NOTICE

How to use, and some performance results
=================

How to use
-----------
You type a following line for encoding:

./vcompress [ID] [FILE]

A pre-defined value needs to be input in [ID], and
the input in [FILE] needs to follow specified formats.
These details can be found in following section.
For decoding, you say a line below:

./vcompress -d [FILE] <[OUT]>

When finishing decoding, it shows performance results
such as the compression ratio and the speed. To write
the decoded integers in local filesystems, a output
filename needs to be filled in [OUT].


For example, if you would like to compress a file, named
XXX, following specified formats with N Gamma(ID: 0), you
just type a followin line:

./vcompress 0 XXX

This line generates two files; XXX.gamma and XXX.pos.
And then, the corresponding decoding command is:

./vcompress -d XXX

Anyone who would like to do a quikc trial can access
a following web site to have sample data sets.

http://integerencoding.isti.cnr.it/


Implemented encoder IDs
-----------
The library uses following IDs:

ID  Name
---
0   N Gamma
1   FU Gamma
2   F Gamma
3   N Delta
4   FU Delta
5   FG Delta
6   F Delta
7   Variable Byte
8   Binary Interpolative
9   Simple 9
10  Simple 16
11  PForDelta
12  OPTPForDelta
13  VSEncodingBlocks
14  VSE-R
15  VSEncodingRest
16  VSEncodingBlocksHybrid
17  VSEncodingSimple


An input/output file format
-----------
In the implemented encoders in the library, an input/output format
follows a rule below:

[total of int., a first int., a second int., ..., a last int.]
[total of int., a first int., a second int., ..., a last int.]
...

Each entry occupies a 32-bit word, and these integers MUST appear
in an ascending order.


Performance
-----------
Some performance results(left and right values are mis and bpi,
respectively) of these encoders are shown here. gov2 was used as
a test set, which is one of TREC test collections in the Terabyte
Track. This test was done on a machine with a Intel Core i5-2500
(3.3GHz) processor and 8GiB of memory.
(git tag benchmark-20120928):

CoderName                           mis          bpi
---
N Gamma                           59.14         5.05
FU Gamma                          93.61         5.05
F Gamma                           89.43         5.05
N Delta                           68.82         4.70
FU Delta                          88.69         4.70
FG Delta                         103.06         4.70
F Delta                          122.26         4.70
Variable Byte                    165.28         9.18
Binary Interpolative              71.71         3.95
Simple 9                         645.52         5.41
Simple 16                        642.80         5.17
PForDelta                       1043.99         6.00
OPTPForDelta                     828.32         5.46
VSEncodingBlocks                 973.65         4.51
VSE-R                            182.61         4.24
VSEncodingRest                   520.72         4.34
VSEncodingBlocksHybrid           808.90         4.40
VSEncodingSimple                1712.34         6.13

, where 'mis' represents these average decompression speeds
expressed in millions of integers per second, and 'bpi' does
these compression performance expressed in bits per integers.