trmv! gives warning for moderately large matrices

[dlfivefifty]

julia> using LinearAlgebra

julia> BLAS.vendor()
:openblas64

julia> n = 10000; A = randn(n,n); b = randn(n); BLAS.trmv!('U', 'N', 'N', A, b);

WARNING unrolling of the trmv_U loop may give wrong results

The offending line seems to be https://github.com/xianyi/OpenBLAS/blob/e11126b26ada8d97b4a522e461ca92311653bfc6/driver/level2/trmv_U.c#L66

See discussion https://discourse.julialang.org/t/why-lapack-trtrs-not-blas-trsv/14239/3

[martin-frbg]

This is related to #1332. Is it the same hardware as in #1698 ? I did not expect that loop to be entered on "common" hardware after the hack, but clearly your definition of "moderately large" is different from mine. Probably best to fully disable the suspicious code block until the actual issue is found and fixed. (Too bad that these issues with original codes from GotoBLAS come up only now.)

[martin-frbg]

Actually BLAS-Tester results (both with the smallish default problem sizes and larger ones like ./xdl2blastst -R trmv -U 1 U -n 15000) suggest that the warning is unnecessary as the results appear to match the reference values even with unrolling of the gemv loop (as long as trmv is not allowed to be parallelized again - this conclusion from #1332 remains valid).