Releases: SpiNNakerManchester/SpiNNFrontEndCommon
Releases · SpiNNakerManchester/SpiNNFrontEndCommon
Release 7.0.0
Get this release from PyPI.
pip install SpiNNFrontEndCommon "1!7.0.0"
Key Changes
- The range of supported Python versions has changed; we now support (and test with) Python 3.8 – 3.11.
- We have changed the software license from GPL v3 to Apache v2.
- We use SQLite databases to store a lot more intermediate and provenance data now, which is significantly less troublesome than the large piles of binary files we used to use. (A consequence of this is that the provenance data is now far more machine-readable than it was previously.)
- We have completely reworked the interaction between the SpiNNaker client tool stack and our Spalloc service. This allows us to integrate our internal resource management more completely with the Human Brain Project/EBRAINS central resource management, as well as enabling much greater manageability of the SpiNNaker1M machine and allowing more ways to allocate resources (such as via specific sections of the machine).
- Internal simplifications (mostly to help ourselves deliver the software):
- We have rebuilt the way the internal structure of the processing workflow is done, greatly simplifying how things work. This should only matter to users who wish to plug in their own algorithms into the stack; if this includes you, please contact the team for help.
- We have rebuilt the way we manage much of the state in the system. This removes the need for complex value injection systems and instead allows us to keep the code simpler and easier to understand.
What's Changed
- java_jar_path by @Christian-B in #783
- Tags passed as Tags object only by @Christian-B in #787
- nuked TDMAAgendaBuilder by @Christian-B in #789
- Remove dead inputs by @Christian-B in #791
- Making the configs truly global by @Christian-B in #790
- test decode the xml files by @Christian-B in #795
- find and check all get_config calls by @Christian-B in #797
- Router max entries from Machine by @Christian-B in #798
- Fix extra monitor memory leak by @rowleya in #799
- Cfg stack by @Christian-B in #800
- More provenance by @dkfellows in #681
- removing very old numpy instructions with broken links by @Christian-B in #780
- _database_socket_addresses is a set by @Christian-B in #785
- extract_iobuff False by default by @Christian-B in #801
- Fix the provenance by @rowleya in #803
- renamed _state (not Pynn state) to status by @Christian-B in #811
- Move the reports to FEC by @Christian-B in #812
- nuked "unrecognized arguments: -q" by @Christian-B in #813
- Reset fix can of worms by @Christian-B in #814
- Simplify Simulator stack by @Christian-B in #810
- Spinn4 lost 4 cores by @Christian-B in #819
- Add tolerance of 1 core in Json comparison by @Christian-B in #821
- report provenace right away by @Christian-B in #808
- PacmanAlgorithmFailedToCompleteException never used so why check for it by @Christian-B in #820
- Test objects by @Christian-B in #775
- remove turn_off_board_after_discovery by @Christian-B in #823
- Graph view by @Christian-B in #822
- Too broad an exception is not a good idea by @Christian-B in #824
- Stop passing config values through the simulator by @Christian-B in #792
- Fix app master pop limits by @andrewgait in #827
- report directories as global_variables by @Christian-B in #826
- Bye bye outer_table_compression_target_length setting by @Christian-B in #825
- Split synapse neuron models by @rowleya in #815
- Synapse neuron splitting by @rowleya in #725
- No memory by @Christian-B in #829
- Max mem regions is now 32 by @andrewgait in #832
- Compressors sdram by @rowleya in #843
- Add app ID to database by @rowleya in #845
- Stop ResourceTracker from looking at the Whole Machine by @Christian-B in #830
- should only report finished when you actually finish successfully. by @alan-stokes in #520
- fix prov call on error by @Christian-B in #869
- Spalloc avoid boards by @Christian-B in #870
- remove extract_iobuf_during_run flag by @Christian-B in #875
- Chip power monitor sync fix by @andrewgait in #880
- Provenance straight to and from the database. by @Christian-B in #881
- No executor by @Christian-B in #847
- remove abilty to set app_id in cfg by @Christian-B in #863
- make sure command sender added with new graph by @Christian-B in #862
- Stop using @Inject by @Christian-B in #882
- Cfg by @Christian-B in #888
- put recover_from_error back in by @Christian-B in #892
- App graph is never None but may be empty by @Christian-B in #893
- No executor algorithms now functions and No xml by @Christian-B in #894
- Key translation by @rowleya in #901
- fix call in test by @Christian-B in #904
- Get machine2 by @Christian-B in #906
- test and support python 3.10 instead of 3.6 by @Christian-B in #908
- do the right precalculation by @Christian-B in #907
- write version to database ASAP by @Christian-B in #916
- Scamp connection cleanup. by @Christian-B in #915
- range compress by @Christian-B in #914
- Category timing into provenance database by @Christian-B in #912
- Add tags where malloc is used by @rowleya in #918
- Dont save config values in ASB by @Christian-B in #909
- Where is chip by @Christian-B in #911
- access missing/ dead processors by @Christian-B in #920
- dont error if db never set by @Christian-B in #902
- No Monitors list use dict.values() by @Christian-B in #921
- fix when to use machine graph by @Christian-B in #919
- Use the Data Spec database more by @Christian-B in https://github.com/SpiNNakerManchester/SpiNNFrontEndCom...
Release 6.0.0
Get this release from PyPI.
pip install SpiNNFrontEndCommon "1!6.0.0"
Release 5.1.0
Get this release from PyPI.
pip install SpiNNFrontEndCommon "1!5.1.0"
This is a minor release that fixes a number of bugs.
Release 5.0.0
August 9 2019
4.0.0
Added .sql and simplified