1.8 -ffast-math

Setting this option tells the compiler to perform more aggressive floating-point optimizations.

It results in behavior that is not fully compliant with the ISO C or C++ standard. However, numerically robust floating-point programs are expected to behave correctly.
Non-ConfidentialPDF file icon PDF versionARM DUI0774E
Copyright © 2014-2016 ARM. All rights reserved.