Which conditional expression is faster?

[RussMartin]

Which of these is faster to execute?

IF A+B+C=0 THEN . . .

or

IF A=0 AND B=0 AND C=0 THEN . . .

[mackrackit]

This is faster
IF A+B+C=0
The generated ASM


ADD?BBW _A, _B, T1
ADD?WBW T1, _C, T1
CMPNE?WCL T1, 000h, L00001


This
IF A=0 AND B=0 AND C=0 THEN
Generates


CMPEQ?BCB _A, 000h, T1
CMPEQ?BCB _B, 000h, T2
LAND?BBW T1, T2, T2
CMPEQ?BCB _C, 000h, T3
LAND?WBW T2, T3, T3
CMPF?WL T3, L00003

[mister_e]

You will have problem with the first one one day or another (put some values in and figure a roll-over), but the second one is bullet proof.

If you want to measure it, you may use the regular suggestions
Set a pin before, and clear it after the IF-THEN or code block. Measure the delay with a scope or frequency meter
Same as above, but use an internal Timer and output the data over serail port
Use MPLAB stopwatch

http://www.picbasic.co.uk/forum/showpost.php?p=39033&postcount=9 <-- really handy
http://www.picbasic.co.uk/forum/showpost.php?p=39049&postcount=13

[Acetronics2]

Which of these is faster to execute?

IF A+B+C=0 THEN . . .

or

IF A=0 AND B=0 AND C=0 THEN . . .


Hi, RUSS

A+B+C = 0 has many solutions AS ... you use an ADDITION and not a Bitwise AND !!!

253 + 2 + 1 = 0 .... for Bytes !!!
1 + 1 + 0 = 0 ... for Bits !!!

so, It can't be equivalent to the second line ... Too Bad !


Now, let's suppose A, B, C are BITs ...



so



IF A | B | C THEN Z = 1...


Will be the fastest ... with PBP.`( 17 asm Lines ...)



IF (NOT A & NOT B & NOT C ) THEN Z = 1


is somewhat longer ...

BUT, if A,B,C are inputs ... would be faster to mask the other inputs and test if result is 0




IF ( PORTx & %00000111 ) THEN Z = 1



Only 12 asm lines ...

Alain

[RussMartin]

Thanks to each and all of you. I appreciate especially the warnings about pitfalls in the summation test; I'll remember those.

I failed to provide the limits/boundary conditions. I should have qualified the question and put it this way:

Which of these is faster to execute?

For A, B, C < 2 (and of course not negative!),

IF A+B+C=0 THEN . . .

or

IF A=0 AND B=0 AND C=0 THEN . . .

It looks as if Dave assumed the limits I had in mind.

The reason for my question was that, back in the bad old days, the syntax of conditional statements in FORTRAN IIb (sometimes called "Weather Bureau FORTRAN") was extremely limited, so doing a test like this was easier, if not necessarily faster.

I happened to try it with PBP in an application, and it worked. I just didn't know how economical it was in terms of generated code. Hence my question.