Improve sparsetb #17
Open
Improve sparsetb #17
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
in place of three matrices X,mloop10/01/00 only use two X,mloop1/0
- use stricter candidate criterion (i.e. optimal branch is non-decomposign). without recomputation, maintaining less candidates (due to new criterion) requires more trace arrows, such that space consumption gets worse
Failing example: ACGAUGCAUGCGACCAGCGGCACGAUCGAGCGCGAUCGAUCGAUCGAGCAACGAGCUAGCGCGACGAUCGUACGGU
- use M.get notation for all get methods that do no additional computation - rename all remaining get_* methods (which perform more computation, e.g. initialization, than the Matrix calls get method) to calc_*
…uggy \time -f "%U %M" ./CCJ CAUGCGACCAGCGGCACGAUCGAGCGCGAUCGAUCGAUCGAGCAACGAGCUAGCGCGACGAUCGUA P(2,64): tracing PK(2,24,32,56) e:-977 and PK(25,31,57,64) e:-472 recompute_PX 2 22 33 56 0 Command terminated by signal 11
- simplify candidate lists
* use SimpleMap (as originally intended) for single lists
* use SimpleMap as well as map storing the single lists;
use extra index for space/time trade-off
- simplify ta_key_pair(): no need to store the indices k and l
- add recomputation for e_intP
- add check for sufficient TURN in calc_PMiloop
NOTE: this commit turns off garbage collection;
recomputation of PX and gc is currently not compatible, ie.
GC removes trace arrows that are necessary for recomputing.
|
I added my rudimentary unit tests for the matrices classes and class Index4D; currently, one can test with g++ matrices_test.cpp && ./a.out or similar; of course it would be nice to have this better integrated with cmake (and/or a testing framework) |
|
Hi @IanWark, this fixed the wrong energy/trace of @HosnaJabbari, I still think the other differences (all due to< TURN constraints) are actually due to a bugfix ,e.g. |
When tracing back to candidate, the target energy must be retrieved from candidate. To simplify handling, set the correct target energy already in the generic decomposition method.
-- use Index4D class and MType -- change interface, such that source information can be omitted in tas -- store only two indices depending on type -- deactivate currently incompatible garbage collection code
…in conversion to uppercase
-- two mixed-up indices in POiloop trace back -- extend can_pair to rule out j-i<=TURN
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hi @IanWark, hi @HosnaJabbari
I made progress with the new recomputation-based trace back, replacing most of the trace arrows.
We are now a lot faster than pknots and use less space :) as it should be expected after all our theoretical work.
This is at a point, where finding bugs became difficult, so @IanWark could you please test on (smaller) benchmark instances?
(I rewrote this branch and opened a new PR)