> At least in Python, trips in and out of ADOL-C almost count for nothing as they're just passing pointers around.
This might be the case in Julia as well. But I think you'll agree that ADOL-C and friends are much more opaque and less hackable than Julia code. > At least on OSX, both can be installed with brew with one command. Yep, Homebrew is pretty awesome, and Julia leverages that awesomeness appropriately (Homebrew.jl). If you have a Mac. We don't all have Macs ;-) > And I don't mind dependencies if they provide state-of-the-art tools. Neither do I, but the flipside of this is segfaults and linker errors and "oh great, Apple changed the C++ standard library," etc. etc. etc... > We're probably coming from different angles. My goal is to assemble numerical methods rather than to provide a modeling environment. Yes, probably. I've spent several years of my PhD trying to come up with ways to do better than Ipopt, without a whole lot of success (probably because I spend too much time getting distracted by the mailing lists of open-source projects...). I've gotten much more traction in creating modeling environments that make Ipopt easier to use by non-experts.
