Graphics Processing Units (GPUs) have experienced a remarkable transformation, evolving from their origins as fixed-function graphics accelerators into versatile, general-purpose computing platforms that now underpin a wide range of modern technologies. Their capacity to exploit massive parallelism and deliver high throughput has established them as indispensable elements of heterogeneous embedded platforms, spanning from resource-constrained edge devices to large-scale High-Performance Computing (HPC) infrastructures. In real-time embedded domains, GPUs are increasingly employed in computationally intensive workloads such as multi-object tracking, traffic monitoring, and video surveillance, where accuracy and responsiveness are simultaneously required. Beyond these domains, emerging applications such as autonomous driving rely heavily on General-Purpose computing on Graphics Processing Units (GPGPUs) to meet stringent performance and reliability demands. As these trends indicate, the integration of real-time guarantees into GPU-based systems has shifted from being merely advantageous to becoming an essential requirement for future computing platforms.