Commit Graph

2221 Commits

Author SHA1 Message Date
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
a7d476e05b Next attempt! 2014-07-11 19:52:05 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
f88fd37e63 Reverted to pre-shared_ptr edition. 2014-07-10 20:53:06 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
7aaaafc9ec Removed more stuff. Performance at 48%. This really sucks. 2014-07-10 20:47:09 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
e479fd1c22 Removed the embedded copy of the value in the iterator. Now at 39% performance. 2014-07-10 20:01:57 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
681e7fe888 Well it works ... but at 33% of a standard unordered_map. This is becoming frustrating :( 2014-07-10 19:58:16 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
c1a18c1f6a Bumped to GCC 4.9 as libstdc++ 4.8 can't cope. 2014-07-10 19:33:21 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
9c32acfc24 Next attempt, this time at a shared_ptr based map. 2014-07-10 19:28:07 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
2b50067d06 Works, though slowly. 2014-07-08 18:24:43 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
6548c1dd8f Turned GCC optimisation back on 2014-07-07 22:35:01 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
a328d6c8db Got it working properly on MSVC thankfully. Performance is about 20x slower than single threaded unordered_map. 2014-07-07 22:31:12 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
cd10fc9b84 Added more debug stuff. I am mystified why this is pushing all the nodes to the final buckets. Moving to Windows, it has a much better debugger. 2014-07-07 19:35:36 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
f2ae73311a wip, at least it runs without crashing, though it's expanding some buckets hugely more than it should. 2014-07-05 20:00:31 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
9e67dab2aa Refactor complete, in the sense it compiles. 2014-07-03 19:50:51 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
3599cf485a Wip of a complete rearchitecture of concurrent_unordered_map into a pure atomics implementation. 2014-07-02 20:10:07 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
c7c7e842f1 WIP on concurrent_unordered_map 2014-06-28 21:58:08 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
d53bd7b8d9 Added test for transacted reads but locked writes. 2014-06-27 18:47:49 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
0fe2141136 Added malloc benchmark test. 2014-06-26 19:10:45 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
f200224edc Renamed whole project to boost.spinlock.
Added testing of transactional GCC support.

Added more tests.

Now benchmarks locked/transacted unordered_map.
2014-06-25 19:06:13 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
7ed7fb0484 Stopped hanging oneself around a cache line. 2014-06-23 20:23:26 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
7fc8a33335 Added a benchmark. 2014-06-23 20:15:18 +01:00
Niall Douglas (s [underscore] sourceforge {at} nedprod [dot] com)
5d9881f064 First import from Boost.AFIO 2014-06-23 19:26:59 +01:00