Binary floating point cannot store most decimal fractions exactly: 0.1 + 0.2 is 0.30000000000000004, and 101.15 is really 101.1500000000000056843... Trading systems therefore keep prices as integers in a fixed unit, the tick. With four decimal places, 101.15 is stored as 1011500. Adding and comparing such prices is exact integer arithmetic, and equal means equal.
The work moves to the edges. Parsing "101.15" must give 1011500 without going through double: read the whole part, then the fraction digit by digit, pad missing digits with zeros and reject extra ones. Formatting goes the other way and must print the sign itself, because -0.25 has a whole part of 0. Multiplying a price by a quantity can overflow long long sooner than you think; __builtin_mul_overflow(a, b, &out) checks and multiplies in one step, before any undefined behavior. And rounding to a coarser tick size needs a stated rule, because C++ division rounds toward zero, which is not "down" for negative prices.
Integer ticks, and an overflow check that runs before the overflow could happen.