The Uintah Computational Framework was developed to provide an environment for solving fluid-structure interaction problems on structured adaptive grids on large-scale, long-running, data-intensive problems. Use EdrawMax to create the CPU-GPU task scheduler to understand how the Data Warehouse takes the data to compute the tasks. As shown in the below image, these GPUs have two copy engines and support multiple kernels running concurrently. Using these features, GPU tasks can be copying data to-and-from the device and running multiple kernels simultaneously. In order to exploit these features, the CPU-GPU scheduler creates and manages queues of CUDA Streams, one for each device on-node. Streams provide a means to perform multiple operations simultaneously in that operations from different streams can be interleaved.
