testcase for -ffast-math GE/LT swapping

[Geoff Keating]

> From: Brad Lucier <lucier@math.purdue.edu>
> Date: Thu, 28 Dec 2000 19:39:58 -0500 (EST)

> So I don't believe that this is a valid test case for LT => GE
> transformations in the compiler.

In the future, it may no longer test what it's supposed to be testing,
but I can assure you that right at present it fails on powerpc targets
because the backend is not expecting LT to be inverted into GE.

-- 
- Geoffrey Keating <geoffk@geoffk.org>