Using Branch Prediction and Speculative Execution to Forecast Space Weather with a Cluster of Standard PCs

Branch prediction and speculative execution consist of making probabilistic predictions about the likely near-term evolution of the near-Earth space, and distributing among the cluster machines simulations that assume each of the probabilistically predicted outcomes as initial conditions. As the near-Earth space evolves and real-time satellite data get assimilated into the algorithm, some of the speculatively executed simulations will be proved wrong. At that point the machines that were executing them will be reassigned either to new lines of speculative simulation, or to increase the processing power devoted to more promising simulations already executing. Branch prediction and speculative execution have been very successful in the design of microprocessors, allowing CPUs to attain average processing speeds much higher than linear code execution would permit.