Huffman code decoding is performed using a multi-level table lookup.
The fastest way to decode is to simply build a lookup table whose
size is determined by the longest code. However, the time it takes
to build this table can also be a factor if the data being decoded
is not very long. The most common codes are necessarily the
shortest codes, so those codes dominate the decoding time, and hence
the speed. The idea is you can have a shorter table that decodes the
shorter, more probable codes, and then point to subsidiary tables for
the longer codes. The time it costs to decode the longer codes is
then traded against the time it takes to make longer tables.
This results of this trade are in the variables lbits and dbits
below. lbits is the number of bits the first level table for literal/
length codes can decode in one step, and dbits is the same thing for
the distance codes. Subsequent tables are also less than or equal to
those sizes. These values may be adjusted either when all of the
codes are shorter than that, in which case the longest code length in
bits is used, or when the shortest code is *longer* than the requested
table size, in which case the length of the shortest code in bits is
There are two different values for the two tables, since they code a
different number of possibilities each. The literal/length table
codes 286 possible values, or in a flat code, a little over eight
bits. The distance table codes 30 possible values, or a little less
than five bits, flat. The optimum values for speed end up being
about one bit more than those, so lbits is 8+1 and dbits is 5+1.
The optimum values may differ though from machine to machine, and
possibly even between compilers. Your mileage may vary.