Print the float in C, avoiding allowing the variational parameter to double

Question

How can I print (i.e. stdout) a float in C without if it will be passed to double when passed to printf?

The problem here is that variable functions in C allow all float parameters to double, which leads to two unnecessary conversions. For example, if you enable -Wdouble-promotion in GCC and compile

float f = 0.f; printf("%f", f); 

You'll get

 warning: implicit conversion from 'float' to 'double' when passing argument to function 

I have relatively little processing power to play with (ARM Cortex-M3 at 72 MHz), and I definitely have a bottleneck in the output of ASCII floating point data. Since the architecture lacks a hardware FPU to begin with, converting between single and double precision does not help.

Is there a way to print float in direct C more efficiently?

Answer 1

Avoiding progress will not save you, since the internal double (or more likely long double ) printf arithmetic will consume at least 1000x as much time. Accurate printing of floating point values ​​is not easy .

If you still don't need accuracy, and you just need to quickly print out the approximate values, you can flip your own print cycle. While your values ​​are not too large to match the integer type, first convert and print the non-fractional part as an integer, and then subtract that off and the cycle multiplied by 10, and cross out the integer part to print the fractional part one digit at a time (buffer in line to improve performance).

Or you could just do something like:

 printf("%d.%.6d", (int)x, (int)((x-(int)x)*1000000)); 

Answer 2

Unfortunately, printf does not support simple float: s support.

This means that you will need to write your own print function. If you don't want the full expressive power of printf , you can easily convert the floating point value to an integral part and the part representing the number of decimal places, and print both values ​​using integers.

If, on the other hand, you just want to get rid of the warning, you can explicitly point the float to double .

Answer 3

I think this does not matter - printf is already such a mysterious problem that these conversions should not matter. Converting floating point time to double should be much less than converting any number to ascii (you should / could profile your code to get the final answer). The only remaining solution would be to write your own custom output procedure, which converts float-> ascii, and then uses puts (or similar).

Answer 4

• First approach: use ftoa instead of printf. Profile.

• To increase the flexibility of output, I would go to the source code of your stdlib compiler, perhaps in some derivative of gcc, find the printf implementation and copy the appropriate code to convert double → ascii. Rewrite it in float -> ascii.

• Then manually change one or two inappropriate call sites to a new (non-invariant) version and profile.

• If it solves your problem, you might consider rewriting your own printf based on the version from stdlib in which you pass float * instead of float. This should get rid of automatic advertising.