128 integer arithmetic

[DirkStarscream]

I am currently working on a couple of assemblers, one for intel/amd processors and one for powerpc processors. So far I have implemented the ability to do 128 bit integer arithmetic across 64 bit registers on the 64bit versions. Thing is, its taking up a fair chunk of space, so I was wondering, does anyone actually bother using 128 bit integer arithmetic? If so I guess I will have to leave it in, but otherwise I can save myself a headache and scrap it
thanks

[CronoMan]

try using the MMX registers http://forums.3drealms.com/ubbthreads/images/graemlins/smile.gif

[Drazula]

Yes, graphical calculations that propagate rounding errors would use 128 bit arithmetic to keep the rounding errors small.

[MentalSentinel]

Just what the hell is 128 bit Arithmetic? http://forums.3drealms.com/ubbthreads/images/graemlins/confused.gif

[DirkStarscream]

128 bit integer arithmetic is simply arithmetic that uses 128bit wide operands, presumably across 2 64bit registers. 386+ processors in 32bit mode would allow 64bit integer arithmetic across 2 32bit registers. the AMD64 processors would allow easy 128bit arithmetic across 2 64bit registers. Needless to say, 256bit arithmetic would require an integer to be stored across 4 64bit registers.

[DudeMiester]

Drazula said:
Yes, graphical calculations that propagate rounding errors would use 128 bit arithmetic to keep the rounding errors small.



But that would likely only be floating point numbers. I don't know of any needs for 128 bit integer arithmatic, not that there arn't any.

[Drazula]

No one uses fixed point arithmetic anymore?

[DudeMiester]

I'm sure they do, but I'm also sure they use floating point much more often.

[MentalSentinel]

DudeMiester and Drazula, where the hell did you learn such things? http://forums.3drealms.com/ubbthreads/images/graemlins/tinyted.gif

[Drazula]

Any software texture mapping, Coding articles by Michael Abrash, Doom Source, etc http://forums.3drealms.com/ubbthreads/images/graemlins/smile.gif.

[DudeMiester]

MentalSentinel said:
DudeMiester and Drazula, where the hell did you learn such things? http://forums.3drealms.com/ubbthreads/images/graemlins/tinyted.gif



http://steamsteamlol.ytmnd.com/

I'm not kidding.

[Jaapio]

DudeMiester said:

Drazula said:
Yes, graphical calculations that propagate rounding errors would use 128 bit arithmetic to keep the rounding errors small.



But that would likely only be floating point numbers. I don't know of any needs for 128 bit integer arithmatic, not that there arn't any.



If someone wants to use a smaller version of something they can cast I with short, if im correct.