the Quotient is floating point in vhdl

[Altera_Forum]

hi all  

 if i want to divide two numbers and i know the result would be float .  what can i do it and how to specify the output from division  thanks a lot.

[Altera_Forum]

what are the two operands? diving two integers gives you an integer and remainder result (if you use the IP core). If you use correct scaling, you can get a fixed point output. You only get a floating point result if both inputs were floating also.

[Altera_Forum]

hi tricky 

when i try to divide two integer as 8/12 =0.66666 it gave a float point the two operands type are std_logic_vector( 7 downto 0) .should i define the output from result in float

[Altera_Forum]

8/12 = 1 and remainder 4. Not 0.666666. If you use different scaling you can get more accurate results. It will not be floating point unless you convert both 8 and 12 to floating point (32 bit usallly). 

 Why not post the code.

[Altera_Forum]

thanks tricky a lot for helping my design is used to measure the distance between two input data in the following code i used the accum to sum the distances measurements and the counter to count the numbers of data measured and i want to divide the accumulator result on the counter result .if you understand me well , is this possible or i should use the trigger signal to sense the last sample input . if that i said is wrong ,so i will divide accumulator on the counter result. sorry for ad English and i hope you help me thank a lot. i attache the code and the simulation picture . i want to thank kaz for helping me before to make this code thanks kaz very much

[Altera_Forum]

Floating point is most likely not the best solution for your problem, a 8.8 fixed point number sounds like a better idea. You easily get it by multiplying the dividend with 256 before the division. 

 8 fractional bitsis just a first guess, you have to find out the required accuracy for your problem. I notice that your ideal result is already several percent off by truncating the divisor of 12.5 to 12. Of course, this point could be corrected in a fixed point algorithm, too.