|
|
@@ -675,6 +675,27 @@ void Matrix::multiply(const Matrix& m, float scalar, Matrix* dst)
|
|
|
{
|
|
|
GP_ASSERT(dst);
|
|
|
|
|
|
+#ifdef USE_NEON
|
|
|
+
|
|
|
+ asm volatile(
|
|
|
+ "vld1.32 {d0[0]}, [%0] \n\t"
|
|
|
+ "vld1.32 {q4-q5}, [%1]! \n\t"
|
|
|
+ "vld1.32 {q6-q7}, [%1]! \n\t"
|
|
|
+
|
|
|
+ "vmul.f32 q8, q4, d0[0] \n\t"
|
|
|
+ "vmul.f32 q9, q5, d0[0] \n\t"
|
|
|
+ "vmul.f32 q10, q6, d0[0] \n\t"
|
|
|
+ "vmul.f32 q11, q7, d0[0] \n\t"
|
|
|
+
|
|
|
+ "vst1.32 {q8-q9}, [%2]! \n\t"
|
|
|
+ "vst1.32 {q10-q11}, [%2]! \n\t"
|
|
|
+ :
|
|
|
+ : "r"(&scalar), "r"(m.m), "r"(dst->m)
|
|
|
+ : "q0", "q4", "q5", "q6", "q7", "q8", "q9", "q10", "q11", "memory"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
+
|
|
|
dst->m[0] = m.m[0] * scalar;
|
|
|
dst->m[1] = m.m[1] * scalar;
|
|
|
dst->m[2] = m.m[2] * scalar;
|
|
|
@@ -691,6 +712,8 @@ void Matrix::multiply(const Matrix& m, float scalar, Matrix* dst)
|
|
|
dst->m[13] = m.m[13] * scalar;
|
|
|
dst->m[14] = m.m[14] * scalar;
|
|
|
dst->m[15] = m.m[15] * scalar;
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void Matrix::multiply(const Matrix& m)
|
|
|
@@ -700,7 +723,46 @@ void Matrix::multiply(const Matrix& m)
|
|
|
|
|
|
void Matrix::multiply(const Matrix& m1, const Matrix& m2, Matrix* dst)
|
|
|
{
|
|
|
- GP_ASSERT(dst);
|
|
|
+ GP_ASSERT(dst);
|
|
|
+
|
|
|
+#ifdef USE_NEON // if set, neon unit is present.
|
|
|
+
|
|
|
+ asm volatile
|
|
|
+ (
|
|
|
+ "vld1.32 {d16 - d19}, [%1]! \n\t" // load first eight elements of matrix 0
|
|
|
+ "vld1.32 {d20 - d23}, [%1]! \n\t" // load second eight elements of matrix 0
|
|
|
+ "vld1.32 {d0 - d3}, [%2]! \n\t" // load first eight elements of matrix 1
|
|
|
+ "vld1.32 {d4 - d7}, [%2]! \n\t" // load second eight elements of matrix 1
|
|
|
+
|
|
|
+ "vmul.f32 q12, q8, d0[0] \n\t" // rslt col0 = (mat0 col0) * (mat1 col0 elt0)
|
|
|
+ "vmul.f32 q13, q8, d2[0] \n\t" // rslt col1 = (mat0 col0) * (mat1 col1 elt0)
|
|
|
+ "vmul.f32 q14, q8, d4[0] \n\t" // rslt col2 = (mat0 col0) * (mat1 col2 elt0)
|
|
|
+ "vmul.f32 q15, q8, d6[0] \n\t" // rslt col3 = (mat0 col0) * (mat1 col3 elt0)
|
|
|
+
|
|
|
+ "vmla.f32 q12, q9, d0[1] \n\t" // rslt col0 += (mat0 col1) * (mat1 col0 elt1)
|
|
|
+ "vmla.f32 q13, q9, d2[1] \n\t" // rslt col1 += (mat0 col1) * (mat1 col1 elt1)
|
|
|
+ "vmla.f32 q14, q9, d4[1] \n\t" // rslt col2 += (mat0 col1) * (mat1 col2 elt1)
|
|
|
+ "vmla.f32 q15, q9, d6[1] \n\t" // rslt col3 += (mat0 col1) * (mat1 col3 elt1)
|
|
|
+
|
|
|
+ "vmla.f32 q12, q10, d1[0] \n\t" // rslt col0 += (mat0 col2) * (mat1 col0 elt2)
|
|
|
+ "vmla.f32 q13, q10, d3[0] \n\t" // rslt col1 += (mat0 col2) * (mat1 col1 elt2)
|
|
|
+ "vmla.f32 q14, q10, d5[0] \n\t" // rslt col2 += (mat0 col2) * (mat1 col2 elt2)
|
|
|
+ "vmla.f32 q15, q10, d7[0] \n\t" // rslt col3 += (mat0 col2) * (mat1 col2 elt2)
|
|
|
+
|
|
|
+ "vmla.f32 q12, q11, d1[1] \n\t" // rslt col0 += (mat0 col3) * (mat1 col0 elt3)
|
|
|
+ "vmla.f32 q13, q11, d3[1] \n\t" // rslt col1 += (mat0 col3) * (mat1 col1 elt3)
|
|
|
+ "vmla.f32 q14, q11, d5[1] \n\t" // rslt col2 += (mat0 col3) * (mat1 col2 elt3)
|
|
|
+ "vmla.f32 q15, q11, d7[1] \n\t" // rslt col3 += (mat0 col3) * (mat1 col3 elt3)
|
|
|
+
|
|
|
+ "vst1.32 {d24 - d27}, [%0]! \n\t" // store first eight elements of result
|
|
|
+ "vst1.32 {d28 - d31}, [%0]! \n\t" // store second eight elements of result
|
|
|
+
|
|
|
+ : // output
|
|
|
+ : "r"(dst->m), "r"(m1.m), "r"(m2.m) // input - note *value* of pointer doesn't change.
