gcc 5.1 doesn't define char16_t and char32_t (unless in C++ mode),
causing compile failures.

Change-Id: I08dcd13cdf8cd59a4a2f191864bedf4c0d1bb313
Signed-off-by: Bernhard Rosenkränzer <Bernhard.Rosenkranzer@linaro.org>
Signed-off-by: Chet Kener <Cl3Kener@gmail.com>

Because NEON_UNALIGNED_ACCESS has never been defined, it has gone unused.
This change changes it to check for __ARM_FEATURE_UNALIGNED, defined by GCC.
This has improved speed by roughly 34% on my testcases.

Also, memcmp checks if HAVE_32_BYTE_CACHE_LINE is defined and adjusts for it;
this is true for Cortex-A9, so define it in cortex-a9 makefile.

Change-Id: I3c76ce7fa2796872e0171d5502b0ebd6e2893339

Create malloc benchmark with three methods.
Record each iteration's time and include the standard deviation
to measure the 95 percent confidence interval. The margin of
error will be removed, and each operation's time will be more stable.

Change-Id: Ie0135fb981035b7abe1fc4561e7fd1b34744fa3d
Signed-off-by: Neal-liu <nealliu.tw@gmail.com>

Since I'm using mmm bionic/libc/ for testing, this was being built.
I don't want it built.

Change-Id: Ia86b605719ffebfd4eb7bdfde10b158bab490554

Change-Id: Ibc401bf2a3b2fdf3951d3f73348f2f8519e787f9

Conflicts:
	benchmarks/string_benchmark.cpp

These aren't exactly 'accurate' but works in a relative sense.

Change-Id: If736e616b74d6ea3bd36a1ecd626476a7adc2ad5

Conflicts:
	benchmarks/string_benchmark.cpp

Not totally sure on this code, but it works.

Change-Id: I421cc7b02f9a4ff15c82674abe50a2dc332c25ed

Conflicts:
	benchmarks/string_benchmark.cpp

Change-Id: I12dffced500e127cc487c056bc5eee05b37907a6

bionic: linker: Load shim libs *before* the self-linked libs

Cm 13.0

Cm 13.0

bionic: Teach fnmatch(3) to handle UTF-8 characters in patterns