summaryrefslogtreecommitdiff
path: root/include/cglm/mat2.h
diff options
context:
space:
mode:
Diffstat (limited to 'include/cglm/mat2.h')
-rw-r--r--include/cglm/mat2.h363
1 files changed, 363 insertions, 0 deletions
diff --git a/include/cglm/mat2.h b/include/cglm/mat2.h
new file mode 100644
index 0000000..9248460
--- /dev/null
+++ b/include/cglm/mat2.h
@@ -0,0 +1,363 @@
+/*
+ * Copyright (c), Recep Aslantas.
+ *
+ * MIT License (MIT), http://opensource.org/licenses/MIT
+ * Full license can be found in the LICENSE file
+ */
+
+/*
+ Macros:
+ GLM_MAT2_IDENTITY_INIT
+ GLM_MAT2_ZERO_INIT
+ GLM_MAT2_IDENTITY
+ GLM_MAT2_ZERO
+
+ Functions:
+ CGLM_INLINE void glm_mat2_make(float * restrict src, mat2 dest)
+ CGLM_INLINE void glm_mat2_copy(mat2 mat, mat2 dest)
+ CGLM_INLINE void glm_mat2_identity(mat2 m)
+ CGLM_INLINE void glm_mat2_identity_array(mat2 * restrict mats, size_t count)
+ CGLM_INLINE void glm_mat2_zero(mat2 m)
+ CGLM_INLINE void glm_mat2_mul(mat2 m1, mat2 m2, mat2 dest)
+ CGLM_INLINE void glm_mat2_mulv(mat2 m, vec2 v, vec2 dest)
+ CGLM_INLINE void glm_mat2_transpose_to(mat2 mat, mat2 dest)
+ CGLM_INLINE void glm_mat2_transpose(mat2 m)
+ CGLM_INLINE void glm_mat2_scale(mat2 m, float s)
+ CGLM_INLINE void glm_mat2_inv(mat2 mat, mat2 dest)
+ CGLM_INLINE void glm_mat2_swap_col(mat2 mat, int col1, int col2)
+ CGLM_INLINE void glm_mat2_swap_row(mat2 mat, int row1, int row2)
+ CGLM_INLINE float glm_mat2_det(mat2 m)
+ CGLM_INLINE float glm_mat2_trace(mat2 m)
+ CGLM_INLINE float glm_mat2_rmc(vec2 r, mat2 m, vec2 c)
+ */
+
+#ifndef cglm_mat2_h
+#define cglm_mat2_h
+
+#include "common.h"
+#include "vec2.h"
+
+#ifdef CGLM_SSE_FP
+# include "simd/sse2/mat2.h"
+#endif
+
+#ifdef CGLM_NEON_FP
+# include "simd/neon/mat2.h"
+#endif
+
+#ifdef CGLM_SIMD_WASM
+# include "simd/wasm/mat2.h"
+#endif
+
+#define GLM_MAT2_IDENTITY_INIT {{1.0f, 0.0f}, {0.0f, 1.0f}}
+#define GLM_MAT2_ZERO_INIT {{0.0f, 0.0f}, {0.0f, 0.0f}}
+
+/* for C only */
+#define GLM_MAT2_IDENTITY ((mat2)GLM_MAT2_IDENTITY_INIT)
+#define GLM_MAT2_ZERO ((mat2)GLM_MAT2_ZERO_INIT)
+
+/*!
+ * @brief Create mat2 (dest) from pointer (src).
+ *
+ * @param[in] src pointer to an array of floats (left)
+ * @param[out] dest destination (result, mat2)
+ */
+CGLM_INLINE
+void
+glm_mat2_make(const float * __restrict src, mat2 dest) {
+ dest[0][0] = src[0];
+ dest[0][1] = src[1];
+ dest[1][0] = src[2];
+ dest[1][1] = src[3];
+}
+
+/*!
+ * @brief Copy mat2 (mat) to mat2 (dest).
+ *
+ * @param[in] mat mat2 (left, src)
+ * @param[out] dest destination (result, mat2)
+ */
+CGLM_INLINE
+void
+glm_mat2_copy(mat2 mat, mat2 dest) {
+ glm_vec4_ucopy(mat[0], dest[0]);
+}
+
+/*!
+ * @brief Copy a mat2 identity to mat2 (m), or makes mat2 (m) an identity.
+ *
+ * The same thing may be achieved with either of bellow methods,
+ * but it is more easy to do that with this func especially for members
+ * e.g. glm_mat2_identity(aStruct->aMatrix);
+ *
+ * @code
+ * glm_mat2_copy(GLM_MAT2_IDENTITY, mat); // C only
+ *
+ * // or
+ * mat2 mat = GLM_MAT2_IDENTITY_INIT;
+ * @endcode
+ *
+ * @param[in, out] m mat2 (src, dest)
+ */
+CGLM_INLINE
+void
+glm_mat2_identity(mat2 m) {
+ CGLM_ALIGN_MAT mat2 t = GLM_MAT2_IDENTITY_INIT;
+ glm_mat2_copy(t, m);
+}
+
+/*!
