Did you find any measurements that show that, say >0.1% power gain can be acquired when going from 16 to 32 RPM bins (and the ECM uses interpolation) ? If yes, I understand the "need" word. Without interpolation many bins are nice, but with interpolation (that VEMS uses) 16 should be enough (the difference between interpolated and optimal values stays below tuning error) even with equidistant bins.

So ign/fuel tables need to be bigger than 16x16. For exemple, on the Aprilia RSV Denso ecu, the tables are 16x29, with fixed bins at 400rpm.

On GenBoard/UnderDevelopment/AlphaN it is mentioned that VEMS uses TPS at (configurable) low RPM (note: RPM ! not load ) and then MAP above a certain (independently configurable) RPM. Interpolation (blending) is used in between. The explanation is that at higher RPM the MAP is smoothed out (=> usable), while at lower RPM MAP fluctuates during 1 period, and synchronized sampling (similar to what we used for knock) would be needed to use it (often said "MAP impossible to use", but it's rather just hard).

* use the crank-edge that gives even spacing (rising or falling).

* change the firing condition so it only fires (fire means: stores trigger time and asks for RPM recalc with engine.status |= _BV(new_rpm) and calls ign_trigger_here() ) at odd (or even, as you like) engine.igncount values, and uses the engine.igncount>>1 as index to look up h[2]. This is needed, because h[2] is only 8 wide, while 12 is needed. You can also try a quick/dirty hack HW_TABLE_SIZE=12 or 16 - however, this way max dwell time might be limited!

* After tests, if this behavior is enabled by a my_make define or a config bit, we'll merge it to SVN.