Could a New Type of Parallelism Speed Up LLM Inference? – EE Times