This paper proposes an architecture for concurrent scheduling of hard, soft and non real-time threads in embedded systems. It is based on a superscalar in-order processor binary compatible to the Infineon TriCore. The architecture allows a tight static WCET analysis of hard real-time threads. To provide high performance anyway, the absence of speculative elements like branch prediction and out-of-order execution is compensated by multithreading, transforming the processor into an in-order SMT processor. The Priority Controller that manages the scheduling is able (1) to assign fixed portions of time to hard real-time threads, (2) to control the IPC of soft real-time threads and (3) to fairly distribute execution cycles to non real-time threads. It is located within a separate unit outside the pipeline to avoid prolonging the critical path. We evaluate the processor using the EEMBC automotive benchmarks and show that the overlapping of two soft real-time threads can be used to either reduce the clock rate by 23% or to grant each thread 65% of its single-threaded IPC. Even if a hard real-time thread is executed predominantly, the remaining resources can be used by concurrent soft real-time threads which reach a performance of 70% compared to their single-threaded execution.