kanduri authored 5 years ago and Benjamin Cumming committed 5 years ago

Remove managed memory from the GPU stack data type that is used for spike collection on the GPU back end. Now the stack has explicit synchronization points, that are called from the host, for:
* copying stack state from GPU to host memory where it can be interrogated.
* resetting the stack state on the GPU to empty.

Fixes  #810.

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

Also remove vestigial segment::length_constant() virtual function.

Fixes #831.

Addresses #824.

* Add modcc support for `CONSTANT` blocks in NMODL, subject to the following constraints:
    - Any identifier declared in the `CONSTANT` block may only be used after that declaration (including within the `CONSTANT` block itself).
    - Units in the `CONSTANT` block are parsed but not processed.
    - Values assigned to identifiers inside the `CONSTANT` block may only be signed numeric literals, or an already declared constant, possibly preceded by unary minus.

* Collect cable cell parameter setting structures into `include/cable_cell_param.hpp`.
* Restructure electrical specifications (axial resistance, membrane capacitance) and ionic concentrations and reversal potentials on cable cells, so that these can be specified with a global default, per-cell defaults, and per-segment values.
* Allow reversal potentials to be set by a mechanism of a new kind 'revpot', which are prohibited from maintaining state or writing to any shared state other than ionic reversal potentials.
* Specify reversal potential mechanisms as global defaults or per-cell. Reversal potential mechanisms may not be specified at the level of a segment in order to avoid non-linearities arising from the discretization.
* Supply default cable cell parameter data that matches NEURON values (this is _not_ used by default).
* Replace the d_lambda calculation with one that approximates more faithfully the effect of tapered segments, and which will ...

Addresses issues #814 and #815.
* Change the discretization so that a spherical soma comprises its own CV.
* Modify the `div_compartment_integrator` to accommodate the single CV soma, adjusting the discretization of the most proximal branches, adding a half-size CV attached to the soma, so as to leave the remainder of the discretization the same.
* Rename the confusing `ncomp` to `ncv` inside the discretization code.

Implement implicit derivation of mechanisms.

* When a mechanism of the form "foo/param=value,..." or "foo/newion" is
requested from a mechanism catalogue, attempt to perform a derive the
information or override set from the mechanism "foo" using the given
global parameter settings and/or ion remappings given in the name.
* Split mechanism catalogue implementation into a private catalogue
state with non-throwing behaviour and without any implicit mutation,
with mechanism_catalogue methods responsible for throwing any generated
errors, with the aim of avoiding throw/catch in expected program flow.

Implements #807.

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

* correct doc for meter_report and small word fixes
* shift make_report to next section

Remove the dead `cell_connection_shim` from the python wrapper.

Fix the connection printer to take an `arb::cell_connection` argument, to fix a runtime python crash caused by passing an `arb::cell_connection` where a `cell_connection_shim` was expected.

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Wraps the `meter_manager` with
- the constructor
- `start(ctx)`
- `checkpoint(name, ctx)`
- `checkpoint_names`
- `times`

and the `meter_report` with
- `make_meter_report(meter_manager, ctx)`

Fixes #765

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Add docs for Python wrapper to ReadTheDocs:

- Overview, Common Types, Hardware Management, Recipes, Domain Decomposition, Simulations, Metering
- Installing Arbor: Optional Requirements (Python), Buidling and Installing (Python Frontend), and Installation (Python Module)

Missing (, since not implemented yet): 

- probes
- arbor-sup 
- hint maps in domain_decomposition
- reset, events, empty schedule in event_generator
Also does not cover unit testing (since doc is user-centric).

Makes also defaults and units in wrapper consistent.

Fixes  #766

Nora Abi Akar authored 5 years ago and Benjamin Cumming committed 5 years ago

Add missing check of whether to use direct or indirect indexing for simd procedure calls.

fixes #804

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

Address inconsistencies in unit conversions between current updates, initialization, ionic currents etc.

Fixes issues #793 and #798.

* Correct units in comments for ion state iX_ to a current density.
* Remove conversion factor for NMODL current densities [mA/cm²] to Arbor current densities [A/m²] from layout weight: from the point of view of Arbor, mechanisms now write in units that Arbor uses natively.
* Remove weight_ as a magic modcc variable; applying weight_ is now performed directly by code generated by printers.
* Change IndexedVariable interface: remove redundant index_name; remove 'op'. Whether external data is accumulated or assigned is now a property of the corresponding sourceKind.
* Conversion factors for NMODL <-> Arbor are collated in printerutils.cpp, determined by the external data sourceKind.
* Permit direct assignment of external data, as opposed to a weighted sum. This behaviour will be required for reversal potential handling.
* `current_` and `conducitivity_` get a different sourceKind in point mechanisms, allowing us to distinguish current contributions from current density contributions (and applying correct unit scaling accordingly.)
* Add unit test for correct initialization and scaling of ionic currents, together with two corresponding unit test mechanisms.

* Add `::arb::` decoration to `fvm_value_type` and `fvm_index_type` in modcc simd cprinter output.
* Return stream from unwind stack trace printer.

Fixes #801 and also a long standing (but low impact) bug in unwind code.

* Correctly decorate required names from arb:: namespace.
* Put the unit test mechanisms in the 'testing::' namespace, to exercise this functionality.

Building validation data with python2 can raise an import error for `builtins`.

* Catch import error, and import from `__builtin__` instead.

