CHART Forskargrupp

Welcome to the CHART team's webpage! CHART is the Chalmers Heterogeneous Architectures and Runtimes Team. The focus of our work are parallel and heterogeneous computer architectures, and the programming models used to program them. The major goals are maximizing performance and the reduction of energy consumption. At the same time, we want to ensure that performance is scalable, portable and exhibits low variability, Our focus applications are from High Performance Computing (HPC) and Artificial Intelligence (AI) domains. Some of our driving motives have been (1) to explore the application of runtime moldability in order to improve resource utilization and reduce interference, (2) leveraging inter-task locality to improve cache utilization and reduce memory traffic, and (3) extract distant parallelism to enable the exploitation of long SIMD units. In recent years we have increasingly explored how to automatically exploit these techniques from modern programming models such as OpenMP, SYCL and two libraries for machine learning kernels: OneDNN and NNPack. If you have any questions or are interested in joining our team, please do not hesitate to get in touch with us!

Spotlight Updates! 

[2025-11-17] Miquel presented two of our group’s papers at the SC Conference! 

1. “Reproducible Performance Evaluation of OpenMP and SYCL Workloads under Noise Injection", Christoffer Persson, Mathias Prétot, Minyu Cui, Miquel Pericàs, 14th International Workshop on Runtime and Operating Systems for Supercomputers (ROSS 25) held as part of SC25. 
2. “ILAN: The Interference- and Locality-Aware NUMA Scheduler”, Edvin Mellberg, Axel Carlsson, Jing Chen and Miquel Pericas, the 16th International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS25) held as part of SC25. 

[2025/09/12] Congratulations to Christoffer Persson, Mathias Prétot, Minyu Cui, Miquel Pericàs, on getting their paper “Reproducible Performance Evaluation of OpenMP and SYCL Workloads under Noise Injection” accepted on the SC Workshop’25 ROSS! 

[2025/09/12] Congratulations to Edvin Mellberg, Axel Carlsson, Jing Chen and Miquel Pericas on getting their  paper "ILAN: The Interference- and Locality-Aware NUMA Scheduler" accepted on the SC Workshops '25 PMBS! 

Research Projects

DARE - A new era for supercomputing in Europe

EUPiLOT

EPI

eProcessor

Miquel Pericàs is a Professor (biträdande professor) at Chalmers University of Technology. His research interests cover parallel and heterogeneous programming models, and hardware/software codesign with a focus on SIMD architectures. At Chalmers Miquel is responsible for two courses on parallel programming and development and optimization of HPC software. He is currently serving as the Chalmers representative in the RISC-V International foundation. View the full list of publications on his Google Scholar page. 

Bhavishya's research primarily focuses on memory subsystems, with an emphasis on scalable cache coherence mechanisms and NUMA (Non-Uniform Memory Access) architectures. He is also passionate about developing energy-efficient resource management techniques to optimize system performance and sustainability. View the full list of publications on his Google Scholar page. 

Jing’s research focuses on enhancing energy efficiency, performance modeling, power modeling, and task scheduling in high-performance computing systems - CPUs and GPUs. Her work involves developing adaptive energy-aware task scheduling techniques for OpenMP, CUDA, SYCL runtimes by leveraging hardware control knobs and task characteristics to optimize multi-objective energy-performance trade-offs. She is also interested in large language model (LLM) quantization techniques on GPUs. A full list of her publications is available on her Google Scholar page. See her personal page for details.

Sonia's research focuses on the combined exploration of the optimization space for convolutional algorithms and the design space of tightly coupled architectures, with the goal of providing insights for designing future CPUs optimized for CNN inference. So far, she has investigated optimizations for the im2col+GEMM, Winograd, and Direct algorithms, along with tuning of micro-architectural parameters such as vector lengths, L2 cache, and on-chip parallelism on long vector architectures (RISC-VV, ARM-SVE). Now her interest is leading towards exploring dense transformers on multi-core architectures.

Hari's research focuses on performance characterization of the SYCL and OpenMP programming models, with an emphasis on SYCL through detailed micro-benchmarking. Additionally, he explores the feasibility of Process-In-Memory (PIM) processors within modern heterogeneous computing environments, aiming to unlock their potential for enhancing scientific applications.

Publications

Here is a list of the team's published papers, ordered by publication date.

New Job Vacancy!

For the job vacancy please contact the group leader Miquel Pericàs.