Great stuff. The transition document is actually linked to from the blog post, so not exactly secret ;-). Regarding the transition doc, the benchmarks would probably be clearer and more impactful expressed as speedup relative to Matlab. Relative time is a decent metric when the baseline is faster (e.g. typical comparison to C), but when the baseline is slower, speedup is generally a better metric with the additional advantage that bigger is better, which seems to be more intuitive as a general rule.
On Thu, Dec 3, 2015 at 11:15 AM, Tim Holy <tim.h...@gmail.com> wrote: > Very nice. I'd be curious to know what you see as the major shortcomings of > the profiler? Personally, I like the Julia profiler better than the Matlab > one. > That's conditional on using ProfileView, of course. > > Of course, I'm heavily biased since I wrote most of those tools myself. A > list > of shortcomings would be useful perspective. If they're mostly > visualization > rather than performance, might be best to file as an issue over at > ProfileView. > > --Tim > > On Thursday, December 03, 2015 07:06:29 AM Spencer Lyon wrote: > > Ahh nice, > > > > You found the draft of our next post ;) > > > > Good point about matlab.jl. we did make heavy use of that when > transitioning > > to julia. > >