Another test case

[David Edelsohn]

>>>>> Richard Henderson writes:

Richard> Actually, I see no reason I can think not to just make this
Richard> generic.  Does any currently supported platform _not_ use
Richard> ieee754 single precision numbers as float?

	The POWER architecture only provides double-precision
floating-point.  Values are restricted to single-precision when storing to
memory.  PowerPC adds full single-precision FP.

David