Issue695

Title space improvement for LM-Count
Priority feature Status in-progress
Superseder Nosy List jendrik, malte, manuel, salome
Assigned To Keywords
Optional summary

Created on 2016-12-14.15:43:20 by jendrik, last changed by jendrik.

Messages
msg7258 (view) Author: jendrik Date: 2018-06-08.19:02:39
I made a few minor comments, but I'll leave the detailed review to Florian, the 
per-state expert :-)
msg7249 (view) Author: malte Date: 2018-06-08.16:42:08
Results look great! +10 coverage for 32 bits, +24 coverage for 64 bits, and 64
bits has finally caught up with 32 bits in terms of coverage. In terms of time
and memory scores, 64 bits also compares favourably with 32 bits. (For the
memory score that's a bit of a surprise, but we'll take it.)

How do we proceed in terms of code reviews? We may get there faster if someone
else looks at it first, and I then only do a light or no review, but there's a
risk that this will create extra work if we have contradictory feedback.
msg7244 (view) Author: salome Date: 2018-06-08.15:02:46
A first implementation is ready for this and preliminary test show decent results.

Link to pull request:
https://bitbucket.org/manuel_h/fd-issue695/pull-requests/1/issue695/diff

Experiments 32 bit:
http://ai.cs.unibas.ch/_tmp_files/simon/issue695-v1-opt-issue695-base-issue695-v1-compare.html

Experiments 64 bit:
http://ai.cs.unibas.ch/_tmp_files/simon/issue695-v1-opt-64bit-issue695-base-issue695-v1-compare.html
msg5881 (view) Author: jendrik Date: 2016-12-14.15:43:20
Split off from issue213.

Quoting Malte in msg5862:

"The drop for bjolp [between 32- and 64-bit builds] surprised me a bit,
so I had a look how it stores a landmark data, and it uses a
PerStateInformation<vector<bool>>, which is very wasteful because we
pay the vector overhead for every single state. For our state
representation, we moved from SegmentedVector to SegmentedArrayVector
for this reason, and it made a big difference. We should probably do a
similar thing here, i.e., have something like PerStateInformation
optimized for same-size arrays. (And because this is a vector<bool>, we
additionally need to pack individual bits.)"
History
Date User Action Args
2018-06-08 19:02:39jendriksetmessages: + msg7258
2018-06-08 16:42:08maltesetmessages: + msg7249
2018-06-08 15:02:46salomesetmessages: + msg7244
2016-12-20 10:39:59manuelsetstatus: unread -> in-progress
nosy: + salome, manuel
2016-12-14 15:43:20jendrikcreate