 2.21.70 for IBM POWER8 (ppc64le) - First ever ppc64le build

Scottcjn /ppc64le-native-modules

Native ppc64le modules for llama.cpp webui (lightningcss, tailwindcss-oxide) - Built on IBM POWER8

noloader /AES-Intrinsics

AES encryption function using Intel, ARMv8 and Power8 intrinsics

Scottcjn /rpi-inference

RPI (Resonant Permutation Inference) — Zero-multiply text generation. 18K tok/s. 868 KB models. Standalone or as speculative draft engine for LLMs. Runs on N64, POWER8, x86, ARM.

Scottcjn /pse-vcipher-collapse

Non-bijunctive attention collapse for LLM inference — POWER8 hardware AES (vcipher) + AltiVec vec_perm. Hebbian path selection, cross-head diffusion, O(1) KV prefiltering.