+ * @brief Given an array of mat2’s (mats) make each matrix an identity matrix.
+ *
+ * @param[in, out] mats Array of mat2’s (must be aligned (16/32) if alignment is not disabled)
+ * @param[in] count Array size of mats or number of matrices
+ */
+CGLM_INLINE
+void
+glm_mat2_identity_array(mat2 * __restrict mats, size_t count) {
+ CGLM_ALIGN_MAT mat2 t = GLM_MAT2_IDENTITY_INIT;
+ size_t i;
+
+ for (i = 0; i < count; i++) {
+ glm_mat2_copy(t, mats[i]);
+ }
+}
+
+/*!
+ * @brief Zero out the mat2 (m).
+ *
+ * @param[in, out] m mat2 (src, dest)
+ */
+CGLM_INLINE
+void
+glm_mat2_zero(mat2 m) {
+ CGLM_ALIGN_MAT mat2 t = GLM_MAT2_ZERO_INIT;
+ glm_mat2_copy(t, m);
+}
+
+/*!
+ * @brief Multiply mat2 (m1) by mat2 (m2) and store in mat2 (dest).
+ *
+ * m1, m2 and dest matrices can be same matrix, it is possible to write this:
+ *
+ * @code
+ * mat2 m = GLM_MAT2_IDENTITY_INIT;
+ * glm_mat2_mul(m, m, m);
+ * @endcode
+ *
+ * @param[in] m1 mat2 (left)
+ * @param[in] m2 mat2 (right)
+ * @param[out] dest destination (result, mat2)
+ */
+CGLM_INLINE
+void
+glm_mat2_mul(mat2 m1, mat2 m2, mat2 dest) {
+#if defined(__wasm__) && defined(__wasm_simd128__)
+ glm_mat2_mul_wasm(m1, m2, dest);
+#elif defined( __SSE__ ) || defined( __SSE2__ )
+ glm_mat2_mul_sse2(m1, m2, dest);
+#elif defined(CGLM_NEON_FP)
+ glm_mat2_mul_neon(m1, m2, dest);
+#else
+ float a00 = m1[0][0], a01 = m1[0][1],
+ a10 = m1[1][0], a11 = m1[1][1],
+ b00 = m2[0][0], b01 = m2[0][1],
+ b10 = m2[1][0], b11 = m2[1][1];
+
+ dest[0][0] = a00 * b00 + a10 * b01;
+ dest[0][1] = a01 * b00 + a11 * b01;
+ dest[1][0] = a00 * b10 + a10 * b11;
+ dest[1][1] = a01 * b10 + a11 * b11;
+#endif
+}
+
+/*!
+ * @brief Multiply mat2 (m) by vec2 (v) and store in vec2 (dest).
+ *
+ * @param[in] m mat2 (left)
+ * @param[in] v vec2 (right, column vector)
+ * @param[out] dest destination (result, column vector)
+ */
+CGLM_INLINE
+void
+glm_mat2_mulv(mat2 m, vec2 v, vec2 dest) {
+ dest[0] = m[0][0] * v[0] + m[1][0] * v[1];
+ dest[1] = m[0][1] * v[0] + m[1][1] * v[1];
+}
+
+/*!
+ * @brief Transpose mat2 (mat) and store in mat2 (dest).
+ *
+ * @param[in] mat mat2 (left, src)
+ * @param[out] dest destination (result, mat2)
+ */
+CGLM_INLINE
+void
+glm_mat2_transpose_to(mat2 mat, mat2 dest) {
+#if defined(__wasm__) && defined(__wasm_simd128__)
+ glm_mat2_transp_wasm(mat, dest);
+#elif defined( __SSE__ ) || defined( __SSE2__ )
+ glm_mat2_transp_sse2(mat, dest);
+#else
+ dest[0][0] = mat[0][0];
+ dest[0][1] = mat[1][0];
+ dest[1][0] = mat[0][1];
+ dest[1][1] = mat[1][1];
+#endif
+}
+
+/*!
+ * @brief Transpose mat2 (m) and store result in the same matrix.
+ *
+ * @param[in, out] m mat2 (src, dest)
+ */
+CGLM_INLINE
+void
+glm_mat2_transpose(mat2 m) {
+ float tmp;
+ tmp = m[0][1];
+ m[0][1] = m[1][0];
+ m[1][0] = tmp;
+}
+
+/*!
