Abstract: Transmit and receive data are stored in buffer regions defined in first and second memories of a system residing on a network. The buffer regions are pointed to by multiple descriptor rings that are also stored in the memories. In accordance with one aspect of the invention, the two memories reside on separate busses connected to a common buss containing a processor. The processor communicates with one or the other of the memories using a handshaking protocol. In accordance with another aspect of the invention, receive data incoming to the system is scattered among multiple descriptor rings. A further aspect splits a frame among multiple descriptors depending on a characteristic code carried by the frame, e.g., in a frame control field. The size of the first descriptor, smaller than that of the others, is programmed to correspond to the size of the header of each frame. Synchronization between headers and data of a frame is maintained by a frame number stored in each descriptor.

Abstract: A network interface interconnects a processor and a memory residing on a system bus having a predetermined average bus latency to a network containing other processors and memories. The interface, which is implemented using a bus master architecture, includes a network access controller for accessing data on the network, a random access memory, preferably in the form of an SRAM as a buffer between the network and system bus, and a network DMA controller. The Network Access Controller configures the buffer to have at least one logical FIFO for storing data; the size of the logical FIFO is related to the predetermined average bus latency. The DMA controller further controls transfer of data between the system memory and the network through the logical FIFOs. Preferably, the buffer is configured to have (1) a first logical FIFO for storing incoming data from the network and (2) at least one additional logical FIFO for storing outgoing data from said system memory to the network.

Abstract: Data arranged in packets are transferred between a system memory and a network bus through a SRAM configured by software pointers as first in-first out memories for transmitting (transmit FIFO) and for receiving (receive FIFO). The packets of data stored in the transmit and receive FIFOs are demarked from each other and classified by tag and status bits at the end of the last word of each packet. Data to be transmitted on the network bus is transferred from the system memory to the transmit FIFO, and data received from the network is stored in the receive FIFO. To maximize data throughput, when either at least a predetermined amount of data or a complete packet is stored in the transmit FIFO, the data is transmitted to the network while data is being received from the system memory. When at least a predetermined amount of data is stored in the receive FIFO, data is transferred to the system memory while network data is incoming from the network.

Abstract: Transmit and receive data are stored in buffer regions defined in first and second memories of a system residing on a network. The buffer regions are pointed to by multiple descriptor rings that are also stored in the memories. In accordance with one aspect of the invention, the two memories reside on separate buses connected to a common bus containing a processor. The processor communicates with one or the other of the memories using a handshaking protocol. In accordance with another aspect of the invention, receive data incoming to the systems is scattered among multiple descriptor rings. A further aspect splits a frame among multiple descriptors depending on a characteristic code carried by the frame, e.g., in a frame control field. The size of the first descriptor, smaller than that of the others, is programmed to correspond to the size of the header of each frame. Synchronization between headers and data of a frame is maintained by a frame number stored in each descriptor.

Abstract: The standard FDDI priority algorithm is implemented by programming Token Holding Time (THT) threshold values for asynchronous service either in an increasing or decreasing order as a function of token holding time. If the thresholds are programmed in a decreasing order, all higher priority data is sent to the network before any lower priority data is sent. If the thresholds are programmed in an increasing order, highest priority data is sent first, until the unexpired token holding time falls below the threshold value for that priority; the next lower priority level data then is transmitted, and so on. Accordingly, at least some data of all priority assignments pending for transmission are sent to the medium during each token capture.

Abstract: An FDDI system and method for transmitting to an optical medium, upon receipt of token, frames of synchronous data and frames of asynchronous data having different levels of priority. The network on which the FDDI is implemented includes a plurality of processors each having a system for storing the frames of data in queues corresponding to priority, and an output buffer configured to have a plurality of logical FIFOs corresponding to the queues. Data is transferred one queue at a time from the system memory to the output buffer through a single physical FIFO. To prevent the FIFO from "locking-up" as a result of any residual data remaining therein following each transfer of a frame to the output buffer, storage remaining available for a particular queue of the output buffer to be transmitted to the medium is detected. Data is transferred from the system memory to the FIFO memory only if the storage remaining available is at least equal to the storage capacity of the FIFO memory.