Рефакторинг и оптимизация вычислений / Refactoring and optimization of computations

basic-geometry-dev/main.c

@@ -52,7 +52,7 @@ BgFP32Versor * make_random_versors(const unsigned int amount)
 
 void print_versor(const BgFP32Versor* versor)
 {
-    printf("(%f, %f, %f, %f)\n", versor->s0, versor->x1, versor->x2, versor->x3);
+    printf("Versor (%f, %f, %f, %f); Delta = %e\n", versor->s0, versor->x1, versor->x2, versor->x3, bg_fp32_versor_get_modulus(versor) - 1.0f);
 }
 
 void print_vector(const BgFP32Vector3* vector)
@@ -125,11 +125,11 @@ int main()
     const unsigned int amount = 1000000;
 
 #ifdef _WIN64
-    ULONGLONG now;
+    ULONGLONG now, start, end;
     now = GetTickCount64();
     srand((unsigned int)(now & 0xfffffff));
 #else
-    struct timespec now;
+    struct timespec now, start, end;
     clock_gettime(0, &now);
     srand((unsigned int)(now.tv_nsec & 0xfffffff));
 #endif // _WIN64
@@ -159,10 +159,14 @@ int main()
     }
 
 #ifdef _WIN64
-    ULONGLONG start, end;
+    end = GetTickCount64();
+    printf("Setup time: %lld\n", end - now);
+
     start = GetTickCount64();
 #else
-    struct timespec start, end;
+    clock_gettime(CLOCK_REALTIME, &end);
+    printf("Time: %lf\n", (end.tv_sec - now.tv_sec) * 1000.0 + (end.tv_nsec - now.tv_nsec) * 0.000001);
+
     clock_gettime(CLOCK_REALTIME, &start);
 #endif // _WIN64
     for (int j = 0; j < 1000; j++) {

basic-geometry/matrix2x2.c

@@ -1 +1,105 @@
 #include "matrix2x2.h"
+
+// ================= Inversion ================== //
+
+int bg_fp32_matrix2x2_invert(BgFP32Matrix2x2* matrix)
+{
+    const float determinant = bg_fp32_matrix2x2_get_determinant(matrix);
+
+    if (-BG_FP32_EPSYLON <= determinant && determinant <= BG_FP32_EPSYLON) {
+        return 0;
+    }
+
+    const float r1c1 = matrix->r2c2;
+    const float r1c2 = -matrix->r1c2;
+
+    const float r2c1 = -matrix->r2c1;
+    const float r2c2 = matrix->r1c1;
+
+    const float multiplier = 1.0f / determinant;
+
+    matrix->r1c1 = r1c1 * multiplier;
+    matrix->r1c2 = r1c2 * multiplier;
+
+    matrix->r2c1 = r2c1 * multiplier;
+    matrix->r2c2 = r2c2 * multiplier;
+
+    return 1;
+}
+
+int bg_fp64_matrix2x2_invert(BgFP64Matrix2x2* matrix)
+{
+    const double determinant = bg_fp64_matrix2x2_get_determinant(matrix);
+
+    if (-BG_FP64_EPSYLON <= determinant && determinant <= BG_FP64_EPSYLON) {
+        return 0;
+    }
+
+    const double r1c1 = matrix->r2c2;
+    const double r1c2 = -matrix->r1c2;
+
+    const double r2c1 = -matrix->r2c1;
+    const double r2c2 = matrix->r1c1;
+
+    const double multiplier = 1.0 / determinant;
+
+    matrix->r1c1 = r1c1 * multiplier;
+    matrix->r1c2 = r1c2 * multiplier;
+
+    matrix->r2c1 = r2c1 * multiplier;
+    matrix->r2c2 = r2c2 * multiplier;
+
+    return 1;
+}
+
+// ================ Set Inverted ================ //
+
+int bg_fp32_matrix2x2_set_inverted(const BgFP32Matrix2x2* from, BgFP32Matrix2x2* to)
+{
+    const float determinant = bg_fp32_matrix2x2_get_determinant(from);
+
+    if (-BG_FP32_EPSYLON <= determinant && determinant <= BG_FP32_EPSYLON) {
+        return 0;
+    }
+
+    const float r1c1 = from->r2c2;
+    const float r1c2 = -from->r1c2;
+
+    const float r2c1 = -from->r2c1;
+    const float r2c2 = from->r1c1;
+
+    const float multiplier = 1.0f / determinant;
+
+    to->r1c1 = r1c1 * multiplier;
+    to->r1c2 = r1c2 * multiplier;
+
+    to->r2c1 = r2c1 * multiplier;
+    to->r2c2 = r2c2 * multiplier;
+
+    return 1;
+}
+
+int bg_fp64_matrix2x2_set_inverted(const BgFP64Matrix2x2* from, BgFP64Matrix2x2* to)
+{
+    const double determinant = bg_fp64_matrix2x2_get_determinant(from);
+
+    if (-BG_FP64_EPSYLON <= determinant && determinant <= BG_FP64_EPSYLON) {
+        return 0;
+    }
+
+    const double r1c1 = from->r2c2;
+    const double r1c2 = -from->r1c2;
+
+    const double r2c1 = -from->r2c1;
+    const double r2c2 = from->r1c1;
+
+    const double multiplier = 1.0 / determinant;
+
+    to->r1c1 = r1c1 * multiplier;
+    to->r1c2 = r1c2 * multiplier;
+
+    to->r2c1 = r2c1 * multiplier;
+    to->r2c2 = r2c2 * multiplier;
+
+    return 1;
+}

