List view
This milestone focuses on the code generation component of the NMODL framework. Up to this version the code generation had many deficiencies in terms of its design that made it awkward to add new backends. Especially towards v0.2 many small changes needed to be added quickly without taking a step back and reconsidering the overall design. The redesign of the code generator backends should be done hand-in-hand with data-structure updates in CoreNEURON to take these into account
No due date•3/9 issues closedThe goal is to complete the integration of accelerator backends and to implement a number of low-hanging-fruit type of improvements in the framework: - integrate CUDA/OpenACC codegen backends - improve CUDA unified memory support - test and benchmark accelerator backends
Overdue by 6 year(s)•Due by June 28, 2019•19/25 issues closed