Difference between revisions of "BEAR4 compression"
From ScienceZero
| (One intermediate revision by the same user not shown) | |||
| Line 1: | Line 1: | ||
WORK IN PROGRESS... | WORK IN PROGRESS... | ||
| − | BEAR4 is designed for microcontrollers that have little RAM. It uses the data decompressed so far as the dictionary, compared to typical LZ compression this decreases efficiency since the index range must be larger and the need for storing the count. Most of the loss is recovered by only reading the number of bits required for the current range. | + | BEAR4 is designed for microcontrollers that have little RAM. It uses the data decompressed so far as the dictionary, compared to typical LZ compression this decreases efficiency since the index range must be larger and the need for storing the count. Most of the loss is recovered by only reading the number of bits required for the current index/count range. |
Header | Header | ||
BE4R - file ID | BE4R - file ID | ||
| + | w - word size (1-64) | ||
s - symbol size (1-64) | s - symbol size (1-64) | ||
h - symbol shift (0-63) | h - symbol shift (0-63) | ||
| Line 10: | Line 11: | ||
c - maximum count size (1-16) | c - maximum count size (1-16) | ||
| + | Data | ||
0 <symbol> | 0 <symbol> | ||
1 <index> <count> | 1 <index> <count> | ||
| + | Pseudocode | ||
do | do | ||
indexSize = min(log2(decompressedSize), maximumIndexSize) | indexSize = min(log2(decompressedSize), maximumIndexSize) | ||
Latest revision as of 16:46, 4 February 2016
WORK IN PROGRESS...
BEAR4 is designed for microcontrollers that have little RAM. It uses the data decompressed so far as the dictionary, compared to typical LZ compression this decreases efficiency since the index range must be larger and the need for storing the count. Most of the loss is recovered by only reading the number of bits required for the current index/count range.
Header
BE4R - file ID
w - word size (1-64)
s - symbol size (1-64)
h - symbol shift (0-63)
i - maximum index size (1-32)
c - maximum count size (1-16)
Data
0 <symbol> 1 <index> <count>
Pseudocode
do
indexSize = min(log2(decompressedSize), maximumIndexSize)
countSize = min(indexSize, maximumCountSize)
read classBit
if classBit = 0 then
read symbol
output symbol
else
read index, count
for each count
read symbolFromDecompressed
output symbol
next
endif
loop
