You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now the numpy.linalg.solve step is a slow down for any of these methods.
If the user solves the for M^-1 * F analytically then we don't need that. This can speed up things right now.
Later on we need to push the solve step to a low level in the generators so there is no python overhead. I'm not sure that the full mass matrix and forcing vector need to be used like this either, especially when the q dots equal the u's, then there is no reason to solve the full system. Finally, we need to do some learning and thinking about the Jain method in which you supposedly don't even have to form the mass matrix and do this solver.
The text was updated successfully, but these errors were encountered:
Issue by moorepants from Wednesday Feb 12, 2014 at 22:57 GMT
Originally opened as https://github.com/pydy/pydy-code-gen/issues/21
Right now the numpy.linalg.solve step is a slow down for any of these methods.
If the user solves the for M^-1 * F analytically then we don't need that. This can speed up things right now.
Later on we need to push the solve step to a low level in the generators so there is no python overhead. I'm not sure that the full mass matrix and forcing vector need to be used like this either, especially when the q dots equal the u's, then there is no reason to solve the full system. Finally, we need to do some learning and thinking about the Jain method in which you supposedly don't even have to form the mass matrix and do this solver.
The text was updated successfully, but these errors were encountered: