Hi Alex, Alex Nitz <[email protected]> writes: > I would like to be able to do in-place multiplication of a gpuarray by > another gpuarray. Currently, only mutiplication by scalars is allowed. It > seems like all the functionality is already implemented but not used, is > there some issue that might be blocking this feature?
No, it was just missing AFAICT. > I've attached a patch that enables in-place multiplication by another > gpuarray. I've only tested that it works for the cases that I needed. > (mutiplication by the same dtype) Thanks for the patch, applied. > On another note, is there a preferred method to submit patches or even new > features? I was thinking of implementing the equivalent of numpy.vdot. Is > this something that would be ok to have in pycuda, or should I just keep it > in my own code? It is a simple modification of the existing dot function, > but if the first argument is complex it is conjugated. Email with a git patch set is perfect. Pull requests on github work, too. Functionality-wise, my guideline for what to stick into PyCUDA and what to leave out is something fuzzy around 'should be reasonably general-purpose and commonly used'. vdot seems perfectly fine to me. Andreas _______________________________________________ PyCUDA mailing list [email protected] http://lists.tiker.net/listinfo/pycuda
