Abstract Binary Decision Diagram (BDD) construction and manipulation is an important part of CAD tasks. One of the ways to improve BDD package performance is to perform certain BDD operations in parallel with the GPU. The recent development of GPU frameworks for general-purpose programming, such as OpenCL or Nvidia CUDA, has made GPUs a very powerful and attractive option for developing... Binary Decision Diagram (BDD) construction and manipulation is an important part of CAD tasks. One of the ways to improve BDD package performance is to perform certain BDD operations in parallel with the GPU. The recent development of GPU frameworks for general-purpose programming, such as OpenCL or Nvidia CUDA, has made GPUs a very powerful and attractive option for developing high-performance numerical applications. This paper proposes an efficient implementation of the BDD package that distributes computational workloads over CPUs and GPUs. This implementation takes advantage of various parallelism sources found in the BDD package. The experimental results demonstrate that implementing this solution results in significant computational speedups. Read More Read Less

Abstract In real-time multitasking systems, the scheduling algorithm is one of the most critical factors in meeting response time requirements while maximizing system resources. There are two popular scheduling algorithms: fixed priority (FP), which has been studied in detail before. Still, we haven’t compared them in the context of cache-related pre-emption (CRPD) delays. Memory and cache are divided into several blocks... In real-time multitasking systems, the scheduling algorithm is one of the most critical factors in meeting response time requirements while maximizing system resources. There are two popular scheduling algorithms: fixed priority (FP), which has been studied in detail before. Still, we haven’t compared them in the context of cache-related pre-emption (CRPD) delays. Memory and cache are divided into several blocks with instructions and stored data. When a task is pre-empted, cache blocks from that task can evict blocks from the preempted task. When the task is restarted, if it has to load those evicted blocks again, CRPD is introduced, which affects the task schedulability. Read More Read Less

Abstract We examine the appropriateness of cache configurations with random placement and replacement in a hard real-time system. It has been argued that these cache configurations drastically decrease the information needed for static WCET analysis and enable measurement-based probabilistic timing analysis. We dispute these claims and argue that with prevailing static and measurement-based analysis techniques, cache configurations with deterministic placement and... We examine the appropriateness of cache configurations with random placement and replacement in a hard real-time system. It has been argued that these cache configurations drastically decrease the information needed for static WCET analysis and enable measurement-based probabilistic timing analysis. We dispute these claims and argue that with prevailing static and measurement-based analysis techniques, cache configurations with deterministic placement and leastused replacement are preferred over random configurations. Read More Read Less