| 1 | - switched to GPL v2 or any later version. |
1 | - switched to GPL v2 or any later version. |
|
|
2 | - speed up compression by ~10-15% in common cases |
|
|
3 | by some manual unrolling. |
|
|
4 | - import some compiler tricks from JSON::XS. |
|
|
5 | - tune hash functions depending on ULTRA_FAST or VERY_FAST settings. |
|
|
6 | - for typical binary data (e.g. /bin/bash, memory dumps, |
|
|
7 | canterbury corpus etc.), speed is now comparable to fastlz, but |
|
|
8 | with better compression ratio. with ULTRA_FAST, it's typically |
|
|
9 | 3-8% faster than fastlz while still maintaining a similar ratio. |
|
|
10 | (amd64, ymmv). thanks a lot for the competition :) |
|
|
11 | - undo inline assembly, it is no longer helpful. |
| 2 | |
12 | |
| 3 | 2.1 Fri Nov 2 13:34:42 CET 2007 |
13 | 2.1 Fri Nov 2 13:34:42 CET 2007 |
| 4 | - switched to a 2-clause bsd license with GPL exception. |
14 | - switched to a 2-clause bsd license with GPL exception. |
| 5 | - get rid of memcpy. |
15 | - get rid of memcpy. |
| 6 | - tentatively use rep movsb on x86 and x86_64 (gcc only) for a |
16 | - tentatively use rep movsb on x86 and x86_64 (gcc only) for a |
