 | | Float and int
|
| Does anyone know what the maximum and minimum size of integer which can be stored is? Likewise, what is the floating point accuracy of BrickOS for variables of type "float" and "double" (is there such thing in BrickOS?) ? Both of these questions (...) (22 years ago, 29-Dec-02, to lugnet.robotics.rcx.legos)
| | | | | Re: Float and int
|
| (...) int is 16 bits, big endian, signed. It goes from -32768 to 32767. unsigned int goes from 0 to 65535. char is 8, long is 32, long long is 64, with the standard limits. Floats and doubles are emulated in software, as the H8/300 processor has no (...) (22 years ago, 30-Dec-02, to lugnet.robotics.rcx.legos)
| | | | | Re: Float and int
|
| (...) You should look in the gnu gcc compiler documentation for this. Float support is not OS but compiler specific. As I recall, gcc uses IEEE float format, the specification should be found somewhere in the net. But doing float with the brick is (...) (22 years ago, 30-Dec-02, to lugnet.robotics.rcx.legos)
| | | | | Re: Float and int
|
| (...) Mabie Fixed Point is an answer. I dont know how it works but I have a book that suggests that a decent Fixed Point library can be faster than Floating Point if there isnt an FPU available. (22 years ago, 30-Dec-02, to lugnet.robotics.rcx.legos)
| | | | | RE: Float and int
|
| (...) I'm doing a seminar at BricksWest on fixed point notation and simple algorithms for square root, trig, etc for BrickOS and pbForth users. The focus will be on the mechanics of the algorithms, the actual code implementation will be a "exercise (...) (22 years ago, 30-Dec-02, to lugnet.robotics.rcx.legos)
| | | | | Re: Float and int
|
| (...) For most brick projects I can think of, it is. If you want to travel your robot around, you will have to deal with sin, cos, tan and square root. First thing you can do is working with tables to avoid "on the fly" sin calculation. Second, you (...) (22 years ago, 30-Dec-02, to lugnet.robotics.rcx.legos)
| |
