.png){kind=icon}
.png){kind=icon}
Return
High-performance Computing Heterogeneous Systems and Subsystems
.png){kind=icon}
PRE
AI
DOI：10.1016/j.future.2026.108500.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This editorial introduces the special issue on “High-performance Computing Heterogeneous Systems and Subsystems” published in Future Generation Computer Systems. The issue comprises fifteen cutting-edge contributions addressing key challenges in modern heterogeneous computing systems, from emerging architectures like SmartNICs/DPUs and processing-in-memory (PIM) to advanced programming models, automatic performance tuning, and intelligent scheduling/resource management. Papers explore task-based data-flow integration of multi-accelerator APIs (CUDA, SYCL, Triton), deep reinforcement learning for dynamic GPU reconfiguration (MIG) and objective adaptation, scalable backfilling for heterogeneous clusters, remote execution in multi-chiplet GPUs, and high-precision NTT on PIM for secure FHE. Additional works cover mixed-precision GEMM optimization, graph-parallel accelerators, ML-driven OpenMP autotuning, portable task-graph frameworks, and real-time AI monitoring for energy-efficient scheduling. Together, these articles provide a comprehensive roadmap for exploiting hardware heterogeneity to achieve scalable performance, energy efficiency, and programmability in next-generation HPC and AI infrastructures.
Keywords:
High-performance Computing
Heterogeneous Systems
Subsystems
Performance Tuning
Intelligent Scheduling
Journal
F
IF:
0
Papers:
642
Citations:
0
Organization
View more organizations