The Computer Architecture stack in today’s systems is becoming more heterogenous than ever. Today’s systems cater to safety-critical or real-time applications more often, making high throughput, increased system utilization, reliability and security a necessity rather than a feature. Accelerators are becoming first class computing citizens especially GPUs. It hence, makes sense for these accelerators to gain access to system services without going through a beefy CPU which makes them slower. System services for accelerators is not a novel concept and has been looked at already such as file systems for GPUs, networking for GPUs etc. However providing these services efficiently, i.e. guaranteeing good system utilization by overlapping operational latency with other work and in turn benefitting performance and throughput is an area I am interested in exploring. Asynchronous interfaces in form of asynchronous system services or micro-services help achieve the goal of increased system utilization and throughput. Hence, I am interested in exploring standalone asynchrnous system services for GPUs to help them escape being pulled down by a beefy CPU and achieve benefits of asynchrony.
Asynchrony is non-intuitive. Add to this a complex programming model of GPUs makes the combination tricky. Hence achieving asynchronous system services for GPUs in a systematic manner is interesting. Formalizing the non-intuitive asynchronous model makes it achievable for safety-critical applications as well as eases programming complexity in any application. Hence, I am also looking at formalizing the asynchronous programming interface to make it easy to reason about in terms of programmability, safety and correctness.
I broadly envision building accelerator centric Operating Systems. My primary research interests are Operating Systems, Formal Methods and Computer Architecture.
Before coming to Yale, I was a part of the Compiler Optimization group at NVIDIA. My work had been on working on the compiler backend support for the then upcoming Turing architecture which was released in 2018.
- October 2020 : Awarded MICRO 2020 conference grant.
- October 2020 : Selected to serve on the board of Women in Sciences at Yale (WISAY).
- April 2020 : Awarded travel scholarship for 2020 CRA-WP Grad Cohort for Women to be held in New Orleans, LA.
- August 2019 : Joined Yale Computer Science Ph.D under Prof. Abhishek Bhattacharjee.