Add to pre #479
Add to pre #479
Conversation
…unctionality. I completed steps up to and including the single-threaded backend.
…d when the model generator was compiled. Model generator crashes with segmentation fault.
… spike condition every 0.2 ms.
Fixed a bug where no pointer to revInSyn was added to merged group because it was in role Presynaptic and Dynamics only. Tydied some of the tests
exited on an error
Codecov Report
@@ Coverage Diff @@
## master #479 +/- ##
==========================================
+ Coverage 87.46% 87.51% +0.04%
==========================================
Files 82 82
Lines 17266 17437 +171
==========================================
+ Hits 15102 15260 +158
- Misses 2164 2177 +13
Continue to review full report at Codecov.
|
* incorporate hashes of pre-output synapse groups in neuron group hashes * add a seperate (empty) hash for pre-output synapse group initialisation
| @@ -0,0 +1,20 @@ | |||
| CXXFLAGS +=-std=c++11 -Wall -Wpedantic -Wextra -I $(GTEST_DIR) -isystem $(GTEST_DIR)/include | |||
There was a problem hiding this comment.
these should all just be symlinks to ../../utils/Makefile
There was a problem hiding this comment.
Thanks for fixing. Some of the other feature tests still have their own Makefile such as
decode_shared_matrix_conn_gen_proceduralg_dense_egp/Makefile
decode_shared_matrix_conn_gen_proceduralg_dense/Makefile
var_init_egp/Makefile
decode_matrix_globalg_bitmask_optimised/Makefile
... and a few more. Is that on purpose?
There was a problem hiding this comment.
Thanks for fixing!
Some other feature tests still have their own Makefile such as
decode_matrix_globalg_bitmask_optimised/Makefile
... is that on purpose?
Added some basic documentation. I have briefly tested that the pygenn wrapper works for $(addToPre) and set_pre_target_var. This should now complete the $(addToPre) pull request.
This is the PR for adding additional functionality of
$(addToPre, XXX)to synapsessim_code,synapse_dynamics_code,event_code, andpost_learn_code.The expression is added to the presynaptic neuron's
Isynby default but the target variable can be set to a different input variable of the presynaptic population.This is still work in progress in the sense that there should be additional feature tests, at least testing for the use in event code and post_learn_code in addition to synapse_dynamics and sim_code that are covered by the two tests I created.
One could also add a set of tests for other connectivity schemes than dense and for the different parallelisation strategies to get a good coverage ...
Developer notes:
I re-used the fusePSM flag to decide whether the presynaptic synaptic output should be fused, and did not introduced a new, separate flag.
I am using the same HashDigest for everything as there are no parameters, i.e. not separate fuse has digest.
The collection variable for reverse presynaptic output is "revInSyn".
I am re-using getInSynLocation, i.e. locations will be identical for inSyn and revInSyn.
I am also allowing $addToPre in synapse_dynamics code.
I did not use shared memory for the SIMT code and PostSpan, PostSpanBitmask strategies. I am not sure whether it would be worth it and also do not want to overload shared memory with yet another separate memory structure.
TODO