Alex Arslan

Hm, okay, that's bizarre. You said you encountered the same problem with SCS? Do you have memory allocations for that?

I'm not sure, actually, though it would appear so. Are these version checks actually called somewhere during the solving process?

If I'm not mistaken, it looks like printing banners is done by the C code itself rather than by the Julia wrappers, so I don't think that would be reflected...

So this is interesting: ``` - function has_conic_form(conic_forms::UniqueConicForms, exp::AbstractExpr) 133353998 return haskey(conic_forms.exp_map, (exp.head, exp.id_hash)) - end ``` Here `conic_forms.exp_map` is an ```julia OrderedDict{Tuple{Symbol,UInt64}, OrderedDict{UInt64,Tuple{Union{AbstractArray,Number}, Union{AbstractArray,Number}}}} ``` I would not be...

Unfortunately not without a huge undertaking to restructure the way Convex uses and expresses types. For reference, `Value` in Convex is actually `Union{AbstractArray,Number}`. (As an aside, because of this, Convex...

I ran the code with `@profile` and interrupted it after a bit, then leafed through the results. Anecdotally it's spending a fair amount of time on sparse matrix operations, but...

I tried that as well, but I'll also note that Convex uses sparse matrices in a number of places regardless of whether any particular input is sparse: ``` $ sift...

I have a local branch of Convex that uses FillArrays and LazyArrays in place of SparseArrays where it makes sense to do so. Running the example in this issue, I'm...

Every time I've interrupted the computation, it seems to be in the middle of this: https://github.com/JuliaOpt/Convex.jl/blob/master/src/conic_form.jl#L62-L71. Note in particular the comment that says it's time consuming.

There were discussions about this a while back (though I'm struggling to locate one now), and ultimately it came down to the original package author's personal preference. Apparently observations-as-columns is...