Dominique
Dominique
Thanks @rscohn2! With your action, is it possible to specify the version of the components?
Many thanks! I notice that icc and ifort are not among the possible choices. Unless I'm mistaken, that means the action cannot be used to test macOS builds. Is there...
It's ok to change the mod files that way. The disadvantages are that 1. there are other reformulations that you might want to try, e.g., `x[1] * y[1]
It's waiting for a motivated contributor to make a pull request ;-).
I was working on this too and I observe segfaults. ELFUN, etc., are always double precision. I have to try and write a single precision Fortran driver to see if...
Same here. Decoding the problem with `--single` has no effect. I don't think the problem is in the Julia interface.
The iterate norm is monotonic, but the residual is not monotonic in LSLQ. So I don't think you can just plug it into a trust-region method.
How about `1 - eps(T)` ?
Not so sure if this operation is available in ASL. Perhaps you have to set y = 0.
I think it may be because there's no default formatter for vectors: https://github.com/JuliaSmoothOptimizers/SolverBenchmark.jl/blob/be8d2f1eb4bbd88b95e6c2cf6f458606559f4224/src/formats.jl#L3 and then we get stuck here: https://github.com/JuliaSmoothOptimizers/SolverBenchmark.jl/blob/be8d2f1eb4bbd88b95e6c2cf6f458606559f4224/src/formats.jl#L66 The question is: what are you expecting if some solver...