basic-geometry/matrix2x2.h

@@ -171,51 +171,9 @@ static inline void bg_fp64_matrix2x2_transpose(BgFP64Matrix2x2* matrix)
 
 // ================= Inversion ================== //
 
-static inline int bg_fp32_matrix2x2_invert(BgFP32Matrix2x2* matrix)
+int bg_fp32_matrix2x2_invert(BgFP32Matrix2x2* matrix);
-{
-    const float determinant = bg_fp32_matrix2x2_get_determinant(matrix);
 
-    if (-BG_FP32_EPSYLON <= determinant && determinant <= BG_FP32_EPSYLON) {
+int bg_fp64_matrix2x2_invert(BgFP64Matrix2x2* matrix);
-        return 0;
-    }
-
-    const float r1c1 =  matrix->r2c2;
-    const float r1c2 = -matrix->r1c2;
-
-    const float r2c1 = -matrix->r2c1;
-    const float r2c2 =  matrix->r1c1;
-
-    matrix->r1c1 = r1c1 / determinant;
-    matrix->r1c2 = r1c2 / determinant;
-
-    matrix->r2c1 = r2c1 / determinant;
-    matrix->r2c2 = r2c2 / determinant;
-
-    return 1;
-}
-
-static inline int bg_fp64_matrix2x2_invert(BgFP64Matrix2x2* matrix)
-{
-    const double determinant = bg_fp64_matrix2x2_get_determinant(matrix);
-
-    if (-BG_FP64_EPSYLON <= determinant && determinant <= BG_FP64_EPSYLON) {
-        return 0;
-    }
-
-    const double r1c1 =  matrix->r2c2;
-    const double r1c2 = -matrix->r1c2;
-
-    const double r2c1 = -matrix->r2c1;
-    const double r2c2 =  matrix->r1c1;
-
-    matrix->r1c1 = r1c1 / determinant;
-    matrix->r1c2 = r1c2 / determinant;
-
-    matrix->r2c1 = r2c1 / determinant;
-    matrix->r2c2 = r2c2 / determinant;
-
-    return 1;
-}
 
 // =============== Set Transposed =============== //
 
@@ -243,51 +201,9 @@ static inline void bg_fp64_matrix2x2_set_transposed(const BgFP64Matrix2x2* from,
 
 // ================ Set Inverted ================ //
 
-static inline int bg_fp32_matrix2x2_set_inverted(const BgFP32Matrix2x2* from, BgFP32Matrix2x2* to)
+int bg_fp32_matrix2x2_set_inverted(const BgFP32Matrix2x2* from, BgFP32Matrix2x2* to);
-{
-    const float determinant = bg_fp32_matrix2x2_get_determinant(from);
 
