RTEMS not compiling with -mfloat-abi=hard in ARM926ejs

Joel Sherrill joel at rtems.org
Fri Jan 4 21:59:44 UTC 2019


On Fri, Jan 4, 2019 at 1:53 PM Amarnath MB <amarnath.mb at mistralsolutions.com>
wrote:

> Hi,
>
> I'm working on RTEMS 5.0 BSP for a new SOC based on ARM926EJS which has a
> VFP9-S Floating point coprocessor inside.
> I was able to successfully boot RTEMS and run hello.exe and ticker.exe
> testsuite samples using u-boot.
>
> I have enabled VFP from Uboot using the coprocessor instructions given in
> ARM926EJS RM.
> In order to use the VFP coprocessor in RTEMS I have added
> *CPU_CFLAGS = -mcpu=arm926ej-s -mthumb -mfloat-abi=hard *in the .cfg file
> for the BSP.
> But during the BSP build, it gives error "configure: error: C compiler
> cannot create executables".
> How can I enable hard float support in my BSP?
>

All that means is that compiling something simple like this:

int main(int argc, char **argv)
{
  return 0;
}

With the compiler arguments you have setup failed. There is a config.log in
the directory of the
failure with details but that's essentially the test.

>
> *As a trial, I have built BSP with CPU_CFLAGS = -mcpu=arm926ej-s -mthumb
> and manually added*
> *CPU_CFLAGS = -mcpu=arm926ej-s -mthumb -mfloat-abi=hard in the makefile
> (Makefile.leaf) for application build.*
> *In this case, I'm getting error 'error: o-optimize/init.o uses VFP
> register arguments, o-optimize/rtems-app.exe does not'*
>

Without seeing the full test case and log, this looks like the linker
arguments don't match the compiler arguments for some reason.

Is this with one of the dynamic loading tests?


>
> *As per RTEMS CPU Architecture Supplement Release 5.0.0 (master) Section
> 3.1.3 Floating Point Unit only*
> *VFPv3-D32, **VFPv3-D16 and *
> *FPv4-SP-D16 are supported by RTEMS, but VFP9-S Floating point coprocessor
> is an *
>
> *implementation of the Vector Floating-point Architecture (VFPv2) which is
> not listed in 3.1.3.*
> *Does this state that VFP9-S Floating point coprocessor is not supported
> by RTEMS?*
>
> *For your info floating point operations are working fine in the
> application built using -mfloat-abi=soft.*
>
>
> *Thanks in advance for all your precious time.*
>
>
> *Regards, *
>
> *Amarnath MB*
> _______________________________________________
> users mailing list
> users at rtems.org
> http://lists.rtems.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/users/attachments/20190104/c999189b/attachment-0001.html>


More information about the users mailing list