a + b + c

How floating-point arithmetic order affects precision in C++

This question tests your understanding of floating-point representation and the limits of computer arithmetic. It's a staple at quantitative firms because traders and researchers routinely work with sums of prices, returns, and risk metrics where small rounding errors compound into material P&L mistakes.

The core issue is that floating-point numbers in C++ (and most languages) use fixed-width binary encoding, which means not all real numbers can be represented exactly. When you combine this constraint with the non-associative nature of addition under finite precision, the order in which you perform operations matters. Understanding why the result depends on the sequence—not just whether it does—reveals whether you grasp the mechanics of rounding, accumulation of error, and the difference between mathematical and computational addition.

• IEEE 754 representation and rounding modes
• Associativity and commutativity in finite-precision arithmetic
• Catastrophic cancellation and loss of significance
• Practical mitigation strategies (Kahan summation, higher precision accumulators)