In a device comprising a serial bus and a plurality of devices, register/address mappings and/or unique group identifiers are used to convey additional information in messages/datagrams over the serial bus without explicitly sending such information in the message/datagram. Such register/address mappings may be done beforehand, and in conjunction with group-specific identifiers, may reduce transmission latency by keeping the size of the messages/datagrams small. Since all devices on the serial bus have prior knowledge of such register/address mappings and/or group-specific identifiers, recipient devices are able to infer information from the group-specific identifiers and/or register/address sent in each message/datagram that is not explicitly sent within such message/datagram.

Provided is a multi-master collision prevention system including: a plurality of functional blocks including a plurality of external modules and a plurality of internal modules performing different functions; a plurality of interfaces respectively connected to the plurality of external modules, respectively; a plurality of dedicated registers including priority information of the plurality of functional blocks and connected to the plurality of functional blocks, respectively; a common block selectively connected to the plurality of functional blocks, and configured to function as a master for controlling the common blocks when the plurality of functional blocks are connected to the common block; and a priority determination unit configured to determine a connection between any one of the plurality of functional blocks and the common block.

A crossbar switch comprises two or more data inputs 10, two or more data outputs 100, a buffer 30 between the inputs and the outputs, an arbiter 52 associated with each output and configured to select data from one of the inputs when there is contention at the output, a bypass 32 associated with the buffer so that the buffer can be enabled or disabled, and a buffer controller 60 configured to enable or disable the buffer. The buffer controller further includes an accumulator 70 configured to assess whether a time-based average of the contention rate, or an average injection rate, at the output associated with the buffer, has reached a predetermined threshold. This prevents the buffer being enabled when the contention is only intermittent, which reduces power consumption without significant loss of performance.

Method and system are disclosed for arbitration of parallel multi-event processing. In one embodiment, a parallel multi-event processing system includes a plurality of hardware components, where each hardware component in the plurality of hardware components is assigned with a unique range of addresses, a plurality of hardware engines, where the plurality of hardware engines are configured to access the plurality of hardware components, a controller configured to perform arbitration on one or more requested transactions among the plurality of hardware engines and the plurality of hardware components based on one or more hardware components in the plurality of hardware components to be accessed, and the plurality of hardware components, the plurality of hardware engines, and the controller are configured to perform the one or more requested transactions according to the arbitration.

In an example, there is disclosed a computing apparatus, having: a first master having a first ordinal quality of service (QoS) profile; a second master having a second ordinal QoS profile, wherein the second ordinal QoS profile is higher in order than the first ordinal QoS profile; a slave; a multiplexed interconnect to communicatively couple the first master and second master to the slave with a priority according to the ordinal QoS profiles; and one or more logic elements, including at least one hardware logic element, providing a QoS engine to: determine that the first master has initiated a slave operation via the interconnect; determine that completing the slave operation according to a QoS criterion provided by the second master requires elevated QoS; and promote the first master to a third ordinal QoS profile having an order higher than the second ordinal QoS profile.

A system-on-chip (SoC) to perform a deadlock control on a processor of the SoC, the SoC including the processor including a plurality of central processing unit (CPU) cores, a first bus connected to the processor, a graphic processing unit (GPU) connected to the first bus, a memory controller connected to the first bus, a second bus connected to the processor, an isolation cell including a logic circuit configured to retain a signal value input to the processor according to an isolation signal, and a deadlock controller connected to the first bus and the second bus. The deadlock controller is configured to isolate the processor, which is in a deadlock state, from the first bus by applying the isolation signal on the isolation cell, and to extract, via the second bus, state information of the isolated processor in the deadlock state.

A system-on-chip (SoC) to perform a deadlock control on a processor of the SoC, the SoC including the processor including a plurality of central processing unit (CPU) cores, a first bus connected to the processor, a graphic processing unit (GPU) connected to the first bus, a memory controller connected to the first bus, a second bus connected to the processor, an isolation cell including a logic circuit configured to retain a signal value input to the processor according to an isolation signal, and a deadlock controller connected to the first bus and the second bus. The deadlock controller is configured to isolate the processor, which is in a deadlock state, from the first bus by applying the isolation signal on the isolation cell, and to extract, via the second bus, state information of the isolated processor in the deadlock state.

Method and apparatus for implementing an optimized credit return mechanism for packet sends. A Programmed Input/Output (PIO) send memory is partitioned into a plurality of send contexts, each comprising a memory buffer including a plurality of send blocks configured to store packet data. A storage scheme using FIFO semantics is implemented with each send block associated with a respective FIFO slot. In response to receiving packet data written to the send blocks and detecting the data in those send blocks has egressed from a send context, corresponding freed FIFO slots are detected, and a lowest slot for which credit return indicia has not be returned is determined. The highest slot in a sequence of freed slots from the lowest slot is then determined, and corresponding credit return indicia is returned. In one embodiment an absolute credit return count is implemented for each send context, with an associated absolute credit sent count tracked via software that writes to the PIO send memory, with the two absolute credit counts used for flow control.

In an example, there is disclosed a computing apparatus, having: a first master having a first ordinal quality of service (QoS) profile; a second master having a second ordinal QoS profile, wherein the second ordinal QoS profile is higher in order than the first ordinal QoS profile; a slave; a multiplexed interconnect to communicatively couple the first master and second master to the slave with a priority according to the ordinal QoS profiles; and one or more logic elements, including at least one hardware logic element, providing a QoS engine to: determine that the first master has initiated a slave operation via the interconnect; determine that completing the slave operation according to a QoS criterion provided by the second master requires elevated QoS; and promote the first master to a third ordinal QoS profile having an order higher than the second ordinal QoS profile.

Method and apparatus for implementing an optimized credit return mechanism for packet sends. A Programmed Input/Output (PIO) send memory is partitioned into a plurality of send contexts, each comprising a memory buffer including a plurality of send blocks configured to store packet data. A storage scheme using FIFO semantics is implemented with each send block associated with a respective FIFO slot. In response to receiving packet data written to the send blocks and detecting the data in those send blocks has egressed from a send context, corresponding freed FIFO slots are detected, and a lowest slot for which credit return indicia has not be returned is determined. The highest slot in a sequence of freed slots from the lowest slot is then determined, and corresponding credit return indicia is returned. In one embodiment an absolute credit return count is implemented for each send context, with an associated absolute credit sent count tracked via software that writes to the PIO send memory, with the two absolute credit counts used for flow control.