Add -marm to those sources who aren't thumb friendly.

branches: 1.8.6; 1.8.12;
If SRCS contains any .S files then remove the corresponding .c file from
SRCS and add to LSRCS (for lint).
Change the 'string' Makefiles so that the arch/*/string/Makefile.inc
need only specify the .S files for that architecture and not the .c
files for all the files they don't override.

Remove references to index.c and rindex.c (strchr.c and strrchr.c
now provide the definitions).

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Add arm asm version of memcmp, basically it's strncmp.S without the
check for *src != 0

Add an asm version of strncmp, based on strcmp.
Add RCSID to strcmp, also use numeric label, rather than named label for
loop.

Add an assembler version of strcmp, based on example code from the ARM
ARM. As an example of the performance difference that this provides
a Dhrystone score on my Shark goes from 213k to 261k.

Update the libc version of ffs to make use of an optimsed asm version.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.