+ * @brief Multiply mat2 (m) by scalar constant (s).
+ *
+ * @param[in, out] m mat2 (src, dest)
+ * @param[in] s float (scalar)
+ */
+CGLM_INLINE
+void
+glm_mat2_scale(mat2 m, float s) {
+#if defined(__wasm__) && defined(__wasm_simd128__)
+ glmm_store(m[0], wasm_f32x4_mul(wasm_v128_load(m[0]),
+ wasm_f32x4_splat(s)));
+#elif defined( __SSE__ ) || defined( __SSE2__ )
+ glmm_store(m[0], _mm_mul_ps(_mm_loadu_ps(m[0]), glmm_set1(s)));
+#elif defined(CGLM_NEON_FP)
+ vst1q_f32(m[0], vmulq_f32(vld1q_f32(m[0]), vdupq_n_f32(s)));
+#else
+ m[0][0] = m[0][0] * s;
+ m[0][1] = m[0][1] * s;
+ m[1][0] = m[1][0] * s;
+ m[1][1] = m[1][1] * s;
+#endif
+}
+
+/*!
+ * @brief Inverse mat2 (mat) and store in mat2 (dest).
+ *
+ * @param[in] mat mat2 (left, src)
+ * @param[out] dest destination (result, inverse mat2)
+ */
+CGLM_INLINE
+void
+glm_mat2_inv(mat2 mat, mat2 dest) {
+ float det;
+ float a = mat[0][0], b = mat[0][1],
+ c = mat[1][0], d = mat[1][1];
+
+ det = 1.0f / (a * d - b * c);
+
+ dest[0][0] = d * det;
+ dest[0][1] = -b * det;
+ dest[1][0] = -c * det;
+ dest[1][1] = a * det;
+}
+
+/*!
+ * @brief Swap two columns in mat2 (mat) and store in same matrix.
+ *
+ * @param[in, out] mat mat2 (src, dest)
+ * @param[in] col1 Column 1 array index
+ * @param[in] col2 Column 2 array index
+ */
+CGLM_INLINE
+void
+glm_mat2_swap_col(mat2 mat, int col1, int col2) {
+ float a, b;
+
+ a = mat[col1][0];
+ b = mat[col1][1];
+
+ mat[col1][0] = mat[col2][0];
+ mat[col1][1] = mat[col2][1];
+
+ mat[col2][0] = a;
+ mat[col2][1] = b;
+}
+
+/*!
+ * @brief Swap two rows in mat2 (mat) and store in same matrix.
+ *
+ * @param[in, out] mat mat2 (src, dest)
+ * @param[in] row1 Row 1 array index
+ * @param[in] row2 Row 2 array index
+ */
+CGLM_INLINE
+void
+glm_mat2_swap_row(mat2 mat, int row1, int row2) {
+ float a, b;
+
+ a = mat[0][row1];
+ b = mat[1][row1];
+
+ mat[0][row1] = mat[0][row2];
+ mat[1][row1] = mat[1][row2];
+
+ mat[0][row2] = a;
+ mat[1][row2] = b;
+}
+
+/*!
+ * @brief Returns mat2 determinant.
+ *
+ * @param[in] m mat2 (src)
+ *
+ * @return[out] mat2 determinant (float)
+ */
+CGLM_INLINE
+float
+glm_mat2_det(mat2 m) {
+ return m[0][0] * m[1][1] - m[1][0] * m[0][1];
+}
+
+/*!
+ * @brief Returns trace of matrix. Which is:
+ *
+ * The sum of the elements on the main diagonal from
+ * upper left corner to the bottom right corner.
+ *
+ * @param[in] m mat2 (src)
+ *
+ * @return[out] mat2 trace (float)
+ */
+CGLM_INLINE
+float
+glm_mat2_trace(mat2 m) {
+ return m[0][0] + m[1][1];
+}
+
+/*!
+ * @brief Helper for R (row vector) * M (matrix) * C (column vector)
+ *
+ * rmc stands for Row * Matrix * Column
+ *
+ * the result is scalar because M * C = ResC (1x2, column vector),
+ * then if you take the dot_product(R (2x1), ResC (1x2)) = scalar value.
+ *
+ * @param[in] r vec2 (2x1, row vector)
+ * @param[in] m mat2 (2x2, matrix)
+ * @param[in] c vec2 (1x2, column vector)
+ *
+ * @return[out] Scalar value (float, 1x1)
+ */
+CGLM_INLINE
+float
+glm_mat2_rmc(vec2 r, mat2 m, vec2 c) {
+ vec2 tmp;
+ glm_mat2_mulv(m, c, tmp);
+ return glm_vec2_dot(r, tmp);
+}
+
+#endif /* cglm_mat2_h */