Abstract: A data processing device has a processor (10) operable at selectable ones of a plurality of performance levels. The processor generates a workload data vector indicating a workload of the processor as a function of time. A memory stores a set of reference workload vectors. A pattern matcher (18) detects whether there is a matching reference workload data vector. A performance level controller (19) selects the performance level of the processor based on control information that is stored in combination with the matching workload data vector.

Abstract: A system on a chip comprises a plurality of circuit blocks, a programmable processor and a communication circuit. Design information includes connection data including an identification of the direct mutual connection and first and second circuit blocks coupled by the direct mutual connection. An additional register is added to the system on a chip coupled to the direct mutual connection. Verification programs are used includescomprising instructions for the processor to access registers in the second one of the circuit blocks, to use the connection data, or information derived therefrom to select the first one of the circuit blocks, and to issue the standardized call to the interface program of the selected further one of the circuit blocks.

Abstract: A method for computing the optimal power mode for a system-on-chip (SoC) in which both the clock and Vdd settings are controlled. Information from hardware blocks is synthesized into a global power mode for the entire SoC. The clocks can be disabled or enabled, and Vdd voltages can be disabled, set at a nominal operating level, and set at a retention level in which the state of memory and registers is retained.

Abstract: A system on a chip comprises a plurality of circuit blocks (18), a programmable processor (12) and a communication circuit (16) coupled between the processor (12) and the plurality of circuit blocks (18), the communication circuit (16) being configured to support program controlled access to registers in the circuit blocks (18) from the processor (12), a first and second one of the circuit blocks (18) of the plurality having direct mutual connection (19) for directly passing a signal between the first and second one of the circuit blocks (18) without communication through the communication circuit (12). Design information is used that comprises connection data including an identification of the direct mutual connection (19) and the first and second circuit blocks (18) coupled by the direct mutual connection (19). An additional register is added to the system on a chip coupled to the direct mutual connection (19) to capture and/or control signals at the direct mutual connection (19).

Abstract: A data processing device has a processor (10) operable at selectable ones of a plurality of performance levels. The processor generates a workload data vector indicating a workload of the processor as a function of time. A memory stores a set of reference workload vectors. A pattern matcher (18) detects whether there is a matching reference workload data vector. A performance level controller (19) selects the performance level of the processor based on control information that is stored in combination with the matching workload data vector.

Abstract: A method for computing the optimal power mode for a system-on-chip (SoC) in which both the clock and Vdd settings are controlled. Information from hardware blocks is synthesized into a global power mode for the entire SoC. The clocks can be disabled or enabled, and Vdd voltages can be disabled, set at a nominal operating level, and set at a retention level in which the state of memory and registers is retained.

Abstract: The processing circuit contains a cache management unit which keeps information about a stream of addresses among addresses accessed by the processor. The cache management unit updates a current address for the stream in response to progress of execution of the program. The cache management unit is make selected storage locations in the cache memory available for reuse, a storage location in the cache memory which is in use for the data corresponding to the particular address being made available for reuse dependent on a position of the particular address relative to the current address.