Another Bit Of Help Please....

[iago]

Well, division is about 100x slower than and, so if that code is running a lot it will make a big difference.

[Tuberload]

Wow. I didn't know the performance differences were that large. Once again thanks for the information, it should help the performance of some of my projects a lot.

[iago]

Well, go back to logic gates.  An and is 32 parallel AND gates, which is the same as a single AND gate.

On the other hand, a divide is .. *finds book*.. blast, I'll just link you to a picture:

http://books.elsevier.com/companions/1558604286/pictures/pdf/Chapter4/f0440.pdf
http://books.elsevier.com/companions/1558604286/pictures/pdf/Chapter4/f0441.pdf

Although I'm not totally sure which algorithm is implemented on 80x86 chips, they tend to be slow like that.  And it runs in series, so it takes a hell of a lot longer.

[Tuberload]

So wouldn't thay make using the AND operator 32X faster than the division operator?

[Adron]

Well, division is about 100x slower than and, so if that code is running a lot it will make a big difference.

On a Pentium it seems to be 25x slower. 1 cycle vs 25. In a very tight loop it would probably be around 6 times slower, and in most loops it might make it 2 times slower.

[iago]

So wouldn't thay make using the AND operator 32X faster than the division operator?

No, because division doesn't just do one thing per cycle. If it was 32 and gates in sequence it would be different.

It really depends on the processor.

[Etheran]

Think about what division actually is.  It's like looped subtraction.  so of course that's going to take longer than a simple anding.