This paper deals with real-time control under computational constraints. A robust control approach to control/real-time scheduling co-design is proposed using the $H_\infty$ framework for Linear Parameter Varying (LPV) polytopic systems. The originality consists in a new resource sharing between control tasks according to the controlled plant performances. Here the varying parameters are images of the control performance w.r.t. the sampling frequencies. Then a LPV based feedback scheduler is designed to adapt the control tasks periods according to the plant behavior and to the availability of computing resources. The approach is illustrated with a robot-arm controller design, whose feasibility is assessed in simulation.