-    if (-BG_FP32_EPSYLON <= determinant && determinant <= BG_FP32_EPSYLON) {
+int bg_fp64_matrix2x2_set_inverted(const BgFP64Matrix2x2* from, BgFP64Matrix2x2* to);
-        return 0;
-    }
-
-    const float r1c1 =  from->r2c2;
-    const float r1c2 = -from->r1c2;
-
-    const float r2c1 = -from->r2c1;
-    const float r2c2 =  from->r1c1;
-
-    to->r1c1 = r1c1 / determinant;
-    to->r1c2 = r1c2 / determinant;
-
-    to->r2c1 = r2c1 / determinant;
-    to->r2c2 = r2c2 / determinant;
-
-    return 1;
-}
-
-static inline int bg_fp64_matrix2x2_set_inverted(const BgFP64Matrix2x2* from, BgFP64Matrix2x2* to)
-{
-    const double determinant = bg_fp64_matrix2x2_get_determinant(from);
-
-    if (-BG_FP64_EPSYLON <= determinant && determinant <= BG_FP64_EPSYLON) {
-        return 0;
-    }
-
-    const double r1c1 =  from->r2c2;
-    const double r1c2 = -from->r1c2;
-
-    const double r2c1 = -from->r2c1;
-    const double r2c2 =  from->r1c1;
-
-    to->r1c1 = r1c1 / determinant;
-    to->r1c2 = r1c2 / determinant;
-
-    to->r2c1 = r2c1 / determinant;
-    to->r2c2 = r2c2 / determinant;
-
-    return 1;
-}
 
 // ================= Set Row 1 ================== //
 
@@ -429,60 +345,52 @@ static inline void bg_fp64_matrix2x2_multiply(const BgFP64Matrix2x2* multiplican
 
 static inline void bg_fp32_matrix2x2_divide(const BgFP32Matrix2x2* dividend, const float divisor, BgFP32Matrix2x2* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp32_matrix2x2_multiply(dividend, 1.0f / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
 }
 
 static inline void bg_fp64_matrix2x2_divide(const BgFP64Matrix2x2* dividend, const double divisor, BgFP64Matrix2x2* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp64_matrix2x2_multiply(dividend, 1.0 / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
 }
 
 // ============ Left Vector Product ============= //
 
 static inline void bg_fp32_matrix2x2_left_product(const BgFP32Vector2* vector, const BgFP32Matrix2x2* matrix, BgFP32Vector2* result)
 {
-    bg_fp32_vector2_set_values(
+    const float x1 = vector->x1 * matrix->r1c1 + vector->x2 * matrix->r2c1;
-        vector->x1 * matrix->r1c1 + vector->x2 * matrix->r2c1,
+    const float x2 = vector->x1 * matrix->r1c2 + vector->x2 * matrix->r2c2;
-        vector->x1 * matrix->r1c2 + vector->x2 * matrix->r2c2,
+
-        result
+    result->x1 = x1;
-    );
+    result->x2 = x2;
 }
 
 static inline void bg_fp64_matrix2x2_left_product(const BgFP64Vector2* vector, const BgFP64Matrix2x2* matrix, BgFP64Vector2* result)
 {
-    bg_fp64_vector2_set_values(
+    const double x1 = vector->x1 * matrix->r1c1 + vector->x2 * matrix->r2c1;
-        vector->x1 * matrix->r1c1 + vector->x2 * matrix->r2c1,
+    const double x2 = vector->x1 * matrix->r1c2 + vector->x2 * matrix->r2c2;
-        vector->x1 * matrix->r1c2 + vector->x2 * matrix->r2c2,
+
-        result
+    result->x1 = x1;
-    );
+    result->x2 = x2;
 }
 
 // ============ Right Vector Product ============ //
 
 static inline void bg_fp32_matrix2x2_right_product(const BgFP32Matrix2x2* matrix, const BgFP32Vector2* vector, BgFP32Vector2* result)
 {
-    bg_fp32_vector2_set_values(
+    const float x1 = matrix->r1c1 * vector->x1 + matrix->r1c2 * vector->x2;
-        matrix->r1c1 * vector->x1 + matrix->r1c2 * vector->x2,
+    const float x2 = matrix->r2c1 * vector->x1 + matrix->r2c2 * vector->x2;
-        matrix->r2c1 * vector->x1 + matrix->r2c2 * vector->x2,
+
-        result
+    result->x1 = x1;
-    );
+    result->x2 = x2;
 }
 
 static inline void bg_fp64_matrix2x2_right_product(const BgFP64Matrix2x2* matrix, const BgFP64Vector2* vector, BgFP64Vector2* result)
 {
-    bg_fp64_vector2_set_values(
+    const double x1 = matrix->r1c1 * vector->x1 + matrix->r1c2 * vector->x2;
-        matrix->r1c1 * vector->x1 + matrix->r1c2 * vector->x2,
+    const double x2 = matrix->r2c1 * vector->x1 + matrix->r2c2 * vector->x2;
-        matrix->r2c1 * vector->x1 + matrix->r2c2 * vector->x2,
+
-        result
+    result->x1 = x1;
-    );
+    result->x2 = x2;
 }
 
