The present invention generally relates to compressing and decompressing a data stream, and more particularly to a method for compressing and decompressing a data stream that is more easily implemented in hardware.
Data compression and decompression techniques remove redundancy in a data stream to provide more efficient utilization of memory and communication channel resources. Lossless compression methods are used for data streams that require the original data to be exactly reconstructed from the compressed data stream without any loss of information. One category of lossless compression algorithms, generally termed LZ compression, relies on substitution of a code for a frequently repeating data pattern.
One group of LZ compression algorithms dynamically creates a table of frequently encountered data patterns as the data stream is compressed. Code sequences indicating the location in the table of the frequently encountered data patterns are generated and embedded in the data stream. The table is added by the compression system as part of the compressed data stream for use by the decompression system to decompress the compressed data stream.
Another group of LZ compression algorithms does not add a table of frequently repeated sequences to the data stream. Instead, the decompression system dynamically recreates the repeating data patterns by copying a specified number of data units starting from a buffer location specified by codes embedded in the compressed data stream.
Both groups of LZ compression algorithms send uncompressed literal data that is not part of a repeating pattern from the compression system to the decompression system. A data type flag is typically incorporated into the data stream to indicate whether a particular data unit is code or literal data, resulting in a non-integral data unit size. In compression systems using a byte as the data unit, a data type flag must be added to each data byte by the compression system, resulting in each data byte represented by 9 bits, including 8 bits of data and the data type flag. The decompression system for this compressed data stream must strip out the data type flag and repack the data stream into bytes to reconstruct the original data.
In various embodiments, a method and apparatus are provided for compressing and decompressing a data stream in a manner that is more easily implemented in hardware.
One aspect of the present invention provides a method and system for replacing repeating sequences in the data stream with codes. The codes characterize a repeating sequence by indicating the location and length of an identical sequence in the data stream. Indicators are inserted into the data stream characterizing the length of non-repeating sequences of data.
In another aspect of the invention, a decompression method and system are provided. The decompression method and system removes the indicators inserted by the compression system from the data stream, each indicator characterizes a non-repeating sequence of variable length. Codes inserted into the data stream by the compression system are replaced with the repeating sequences characterized by the codes.
It will be appreciated that various other embodiments are set forth in the Detailed Description and claims which follow.
Various aspects and advantages of the invention will become apparent upon review of the following detailed description and upon reference to the drawings in which:
Various embodiments of the present invention are described in terms of compressing and decompressing a data stream in a manner that facilitates hardware implementation. The present invention provides a compression method and system, and an associated decompression method and system. The compression and decompression methods and systems provided by the invention may be advantageously used for more efficient data storage and/or transmission.
The present invention is believed to be applicable to a variety of processes and systems within the technology areas of data communication and storage. The invention is believed to be particularly beneficial for configuring programmable logic devices. Configuration data for complex programmable logic devices (CPLDs) and field programmable gate arrays (FPGAs) can be compressed and stored in less expensive, lower density storage devices.
Previous compression/decompression techniques use a flag bit to identify each data unit (e.g., a byte) as either a code data unit or a literal data unit. By previous techniques, the compression system inserts a data type flag into the data stream to identify the data type for each data unit. Adding type flags forces the compression system to take bits from the following code or literal data units to incorporate the type flag to make a whole data unit. The remaining bits are added to bits from the next data unit to form a byte, and so on. The decompression system is required to strip out the flags and reassemble the data into integral data units (e.g., bytes). These techniques require circuitry in the compression system to insert the flag bits and form data units, and also circuitry in the decompression system to remove the flag bits and regenerate the data units.
The present invention provides methods of compression and decompression allowing the data to be manipulated by bytes, or by any other logical data unit (e.g., words), thereby simplifying hardware implementation relative to previous techniques. Additionally, the present invention identifies compressed data sequences using an actual buffer or FIFO address rather than a logical address, further reducing the required hardware.
Those skilled in the art will appreciate that the invention may be used in a wide variety of technology areas to conserve memory space and/or make more efficient use of communication channel capacity. The invention may be applied to a variety of programmable devices and adapted to a number of applications to increase the effective memory storage in a programmable device or to enhance communication bandwidth.
Repeating sequences in a data stream are identified 110. The repeating sequences are characterized by codes indicating the length of the repeating sequence and the location of an identical sequence in the data stream 120. The repeating sequences are replaced in the data stream by the codes 130. Non-repeating sequences are characterized using indicators 140 providing the length the non-repeating sequences. The indicators characterizing the non-repeating sequences are inserted into the data stream 150.
The compression method is implemented using two sliding windows to hold the data stream or data block to be compressed. The two sliding windows are implemented, for example, as memory buffers, or as first-in-first-out (FIFO) registers. The first sliding window is denoted the data window and the second sliding window is denoted the look-ahead window. Additionally, a write buffer may be used as a temporary data storage buffer.
The two sliding windows and the write buffer are initialized 210. The initialization may include filling the data and look-ahead windows with data from the data stream and filling the write buffer with data from the data window. Data in the look-ahead window is compared to data in the data window. Sequences in the look-ahead window that repeat an identical sequence in the data window are identified 220. The repeating sequence may be required to have a minimum length, such as 3 bytes. Furthermore, it may be required that the length and location of the repeating sequence can be defined by a predetermined number of data units, for example, 2 bytes.
