I’ve been trying to export area lights to Mitsuba 2, and ran into some issues converting the Power value specified in the UI (W) to the irradiance value (W/m²/sr) Mitsuba 2 expects. Given a planar surface emitting along its normal direction, the conversion should be: irr = P / (area * 2pi).
However, using this formula to compute the irradiance gives different results. As I mentioned here, a 1m² plane with an emission shader does not emit the same amount of light as a 1m² area light with the same power value.
I dug in Cycles code and found out that the eval_fac of KernelLight in the area light case is computed as 1 / (area * 4.0) (code here), I think it should be 1 / (area * 2.0 * pi) to actually be in the correct unit, as claimed by @brecht here: Cycles, unit of light energy ATTN: Brecht - #2 by brecht
Using this conversion factor for exporting area lights in Mitsuba 2 produces the same render as Cycles, confirming that it is indeed this scale factor.