Abstract: Some embodiments provide a method for a parser of a processing pipeline. The method receives a packet for processing by a set of match-action stages of the processing pipeline. The method stores packet header field (PHF) values from a first set of PHFs of the packet in a set of data containers. The first set of PHFs are for use by the match-action stages. For a second set of PHFs not used by the match-action stages, the method generates descriptive data that identifies locations of the PHFs of the second set within the packet. The method sends (i) the set of data containers to the match-action stages and (ii) the packet data and the generated descriptive data outside of the match-action stages to a deparser that uses the packet data, generated descriptive data, and the set of data containers as modified by the match-action stages to reconstruct a modified packet.

Abstract: Some embodiments provide a method for a parser of a processing pipeline. The method receives a packet for processing by a set of match-action stages of the processing pipeline. The method stores packet header field (PHF) values from a first set of PHFs of the packet in a set of data containers. The first set of PHFs are for use by the match-action stages. For a second set of PHFs not used by the match-action stages, the method generates descriptive data that identifies locations of the PHFs of the second set within the packet. The method sends (i) the set of data containers to the match-action stages and (ii) the packet data and the generated descriptive data outside of the match-action stages to a deparser that uses the packet data, generated descriptive data, and the set of data containers as modified by the match-action stages to reconstruct a modified packet.

Abstract: Some embodiments provide a method for a parser of a processing pipeline. The method receives a packet for processing by a set of match-action stages of the processing pipeline. The method stores packet header field (PHF) values from a first set of PHFs of the packet in a set of data containers. The first set of PHFs are for use by the match-action stages. For a second set of PHFs not used by the match-action stages, the method generates descriptive data that identifies locations of the PHFs of the second set within the packet. The method sends (i) the set of data containers to the match-action stages and (ii) the packet data and the generated descriptive data outside of the match-action stages to a deparser that uses the packet data, generated descriptive data, and the set of data containers as modified by the match-action stages to reconstruct a modified packet.

Abstract: Some embodiments provide a method for a hardware forwarding element deparser. The method receives, from a match-action pipeline, (i) packet header field values stored in a set of data containers and (ii) a set of data indicating which packet header fields, of multiple possible packet header fields, to include in a packet constructed from the packet header field values. The method uses the received set of data and a list of data container identifiers for multiple possible packet header fields to generate an ordered list of references to data containers of the set of data containers. Based on the ordered list, the method constructs the packet using the packet header field values stored in the referenced data containers.

Abstract: Some embodiments provide a method for a parser of a processing pipeline. The method receives a packet for processing by a set of match-action stages of the processing pipeline. The method stores packet header field (PHF) values from a first set of PHFs of the packet in a set of data containers. The first set of PHFs are for use by the match-action stages. For a second set of PHFs not used by the match-action stages, the method generates descriptive data that identifies locations of the PHFs of the second set within the packet. The method sends (i) the set of data containers to the match-action stages and (ii) the packet data and the generated descriptive data outside of the match-action stages to a deparser that uses the packet data, generated descriptive data, and the set of data containers as modified by the match-action stages to reconstruct a modified packet.

Abstract: Some embodiments provide a method for a deparser of a processing pipeline. The method receives, from a set of match-action stages of the pipeline, packet header field (PHF) values for a first set of PHFs of a packet processed by the match-action stages. The method also receives, directly from a parser of the pipeline, (i) packet data for the packet prior to any modification by the match-action stages and (ii) descriptive data that specifies locations within the packet data for a second set of PHFs of the packet that are not included in the first set of PHFs. The method constructs a packet from (i) the PHF values received for the first set of PHFs and (ii) the packet data received for the second set of PHFs. The descriptive data is used to extract packet header field values for the second set of PHFs from the packet data.

Abstract: A method of operating a first host within a network of a plurality of hosts. Over the course of a first time interval, requests received at the first host from a second host to send data to destination hosts are monitored. Identities of destination hosts monitored during the first time interval are compared with destination host identities in a record. Then, either data relating to requests which identify a destination host not in the record are stored in a storage buffer. Or the passage of data from the second host to the destination host within the network is limited over the course of the first time interval, so that during the first time interval the second host is unable to send data to more than a predetermined number of hosts not in the record.

Abstract: Many modern microprocessors support parallel processing operations, such as single instruction multiple data (SIMD) operations. The present invention presents a number of ways in which maximum advantage can be taken of these operations to provide an efficient way of processing multiple data channels.

Abstract: A method of operating a first host within a network of a plurality of hosts. Over the course of a first time interval, requests received at the first host from a second host to send data to destination hosts are monitored. Identities of destination hosts monitored during the first time interval are compared with destination host identities in a record. Then, either data relating to requests which identify a destination host not in the record are stored in a storage buffer. Or the passage of data from the second host to the destination host within the network is limited over the course of the first time interval, so that during the first time interval the second host is unable to send data to more than a predetermined number of hosts not in the record.

Abstract: Many modern microprocessors support parallel processing operations, such as single instruction multiple data (SIMD) operations. The present invention presents a number of ways in which maximum advantage can be taken of these operations to provide an efficient way of processing multiple data channels.

Abstract: A data packet switch includes a decision tree for classifying data packets, which can be dynamically modified. To conserve memory resources nodes which are found during modification to have matching effects are combined. If only a subset of paths to a node are relevant to a modification, the node is split. Prior to implementation of the modifications, temporary nodes are inserted before modified nodes to preserve existing paths. These temporary nodes are controlled by a single memory value which can be changed to effect all the modifications to the decision tree simultaneously.

Abstract: A hashing method and apparatus uses a hash function that can be modified in real time by a hash control code. The hash function involves the combining together of multiple bit-shifted versions of a multi-bit input to produce a transformed value from which the hash output is formed. The hash control code is used to set the number of input versions used to produce the transformed value and their respective degrees of bit-shifting. The hashing method and apparatus may be used in executing processor branch instructions where the identity of an item to be accessed occupies a search space that varies in size and degree of population between different branch instructions.

Abstract: A data packet switch includes a decision tree for classifying data packets, which can be dynamically modified. To conserve memory resources nodes which are found during modification to have matching effects are combined. If only a subset of paths to a node are relevant to a modification, the node is split. Prior to implementation of the modifications, temporary nodes are inserted before modified nodes to preserve existing paths. These temporary nodes are controlled by a single memory value which can be changed to effect all the modifications to the decision tree simultaneously.

Abstract: A program-controlling processing unit executes instructions stored in memory. A special instruction type is provided for selectively retrieving an element from memory in dependence on the value of input data subject of the instruction. Each instruction of this type has a header identifying the instruction type, and a body in the form of a hash table having at least one entry with both (i) a check value corresponding to a value of interest of the input data subject of the instruction, and (ii) an element to be used when the input data has said value of interest. Upon execution of such an instruction, the related input data is hashed to produce an offset value that is used to access in memory, relative to the position in memory of the current instruction, a corresponding entry in the hash table of the instruction. If a “hit” results (that is, the check value of the accessed entry matches the input data value), the entry element is operatively output.