In article <200102222329.f1MNTFD26158@buzzard.freeserve.co.uk>,
   Richard Earnshaw <rearnsha@buzzard.freeserve.co.uk> wrote:
> > fpe0 at cpu0: Advanced Risc Machines floating point emulator 1.07M
> > fpe0: FPA11 found
> > 
> > Is this what's required, or do I need to do more work?

> Good start.

> So, 

> Step 1:

> add 

> options ARMFPE

> to your kernel conf file, apply the patch I posted earlier, build a kernel 

> Step 2:

> Try a few floating point operations (write some assembly code).

> It should be pretty obvious from the timing whether or not the FPA is 
> running the instructions or not.

> Note, a quick poke into the FPE code shows that it can try to talk 
> directly to the FPA.

> Step 3

> Try building several programs with FPA instructions, run them in parallel 
> and check that they work correctly.

> R.


I tried this out as I am also interested in getting the floating
point coprocessor working on my A7000+. I applied both patches and
specified 'options ARMFPE' in the config file for my kernel. After
I rebuilt the kernel I found that there was no difference in the
time it took my test program to run. Given that it took about 80
seconds under RISC OS and over 1300 under NetBSD with and without
the changes, it is clear to me that the floating point hardware is
still not being used. Have I missed something?

Dave Daniels