|
|
1 | |
|
|
2 | - include a workaround for a compiler bug on 64 bit windows |
|
|
3 | (microsoft claims to support POSIX, but is far from it). |
|
|
4 | (analysed nicely by John Lilley). |
|
|
5 | |
|
|
6 | 3.1 Fri Nov 30 11:33:04 CET 2007 |
|
|
7 | - IMPORTANT BUGFIX: a too long final literal run would corrupt data |
|
|
8 | in the encoder (this was introduced in 3.0 only, earlier versions |
|
|
9 | are safe). |
|
|
10 | |
|
|
11 | 3.0 Tue Nov 13 22:13:09 CET 2007 |
|
|
12 | - switched to 2-clause bsd with "GPL v2 or any later version" option. |
|
|
13 | - speed up compression by ~10-15% in common cases |
|
|
14 | by some manual unrolling. |
|
|
15 | - import some compiler tricks from JSON::XS, for further speed-ups. |
|
|
16 | - tune hash functions depending on ULTRA_FAST or VERY_FAST settings. |
|
|
17 | - for typical binary data (e.g. /bin/bash, memory dumps, |
|
|
18 | canterbury corpus etc.), speed is now comparable to fastlz, but |
|
|
19 | with better compression ratio. with ULTRA_FAST, it's typically |
|
|
20 | 3-15% faster than fastlz while still maintaining a similar ratio. |
|
|
21 | (amd64 and core 2 duo, ymmv). thanks a lot for the competition :) |
|
|
22 | - undo inline assembly in compressor, it is no longer helpful. |
|
|
23 | - no changes to the decompressor. |
|
|
24 | - use a HLOG of 16 by default now (formerly 15). |
|
|
25 | |
| 1 | 2.1 Fri Nov 2 13:34:42 CET 2007 |
26 | 2.1 Fri Nov 2 13:34:42 CET 2007 |
| 2 | - switched to a 2-clause bsd license with GPL exception. |
27 | - switched to a 2-clause bsd license with GPL exception. |
| 3 | - get rid of memcpy. |
28 | - get rid of memcpy. |
| 4 | - tentatively use rep movsb on x86 and x86_64 (gcc only) for a |
29 | - tentatively use rep movsb on x86 and x86_64 (gcc only) for a |
| 5 | moderate speed improvement. |
30 | moderate speed improvement. |
