More compiler friendly error handling.

[neworderofjamie]

Turns out that compiling the code generated by our CHECK_CUDA_ERRORS macro can cause runner.cc compilation to take a long time and a lot of memory. Switching this out of a simpler macro that just calls std::abort can reduce compile times by up to 10x on a model with 10000 state variables (equivalent to about 100 neuron populations, nearly all-to-all connected):

This simpler error handling macro can now be enabled from C++ with:

void modelDefinition(NNmodel &model)
{
    ...
    GENN_PREFERENCES.generateSimpleErrorHandling = true;
    ...
}

or from Python like:

model = GeNNModel("float", "my_model", generateSimpleErrorHandling=True)

[codecov]

Codecov Report

Base: 88.82% // Head: 88.82% // Decreases project coverage by -0.00% ⚠️

Coverage data is based on head (1383827) compared to base (ec5bbe9).
Patch coverage: 75.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #554      +/-   ##
==========================================
- Coverage   88.82%   88.82%   -0.01%     
==========================================
  Files          76       76              
  Lines       12341    12344       +3     
==========================================
+ Hits        10962    10964       +2     
- Misses       1379     1380       +1     

Impacted Files	Coverage Δ
src/genn/backends/cuda/backend.cc	80.44% <66.66%> (-0.06%)	⬇️
include/genn/backends/cuda/backend.h	97.61% <100.00%> (+0.05%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[tnowotny]

Good work on this. For a moment I thought it might be a good idea to make the simple error handling the default but now I think you have it right as for small models it is not an issue and you would want the more informative output as a default option.

[neworderofjamie]

Yeah, the longer term plan is to consider what error information is actually useful and maybe move some of the error handling out of the generated code