Float64 reduction of Float32 computation idiom using @distributed?

[fghorow]

Thanks for the hint @JeffreySarnoff!

I am fairly new to Julia. However, if I am reading the code for the @distributed macro correctly, the argument [reducer] – once passed to the preduce function – can not be a reduce call, but only the ‘binary op’ part of the call to reduce. (Please correct me if I am wrong – preferably with a working toy example! )

I guess my original question still stands then. What is the idiom within the context of the @distributed macro for performing a Float64 reduction of a Float32 computation?