floating point support for ARM

David Keller dchkeller at gmx.ch
Thu Jun 16 21:35:01 CEST 2005

Hi Ralph,

Thanks for your answer.

> You can use either kernel-fpe (two choices: NWFPE or FastFPE), or 
> soft-float is integrated into recent gcc+binutils.  The latter is 
> highly recommended for performance, since the kernel-mode method 
> incurs exectption and context switch overhead.

We have compiled the hole environment with a soft-float-compiler, but 
the FPE is still on in kernel-config.

>> Actually we've tried the example from 
>> http://www.captain.at/programming/rtai/examples.php, but when loading 
>> the module we get some unresolved symbols (looks like floating-point 
>> support is missing). The same if we try to load rtai_math. We're 
>> using a softfloat-compiler (gcc 3.3.3).
> Normally floating-point is not allowed in the kernel.

But shouldn't RTAI allow to do this in real-time tasks?

>   With soft-float it might be possible (I have never tried) as long as 
> you take care to link statically all the libraries with the FP 
> functions into your module.

Since the hole environment is compiled with soft-float, we didn't link 
it satically. We thought statically linked libraries are only necessary, 
if only a part of the system uses soft-float and another one uses the 
kernels FPE. Please correct me, if this I'm wrong here.

>   The traditional way is to do messy FP stuff in userspace.

But LXRT is missing for ARM, so real-time FP stuff should be done in 
kernel-space, shouldn't it?

Thanks and best regards
David Keller

More information about the Rtai mailing list