This patch fixes the error introduced by the last one.
It also enforces detection of VSX and ZVECTOR2.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch fixes handling of compiler flags that was preventing AVX2 detection.
This patch also adds ENFORCE options to prevent the same problem from occurring.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch changes suffixes of constants to a more portable form.
For example, 1LL is changed to INT64_C(1).

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

With this patch, multiple inline headers can be included from the same source file.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

sleef.h will use `vector ...` types unprotected on POWER, which
requires including AltiVec just like, for example NEON requires
arm_neon.h.  However, unlike NEON, SVE, and SSE, until this patch
SLEEF required including altivec.h prior to including sleef.h.

Co-authored-by: Naoki Shibata <shibatch@users.sourceforge.net>

This patch adds testing for nested builds.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

* Use GNUInstallDirs to determine install paths.

* Add pkg-config file for libm.

Co-authored-by: Naoki Shibata <shibatch@users.sourceforge.net>

Co-authored-by: Nico v. Huene <nico.von.huene@kinexon.com>
Co-authored-by: Naoki Shibata <shibatch@users.sourceforge.net>

This patch adds DISABLE_FFTW cmake option.

This patch adds IBM System/390 support.

With this patch, the library can be built with cmake-3.5.1.

This patch moves LTO testing to Jenkins.
It adds additional processing for inline headers to remove empty lines.
It also updates comments.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This should allow us to build the test executables statically.

With this patch, static libraries with LTO (with gcc) or ThinLTO (with clang) can be generated by specifying -DENABLE_LTO=TRUE option to cmake.

This can be used to detect whether the version of SLEEF being used
supports a feature.  For example

  #if SLEEF_VERSION_MAJOR > 3 || \
      (SLEEF_VERSION_MAJOR == 3 && SLEEF_VERSION_MINOR >= 5)
    /* Use function only available in SLEEF 3.5+ */
  #endif

With this patch, the library can be built with Clang on Windows.

This patch fixes a bug in remainder function.

1. `string(CONCAT)` does not append content to the string but replaces the content of the output variable. Appending should use `string(APPEND)` instead.
2. The contents of the two arguments should be concatenated with a space so
   `string(APPEND)`/`string(CONCAT)` aren't the right/most convinient functions to use either.
   Since there're only two strings input, simply use a string interpolation instead.

CI has been extended to use gcc-10.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch adds "extern" to the reference to variable enableFlushToZero in testerutil.h.
This bug prevented gcc-10 from building the library.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch fixes a stupid bug in handling compiler options.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch adds extern "C" declaration to sleefdft.h, as suggested in issue #298.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch removes bug workarounds which were required when cross-compiling for PPC64.
We do not need cross-compilation to PPC64 anymore, because Travis now has native PPC64 servers.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch updates CI configurations.
Ubuntu bionic and newer compilers will be used for testing at Travis.
Appveyor will utilize VS2019 and self-hosted computers.
Ninja build system is used in replacement of make.

Authored-by: shibatch <shibatch.sf.net@gmail.com>

With this patch, testing will fail if IUT crashes.
This patch also adds ENFORCE_TESTER cmake option which make a build fail if no tester is available.

Authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch simplifies rint algorithm that are used when such an instruction is not available.
This patch also replaces sqrt algorithm with FMA.
These changes do not affect performance with recent vector extensions.

Authored-by: shibatch <shibatch.sf.net@gmail.com>

This patch updates copyright notice.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

With this patch, alternative division and sqrt can be chosen by specifying -DENABLE_ALTDIV=TRUE -DENABLE_ALTSQRT=TRUE as cmake options.
The alternative methods use combinations of FMA operations to compute division and sqrt.
These methods could possibly be beneficial for micro-architectures on which the corresponding instructions are non-pipelined and have long latencies.

Co-authored-by: shibatch <shibatch.sf.net@gmail.com>

* ppc64le: Several improvements for VSX

  - fix the build on gcc
  - fix accuracy tests
  - fix the build with C++ when 'vector' compiler token is combined,
    NOTE: the generated header is using '__vector' instead of 'vector'
  - fix unaligned memory load
  - set rounding to the nearest even, make it equivalent to X86
  - improve the performance of conversions
  - improve the performance of shuffles and reverses
  - cleanup the whole VSX implementation

* Travis CI: Add ppc64le support

This patch allows rounding operations to use _mm512_roundscale_pd or _mm512_roundscale_ps with AVX-512. This should make the whole computation faster.

This patch adds a program for generating rempitab.c.