The CPU the MegaSquirt uses has a hardware divide function. The ATMega328 doesn't come anywhere near the capabilities. The Mega might be able to do huge lookup tables with no interpolation between them, but why reinvent the wheel, per say?
If I were to do it, I'd either wait for the ARM based Arduino, or grab one of the other already available ARM boards and use one of them. The STM32F4 has an actual FPU, that'd put it well above even the CPU in the MegaSquirt3, I'd imagine. Still, writing all the timing code they have in there, good freakin luck with that.