SRSourav Roy

Sourav Roy

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Sourav Roy has a total of 15 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Modeling Energy and Delay for ECC-Enabled Cache Architectures

    by Dinesh Joshi, Aritra Bagchi, Ayushi Agarwal, Garima Modi, Neha Srivastava, Sourav Roy, Preeti Ranjan Panda

    With increasing integration density, modern Systems-on-Chips (SoCs), especially on-chip caches and memories, are critically vulnerable to soft errors. Error Correcting Codes (ECC) are the most widely adopted mitigation technique in safety-critical SoCs, making accurate modeling of ECC overheads essential for effective design decisions. Existing cache modeling frameworks such as CACTI consider only the memory overheads for storing ECC check-bits, and neglect the delay and energy costs of ECC combinational logic, assuming them negligible. We extend CACTI to include the delay and energy overheads due to the ECC logic along with the memory-induced ECC overheads. In addition to the configurations supported by CACTI, we model High Reliability (HR) and High Performance (HP) modes of ECC operation for flexible design exploration. Extensive evaluation across SPEC CPU 2006 and MiBench workloads demonstrates the effectiveness of our modeling, highlighting that neglecting logic overheads underestimates the ECC performance and energy overheads by 12% and 18%, respectively, for SPEC CPU 2006, and 12% and 17%, respectively, for MiBench.

  • Open Access

    FLASH: Deadline-Aware Flexible LLC Arbitration and Scheduling for Hardware Accelerators

    by Ayushi Agarwal, Pulkit Goel, P.J. Joseph, Prokash Ghosh, Sourav Roy, Preeti Ranjan Panda

    Integrating domain-specific hardware accelerators on modern systems on chips (SoCs) has enabled complex applications, such as vision, natural language processing, autonomous driving, and augmented reality, on small form factors. This leads to challenges in the integration of accelerators, with high memory bandwidth requirements and strict deadlines, on the system’s memory hierarchy. The system-level shared cache, or last-level cache (LLC), is a critical resource shared by multi-core processors, GPUs, and hardware accelerators in modern heterogeneous SoCs. It significantly reduces the bottleneck at the off-chip memory and delivers high performance. With the integration of accelerators on the LLC gaining momentum, the on-chip shared cache management becomes vital. If not managed intelligently, the interference between cache requests from the cores and the accelerators can significantly deteriorate their performance. Given the architectural differences between DRAM and cache systems, the off-chip memory management strategies explored by previous works cannot be extended to the LLC. We propose a deadline-aware flexible LLC arbitration and scheduling framework, FLASH , to dynamically partition the LLC bandwidth between the accelerators and multi-core processors to meet the deadline given for the accelerator while minimizing the impact on the performance of the cores. FLASH arbitrates between the requests from the cores and the accelerators and schedules the requests depending on the accelerator’s progress and its chances of meeting the deadline. We evaluate FLASH across different workloads and hardware accelerator configurations to show that it not only achieves significantly better performance for the cores than other static scheduling policies but also significantly reduces the deadline miss rates of the accelerator.


  • Exception Coverage on Automotive Processors

    by Sourav Roy, Nikhil Jain, Neha Srivastava, Ravinder Dasila, Prokash Ghosh, Ajay Sharma, Afshan Anjum, Jeff Scott

    With the advent of modular and independent zonal processing in Software-Defined-Vehicles (SDV), the processor in an automotive system is expected to handle complex exception scenarios elegantly. This paper addresses the need for comprehensive exception coverage during the verification of the automotive processor. Unlike prior research focusing on only two simultaneous exceptions, our work extends to all combinations of simultaneous exceptions. The proposed methodology involves the systematic grouping of exceptions into sub-categories based on similarity while maintaining priority, followed by targeted coverage assessment for each group. A two-pass simulation technique is used to drive multiple external events like interrupts in the second pass to create high-activity time windows with simultaneous exceptions. This hierarchical approach allows for the creation of a manageable and effective exception list, facilitating thorough verification without overwhelming computational resources. Applied to a 32-bit RISC-V processor, our methodology shows a large improvement of 7X in efficiency compared to traditional methods, highlighting its advantage in enhancing safety and trustworthiness of automotive systems. This work takes a major step towards the verification of all combinations of simultaneous exceptions in processors.