Wolfgang Denk writes ...
In message [off-list ref] Dan Malek wrote:
quoted
You have to be careful here. If you choose to use the in-line
floating point software, you must use it for all applications
because the libraries are built to do this as well. You can't
mix programs that use the kernel FPU emulation and also in-line
software floating-point. The kernel works out of the floating
point register area in the TSS, while the software float assigns
real processor registers (and other memory locations) to contain
the floating-point values.
I am aware that I cannot mix this WITHIN ONE application, for
instance compiling the app with soft-float and linking against a
library that uses the kernel FPU emulation.
Are there really problems when running soft-float only and FPU emu
only applications on the same machine? I did not expect any...
I think this may be a problem if the compiler uses registers to pass
arguments to and/or return values from functions. I'd guess if you
can fix that issue then you might be OK.
Matt
** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/