Performance: float(AbstractArray) and complex(AbstractArray)

[catawbasam]

The test_profile() function used to demonstrate https://github.com/timholy/ProfileView.jl shows a performance hit from an anonymous function at abstractarray.jl line 319, which is complex(A::AbstractArray) . map_promote() in that function contains an anonymous function that creates a performance hit. Switching to a named function speeds test_profile() by about 3.3x on my 64-bit Linux system. cc: @tholy

Profile View example with current float(A::AbstractArray) and complex(A::AbstractArray) --
http://nbviewer.ipython.org/gist/catawbasam/1fd6a82605a54b052425

Profile View example with proposed float(A::AbstractArray) and complex(A::AbstractArray) --
http://nbviewer.ipython.org/gist/catawbasam/e1768937952565213561

[ivarne]

Probably you intended to ping @timholy

[catawbasam]

@ivarne : yes. How does that work?

[timholy]

You did it fine except you wrote it as tholy rather than timholy.

Off topic, but thanks for pinging me: before now, I hadn't realized that ProfileView works exactly as intended when viewed through the notebook viewer. When used directly in IJulia, there's still something not updating, and I don't understand why. I'll file an IJulia issue.

[catawbasam]

wrt to ProfileView in IJulia, it seems to work fine in FireFox on 64-bit Linux. On 64-bit Windows 7, IJulia within Chrome does not display the function info properly. And thanks for creating ProfileView. It's a fun tool.

[timholy]

For further discussion of the off-topic IJulia issue: JuliaLang/IJulia.jl#196
(but thanks for the tip, @catawbasam! I'll go check that out.)