 #endif

basic-geometry/matrix2x3.h

@@ -317,26 +317,12 @@ static inline void bg_fp64_matrix2x3_multiply(const BgFP64Matrix2x3* multiplican
 
 static inline void bg_fp32_matrix2x3_divide(const BgFP32Matrix2x3* dividend, const float divisor, BgFP32Matrix2x3* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp32_matrix2x3_multiply(dividend, 1.0f / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
-
-    quotient->r3c1 = dividend->r3c1 / divisor;
-    quotient->r3c2 = dividend->r3c2 / divisor;
 }
 
 static inline void bg_fp64_matrix2x3_divide(const BgFP64Matrix2x3* dividend, const double divisor, BgFP64Matrix2x3* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp64_matrix2x3_multiply(dividend, 1.0 / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
-
-    quotient->r3c1 = dividend->r3c1 / divisor;
-    quotient->r3c2 = dividend->r3c2 / divisor;
 }
 
 // ============ Left Vector Product ============= //

basic-geometry/matrix3x2.h

@@ -275,24 +275,12 @@ static inline void bg_fp64_matrix3x2_multiply(const BgFP64Matrix3x2* multiplican
 
 static inline void bg_fp32_matrix3x2_divide(const BgFP32Matrix3x2* dividend, const float divisor, BgFP32Matrix3x2* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp32_matrix3x2_multiply(dividend, 1.0f / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-    quotient->r1c3 = dividend->r1c3 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
-    quotient->r2c3 = dividend->r2c3 / divisor;
 }
 
 static inline void bg_fp64_matrix3x2_divide(const BgFP64Matrix3x2* dividend, const double divisor, BgFP64Matrix3x2* quotient)
 {
-    quotient->r1c1 = dividend->r1c1 / divisor;
+    bg_fp64_matrix3x2_multiply(dividend, 1.0 / divisor, quotient);
-    quotient->r1c2 = dividend->r1c2 / divisor;
-    quotient->r1c3 = dividend->r1c3 / divisor;
-
-    quotient->r2c1 = dividend->r2c1 / divisor;
-    quotient->r2c2 = dividend->r2c2 / divisor;
-    quotient->r2c3 = dividend->r2c3 / divisor;
 }
 
 // ============ Left Vector Product ============= //

basic-geometry/matrix3x3.c

@@ -22,17 +22,19 @@ int bg_fp32_matrix3x3_invert(BgFP32Matrix3x3* matrix)
     const float r3c2 = matrix->r1c2 * matrix->r3c1 - matrix->r1c1 * matrix->r3c2;
     const float r3c3 = matrix->r1c1 * matrix->r2c2 - matrix->r1c2 * matrix->r2c1;
 
-    matrix->r1c1 = r1c1 / determinant;
+    const float multiplier = 1.0f / determinant;
-    matrix->r1c2 = r1c2 / determinant;
-    matrix->r1c3 = r1c3 / determinant;
 
-    matrix->r2c1 = r2c1 / determinant;
+    matrix->r1c1 = r1c1 * multiplier;
-    matrix->r2c2 = r2c2 / determinant;
+    matrix->r1c2 = r1c2 * multiplier;
-    matrix->r2c3 = r2c3 / determinant;
+    matrix->r1c3 = r1c3 * multiplier;
 
-    matrix->r3c1 = r3c1 / determinant;
+    matrix->r2c1 = r2c1 * multiplier;
-    matrix->r3c2 = r3c2 / determinant;
+    matrix->r2c2 = r2c2 * multiplier;
-    matrix->r3c3 = r3c3 / determinant;
+    matrix->r2c3 = r2c3 * multiplier;
+