long double problems in math.h (intel version)

long double problems in math.h (intel version)

I am running the intel compiler v10.1.012, on Ubuntu 7.10 x86-64 bit edition. If I make a long double and run sizeof on it, it is 128bit, which is what I want. However, the precompiler __LONG_DOUBLE_SIZE__ which math.h (intel version) uses extensively is set to 80bit, which is a bad problem; since it creates headers for everything at the 80bit level. I have tried the simplest solution of undef/define __LONG_DOUBLE_SIZE__ to 128bit however this does not seem to work.

Does someone know of a simple solution which does not require me to rewrite the math.h header to fix the problem.

1 post / 0 new
For more complete information about compiler optimizations, see our Optimization Notice.