18069026. PIM COMPUTING SYSTEM AND PIM COMPUTATION OFFLOADING METHOD THEREOF simplified abstract (SK hynix Inc.)
PIM COMPUTING SYSTEM AND PIM COMPUTATION OFFLOADING METHOD THEREOF
Organization Name
Inventor(s)
Seon Wook Kim of Namyangju (KR)
Chang Hyun Kim of Seongnam (KR)
PIM COMPUTING SYSTEM AND PIM COMPUTATION OFFLOADING METHOD THEREOF - A simplified explanation of the abstract
This abstract first appeared for US patent application 18069026 titled 'PIM COMPUTING SYSTEM AND PIM COMPUTATION OFFLOADING METHOD THEREOF
Simplified Explanation
The abstract describes a Processing-In-Memory (PIM) computing system and a method for offloading PIM computations using a DMA engine. The system includes a CPU, a memory with a PIM unit and a memory array (such as DRAM), and a DMA engine. The DMA engine processes transactions by handling descriptor requests and PIM requests for descriptors stored in memory, in response to memory requests from the CPU. The PIM unit performs PIM operations using data provided to the DMA engine in response to the descriptor requests.
- The patent application describes a PIM computing system and method for offloading PIM computations using a DMA engine.
- The system includes a CPU, a memory with a PIM unit and a memory array (e.g., DRAM), and a DMA engine.
- The DMA engine processes transactions by handling descriptor requests and PIM requests for descriptors stored in memory.
- The PIM unit performs PIM operations using data provided to the DMA engine in response to the descriptor requests.
Potential Applications
- High-performance computing: The PIM computing system can be used in applications that require high computational power, such as scientific simulations, data analytics, and machine learning.
- Edge computing: The system can be deployed in edge devices to offload PIM computations, enabling faster and more efficient processing of data at the edge.
- Internet of Things (IoT): The PIM computing system can be utilized in IoT devices to perform complex computations locally, reducing the need for data transmission and improving response time.
Problems Solved
- Offloading PIM computations: The system provides a method for offloading PIM computations from the CPU to the PIM unit, reducing the burden on the CPU and improving overall system performance.
- Efficient memory access: The DMA engine efficiently handles memory requests, allowing for faster data transfer between the CPU, PIM unit, and memory array.
- Enhanced processing capabilities: The PIM unit performs PIM operations using data provided by the DMA engine, enabling efficient and parallel processing within the memory itself.
Benefits
- Improved performance: Offloading PIM computations and utilizing the PIM unit within the memory array can significantly enhance overall system performance and reduce processing time.
- Reduced CPU load: By offloading computations to the PIM unit, the CPU can focus on other tasks, improving system multitasking capabilities.
- Energy efficiency: The PIM computing system reduces data movement between the CPU and memory, resulting in lower energy consumption and improved energy efficiency.
Original Abstract Submitted
A Processing-In-Memory (PIM) computing system and a PIM computation offloading method thereof perform PIM computation offloading using a DMA engine. The DMA engine is configured to process a transaction by respectively performing descriptor requests and PIM requests for one or more descriptors stored in a memory, in response to a memory request of a CPU. The memory includes a PIM unit and a memory array, which memory array may be a DRAM. In response to the PIM requests, the PIM unit performs PIM operations using information included in data provided to the DMA engine in response to the descriptor requests.
