Matt Joiner
|
a893e685b0
|
Don't allocate pieces map unless necessary when filling responsive requests
|
2014-11-19 20:01:32 -06:00 |
Matt Joiner
|
6942eed15f
|
Lazily hash pieces where possible
|
2014-09-15 03:25:53 +10:00 |
Matt Joiner
|
1cfd6ecdcb
|
Improvements arising from attempting to delay piece hashing until necessary
|
2014-09-14 04:03:23 +10:00 |
Matt Joiner
|
5d35e10706
|
Tighten responsive download strategy partial piece request filling
|
2014-09-11 14:22:29 +10:00 |
Matt Joiner
|
1507d803bd
|
Great complexifying of the responsive download strategy
Should be better after 4 days of experimentation...
|
2014-08-28 10:04:00 +10:00 |
Matt Joiner
|
2d9c22ae1e
|
Misc trivial improvements
|
2014-08-25 22:14:10 +10:00 |
Matt Joiner
|
c8f335182f
|
Improve incomplete piece ordering, responsive download strategy
|
2014-08-25 05:31:34 +10:00 |
Matt Joiner
|
2b079e4a9d
|
Got a nice working algorithm for responsive download strategy
|
2014-08-24 03:08:11 +10:00 |
Matt Joiner
|
189f1e364e
|
Ditch ordering of prioritized chunks
|
2014-08-22 17:35:15 +10:00 |
Matt Joiner
|
95d5d4a30c
|
Improve status and logging
|
2014-08-22 17:33:17 +10:00 |
Matt Joiner
|
74f05db2ea
|
Clean up imports
|
2014-08-21 21:08:56 +10:00 |
Matt Joiner
|
ef7c4f4120
|
Move priority management entirely into the download strategies
|
2014-07-24 13:42:31 +10:00 |
Matt Joiner
|
e092ba2c5f
|
Use low-water mark to populate connection requests
I can't believe I didn't think of this before. It should address massive
performance problems I've had here.
|
2014-07-16 17:11:45 +10:00 |
Matt Joiner
|
4c2d07337d
|
Keep pieces sorted by bytes left
|
2014-07-10 00:26:58 +10:00 |
Matt Joiner
|
147ad31135
|
A CPU optimization to FillRequests when we know no more can be added
|
2014-07-01 00:03:07 +10:00 |
Matt Joiner
|
eb070383b1
|
Move download strategies into their own file
|
2014-06-30 00:22:05 +10:00 |