Architecture-ABI
Introduction
There are currently many official and unofficial ABI -- Application Binary Interfaces -- in use within embedded Linux projects. The following is an attempt to lightly document what the individual architecture ABI consist of, without going into too many details..
Many of the following ABIs have grown organically based on specific processor designs and optimization levels. For the most part the processor optimizations are ignored in these descriptions, unless otherwise mentioned.
Additional references, CPU optimization compatibility and suggested canonical architecture namings may be provided.
ARM
apcs-gnu a.k.a. OABI - The original ARM ABI.
It has two variants, an emulated hardware floating point, and a soft floating point variant. The legacy ABI should no longer be used in embedded projects and has more or less disappeared. The legacy ABI is not compatible with the EABI.
It can be detected by:
From: http://wiki.debian.org/ArmEabiPort
objdump -x contains [APCS-32]
The legacy abi can be enabled by using "-mabi=apcs-gnu"
EABI - The new "Embedded" ABI by ARM ltd.
The new EABI specifies a single ABI that works in various hardware floating point and software floating point configurations as well as Thumb interworking. Overall structure packing is better, a more efficient syscall convention is provided.
ARM Processors are available in both Big Endian and Little Endian varieties. Each variety impacts the ABI and specifies the endian bit and byte order.
The following table lists the current ARM ABIs in common usage:
| ABI | Endian | Compatibility |
| EABI | Big Endian | |
| EABI | Little Endian |
