DirectXMath in C

This is an adaptation of the official DirectXMath C++ library using C.

This was done for educational purposes, in order to learn better about SIMD and DirectX.

How to build (MSVC)

To run the tests, build with the following CMake command:

cmake -B build-test-vectorcall -S . -DXM_USE_TEST=1 -DUSE_VECTORCALL=1 -DCMAKE_BUILD_TYPE=Debug

To not use __vectorcall, just change the -DUSE_VECTORCALL=0

To do the same disabling intrinsics:

cmake -B build-test-nointrin -S . -DXM_NO_INTRINSICS=ON -DXM_USE_TEST=ON -DCMAKE_BUILD_TYPE=Debug

To run the tests, run the following:

cmake --build build-test-vectorcall
ctest --test-dir build-test-vectorcall -C Debug -V

How to build (GCC)

Just add -G "MinGW Makefiles":

cmake -B build-test-gcc -S . -DXM_USE_TEST=1 -DUSE_VECTORCALL=1 -DCMAKE_BUILD_TYPE=Debug -G "MinGW Makefiles"
cmake --build build-test-gcc
ctest --test-dir build-test-gcc -C Debug -V

Installing (to use in other projects)

MSVC:

cmake -S . -B build-install
cmake --build build-install --target INSTALL --config Release

GCC:

cmake -S . -B build-install-gcc -G "MinGW Makefiles" -DCMAKE_BUILD_TYPE=Release
cmake --build build-install-gcc --target install

This will generate the folder xmathc, probably in C:/. This will allow us to use xmathc in other projects.

Usage examples

To allow it to use callconv like the original DirectXMath, I had to create macros that hide the implementation details. In the original implementation, this is cleaner because C++ can pass arguments by reference or by value seamlessly to the caller.

XMMATRIX A_XMMATRIX = XMLoadFloat3x3(&A); // A is XMFLOAT3X3 
XMMATRIX B_XMMATRIX = XMLoadFloat3x3(&B); // B is XMFLOAT3X3

// XM_MAT_MULT macro will handle if it should pass by value or ref depending on adequate callconv 
XMMATRIX C_XMMATRIX = XM_MAT_MULT(A_XMMATRIX, B_XMMATRIX);

XMFLOAT4X4 result;
// XM_STORE_FLOAT4X4 is a macro for the same reason XM_MAT_MULT 
XM_STORE_FLOAT4X4(&result, C_XMMATRIX);