Removes preprocessor check for FMA instructions in matrix multiplication functions.
This simplifies the code and relies on the compiler's ability to optimize the
code based on available hardware support. The assumption is that modern
compilers will automatically utilize FMA instructions if available, and fall
back to alternative implementations if not.
Ensures that AVX2 intrinsics are only included when the
OMATH_USE_AVX2 preprocessor definition is set. This prevents
compilation errors when AVX2 support is not available or
explicitly disabled.
The C standard library function fmodf is not guaranteed to be in namespace std, and in fact is not with a default Ubuntu 24.04 installation, leading to the following compile error:
```console
Change Dir: '/vcpkg/buildtrees/vcpkg-ci-orange-math/x64-linux-dbg'
Run Build Command(s): /vcpkg/downloads/tools/ninja/1.12.1-linux/ninja -v -v -j33
[1/2] /usr/bin/c++ -DOMATH_SUPRESS_SAFETY_CHECKS -DOMATH_VERSION=\"3.5.0\" -isystem /vcpkg/installed/x64-linux/include -fPIC -g -std=gnu++23 -MD -MT CMakeFiles/main.dir/main.cpp.o -MF CMakeFiles/main.dir/main.cpp.o.d -o CMakeFiles/main.dir/main.cpp.o -c /vcpkg/scripts/test_ports/vcpkg-ci-orange-math/project/main.cpp
FAILED: CMakeFiles/main.dir/main.cpp.o
/usr/bin/c++ -DOMATH_SUPRESS_SAFETY_CHECKS -DOMATH_VERSION=\"3.5.0\" -isystem /vcpkg/installed/x64-linux/include -fPIC -g -std=gnu++23 -MD -MT CMakeFiles/main.dir/main.cpp.o -MF CMakeFiles/main.dir/main.cpp.o.d -o CMakeFiles/main.dir/main.cpp.o -c /vcpkg/scripts/test_ports/vcpkg-ci-orange-math/project/main.cpp
In file included from /vcpkg/installed/x64-linux/include/omath/omath.hpp:22,
from /vcpkg/scripts/test_ports/vcpkg-ci-orange-math/project/main.cpp:1:
/vcpkg/installed/x64-linux/include/omath/color.hpp: In member function ‘constexpr omath::Hsv omath::Color::to_hsv() const’:
/vcpkg/installed/x64-linux/include/omath/color.hpp:98:45: error: ‘fmodf’ is not a member of ‘std’; did you mean ‘modf’?
98 | hsv_data.hue = 60.f * (std::fmodf(((green - blue) / delta), 6.f));
| ^~~~~
| modf
ninja: build stopped: subcommand failed.
```
Only the 'sufficient additional overloads' of `fmod` are guaranteed to be in `std`. Since this is clearly intended to call the (float, float) overload, explicitly cast `((green - blue) / delta)` (which is a `double`) to `float` and call the name in `std` as suggested by the diagnostic.
Optimizes matrix multiplication by specializing the algorithm
based on the matrix storage type (row-major or column-major).
This change significantly improves performance by leveraging
memory access patterns specific to each storage order.
