Next-Generation Computing: PyPIM Replaces Processors with Memory
16:11, 13.11.2024
Next-Generation Computing: PyPIM Replaces Processors with Memory
Researchers at Israel’s Technion Institute have developed an innovative software platform, PyPIM, enabling computations to be carried out directly in memory, without the involvement of a processor. This approach, known as Processing-in-Memory (PIM), significantly accelerates computing tasks and reduces energy consumption. The foundation of the new platform is a command library that transforms Python commands into machine code executed directly in memory, allowing developers to adapt to a new way of working with data.
Professor Shahar Kvatinsky, who leads this project, and his team are tackling the “memory wall problem” — a bottleneck caused by the need to transfer data between memory and the processor. Over the past few years, the team has published numerous papers outlining the theoretical underpinnings of such computations and specific examples of PyPIM implementation.
Applications and Development Potential of PyPIM
PyPIM holds great potential for various fields, including artificial intelligence, bioinformatics, finance, and information systems, as it can greatly enhance performance and reduce energy costs. Furthermore, Technion researchers have developed a special modeling tool to support PIM hardware that allows performance measurements and comparisons with traditional computers.
The team’s findings were presented at the IEEE/ACM International Symposium on Microarchitecture in Austin, Texas, and published on arXiv. Among the authors is 21-year-old Orian Leitersdorf, who is soon to become the youngest Ph.D. at Technion.