Test report by @migueldiascosta
SUCCESS
Build succeeded for 2 out of 2 (2 easyconfigs in this PR)
ip-172-31-8-244.us-west-2.compute.internal - Linux RHEL 7.6, AMD EPYC 7571, Python 2.7.5
See https://gist.github.com/a4196f02488be8466df95e07ed555fc9 for a full test report.

I wonder what the benefit of this blis is over the upstream blis? As far as I know all AMD changes were merged back but that was >3 months ago. Would be good to see a diff.
(just trying to keep complexity down if this also is going to end up in toolchains).

@bartoldeman more than the current diff, it would be good to know if they will always merge back

Currently the diff is minor, "Improved single thread Triangular Solve Matrix (TRSM) performance for small matrices on AMD EPYC Family processors", flame/blis@master...amd:master

I agree that we should only consider an amd toolchain if there is stronger reason for it, but for now the easyconfigs don't hurt (?)

Also, I'm happy with leaving this (and #7164 and #7165) as WIP until more is known

@migueldiascosta ping on the versionsuffix suggested by @boegel ?

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node3118.skitty.os - Linux centos linux 7.6.1810, Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz, Python 2.7.5
See https://gist.github.com/c310f3b9ae241e146e299570d7550101 for a full test report.

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2039.delcatty.os - Linux centos linux 7.6.1810, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/1229511ab29a220801f676f9cce6ae36 for a full test report.

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2406.golett.os - Linux centos linux 7.6.1810, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz, Python 2.7.5
See https://gist.github.com/b56316ba2eeee8055e279569dd6d66e3 for a full test report.

Going in, thanks @migueldiascosta!