On Tuesday, 17 May 2016 at 14:05:54 UTC, Matthias Bentrup wrote:
If you try to make compile-time FP math behave exactly like run-time FP math, you'd not only have to use the same precision in the compiler, but also the same rounding mode, denormal handling etc., which can be changed at run time (http://dlang.org/phobos/core_stdc_fenv.html), so the exact same piece of code can yield different results based on the run-time FP-context.
Yes, heading rounding mode is an absolute requirement. If the system don't heed the rounding mode then it will change the semantics and break the program when computing upper vs lower bounds.
I've never liked the stateful approach to rounding. If I was to design a language I probably would make it part of the type or operator or something along those lines.