Support for recording spikes generated by a simulation in the Python wrapper
* Implement a `spike_recorder` that holds a shared pointer to a `std::vector` of spikes, and a callback for the `arb::simulation` spike recording API.
* Add `python/example/ring.py` that creates a ring network, then records and prints spikes.
* Some fixes to get the full `recipe` -> `domain_decomposition` -> `simulation` -> `spikes` workflow to work
  * always use default `global_parameters`: user customization of global parameters for cable cells can wait until the ion species interface is finished.
  * change the Python recipe interface for `recipe::connections_on` to use `pybind11::objects` because of shim.
* Some small improvements to error and help messages.

Fixes #764 

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

* Make global parameters and ion rebindings part of the instantiate interface, rather than insist that all concrete mechanisms implement these as methods.
* Mechanism catalogue instance() returns a pair, comprising the concrete mechanism for the requested backend, together with the override data.
* Extend catalgoue derive() method to take a list of old ion name -> new ion name remappings for a mechanism.
* Add exceptions for ion remapping errors, and check for these errors.
* Add convenience function for reparameterizing a mechanism with a single ion dependency over other ions. (This will be used for the future nernst pre-supplied mechanism.)
* Add unit tests for: chained renamings of ion names across multiple derivations; correct shared state ion assignment after renamings; ion remapping exceptions; parameterize_over_ion.

Eric Müller authored 5 years ago and Benjamin Cumming committed 5 years ago

* cmake/FindUnwind.cmake used the wrong variable (caused by commit ad1c78ab)
* CMakeLists.txt seems to use wrong syntax for providing a compiler define
  of `WITH_UNWIND` (used in `arbor/util/unwind.cpp`)

Fixes #782 
* consistent formatting of strings returned by `__repr__` and `__str__` for wrapped types
* rename enum wrappers to be less verbose, where it makes sense.

Add full wrappers for

- `lif_cell`
- `spike_source_cell`
- `benchmark_cell

And an opaque wrapper for cable_cell with a helper function that generates a random morphology identical to the ones in the nsuite benchmarks.

Fixes #767.

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Wrap `arb::simulation` type and interface for running `run()`, resetting `reset()`, and `set_binning_policy(policy, bin_interval)`. 

Wrap `enum binning_kind`, used to set event binning on `simulation` API.

Fixes #763 

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

Extended USEION VALENCE support.

* Augment modcc parser to accept identifier or signed integer after VALENCE.
* Treat identifer after VALENCE as global scalar, tied to ionic charge.
* Add ionic charge field to mechanism ion parameter pointers, replacing scalar in shared state with a length-1 array.
* Verify presence of ion name and check ion charge against numeric valence in mechanism info when discretizing in `fvm_build_mechanism_data`.
* Pass ion info, coalesce_synapse global state to `fvm_build_mechanism_data` just via cable cell global properties.
* Add more unit-test only mechanisms (test_ca_read_valence, test_cl_valence).
* Add unit tests for valence verification and reading into mechanism state (multicore only).
* Update existing unit tests for new interfaces.

Implements #780.

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

* Change occurances of 'ionKind' to corresponding string values in test_parser.cpp.
* Add unit-modcc to travis script.

Fixes #775.

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Domain decomposition in python wrapper
* functions: `partition_load_balance`
*  types: `group_description` and `domain_decomposition`

Also includes
* remove comment that `gid` in a `group_description` must be sorted.
* a host of small clean ups of python wrapper

Fixes #761

clean up stringification in python wrappers

move python wrapper for schedules to their own files

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Fixes #760

Wraps arbor recipe (without probes, i.e. num_probes, probe_info, get_probe) including 

- cell_connection, 
- gap_junction_connection, 
- recipe with
  - num_cells
  - cell_description
  - cell_kind
  - num_sources
  - num_targets
  - num_gap_junctions_sites
  - event_generators
  - connections_on
  - gap_junctions_on
  - global_properties
- enum cell_kind in `identifiers.cpp`

Addresses #757.

* Check if the multiplicity vector is empty before trying to use its contents. 
* Add unit test to catch the problem

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

Fix a memory error triggered on the GPU when assertions were enabled. An assertion was testing a value in shared memory, that should have been stored in host memory.

Python: mpi4py communicators as first class citizens

Sam Yates authored 5 years ago and Benjamin Cumming committed 5 years ago

* Replace ionKind enumerations with corresponding string values.
* Rename `ion.hpp` as `ion_info.hpp`.
* Re-jig 'nonspecific' current handling in modcc: make explicit the external source of the corresponding current variable, and treat as non-indexed those locals with an external variable that has no data source, as opposed to special-casing the `ionKind::nonspecific` value.
* Add another field in ion dependency information that captures a write to an ionic reversal potential.

Fixes #748.

akuesters authored 5 years ago and Benjamin Cumming committed 5 years ago

 + Included conversion from `pybind11::object()` to `arb::util::optional` for `arbor.proc_allocation().gpu_id` and exception handling for `arbor.context(mpi=comm)`.

+ cleaned-up `test_event_generator.py`

+ for discussion: How do we want to handle the strings? Include them in appropriate `.cpp` or collect them in `strings.c/hpp`?

Fixes #746.

* Make `one_probe` an `inline` function rather than `static inline` in order to work around a linker issue (see issue #746).

* Fixes segmentation faults; output compiler errors instead.
* Consume units if provided in function/procedure prototypes or after function prototype.
* Consume UNITSON/UNITSOFF.
* Consume and store VALENCE as an ion property.