( ESNUG 317 Item 2 ) --------------------------------------------- [5/13/99]
From: Erik Runeson <erikr@iar.se>
Subject: Bug in SPARC floating-point implementation?
I just discovered what I consider a bug, or at least a deviation from the
IEEE 754 Standard, in the floating-point hardware of an Ultra-1 Sparc
station. When comparing two quiet NaNs with the predicates equal (==) and
not equal (!=), the Invalid Operation flag was raised. Now, the standard
very clearly states that these predicates should not raise Invalid Operation
when the operands are unordered.
Does anyone know why Sun has choosen not to follow the standard for
this case?
- Erik Runeson
IAR Systems AB Uppsala, Sweden
|
|
