Thomas Witzel wrote:
Is there any resource where I can read up on what is being done in the gcc development in this regard to learn more about whats going on ?
-fcx-limited-range appears to be the relevant specific option implied by -ffast-math. The question appears to be how much protection you want against problems associated with over/underflow; if you want such protection, do you care about the details. Promotion of float to double would take care of many of the issues with float over/underflow, even with -fcx-limited-range set. Arguments in favor of -mfpmath=387 go back at least to the mid-point of my career.
