Open YuryPakhomov opened 8 months ago
PRs are welcome, as are references to papers explaining the theoretical parallelisation.
A quick review of the README reminds me that this thing can use LAPACK if installed. Some BLAS implementations use multi-cores and/or GPUs. Have you experimented with that? Can you at least provide example data/code that demonstrates what you'd like to operate in parallel?
This is not issue, but request to improve Levmar module. Now the module uses only one CPU core and sequentially calculates the gradient - partial derivatives for all parameters. This operation can be paralleled theoretically due to independences of each partial derivative. Is it possible to implement this in the module?