Ioannis Magkanaris

[LLVM] Add debug symbols to GPU code

2

Currently debug symbol addition fails to compile with CUDA compiler. According to [a mail](https://lists.llvm.org/pipermail/llvm-dev/2017-November/118841.html) from the llvm mailing list there are multiple issues that need to be addressed to make...

Update the status of the implemented solvers in https://github.com/BlueBrain/nmodl/blob/master/docs/language.rst and check which ones are still related to the CoreNEURON solver to make the change from MOD2C to NMODL as default...

Check the `LONGITUDINAL_DIFFUSION` construct implementation and documentation in `NEURON` and add related unit test in `NMODL`.

Check the `DISCRETE` block implementation and documentation in `NEURON` and add related unit test in `NMODL`.

Take for example: ``` NEURON{ POINT_PROCESS NMDA RANGE B } PARAMETER { mg = 1. (mM) : external magnesium concentration Cdur = 1. (ms) : transmitter duration (rising phase) Alpha...

state_discontinuity functions not defined

4

`state_discontinuity` functions from mod files in https://github.com/Neurosim-lab/netpyne/tree/coreneuron/examples/coreneuron/m1/mod is not defined in `cpp` and `ISPC` backends. `my_exp2syn.mod` ``` x86_64/corenrn/mod2c/my_exp2syn.ispc:69:9: Error: Undeclared symbol "state_discontinuity". state_discontinuity(inst->B[id], inst->B[id] + (*weight) * inst->factor[id]); ^^^^^^^^^^^^^^^^^^^ ```...

In the vectorized kernel loop and the epilogue loop we do some variables transformantions and renaming that might lead to bugs in the generated code. First of all we convert...

Atomic SIMD instructions should be handled by doing a reduction loop at each loop iteration of the vectorised code similar to the external loop that exists in the ISPC generated...

Currently class `Platform` is used to store information for all platforms (CPU and GPU). Instead of this we should define new derived classes `CPUPlatform` and `GPUPlatform` from `Platform` that include...

[LLVM] Atomic operations in GPU generated code

2

Some mod files (mostly for synapses) update the current of the cell in parallel and thus this update needs to be done atomically or using a scatter-gather policy like we...