single precision mode and difference between Matlab and cortex-m

Hi,

I have a kallman function that it wrote in double precision and I want to convert it to single precision, is there a normal way?

I find out I have to use Single Fcn but I think if I wrote Single Fcn for every vector it's really weird.

But why I want to use single-precision!!! I should implement this filter on stm32f4 with DSP CMSIS and compare data!

another question is that why there is a difference between float in stm32f4 and single in Matlab?

let's assume a variable

>>a=0.06;
>>x=single(a);
 single
    0.0600

but in microcontroller

    float a=0.06;

when I watch a:

0.0599999986

Answer 1

[Jan]

This is the expected behaviour.

There is not exact representation of 0.06 in IEEE 754 32 bit floating point numbers, neither in Matlab nor in C. See:

a = single(0.06)
fprintf('%.16g', a)
>> 0.05999999865889549
%  0.0599999986   % From your output

The 0.06 displayed my Matlab is a rounded output.

"is there a normal way?" - It is hard to guess, what you call "normal". Matlab's default type is double, so if you want single precision, you have to include the corresponding command.

[mohammad.H Bagherian]

thanks for your share,

i think a normal ways is one line command to change precision. for example write

precision single

or somthing simlar to it... is there any command?

[Jan]

No, not in Matlab.