Jacob Wahlgren

My research in High Performance Computing Systems is at the intersection of hardware and software. Motivated by the massive cost and energy usage of large-scale parallel computing, I explore topics such as disaggregated memory to reduce idle resources, unified memory for efficient GPU systems, and SmartNICs to accelerate distributed communication.

I am part of the ScaLab research group led by Prof. Ivy Peng.

My research interests include:

• Disaggregated memory
• Tiered memory
• High-performance computing
• Operating systems
• Computer architecture

News

• October 2025:I presented "Dissecting CPU-GPU Unified Physical Memory on AMD MI300A APUs" at IISWC (Irvine, USA), and was nominated for the Best Paper award!
• July 2025: I am attending ATPESC 2025, the Argonne Training Program on Extreme-Scale Computing, a two-week intensive summer school in Illinois, USA hosted by Argonne National Laboratory on computational science and HPC (acceptance rate 32%).
• September 2024: The paper "Multi-level Memory-Centric Profiling on ARM Processors with ARM SPE" lead by our master's student Samuel Miksits was accepted to MEMO'24 International Workshop on Memory System, Management and Optimization held in conjunction with SC (Atlanta, USA). It is a continuation of my previous work on heterogeneous memory profiling.
• September 2024: Our paper "Disaggregated Memory with SmartNIC Offloading: a Case Study on Graph Processing" was accepted to SBAC-PAD 2024 (Hilo, USA)!
• November 2023: I presented a poster "Exploring Userspace Memory Mapping for RDMA-Enabled Network-Attached Memory" at SC 2023 (Denver, USA).
• June 2023: Our paper "A Quantitative Approach for Adopting Disaggregated Memory in HPC Systems" was accepted to SC 2023 (Denver, USA)! Acceptance rate: 24%.
• May 2023: I am honored to receive the Research Poster Award, 2nd place for our poster "HPC Applications on Disaggregated Memory Systems" at ISC 2023 (Hamburg, Germany)!
• September 2022: Our paper "Evaluating Emerging CXL-enabled Memory Pooling for HPC Systems" was accepted to the IEEE/ACM Workshop on Memory Centric High Performance Computing (MCHPC'22) held in conjunction with SC (Dallas, USA).