Description of problem: Please BuildRequire: openblas instead of atlas where available (x86_64 %{ix86} armv7hl %{power64} aarch64 s390x, see %{openblas_arches} macro defined in openblas-srpm-macros package). Currently, it's available on all primary Fedora arches, so you can technically skip using the macro. Version-Release number of selected component (if applicable): ergo-3.7-2.fc29 Additional information: OpenBLAS selects best code path for the target architecture at runtime unlike ATLAS. A lot of benchmarks around the net suggest openblas is faster than ATLAS in most workloads: https://github.com/tmolteno/necpp/issues/18 http://blog.nguyenvq.com/blog/2014/11/10/optimized-r-and-python-standard-blas-vs-atlas-vs-openblas-vs-mkl/ https://freddie.witherden.org/pages/blas-gemm-bench/ http://nghiaho.com/?p=1726 Many Fedora packages have switched to openblas already (e.g. scalapack and all dependent ones and numpy).
It used to be OpenBLAS. But the tests fail with it.
Have you reported the failures upstream? What did upstream say?
Upstream told me to use some other BLAS library than OpenBLAS and that made the problems go away.
As the maintainer of OpenBLAS I'm not sure I agree with the push, since OpenBLAS is *not* as stable as one would hope. There's constantly problems on some processor model with incorrect results, and upstream also makes large changes with insufficient testing, that then blow up in Fedora. Like the recent update to 0.3.2...