Researchers have developed a new framework for partitioning machine learning workloads between central processing units (CPUs) and Computing-in-Memory (CIM) accelerators. This framework addresses limitations in existing approaches by considering Resistive Random Access Memory (RRAM) constraints, parallelism, and the CPU's role as a complementary resource. The proposed Integer Linear Programming (ILP)-based method minimizes inference latency and has demonstrated significant speedups, achieving up to 30.9x over CPU-only execution on an edge CPU and 7.3x over a high-performance CPU. AI
IMPACT This research could lead to more efficient AI hardware architectures and faster inference times for ML applications.
RANK_REASON The cluster contains an academic paper detailing a new method for optimizing ML workloads.
- central processing unit
- CIM Accelerators
- Design space exploration
- integer linear programming
- machine learning
- Resistive Random Access Memory
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
