Add chicoma-gpu

[altheaden]

In this PR I added the necessary files to add chicoma-gpu to mache. The files for the nvidia and gnu compilers are symlinks to the existing chicoma-cpu files, while the nvidiagpu and gnugpu files are newly created.

Checklist

• Testing comment in the PR documents testing used to verify the changes

closes #180

[altheaden]

@xylar will do testing on chicoma since I do not have access. It's likely that some changes still need to be done to the nvidiagpu and gnugpu files, but I did try to update them via what's in the config_machines.xml file to get things started.

[altheaden]

I realized after committing this that there is some documentation referencing the supported machines. I'll update that to include this machine as well and push that to this branch tomorrow.

[altheaden]

@xylar In regards to adding new machines to discover.py, do we want to add another regex to find chicoma-gpu from its hostname? I noticed that we don't have an entry for pm-gpu, so I'm unsure if we need a separate one from chicoma-cpu. If we do, I need to know what the hostnames look like for chicoma-gpu, I've looked at the documentation that I have access to but I can't find anything about it.

[xylar]

@altheaden, no we can't do that because chicoma-cpu and chicoma-gpu (just as pm-cpu and pm-gpu) have the same login nodes so the hostname at least on the login nodes can't tell us which is which. We default to the -cpu version and make the user specify if they want the -gpu version.

[xylar]

Thanks for asking, though!

[altheaden]

@xylar thanks! I realized that the documentation was referring to hostnames, so nothing needs to be updated for this PR after all, and my work here is done as far as I can tell.

[xylar]

@altheaden, I'm going to put this on hold until I get #188 working (hopefully not too long) since those changes will also apply here.

[xylar]

This isn't a high priority for now.