This Assignment Agreement has to be filled in, signed and sent to
optimized-routines-assignment@arm.com by Contributors before their
contributions can be accepted into optimized-routines.

Use L(name) for all assembler labels.

Cleanup string functions to use asmdefs.h, ENTRY and END instead of
defining macros in each file.

Modify integer and SIMD versions of memcpy to handle overlaps correctly.

Make __memmove_aarch64 and __memmove_aarch64_simd alias to
__memcpy_aarch64 and __memcpy_aarch64_simd respectively.

Complete sharing of code between memcpy and memmove implementations is
possible without noticeable performance penalty. This is thanks to
moving the source and destination buffer overlap detection after
the code for handling small and medium copies which are overlap-safe
anyway.

Benchmarking shows that keeping two versions of memcpy is necessary
because newer platforms favor aligning src over destination for large
copies. Using NEON registers also gives a small speedup. However,
aligning dst and using general-purpose registers works best for older
platforms. Consequently, memcpy.S and memcpy_simd.S contain memcpy
code which is identical except for the registers used and src vs dst
alignment.

Mention releases.

Create a new memcpy implementation for targets with the NEON extension.

__memcpy_aarch64_simd has been tested on a range of modern
microarchitectures. It turned out to be faster than __memcpy_aarch64 on
all of them, with a performance improvement of 3-11% depending on the
platform.

Include asmdefs.h in memcpy.S to avoid duplicate macro definitions.

Add macro for defining labels in asmdefs.h.

Change the default routine entry point alignment to 64 bytes.

Define a new macro which allows controlling the entry point alignment.

Add include guard to asmdefs.h.

Don't include the makefile fragments of subprojects that aren't built.

With this the build fails more reasonably when SUBS is set incorrectly.

Reorganise the makefiles so subprojects can be more separately used and
maintained.  Still kept the single toplevel Makefile and config.mk.

Subproject Dir.mk is expected to provide all-X, check-X, clean-X and
install-X targets where X is the subproject name and it may use generic
make variables set in config.mk, like CFLAGS_ALL and CC, or subproject
specific variables like X-cflags.

Use .d rather than .2d for element mov instructions in string routines
so the assembly compiles with clang too.

When defined as 0 the vector math code is not built and not tested.

The math_errhandling checks are incorrect in general: it is defined
by the libc math.h which is not appropriate for optimized-routines
provided functions that we are testing.

However even if we want to test a libc implementation, ISO C allows
the setting of errno even if !(math_errhandling&MATH_ERRNO), so
relax the checks.

Vector functions are only used on aarch64, so only define them there.

  math/test/mathbench.c:95:1: warning: '__v_dummyf' defined but not used [-Wunused-function]

gcc-9 started warning if alias symbols have different attributes:

  math/expf.c: At top level:
  math/expf.c:89:21: warning: '__expf_finite' specifies less restrictive attributes than its target 'expf': 'leaf', 'nothrow', 'pure' [-Wmissing-attributes]

so copy the attributes when creating the aliases.

Compilers (incorrectly) warn about unused volatile variables:

  math/math_config.h: In function 'force_eval_float':
  math/math_config.h:188:18: warning: unused variable 'y' [-Wunused-variable]

silence them.

Compiler checks and realated macros need to be done earlier so they are
usable for the static inline functions.

Fix the Makefile so the documented mechanism in the README still works.

Same design as in expf. Worst-case error of __v_exp2f and __v_exp2f_1u
is 1.96 and 0.88 ulp respectively.

It is not clear if round/convert instructions are better or +- Shift.
For expf the latter, for exp2f the former seems more consistently
faster, but both options are kept in the code for now.

Use heredoc instead of pipe when iterating over test cases to avoid
creating a subshell that would break the PASS/FAIL accounting.

Increase the upper bound on medium cases from 96 to 128 bytes.
Now, up to 128 bytes are copied unrolled.

Increase the upper bound on small cases from 16 to 32 bytes so that
copies of 17-32 bytes are not impacted by the larger medium case.

Implicit function declaration is always a bug, but compilers don't
turn it into an error by default for historical reasons, so add it
to the default config.

This is a simple fix to the v_powf code, but in general the vector
code may not work on arbitrary targets even when compiled with
scalar types (s_powf.c), so in the long term may be all s_* should
be disabled for non-aarch64 targets (requires test system and header
changes too).

Worst-case error is 1.67 ulp, the polynomial was generated by sollya.
Uses a 128 entry (2KB) lookup table. Special cases fall back to scalar
log call.

Worst-case error is 3.5 ulp, the polynomial was generated by sollya.
For large (>2^23) and special inputs the code falls back to scalar
sin and cos.

Essentially the scalar powf algorithm is used for each element in the
vector just inlined for better scheduling and simpler special case
handling. The log polynomial is smaller as less accuracy is enough.

Worst-case error is 2.6 ulp.

The polynomials were produced by searching the coefficient space using
heuristics and ideas from https://arxiv.org/abs/1508.03211

The worst-case error is 1.886 ulp, large inputs (> 2^20) and other
special cases use scalar sinf and cosf.

The polynomial was produced by searching the coefficient space using
heuristics and ideas from https://arxiv.org/abs/1508.03211

The worst-case error is 3.34 ulp, subnormal range inputs and other
special cases use scalar logf.

Vector math routines are added to the same libmathlib library as scalar
ones. The difficulty is that they are not always available, the external
abi depends on the compiler version used for the build. Currently only
aarch64 AdvSIMD is supported, there are 4 new sets of symbols:

  __s_foo is a scalar function with identical result to the vector one,
  __v_foo is a vector function using the base PCS,
  __vn_foo uses the vector PCS and
  _ZGV*_foo is the vector ABI symbol alias of vn_foo

for a scalar math function foo.

The test and benchmark code got extended to handle vector functions.

Vector functions aim for < 5 ulp worst case error, only support nearest
rounding mode and don't support floating-point exceptions. Vector
functions may call scalar functions to handle special cases, but for a
single value they should return the same result independently of values
in other vector lanes or the position of the value in the vector.

The __v_expf and __v_expf_1u polynomials were produced by searching the
coefficient space with some heuristics and ideas from
https://arxiv.org/abs/1508.03211
Their worst case error is 1.95 and 0.866 ulp respectively.

The exp polynomial was produced by sollya, it uses a 128 element (1KB)
lookup table and has 2.38 ulp worst case error.

Not all symbols referenced by mathbench may be available in libc so
link to libmathlib too to resolve the missing symbols.

Fix it to be python3 compatible and plot the exact and approximated
values too.

The ulp tool compares output of a math function to a larger precision
implementation of the same function.

But when the input argument is converted to a larger precision number
the signaling nan property is lost, so ensure that the conversion
happens inside the critical region where fenv exceptions are checked
and then the conversion itself will raise the invalid exception, which
is the correct behaviour in most cases.

The volatile barrier is not perfect and the snan behaviour is not
always signaling, but this should give more reliable results in most
cases than before.

fenv support is not reliable in clang so provide a mechanism to
disable fenv status checks and only check the result values.

Users may want different CFLAGS for math and string subprojects, expose
a mechanism for this in config.mk.

Allows optimizing the code in shared libraries differently.
Has significant effect on literal loads in simd code.

Make ulp and runulp.sh fail on error.

Only increment once per fgets.

Make mathtest fail on error so make check fails too.

Without printing anything on success it is unclear if the right set
of functions got hooked up in the test code.