On Wednesday, 18 May 2016 at 19:30:12 UTC, deadalnix wrote:
I'm confused as to why the compiler would be using soft floats
instead of hard floats.
Cross compilation.
Ah, looking back on the discussion, I see the comments about
cross compilation and soft floats. Making more sense now...
So if compiling on x86 for x86, you could just use hard floats,
but if compiling on x86 for some other system, then use soft
floats to mimic what the result would be as if you had compiled
on that system. Correct?
But what if you are compiling for a system whose float behavior
matches the system you're compiling on? So for instance, suppose
you are only using 32bit floats and not allowing anything fancy
like 80bit intermediate calculations. And you're compiling for a
system that treats floats the same way. Then, you could
theoretically use hard floats in the compiler and the results
would be the same.
