Source code for floating point arithmetic?

Floating point code is under gcc/libgcc
On Sep 30, 2012 10:01 PM, "Matthew Mucker" <
<email address hidden>> wrote:

> New question #209985 on GCC ARM Embedded:
> https://answers.launchpad.net/gcc-arm-embedded/+question/209985
>
> I downloaded the source code package and extracted a few of the embedded
> archives. I'm looking for the source code for floating point arithmetic for
> the Cortex-M0 core, and I haven't been able to find it. I'm probably just
> looking in the wrong place. If someone could point me to the correct source
> code file for these functions, I'd be very grateful.
>
> Thanks,
>
> -Matthew
>
> --
> You received this question notification because you are an answer
> contact for GCC ARM Embedded.
>

Joey, thank you for the reply. However, I'm still a bit confused. In gcc/libgcc, there are only three .c files:

-fixed-bit.c
-generic-morestack.c
-generic-morestack-thread.c

If I go into config/arm, there is a file bpabi-lib.h, which has a number of #defines that appear to be related to floating point math, but I still can't find the implementation of floating point arithmetic.

Hi Matthew,
I am not an expert in software floating point of GCC, but maybe can help a little here.
For our release, which is in GCC version 4.6, there are two version of implementations of software floating point:
1. the traditional implementation in files like gcc/config/fp-bit.c.
2. the soft-fp implementation which is extracted from GNU C library and used in GCC, it resides in directory gcc/config/soft-fp.
3. Additionally, ARM target implement its own software floating point arithmetic in assembly code, which resides in gcc/config/arm/leee754-{s,d}f.S something like that.

For GCC trunk, which is version 4.8, these files and directory has been relocated to directory libgcc/ in similar directory structure.

In Trunk:
For Thumb2 and ARM, I believe the 3rd is used for software-fp arithmetic.
For Thumb1, I think soft-fp is used.

Again, I am not familiar with software floating point in GCC and not sure about:
1. The 3rd implementation could be a supplement to the 1st/2nd.
2. Same story for 4.6 as in TRUNK.

You can build the toolchain and check the .dep files in libgcc/ to see which source file is used.

Or maybe someone with expertise in this field can answer this question better.

Thanks.