Excalibur notes 12 April. Phase 1b application. Call announced mid may. Deadline w/c 27th Sept. (so after the cross-cutting results are announced). Cross-cutting calls we’re supporting/involved in: Met office call: - Task parallelism (Tobias) - Parallel in time (Dave) UKRI - Code coupling (Ben/Stephen) - Graphics (Dave) - Future paradigms (Rgb/Dirac) - Quantum computing for HPC (Rgb) Contract. NDA Abouzied. Tasks on the GPU. We discussed the best strategy. Abouzied is talking to Matthieu to see how best to do this. Moving to openCL rather than Cuda if possible. Edo. Plots of timesteps - need to figure out a way to compensate for the length of the run. Draper. Single core band width not used well at times, so this can be sped up by parallelising. However, this shows that the ultimate performance is still limited by the amount of data being sent. Could we avoid sending complete particles - this is challenging because it makes it difficult to programme new tasks - since the task need to determine what data is exchanged. An issue of separation of concerns. Key point is that this is not a limitation in the longer steps - its an issue that arises from using multi-dt and having fewer particles per step. Peter’s going to now focus on whether this can be improved by exploring the fabric design.