We address the problem of synthesising real-time embedded controllers taking into account constraints deriving from the implementation platform, thus exploring the relation between the processor's time (or "attention") devoted to different control tasks and the overall robustness of the resulting design. Assuming a time-triggered model of computation for tasks controlling a set of independent systems and a real-time preemptive scheduling policy managing a single CPU processor board, we deal with two problems: 1) deciding whether a performance specification can be attained on a candidate platform, 2) optimising performance on a platform. The considered performance metric is the minimum stability radius attained over the different feedback loops.