search

mumax / 3

GPU-accelerated micromagnetic simulator
Other
450 stars 150 forks source link

Add Helical initial magnetisation function #202

Closed rpep closed 5 years ago

rpep commented 5 years ago

Hi there,

This function just adds a function which allows a helical initial magnetisation propogating along (1, 0, 0). I had some trouble with gofmt not working on the machine I compile on, so I had to disable it - if there are any changes that I'd need to make to get this accepted then please let me know.

Best wishes,

Ryan Pepper University of Southampton

JeroenMulkers commented 5 years ago

This is definitely useful, but it might be even better if we could generalize this so that it works for ever (3D) wave vector.

I am also not sure if the implementation is fully correct. This probably need some testing before we accept the pull request.

rpep commented 5 years ago

OK, let me edit it and come back to you on that, shouldn't be too difficult. I'll put together an example as well.

rpep commented 5 years ago

I've just tried messing around locally on my own machine; I can't currently build against the Master branch even after upgrading to CUDA 10.0 - I get the error:

No binary for GPU. Your nvidia driver may be out-of-date

Any ideas? I'm using Ubuntu 18.04

godsic commented 5 years ago

Which gpu and driver version are you using?

kkingstoun commented 5 years ago

You have to update the driver.

rpep commented 5 years ago

Not sure what I was using previously (although with that driver version the 9.0 toolkit worked OK) as I've now uninstalled it in trying to fix it; are you guys building against drivers in a PPA for Mumax generally, or from the binary distributions on NVidia's website?

rpep commented 5 years ago

The card is a 750 Ti, so fairly old, although it's worked fine up til now.

godsic commented 5 years ago

The card is still supported. I would strongly suggest you to setup NVidia own repository as it provides both cuda and appropriate driver.

rpep commented 5 years ago

I've just done that; seems to work fine now - thanks a lot. I'd used the local runfile installer previously and I suppose there were some conflicting files lying around.

godsic commented 5 years ago

@rpep Please let us know when the pull request is ready to go.

rpep commented 5 years ago

Will do, thanks.

JeroenMulkers commented 5 years ago

@rpep Small note: when you specify the wave vector q, you do not have to specify the wavelength L since q = 2*pi/L.

Your idea to add the helical state to the available magnetization configurations, inspired me to implement also the general conical configuration (this includes the helical state, cycloidal state, and anything in between). See commit 29a6475.

I believe it is good practice to implement only the general case (conical state) and make a user friendly wrapper for the special case (helical state). Please, have a look at commit 29a6475 and let me know if you agree?

rpep commented 5 years ago

Aha, I was thinking about that too. I wasn't sure whether it was a better interface for people to think about the helical length separate to the direction, because it's somewhat natural for people to state things like [110], [101], etc in papers and that would line up more consistently with that.

I agree; and your way looks much simpler than my way! I'll close this now; thanks!