Packet processing match and action unit with a VLIW action engine
An embodiment of the invention includes receiving packet header vectors where a header vector includes a number of packet header words. Match operations are performed on the packet header words. At least one packet header word is modified based on the match operations. At least one processor is used for each packet header word to perform the packet match operations and modify at least one packet header word. Instructions are received from an instruction word where a VLIW instruction word includes all of the instruction words. Each processor performs an operation in response to the instruction word.
1. An apparatus for packet processing in a switching network, the apparatus comprising:
processing circuitry coupled to memory arrays and receiving circuitry;
the receiving circuitry operable to receive packet header vectors comprising respective packet header words, wherein the packet header vectors are associated with ternary values that are representative of “1”, “0”, and “don't care”;
the processing circuitry comprising a plurality of processors, including at least first and second processors coupled to a respective first and second end destinations, the first and second processors operable to modify respective first and second packet header words in response to:
at least one packet match operation based on the ternary values; and
at least one very long instruction word (VLIW) that provides an instruction to modify the first and second packet header words without specifying the end destination.
2. The apparatus of claim 1 where memory arrays comprise a separate action instruction memory array and a separate action data memory array.
3. The apparatus of claim 1 wherein instructions are received from multiple flow tables.
4. The apparatus of claim 3 wherein one table in the multiple flow tables is an Ether table.
5. The apparatus of claim 3 wherein one table in the multiple flow tables is a MPLS table.
6. The apparatus of claim 3 wherein one table in the multiple flow tables is a VLAN table.
7. The apparatus of claim 3 wherein one table in the multiple flow tables is a GRE table.
8. The apparatus of claim 3 wherein one table in the multiple flow tables is an exact match table.
9. The apparatus of claim 8 wherein the exact match table comprises a hash table.
10. The apparatus of claim 9 wherein the hash table is implemented with static random access memory.
11. The apparatus of claim 9 wherein the hash table is implemented using a cuckoo hash table.
12. The apparatus of claim 11 wherein the cuckoo hash table is implemented using static random access memory.
13. The apparatus of claim 1 wherein the memory arrays are implemented with ternary content addressable memory.
14. The apparatus of claim 1 wherein the first and second processors operate in parallel.
15. A method for packet processing in a switching network, the method comprising:
receiving packet header vectors wherein each packet header vector comprises a plurality of packet header words, wherein the packet header vectors are associated with ternary values that are representative of “1”, “0”, and “don't care”;
modifying, by a processor associated with an end destination, at least one packet header word in response to:
at least one packet match operation; and
at least one very long instruction word (VLIW) that provides an instruction to modify the at least one packet header word without specifying the end destination.
16. The method of claim 15 wherein the instructions are received from multiple flow tables.