GPGPU Microarchitecture Peformance Enhancement
Energy efficient Instruction scheduling mechanisms for GPUs.
I am an Associate Professor at Pillai University and Pillai HOC College of Engineering and Technology under Mahatma Education Society. Iam currently in my final year as research scholar in Computer Science and Engineering at IIT Bombay, in Prof. Virendra Singh’s Computer Architecture and Dependable Systems Lab (CADSL). My research interests lie in GPGPU Architecture and Performance, Microarchitecture Performance and OS-Microarch Interactioins. I am primarily conducting research in the area of GPU microarchitecture enhancements to accelerate non traditional, scientific and AI GPGPU workloads. The research involves designing novel out of order scheduling techniques and front end pipeline redesign to extract Instruction Level Parallelism from applications with limited data level parallelism.
News
- BB8 accepted at ASP-DAC '27 - breaking head-of-line stalling in GPGPUs via Instruction Bypassing.
- PRISM accepted at ICCD 2026 - priority-ranked instruction scheduling mechanism for GPU scoreboard stalls.
- HOOP accepted at SBAC-PAD 2026, Madrid - hint-driven out-of-order issue without per-SM scoreboard logic.
- Best Paper Award at ACM Computing Frontiers 2026 in Catania for FlIP.
Publications
2027
BB8: Blocking Buffer for breaking head-of-line stalling in GPGPUs via Instruction Bypassing.Accepted
2026
PRISM: Priority Ranked Instruction Scheduling MechanismAccepted
HOOP: Hint-based Out-of-Order Processing for GPGPUs via Per-Kernel Dependency AnalysisAccepted
Bypassing Blocking Instructions to Enable Out-of-Order Execution in GPGPUs
Selected work
PRISM: Priority Ranked Instruction Scheduling MechanismAccepted
HOOP: Hint-based Out-of-Order Processing for GPGPUs via Per-Kernel Dependency AnalysisAccepted