GCC or similar compilers perform optimizations aimed at improving the numerical stability of floating point operations.
It is known that it would seem that just operations, such as adding or calculating the norm of a vector, are numerically unstable if they are implemented in an obvious way, and, on the other hand, compilers sometimes destroy workarounds for these problems for the sake of speed optimization.
What is the current state of compiler output optimization for numerical stable computing? Anything better than expected?
source
share