RT-Swap accepted at IEEE RTAS 2024

Our paper “RT-Swap: Addressing GPU Memory Bottlenecks for Real-Time Multi-DNN Inference” was accepted at the 30th IEEE Real-Time and Embedded Technology and Applications Symposium.

RT-Swap extends usable GPU capacity with CPU memory and coordinates memory-object swapping with real-time DNN scheduling. The framework enables workloads whose combined memory demand exceeds physical GPU memory while preserving their timing guarantees.

The paper is available from IEEE Xplore.

Source: RTCL@DGIST announcement