Pushing the Envelope of LLM Inference on AI-PC and Intel GPUs

TL;DR

This paper develops optimized 1-bit and 2-bit microkernels for CPUs and Intel GPUs, significantly improving the efficiency and speed of ultra-low-bit LLM inference on AI PCs and GPUs, enabling resource-efficient deployment.

Contribution

It introduces novel microkernels for ultra-low-bit LLM inference on CPUs and GPUs, achieving state-of-the-art performance and end-to-end speedups over existing runtimes.

Findings

2.2x faster than bitnet.cpp for 2-bit models

Up to 7x speedup over 16-bit inference

4x-8x reduction in GEMM time on Xe GPUs

Abstract

The advent of ultra-low-bit LLM models (1/1.58/2-bit), which match the perplexity and end-task performance of their full-precision counterparts using the same model size, is ushering in a new era of LLM inference for resource-constrained environments such as edge devices and AI PCs. While these quantization advances promise models that are more cost-effective in terms of latency, memory, throughput, and energy consumption, the computational efficiency of state-of-the-art (SOTA) inference runtimes (e.g., bitnet.cpp) used to deploy them remains underexplored. In this work, we take a bottom-up approach: we first design and implement 1-bit and 2-bit microkernels optimized for modern CPUs, achieving peak computational efficiency across a variety of CPU platforms. We integrate these microkernels into a state-of-the-art LLM inference framework, namely PyTorch-TPP, and present end-to-end inference results with 2-bit models that outperform the current SOTA runtime bitnet.cpp by up to 2.2x, and deliver up to 7x speedup compared to the 16-bit model inference. We then extend this work to Intel GPUs where we design and implement mixed precision, 2-bit GEMM kernels, and show their performance to be close to optimal. We integrated our optimized Xe2 kernels in the vLLM framework as a quantization plugin and evaluated end-to-end LLM inference results for a range of LLM models and Xe2 GPUs. Depending on the model and platform, we see a 4x - 8x reduction in GEMM time compared to the BF16 case, and we get up to 6.3x speedup in end-to-end latency compared to the BF16 execution. Our optimized runtime advances the state of LLM inference on AI PCs and Intel Xe GPUs, paving the way for efficient deployment of ultra-low-bit LLM models.