Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.22512 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915516682076160 |
|---|---|
| author | Ahadi, Soroush Modarressi, Mehdi Daneshtalab, Masoud |
| author_facet | Ahadi, Soroush Modarressi, Mehdi Daneshtalab, Masoud |
| contents | Large language models demand massive computational power and memory resources, posing significant challenges for efficient deployment. While quantization has been widely explored to reduce model size and computation, this paper demonstrates an additional benefit: quantization increases parameter locality, creating opportunities for computation reuse. Building on this insight, we propose AxLLM, a hardware accelerator architecture designed for quantized models. Axllm introduces a novel redundancy elimination technique that caches and reuses multiplication results for repeated weight values, substantially reducing redundant operations. The architecture features dual multiply and reuse pipelines, efficiently supporting both base models and LoRA fine-tuned models without altering parameters, retraining, or requiring offline preprocessing. Experimental results show that AxLLM achieves up to 90% reduction in computations, delivering 28% lower energy consumption and a 1.7x speedup over baseline execution. These results highlight Axllm as a scalable and efficient solution for accelerating LLMs on specialized hardware. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_22512 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AxLLM: accelerator architecture for large language models with computation reuse capability Ahadi, Soroush Modarressi, Mehdi Daneshtalab, Masoud Hardware Architecture n/a Large language models demand massive computational power and memory resources, posing significant challenges for efficient deployment. While quantization has been widely explored to reduce model size and computation, this paper demonstrates an additional benefit: quantization increases parameter locality, creating opportunities for computation reuse. Building on this insight, we propose AxLLM, a hardware accelerator architecture designed for quantized models. Axllm introduces a novel redundancy elimination technique that caches and reuses multiplication results for repeated weight values, substantially reducing redundant operations. The architecture features dual multiply and reuse pipelines, efficiently supporting both base models and LoRA fine-tuned models without altering parameters, retraining, or requiring offline preprocessing. Experimental results show that AxLLM achieves up to 90% reduction in computations, delivering 28% lower energy consumption and a 1.7x speedup over baseline execution. These results highlight Axllm as a scalable and efficient solution for accelerating LLMs on specialized hardware. |
| title | AxLLM: accelerator architecture for large language models with computation reuse capability |
| topic | Hardware Architecture n/a |
| url | https://arxiv.org/abs/2509.22512 |