Fast decoding of compressed data
Techniques described herein relate to methods for decompressing a compressed data stream. An example method includes reading groups of data from the compressed data stream and, for each group of data read from the compressed data stream, identifying symbols in the group of data, the symbols having a type corresponding to Literal bytes or Copy bytes of data in the compressed data stream, processing each symbol in one of a plurality of processing stream to produce decompressed data bytes, the processing stream used to process the symbol being selected according to the symbol's type, and merging the uncompressed data bytes from the different processing streams into a decompressed output data stream, where the uncompressed data bytes from the different processing streams are merged to preserve their order corresponding to an order or corresponding symbols in the compressed data stream.
1 . A method of decompressing a compressed data stream, the method comprising:
reading groups of data from the compressed data stream;
for each group of data read from the compressed data stream:
identifying symbols in the group of data, the symbols having a type corresponding to Literal bytes of data or Copy bytes of data in the compressed data stream;
processing each symbol in one of a plurality of processing stream to produce decompressed data bytes, wherein the processing stream is used to process the symbol is selected according to the symbol's type; and
merging the uncompressed data bytes from different processing streams of the plurality of processing streams into a decompressed output data stream having uncompressed data bytes, wherein the uncompressed data bytes from the different processing streams are merged to preserve their order corresponding to an order of corresponding symbols in the compressed data stream.
2 . The method of claim 1 , wherein the groups of data are read from the compressed data stream in a fixed-size load.
3 . The method of any one of claim 1 , further comprising identifying a tag mask corresponding to the fixed-size load.
4 . The method of claim 3 , wherein each bit of the tag mask identifies whether a corresponding byte of compressed data identifies a Literal byte of compressed data or a Copy byte of compressed data.
5 . The method of claim 4 , further comprising, for Copy data having an offset value above a first threshold value, reading the data out from a remote memory.
6 . The method of claim 5 , further comprising, for Copy data having an offset value below the first threshold value but above a second threshold value, reading the data out from an on-chip cache memory.
7 . The method of claim 6 , further comprising, for Copy data having an offset value below the second threshold value, reading the data may be read out of a data pipeline used by the processor that performs the decompression of the received data.
8 . The method of claim 4 , further comprising, for Literal data, reading the data out from a processor.