If a repeating sequence is found 245, the look-ahead window contains a sequence that repeats a sequence in the data window. The write buffer is cleared 250. Clearing the write buffer requires writing an indicator describing the number of data units in the write buffer to the compressed data stream. These data units are considered a non-repeating sequence. The indicator containing the non-repeating sequence length is described in whole data units. For example, an 8-bit byte can have the most significant bit set to 0 to indicate that a literal sequence is being defined, and the lower 7 bits used to specify the length of the data literal sequence. The contents of the write buffer are copied to the compressed data stream and the write buffer is emptied.
A code characterizing the repeating sequence in the look-ahead window is written to the compressed data stream 260. The code includes 1) a code flag indicating a repeating data sequence has been compressed, 2) an address in the data window where the repeating sequence may be located, and 3) the length of the repeating sequence. The repeating sequence is replaced in the data stream by the code. The code is represented in whole data units. For example, two 8-bit bytes can be used to represent the compressed data code: the most significant bit of the first byte can be set to 1 to indicate that a repeating data sequence has been found; the remaining 7 bits to indicate the length of the repeating sequence, and the next byte to indicate the position of the repeating data sequence within the data window.
If a repeating sequence is not found 245, and there is still more data to compress 265, the write buffer is checked to determine if the write buffer is full 225. If the write buffer is not full 225, new data is shifted into the look-ahead window, the look-ahead data is shifted into the data window and the write buffer is filled with data from the data window 240. If the write buffer is full, the write buffer is cleared 230 as described above.
The process described in blocks 210–265 continues until there is no more data to compress 265. If there is no more data to compress 265, the write buffer is cleared 270 and the process is complete.
A decompression method in accordance with an embodiment of the invention is conceptually illustrated, in broad and general terms, in the diagram of
If the data unit is code, then the sequence length and sliding window location of the repeating data sequence are extracted from the code units 450. The decompressor replaces the codes by copying the specified number of data units starting from the specified sliding window location to the sliding window 460. After reading the specified number of literals from the compressed data stream or copying the specified number of data units from the sliding window, the next data unit is checked to determine if the next data unit is an indicator or a code. The process defined by blocks 420–460 is repeated until there is no more compressed data 410 and the process ends.
Every data unit loaded into the decompressor sliding window is decompressed data. The decompressed data does not have to be repacked into data units because indicators and code bytes were inserted as whole data units.
An example of the compression process of the present invention is illustrated in
In this example, a 5-byte sequence, AEAAA 522, located in the look-ahead window 520 at look-ahead window positions 0–4 repeats a sequence 511 in the data window located at data window positions 6–10. The sequence AEAAA 522 in the look-ahead window 520 represents a repeating sequence that may be encoded using a two byte code including the location and the length of the repeating sequence. In this example, the repeating sequence may be characterized by the n-tuple (6,5) 530, where 6 represents the location of the repeating sequence in the data window, and 5 indicates the length of the repeating sequence.
An example of decompression in accordance with an embodiment of the invention is illustrated by a compressed data stream in
Turning again to
After the 20 byte literal sequence 620 is shifted into the decompression FIFO, the next byte 631 is processed. The MSB of byte 631 is a 1, which identifies this as the first byte of a two-byte code 630. The code characterizes a repeating sequence that may be identified by a length and a location of an identical sequence in the FIFO.
The length of the repeating sequence characterized by the code 630 is contained in the 7 LSBs of the first code byte 631. The location of the repeating sequence in the FIFO is contained in the 8 bits of the second code byte 632.
In this example, the 7 LSBs of the first code byte 631 contains 000100, decimal 4 to which 1 is added in one implementation to get decimal 5, indicating that the repeating sequence is 5 bytes in length. The 8 bits of the second code byte 632 identify the location of the repeating sequence in the decompression FIFO 700 illustrated in
A block diagram of a system for compressing a data stream in accordance with an embodiment of the present invention is illustrated in
Other aspects and embodiments of the present invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. It is intended that the specification and illustrated embodiments be considered as examples only, with a true scope and spirit of the invention being indicated by the following claims.
This application is a continuation of Ser. No. 10/175,710, now U.S. Pat. No. 6,744,388.
Number | Name | Date | Kind |
---|---|---|---|
4906991 | Fiala et al. | Mar 1990 | A |
5237460 | Miller et al. | Aug 1993 | A |
5371499 | Graybill et al. | Dec 1994 | A |
5414425 | Whiting et al. | May 1995 | A |
5608396 | Cheng et al. | Mar 1997 | A |
5729223 | Trissel | Mar 1998 | A |
5798718 | Hadady | Aug 1998 | A |
5974180 | Schwendeman | Oct 1999 | A |
6446145 | Har et al. | Sep 2002 | B1 |
6571016 | Mehrotra et al. | May 2003 | B1 |
6671420 | Feather et al. | Dec 2003 | B1 |
6737992 | Wolthuis | May 2004 | B1 |
Number | Date | Country | |
---|---|---|---|
Parent | 10175710 | Jun 2002 | US |
Child | 10847967 | US |