Forum Discussion
13 Replies
- Altera_Forum
Honored Contributor
do you mean fixed point or floating point? fixed point is just integer arithmatic with an offset. floating point is usually 32 bits. ALtera provides floating point cores.
- Altera_Forum
Honored Contributor
I mean floating point, witch is indicated in this link
http://en.wikipedia.org/wiki/fast_inverse_square_root#overview_of_the_code !!??? - Altera_Forum
Honored Contributor
Then you need to use the Altera floating point IP cores from the megawizard
- Altera_Forum
Honored Contributor
yes but I want to use number bigger than 127 for example 200.5 or -241.2. so what can I do thank you
- Altera_Forum
Honored Contributor
I dont know where you got this idea of limitation from - but Altera IP cores can go up to 32 bit floating point, meaning they can go from -3.4028234 × 10^38 to 3.4028234 × 10^38
Thats a rather large range. - Altera_Forum
Honored Contributor
I have found it in the some link "The next 8 bits form the exponent, which is biased in order to result in a range of values from −128 to 127..."
I will try with your idea thank you very much. - Altera_Forum
Honored Contributor
The range of values from −128 to 127 is for exponent part of your floating number, not the number itself.
I also head for synhtetisable floating point. I have found IEEE packages that work in ModelSim but not in Quartus yet (version <= 12.0sp1). - Altera_Forum
Honored Contributor
ok thank you very much mmTsuchi I haven't concentrate for this :)
I have found also a package, hoping that will be synthetisable with quartus, I m working with 9.0 version thank you - Altera_Forum
Honored Contributor
The '93 compatible versions of this library will work just fine in quartus. But you'll get some terrrible Fmax figures because there is no way to pipeline them.
- Altera_Forum
Honored Contributor
can you explain more please, Tricky ??!! because I don't understand ""get some terrrible Fmax figures""
thanks