The Alpha edition of Visual C++ supports both the /Op and /Op– options, which you use to improve or override floating-point precision. The /Op option disables arithmetic identities, so for example a+(b+c) will not be changed to (a+b)+c. This differentiation is only important with numerically unstable algorithms. On x86 platforms, the default is /Op–, which allows some loss of floating-point precision in favor of faster code. The Alpha default is /Op–.
Tip Because /Op– performs different optimizations across platforms, use /Op to ensure consistent results.