Hardware floating point on STM32F4

[Chris Johns]

On 22/02/2014 10:32 am, Andrei Chichak wrote:
> Hello gurus,
>
> Something in Eclipse lead me to believe that I’m using soft float on my STM32F407 on my ST discovery board.
>
>    #if (__FPU_PRESENT == 1) && (__FPU_USED == 1)
>
> indicated that __FPU_USED was 0, and digging further lead me to the flag __SOFTFP__ is set. Hmmmm.
>
> Putting in some cheesy float code showed that I was calling a bunch of aeabi float library code instead of chucking in the hardware floating point commands. I have defined __FPU_PRESENT=1, but I think I may need a differently compiled compiler.
>
> The tools are running on my Mac and I put the whole thing together with Source Builder with the 4.11/rtems-arm.bset from Chris Johns about October 2013.
>

Nice to hear.

> Question: How do I go about getting the C compiler to generate hardware floating point instructions? Load a different compiler (how), or just throw another compiler flag?

A google for "gcc flags cortex m4 fpu" gave me this link ...

   https://groups.google.com/forum/#!topic/mbed-devel/didoeCqf3fI

Maybe give these a try. You may need to update the BSP flags and rebuild 
RTEMS as the ABI may change from soft to hard floating point.

I suggest you check the generated code to make sure the instructions are 
using the FPU and then run the paranoia test in the samples directory of 
the testsuite.

Chris