[Yoshi]

You're welcome.

Originally Posted by dcb

Any thoughts you'd be willing to share on the trip vs instant mpg discrepancy?

I've no time to figure it out yet, but I think you could.
I believe you can refer raw numbers for the variables and you'll find the cause is whether distance data or gallons, instead of the result of mpg value.

Or better yet, a generic way to multiply and divide 32 bit numbers while not overflowing but retaining max resolution (without a bunch of "if" statements)?

I know 32 bit data is better than 16 bit, but...
We observe approx 1% variance on the VSS and injector pulse width, therefore 1% accuracy is my target. I believe 0.1% internal calc error is acceptable, then 5 digit decimal value (or 16 bit data) is enough for our instrument.

Again, my suggestion is to place cli()/sei() codes for another interrupt driven variables, such as vssPulse and injector pulses.
Following is an example in instruction level. A 32 bit variable is sent to registers as follows and interrupts can be occurred any timing between instructions.

; a 32 bit variable is 0x0000FFFF
        LDS     R14,0x0244 ; R14<-0xFF byte 0 of 0x0000FFFF
        LDS     R15,0x0245 ; R15<-0xFF byte 1 of 0x0000FFFF
; an interrupt occurred, then it became 0x00010000
        LDS     R16,0x0246 ; R16<-0x01 byte 2 of 0x00010000
        LDS     R17,0x0247 ; R17<-0x00 byte 3 of 0x00010000
; then R17..R14=0x0001FFFF

Regards,
Yoshi