Pattern matching
View Patent ↗A method for processing data for pattern matching includes: receiving a first sequence of data values; and generating a second sequence of data values based on the first sequence and one or more patterns and history of data values in the first sequence, wherein the second sequence has fewer data values than the first sequence and all subsequences in the first sequence that match at least one of the one or more patterns are represented in the second sequence.
1. A method for processing data for pattern matching, the method comprising:
receiving a first sequence of data values; and
generating, by at least one processor, a second sequence of data values based on the first sequence, one or more patterns, and a history of data values in the first sequence, with the second sequence having fewer data values than the first sequence and all subsequences in the first sequence that match at least one of the one or more patterns being represented in the second sequence, with generating the second sequence of data values comprising
generating data values according to an associated buffer state corresponding to at least some of the first sequence of data values, the buffer state corresponding to fewer than or equal to a maximum number of data values, and
outputting at least one of the data values corresponding to the buffer state to prevent overflow of the buffer state beyond the maximum number of data values.
2. The method of claim 1 , wherein generating the second sequence comprises replacing an original subsequence of data values in the first sequence with a shorter subsequence of data values that corresponds to the same transition vector of a pattern-matching finite state machine associated with the one or more patterns as the original subsequence.
3. The method of claim 1 , wherein the subsequences in the first sequence that match at least one of the one or more patterns are included in the second sequence.
4. The method of claim 1 , wherein a given subsequence in the first sequence that matches at least one of the one or more patterns is represented in the second sequence by one or more data values that correspond to the same transition vector of a pattern-matching finite state machine associated with the one or more patterns as the given subsequence.
5. The method of claim 4 , wherein the given subsequence in the first sequence that matches at least one of the one or more patterns is represented in the second sequence by a single character that correspond to the same transition vector of the pattern-matching finite state machine as the given subsequence.
6. The method of claim 1 , wherein the buffer state corresponds to data values stored in a buffer.
7. The method of claim 1 , wherein the buffer state corresponds to data values associated with a state of a finite-state machine.
8. The method of claim 1 , wherein generating the second sequence of data values comprises evaluating data values corresponding to the buffer state for possible reduction.
9. The method of claim 1 , wherein generating the second sequence of data values comprises processing the first sequence according to a finite-state machine.
10. The method of claim 9 , wherein the number of states in the finite-state machine are determined according to an amount of compression of the second sequence relative to the first sequence.
11. The method of claim 1 , wherein at least some of the patterns are specified by a regular expression.
12. A system, comprising:
a plurality of processor cores interconnected by an interconnection network; and
information for configuring the system to execute instructions to
receive a first sequence of data values,
generate a second sequence of data values based on the first sequence, one or more patterns, and a history of data values in the first sequence, with the second sequence having fewer data values than the first sequence and all subsequences in the first sequence that match at least one of the one or more patterns being represented in the second sequence, with generating the second sequence of data values comprising
generating data values according to an associated buffer state corresponding to at least some of the first sequence of data values, the buffer state corresponding to fewer than or equal to a maximum number of data values, and
outputting at least one of the data values corresponding to the buffer state to prevent overflow of the buffer state beyond the maximum number of data values, and
match the second sequence of data values to the one or more patterns in a plurality of processor cores.
13. The system of claim 12 , further comprising a memory for storing the information for configuring the system.
14. The system of claim 12 , wherein generating the second sequence comprises replacing an original subsequence of data values in the first sequence with a shorter subsequence of data values that corresponds to the same transition vector of a pattern-matching finite state machine associated with the one or more patterns as the original subsequence.
15. The system of claim 12 , wherein the subsequences in the first sequence that match at least one of the one or more patterns are included in the second sequence.
16. The system of claim 12 , wherein a given subsequence in the first sequence that matches at least one of the one or more patterns is represented in the second sequence by one or more data values that correspond to the same transition vector of a pattern-matching finite state machine associated with the one or more patterns as the given subsequence.
17. The system of claim 12 , wherein the interconnection network comprises a two-dimensional network.
18. The system of claim 12 , wherein the interconnection network comprises a bus network, a ring network, a mesh network, or a crossbar switch network.
19. The system of claim 12 , wherein each of the plurality of processor cores corresponds to a tile on an integrated circuit, each tile comprising:
a computation unit; and
a switch including switching circuitry to forward data received over data paths of the interconnection network from other tiles to the computation unit and to switches of other tiles, and to forward data received from the computation unit to switches of other tiles.
20. The system of claim 19 , wherein the computation unit comprises a pipelined processor and the switch is coupled to a plurality of pipeline stages of the pipelined processor.
21. The system of claim 20 , wherein at least one port of the switch is mapped to a register name space of the pipelined processor.
22. The method of claim 1 , further comprising performing pattern matching to match the second sequence of data values to the one or more patterns.
23. The system of claim 12 , wherein at least a first processor core is configured to break down the first sequence of data values into segments and distribute the segments to different respective processor cores including a second processor core configured to generate the second sequence of data values.
24. The system of claim 23 , wherein the first of the processor cores is configured to distribute the segments based on respective sets of patterns to be matched.
25. The system of claim 23 , wherein the different respective processor cores are configured to provide data values to the plurality of processor cores configured to match the second sequence of data values to the one or more patterns.
26. The system of claim 25 , wherein at least a third processor core is configured to collect outputs of multiple of the plurality of processor cores configured to match the second sequence of data values to the one or more patterns.
27. A method for processing data for pattern matching in an integrated circuit comprising a plurality of processor cores interconnected by an interconnection network, the method comprising:
receiving a first sequence of data values;
breaking down the first sequence of data values into segments in at least one processor core and distributing the segments to different respective compressor processor cores;
for each of the compressor processor cores, generating a second sequence of data values based on the first sequence, one or more patterns, and history of data values in the first sequence, with the second sequence having fewer data values than the first sequence and all subsequences in the first sequence that match at least one of the one or more patterns being represented in the second sequence, with generating the second sequence of data values comprising
generating data values according to an associated buffer state corresponding to at least some of the first sequence of data values, the buffer state corresponding to fewer than or equal to a maximum number of data values, and
outputting at least one of the data values corresponding to the buffer state to prevent overflow of the buffer state beyond the maximum number of data values;
for each of multiple pattern matching processor cores coupled to a respective compressor processor core, performing pattern matching on the second sequence of data values to match the second sequence of data values to the one or more patterns of the respective compressor processor core; and
collecting outputs of multiple of the pattern matching processor cores in at least one processor core.