You can disassemble the components of an IEEE754 floating point number and put the logic together yourself:
#include <cstring>
float my_ceil(float f)
{
unsigned input;
memcpy(&input, &f, 4);
int exponent = ((input >> 23) & 255) - 127;
if (exponent < 0) return (f > 0);
// small numbers get rounded to 0 or 1, depending on their sign
int fractional_bits = 23 - exponent;
if (fractional_bits <= 0) return f;
// numbers without fractional bits are mapped to themselves
unsigned integral_mask = 0xffffffff << fractional_bits;
unsigned output = input & integral_mask;
// round the number down by masking out the fractional bits
memcpy(&f, &output, 4);
if (f > 0 && output != input) ++f;
// positive numbers need to be rounded up, not down
return f;
}