Device for processing a stream of data words
View Patent ↗State of the art processor systems, esp. in embedded systems, are not able to process data under real-time conditions especially with throughput rates near 10 Gbps. So, when using interfaces like PCI Express (PCIe) or Infiniband or 10 G-Ethernet for 10 Gbps data throughput, special data-paths have to process the high throughput rate data. But tasks like connection management or time uncritical control messaging are better manageable by a processor. According to the invention it is proposed a kind of multiplexer architecture that is needed to split between control and data-path access for a PCI Express based architecture.
1. Device for processing a stream of data words received from a data source at one input, wherein the data words are buffered in a first FIFO memory, further comprising an interface for outputting the data words in completion packets of a defined format, and comprising a processor for generating request packets for controlling an external device via the interface, wherein the request packets are buffered in a second FIFO memory, wherein said completion packets and request packets are transferred to the interface via the same data lines, further comprising a multiplexer unit, that mixes completion packets and request packets for output via the interface in a manner according to a priority scheme, where a currently being transferred completion packet with a preset completion packet length is continued to be sent until the last byte of the currently being transferred completion packet is sent and a request packet ready to be sent is transferred with high priority over the bus, thereby interrupting the transfer of further completion packets ready to be sent for the time needed to send the request packet ready to be sent.
2. Device according to claim 1 , wherein the first FIFO memory is adapted to provide a first packet indication signal to the multiplexer unit, in case the number of collected data words reaches a threshold corresponding to the preset completion packet length.
3. Device according to claim 1 , wherein the second FIFO memory is adapted to provide a second packet indication signal to the multiplexer unit, in case the number of collected data words reaches a threshold corresponding to a preset request packet length.
4. Device according to claim 1 , wherein the interface is a PCI Express interface and the multiplexer unit comprises a completion header generator for generating a PCI Express completion header for the completion packets to be sent to the interface.
5. Device according to claim 4 , wherein the multiplexer unit is equipped with a PCI Express completion header generator and a completion packet and request packet multiplexer.
6. Device according to claim 4 , wherein the multiplexer unit comprises a separation unit that separates requests for sending completion packets from request packet responses and directs the requests for sending completion packets to a finite state machine for building corresponding completion packets on one hand and the request packets responses to the processor on the other hand.
7. Device according to claim 6 , wherein the separation unit includes two correspondingly adapted address and transfer type decoders.
8. Device according to claim 1 , wherein the multiplexer unit is equipped with a finite state machine that is adapted to work according the rule that if the second indication signal is detected, the state machine waits for the indication that the last data word from a packet in the first FIFO memory has been processed in the multiplexer before it starts to transfer the request packet from the second FIFO memory.