I was using a 14:1 geared extruder with 1/32 microstepping (DVR2588 driver), and I noticed that the printer was not reaching full speed. I kept trying to speed it up with M220 to see if the smoother drive would help prevent skipping, but it never got faster.
The extruder step rate seems to have been the problem. I just switched the machine to a direct-drive 1.75mm extruder, decreasing the motor steps to extrude the same volume by a factor of about 4.5, and the machine prints a LOT faster… except when minimum layer time comes into play, which now is almost always.
I can haz Cortex M-series?
It’s being worked on. Both on current hardware and on Arms. 
@Rob_Giseburt I don’t think I asked you, but how does the Motate system compare in terms of the speed that it can pulse steppers to the “fastio” library used by current firmware? I know Motate is a lot faster than the call in the built-in arduino libraries, but presumably fastio is too.
I don’t know. Let me look and get back with you on that.
It appears that it’ll be approximately the same. (I would have to disassemble to see if the optimizer picks up the cues right, but it looks like it should.)
The code in fastio is very specific to that firmware, vs @Motate pin code being very generic and reusable. Fastio code is also very dense … not nearly as easy to read and parse. This will make debugging much harder. (It is a bunch of macros, so the generated code is harder to determine. Motate also uses macros, but only to generate a template that is much clearer.)
Motate will also include much more that just pin handling, of course. It is also cross platform. So far three: AVR ATMega, one flavor of ARM (of the Due), and AVR ATXMega.
All of that is what I would have guessed, thanks 
Smoothie is coming out soon, and happily drives my printers at ludicrous speed (1.25m/s feedrates with 1/16 microstep!)… whether or not your mechanics can keep up is another matter