High throughput decoding of variable length data symbols
View Patent ↗A method of decoding data includes: receiving an encoded data stream transmitted as a plurality of variable length symbols; dividing the data stream into a sequence of blocks, each block having a sequence of adjacent bit positions starting a first bit position and ending with a last bit position; pre-processing each block prior to fully decoding each block, wherein pre-processing includes, for each block, selecting a bit position in a current block and determining a starting position of a first symbol in an adjacent block based on the selected bit position, wherein determining is initially performed for the last bit position in the current block, and is repeated sequentially for each preceding bit position through and including the first bit position; and fully decoding each block by decoding a first block starting at the first bit position and decoding each adjacent block starting at the starting position.
1. A method of decoding data, comprising:
receiving a data stream, the data stream encoded by an encoding algorithm and transmitted as a plurality of variable length symbols;
dividing the data stream into a sequence of blocks, each block having a sequence of adjacent bit positions starting with a first bit position and ending with a last bit position;
pre-processing each block prior to fully decoding each block, wherein pre-processing includes, for each block, selecting a bit position in a current block and determining a starting position of a first symbol in an adjacent block based on the selected bit position, wherein determining is initially performed for the last bit position in the current block, and is repeated sequentially for each preceding bit position through and including the first bit position; and
fully decoding each block by decoding a first block starting at the first bit position and decoding each adjacent block starting at the starting position.
2. The method of claim 1 , wherein the encoding algorithm is a compression algorithm, and decoding includes expanding data represented by each symbol to recreate an original data stream.
3. The method of claim 1 , wherein determining the starting position includes identifying a symbol and determining the starting position based on a length of the symbol.
4. The method of claim 1 , wherein determining the starting position includes storing the starting position in a table indexed to the current block and the selected bit position.
5. The method of claim 1 , wherein fully decoding each block includes feeding each block into a pipeline to decode the block beginning at a starting position determined from at least one look-up table.
6. The method of claim 4 , wherein determining the starting position includes:
determining the length of the identified symbol that would begin at the selected bit position;
calculating a position of a following symbol by adding the length of the identified symbol to the selected bit position.
7. The method of claim 6 , wherein calculating the starting position includes:
responsive to the position of the following symbol being in the adjacent block, recording the position as the starting position in the table; and
responsive to the position of the following symbol being in the selected block, reading the starting position of the first symbol in the next block from an entry in the table corresponding to the position, and recording the starting position in the table.
8. The method of claim 1 , wherein two or more of the blocks are pre-processed in parallel.
9. The method of claim 1 , wherein the plurality of variable length symbols includes symbols from different alphabets, and determining the starting position includes:
determining for each alphabet the length of an identified symbol that would begin at the selected position;
calculating the starting position associated with the first symbol in the next block based on the length and the alphabet of the identified symbol.
10. The method of claim 9 , wherein calculating the starting position includes storing the starting position in a table indexed to the selected bit position and the alphabet of the identified symbol.