From: Dave Hylands <dhylands@gm...>  20050314 19:34:44

Hi Paul, > I am new to the ARM with fixed point precision. Can anyone > give me a hint, what limitations I will meet with Python on > the gumstix for floating point calculations? > > I understood so far that there is a floating point emulation. > Where does it happen? There are two ways that floating can be done on the ARM. The first is where floating point is emulated in the kernel. The second is where the application is linked against a softfloat library. The softfloat is also an emulation, but because it's a usermode emulation you have to take the penalty for user>kernel>user. If you're doing extensive floating point, then using softfloat will give you better performance. If you're just doing occaisonaly stuff, then using the kernel emulation is fine. > Is there double precision or only single float? You get both. > Is there any comparison for speed re x86 with floating point > processor? I don't know.  Dave Hylands Vancouver, BC, Canada http://www.DaveHylands.com/ 