3 Nov
2010
3 Nov
'10
5:43 p.m.
On Wed, Nov 3, 2010 at 2:55 AM, Julian Brown julian@codesourcery.com wrote:
On Mon, 1 Nov 2010 15:57:11 +0200 Ira Rosen IRAR@il.ibm.com wrote:
- According to gcc.dg/vect/vect.exp the only flag that is used for
NEON (in addition to target independent flags) is -ffast-math. Is that enough?
Note that the NEON unit doesn't implement the deeper parts of IEEE754 floating point like NaN and exceptions. These are needed for general purpose applications but not, say, radial fills in graphics. Turning on -ffast-math allows the compiler to send floating point operations via NEON instead of VFP.
-- Michael