摘要:
A computing device includes a hardware data processing unit having at least one input buffer, a plurality of output buffers, a data transfer unit, and a software control unit, the data transfer unit configured to transfer data from the input buffer to the plurality of output buffers, and the software control unit configured to control the data transfer unit.
摘要:
In order to be able to use a smaller routing table (4) and, thus, to reduce the costs and power consumption and to improve the performance of an IP router, it is proposed to extract a destination address identifier (ADR) from a data packet to be forwarded by the IP router, compress the extracted destination address identifier (ADR) by using a lossless data compression algorithm, and compare the compressed destination address identifier with entries stored in the routing table (4) so as to find a correspondence between the destination address identifier and one of the entries of the routing table (4). Each entry of the routing table (4) corresponds to a possible or available forwarding address of the IP router, the forwarding addresses having been compressed with the same data compression algorithm as the destination address identifier. After having found a correspondence between the destination address identifier and one of the compressed forwarding addresses stored in the routing table (4), a switch (6) of the IP router switches the respective data packet to one of its output links (OUT) which is associated with the respective forwarding address matching the destination address identifier (ADR).
摘要:
The invention relates to a heterogeneous parallel multithread processor (1) with shared contexts which has a plurality (M) of parallel-connected standard processor root unit types (2p; pε[1, . . . , M]), where each respective standard processor root unit type (2p) has at least one or more (K) parallel-connected standard processor root units (2pq; qε[1, . . . , K]) for instruction execution of program instructions from various threads (T), each standard processor root unit type (2p) having N local context memories (32pt) which each buffer-store part of a current processor state for a thread. The multithread processor (1) also has a plurality (N) of global context memories (3t; tε[1, . . . , N]) which each buffer-store part of a current processor state for a thread, and a thread control unit (4) which can connect any standard processor root unit (2pq) to any global context memory (3t).
摘要:
A method for compressing a data packet is proposed, the data packet comprising at least a first data block and a second data block, the first data block referring to the second data block. In the method, the second data block is compressed and it is noted in the data packet that the second data block has been compressed. In one embodiment, the method is suitable for IPv6 data packets, the second data block then being, for example, a routing header.
摘要:
The invention relates to a heterogeneous parallel multithread processor (1) with shared contexts which has a plurality (M) of parallel-connected standard processor root unit types (2p; p∈ [1, . . . , M]), where each respective standard processor root unit type (2p) has at least one or more (K) parallel-connected standard processor root units (2pq; q∈ [1, . . . , K]) for instruction execution of program instructions from various threads (T), each standard processor root unit type (2p) having N local context memories (32pt) which each buffer-store part of a current processor state for a thread. The multithread processor (1) also has a plurality (N) of global context memories (3t; t∈ [1, . . . , N]) which each buffer-store part of a current processor state for a thread, and a thread control unit (4) which can connect any standard processor root unit (2pq) to any global context memory (3t).
摘要:
A processor comprises checking and control devices, first register, and register bank. The control device checks a condition or a subcondition of the condition within a first time unit based on a first subcondition checked within a second time unit preceding the first time unit, a second subcondition checked within a third time unit preceding the second time unit, and a single condition. The first register is coupled to the control device for storing the checked condition and the output of the first register is coupled to the control device for providing the stored, checked subcondition as a checked, first subcondition. The input of the register bank is coupled to the first register for receiving the stored, checked subcondition, the second register stores the received, checked subcondition as the checked, second subcondition, and the register bank is coupled to the control device for providing the checked, second subcondition.
摘要:
The present invention relates to a parallel multithread processor (1) with split contexts, with M parallel-connected standard processor root units (2) being provided for instruction execution of program instructions for different threads (T), and with N context memories (3) being provided, which each temporarily store a current state of a thread (T), and with a thread monitoring unit (4) being provided, by means of which each standard processor root unit (2) can be connected to each context memory (3). The invention accordingly provides a processor architecture in which a number N of different context memories (3) and corresponding threads (T) are effectively fully networked with a number M of standard processor root units (2). This means that use is made not only of paralleling of the standard processor root units (2), but also of the threads (T) and of the context memories (3).
摘要:
In a method for forwarding data packets in a network a circuit comprises a data storage and a control device. Each data packet has a destination address and the data storage comprises T data sub-storages for storing all network addresses which are coded by a particular coding method on the basis of their respective key bit length in precisely one of the data sub-storages. The t-th data sub-storage is divided into blocks having D data elements of identical data element bit length, where tε[1, . . . , T] and D is the smallest common multiple of tε[1, . . . , T]. The control device receives data packets, each having a destination address, codes a destination address of a received data packet using the particular coding method to produce a coded key, and compares the coded key on the basis of its key bit length with the coded addresses stored in the corresponding data sub-store block by block in order to forward the respective data packet via an output to the destination address whose associated coded address matches the coded key from the respective received data packet.
摘要:
The present invention relates to a parallel multithread processor (1) with split contexts, with M parallel-connected standard processor root units (2) being provided for instruction execution of program instructions for different threads (T), and with N context memories (3) being provided, which each temporarily store a current state of a thread (T), and with a thread monitoring unit (4) being provided, by means of which each standard processor root unit (2) can be connected to each context memory (3). The invention accordingly provides a processor architecture in which a number N of different context memories (3) and corresponding threads (T) are effectively fully networked with a number M of standard processor root units (2). This means that use is made not only of paralleling of the standard processor root units (2), but also of the threads (T) and of the context memories (3).
摘要:
A data moving processor includes a code memory coupled to a code fetch circuit and a decode circuit coupled to the code fetch circuit. An address stack is coupled to the decode circuit and configured to store address data. A general purpose stack is coupled to the decode circuit and configured to store other data. The data moving processor uses data from the general purpose stack to perform calculations. The data moving processor uses address data from the address stack to identify source and destination memory locations. The address data may be used to drive an address line of a memory during a read or write operation. The address stack and general purpose stack are separately controlled using bytecode.