Chris Elrod

> > Add a test dependency on something like StaticArrays.jl (or StrideArrays after it is registered). Testing with mutable statically sized arrays (meaning `StaticArrays.MArray`) would let us hit `inlineloopmul!`. It...

> > Yes, that's fine. Note that Octavian should be a lot faster for 10x10: > > Yeah, I remember that we start beating StaticArrays pretty early. What's the biggest...

> I'm not really sure on how to proceed here Have any questions/anything blocking you?

``` coverage=true/Julia nightly/3 threads/ubuntu-latest/x64/pull_request succeeded 2 minutes ago in 6m 50s ``` Yeah, that's way faster. I'd be in favor of disabling coverage on CI except for nightly, honestly.

> I noticed after submitting this that there's another CI config file for nightly. This PR might be duplicating runs? The CI situation seems a little weird. Seems nightly doesn't...

> By the way, the new codecov option will be on 1.8 but needs manual override, which needs exposing in `julia-actions/julia-runtest` for that to work. Does this mean we need...

With coverage enabled, Julia 1 takes anywhere from half an hour to over 4 hours: https://github.com/JuliaLinearAlgebra/Octavian.jl/runs/5476242675?check_suite_focus=true

It is odd, however, that codecov is reporting

https://github.com/JuliaLang/julia/pull/44625#event-6272201431 Do you know when/how often the nightly builds are rebuilt?

Setting `UBSAN_OPTIONS='print_stacktrace=1'`: ```stacktrace /home/chriselrod/.cache/CPM/googletest/a36d6fcbd7356d396ea479e6b11f4b9760ad0c04/googletest/src/gtest.cc:2727:32: runtime error: member call on address 0x000001b19500 which does not point to an object of type 'TestFactoryBase' 0x000001b19500: note: object has invalid vptr 00 00 00...