| Idea Name: |
Vector Processors |
| Slogan: |
Enabling Edge Computing for AI |
| Supervisor Name: |
Muhammad Imran |
| Supervisor Designation: |
Assistant Professor |
| Supervisor School: |
SEECS |
| Supervisor Department: |
Department of Electrical Engineering |
| Contact number: |
+923344921069 |
| Email ID: |
[email protected] |
| Abstract: |
| Vector Processors are general-purpose low-power computing accelerators. When paired with a regular processor, they can accelerate data-intensive applications such as Machine Learning and Signal Processing. Our vector processor is based on RISC-V vector ISA, which is targeted to enable machine learning on edge computing devices, opening an avenue of novel AI applications to improve quality of life. |
| What is the unmet need in society that your idea will fulfill ? |
| Vector processors are in high demand for powerful and efficient computing systems. In Pakistan, we see a huge untapped potential to develop indigenous semiconductor products that can boost our economy, while achieving self-sufficiency in a global market where few players control technology access. |
| Who needs it ? How many would benefit ? |
| Today, AI and IoT are enabling useful applications in smart cities, security, and environment. An energy-efficient vector processor can be utilized in many such applications. It is also a steppingstone to build our capacity in rapidly growing global semiconductor market of $500 Billion by revenue. |
| How will the solution works |
| A Vector Processor enhances the performance by amortizing the instruction fetch and decode overhead, utilizing multiple functional units and time-multiplexing the processing of entire array of data elements. It is designed to operate on large vectors of data, reducing the branching undercuts of “for loops” in regular processors. With the ability to disable unused functional units, it can also reduce power-consumption, which allows it to be used in small mobile devices running on batteries. |
| Who are your competitors ? How is your solution different |
| Globally, there are many academic institutes and industries that are developing AI processing hardware; however, we don’t see any immediate competitor in national market. We are incorporating features in our design that will enable better performance over existing implementations. |
| Status: |
new |
| Entry Date & Time: |
2022-12-28 (1340) |
