Update machine files to match new directory names at NERSC

[jonbob]

This PR updates the machine files to reflect directory name changes from "acme" to "e3sm" on cori-knl and cori-haswell. It was cherry-picked from #3516 with some other slight modifications.

[BFB]

[darincomeau]

@jonbob thanks for bringing this in. I built and ran a GMPAS-DIB-IAF-ISMF.T62_oEC60to30v3wLI 5 day test successfully - any other testing you'd like me to do?

[jonbob]

How'd you beat me through the queues?? No, if you're happy then I'm happy, @darincomeau !

[stephenprice]

If this is the new 1.2 tag I'm happy to try out a B-case test at some point.

[darincomeau]

@stephenprice I think what you're really after is #3513, which should be the last of the code changes needed to bring into 1.2.

[jonbob]

@stephenprice - I'll aim to get both PR's in as soon as possible, so hopefully we'll have something for you to test on Monday

[stephenprice]

Great. I'll look for it. Thanks

[jonbob]

@ndkeen - whenever you have a chance, could you please make sure this looks right to you? The cherry-pick of your commit wasn't clean and I just want to be sure

[jonbob]

@darincomeau - what compiler did you use for your test? So far my tests are passing with intel but failing with gnu...

[darincomeau]

@darincomeau - what compiler did you use for your test? So far my tests are passing with intel but failing with gnu...

I only used intel - I'll try gnu now.

[jonbob]

@darincomeau - for what it's worth, I don't think the gnu problem is related to this PR...

[jonbob]

And it may be an issue mostly for CAM -- I tried a SMS.ne4_oQU240.A_WCYCL1850.cori-knl_gnu test and it fails when it's initializing the atm

[ndkeen]

Looks good. I see I also missed the "acme" in config_compilers which would affect ALBANY builds. So will have to make another PR to correct that on master.

[darincomeau]

And it may be an issue mostly for CAM -- I tried a SMS.ne4_oQU240.A_WCYCL1850.cori-knl_gnu test and it fails when it's initializing the atm

Ah yeah I was only doing a G-case. My gnu case compiled, just waiting in the queue, but I suppose that won't be all that helpful.

[jonbob]

Thanks @ndkeen

[jonbob]

@darincomeau - it will still be useful to see if gnu has problems beyond CAM. Thanks!

[darincomeau]

Yeah the G-case with gnu did indeed die, I think at the first time step. I don't see any errors in the component logs, but from the e3sm.log:

 133:
 133: Backtrace for this error:
 134: #0  0x20f205af in ???
 134:   at /home/abuild/rpmbuild/BUILD/glibc-2.26/nptl/../sysdeps/unix/sysv/linux/x86_64/sigaction.c:0
 134: #1  0x204f3c2d in ???```

[jonbob]

Yeah, I saw similar things. But I'm pretty sure it has nothing to do with these paths -- so I'll get this PR in and maybe we can open an issue about gnu

[jonbob]

Oops, I put this in the wrong PR:
@ndkeen has an open issue about GNU being broken on cori (#3270)

[jonbob]

merged to maint-1.2