hello:
this is FYI from Nick Clifton in response to my query regarding
deprecation of -mno-fpu for ARM GAS. in the response, the list
was not CC'd by mistake.
Aaron
-------- Original Message --------
Subject : Re: [Fwd: QUERY : why is -mno-fpu deprecated when we need
it for integer only software?]
Date : Wed, 23 Jul 2008 12:29:50 +0100
From : Nick Clifton <[EMAIL PROTECTED]>
To : Aaron P. D'Souza <[EMAIL PROTECTED]>
Hi Aaron,
ARM: Do not pass -mno-fpu to the assembler
http://gcc.gnu.org/ml/gcc-patches/2003-04/msg00593.html
why was that?
Well the posting explained it quite clearly I though. Basically the
-mno-fpu option to the assembler was insufficient to describe the
fact that although no hardware floating point instructions were being
used it is still possible that software floating point functions were
being called from the assembler source file. Since there are two
incompatible ABI conventions for software floating point on the ARM,
the -mno-fpu option was changed so that the assembler could mark the
object files it creates with a flag saying which kind of software
floating point *might* be used by them. The linker can then check
and warn the user if they attempt to link together object files that
use different software floating point conventions.
arm-elf-as -mno-fpu -march=armv4t -mcpu=arm7tdmi -mthumb-interwork \
-o FirstAsm.o FirstAsm.s
Assembler messages:
option `-mno-fpu' is deprecated: use either -mfpu=softfpa or
-mfpu=softvfp
i am currently developing software that does not use any FP for a
processor that has the ARM7TDMI core with no FPU. we do not want to
mess with soft and hard FPU libraries when we do not need them.
In which case you could use -mfpu=softvfp quite safely as a
replacement for -mno-fpu as it will have no effect on your binaries.
we are concerned that by using -msoft-float the calling convention
is changed and that we have to recompile libgcc.a all over again.
There should be no need to recompile libgcc. Just change your build
script to use the -mfpu=softvfp switch instead of -mno-fpu.
why mess with FP when we dont need it at all? the -mno-fpu solves
the problem quite neatly.
So does using -mfpu=softvfp since you are not actually going to
encounter any software floating point ABI issues anyway.
Cheers
Nick
--
{
assisted: Corel, uSoft, HP, Yahoo, Adobe, Macromedia, Google,
Countrywide
missed: MCI/Worldcom, AT&T
}
