New Float number

Question

i want to make a new float type with a size of 128-bit , the exponent would have 4-bytes(32-bit) and the fraction would have 12-bytes(96-bit), how i can make this type in c++, where i will be able to do input, output ,+,-,*,/ operations.

[I am thinking of c++ class but i can't figure how i will mange up between exponent and fraction with the specified length]

I have tried "bitset" std class, but it has a huge size in comparison with what i need for example "std::bitset<16> foo;" give a size of 8-bytes.

Any Suggestions !

_'"std::bitset<16> foo;" give a size of 8-bytes.'_ Did you check for size of a pointer?? Show some code please! — πάντα ῥεῖ, Nov 26 '13 at 22:40
Might interess you: http://stackoverflow.com/questions/4798777/is-there-a-c-equivalent-to-javas-bigdecimal — Sebastian Hoffmann, Nov 26 '13 at 22:41
#include #include using namespace std; int main() { bitset<32> expo; bitset<96> frac; cout << sizeof(expo) << endl; cout << sizeof(frac); return 0; } — Yazan Houshieh, Nov 26 '13 at 22:46
i am expecting 4 from the first print statement and 12 from the other but what i get is 8 and 16 ?! — Yazan Houshieh, Nov 26 '13 at 22:47

score 1 · Accepted Answer · answered Nov 26 '13 at 22:47

1

You have to build your own data type and implement all the operations you want to do with it. This is exactly how a compiler emulates a 64 bit dataset when the architecture supports only 32 bit data native.

So a class/struct with two members is the right starting point. You have to implement the basic operations now: +,-,*,/. Add and sub are easy, you need to pay more attention to * and /. What I can tell you is that usually compiles implement these using repeated add and sub operations.

answered Nov 26 '13 at 22:47

Pandrei

4,683
3
22
42

thank you, can you help me with a paper or a website pages that talks about this,, i have searched a lot with finding an appropriate one – Yazan Houshieh Nov 26 '13 at 22:52
@user3038489: Search for "IEEE floating point format" for some examples of different floating point formats. – Thomas Matthews Nov 26 '13 at 22:54

New Float number

1 Answers1