|
|
|
+ : "memory", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
|
|
|
// Support the case where m1 or m2 is the same array as dst.
|
|
|
float product[16];
|
|
|
@@ -726,6 +788,8 @@ void Matrix::multiply(const Matrix& m1, const Matrix& m2, Matrix* dst)
|
|
|
product[15] = m1.m[3] * m2.m[12] + m1.m[7] * m2.m[13] + m1.m[11] * m2.m[14] + m1.m[15] * m2.m[15];
|
|
|
|
|
|
memcpy(dst->m, product, MATRIX_SIZE);
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void Matrix::negate()
|
|
|
@@ -737,6 +801,26 @@ void Matrix::negate(Matrix* dst) const
|
|
|
{
|
|
|
GP_ASSERT(dst);
|
|
|
|
|
|
+#ifdef USE_NEON
|
|
|
+
|
|
|
+ asm volatile(
|
|
|
+ "vld1.32 {q0-q1}, [%1]! \n\t" // load m0-m7
|
|
|
+ "vld1.32 {q2-q3}, [%1]! \n\t" // load m8-m15
|
|
|
+
|
|
|
+ "vneg.f32 q4, q0 \n\t" // negate m0-m3
|
|
|
+ "vneg.f32 q5, q1 \n\t" // negate m4-m7
|
|
|
+ "vneg.f32 q6, q2 \n\t" // negate m8-m15
|
|
|
+ "vneg.f32 q7, q3 \n\t" // negate m8-m15
|
|
|
+
|
|
|
+ "vst1.32 {q4-q5}, [%0]! \n\t" // store m0-m7
|
|
|
+ "vst1.32 {q6-q7}, [%0]! \n\t" // store m8-m15
|
|
|
+ :
|
|
|
+ : "r"(dst->m), "r"(m)
|
|
|
+ : "q0", "q1", "q2", "q3", "q4", "q5", "q6", "q7", "memory"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
+
|
|
|
dst->m[0] = -m[0];
|
|
|
dst->m[1] = -m[1];
|
|
|
dst->m[2] = -m[2];
|
|
|
@@ -753,6 +837,8 @@ void Matrix::negate(Matrix* dst) const
|
|
|
dst->m[13] = -m[13];
|
|
|
dst->m[14] = -m[14];
|
|
|
dst->m[15] = -m[15];
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void Matrix::rotate(const Quaternion& q)
|
|
|
@@ -940,11 +1026,38 @@ void Matrix::transformVector(const Vector3& vector, Vector3* dst) const
|
|
|
void Matrix::transformVector(float x, float y, float z, float w, Vector3* dst) const
|
|
|
{
|
|
|
GP_ASSERT(dst);
|
|
|
-
|
|
|
+
|
|
|
+#ifdef USE_NEON
|
|
|
+
|
|
|
+ asm volatile(
|
|
|
+ "vld1.32 {d0[0]}, [%0] \n\t" // load x
|
|
|
+ "vld1.32 {d0[1]}, [%1] \n\t" // load y
|
|
|
+ "vld1.32 {d1[0]}, [%2] \n\t" // load z
|
|
|
+ "vld1.32 {d1[1]}, [%3] \n\t" // load w
|
|
|
+ "vld1.32 {d18 - d21}, [%4]! \n\t" // load first 8 elements of matrix m0-m7
|
|
|
+ "vld1.32 {d22 - d25}, [%4]! \n\t" // load second 8 elements of matrix m8-m15
|
|
|
+
|
|
|
+ "vmul.f32 q13, q9, d0[0] \n\t" // Q5 = (m0-m3)*x
|
|
|
+ "vmla.f32 q13, q10, d0[1] \n\t" // Q5 += (m4-m7)*y
|
|
|
+ "vmla.f32 q13, q11, d1[0] \n\t" // Q5 += (m8-m11)*z
|
|
|
+ "vmla.f32 q13, q12, d1[1] \n\t" // Q5 += (m12-m15)*w
|
|
|
+
|
|
|
+ "vst1.32 {d26[0]}, [%5]! \n\t" // store dst->x
|
|
|
+ "vst1.32 {d26[1]}, [%5]! \n\t" // store dst->y
|
|
|
+ "vst1.32 {d27[0]}, [%5]! \n\t" // store dst->z
|
|
|
+ :
|
|
|
+ : "r"(&x), "r"(&y), "r"(&z), "r"(&w), "r"(m), "r"(dst)
|
|
|
+ : "q0", "q9", "q10","q11", "q12", "q13", "memory"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
+
|
|
|
dst->set(
|
|
|
x * m[0] + y * m[4] + z * m[8] + w * m[12],
|
|
|
x * m[1] + y * m[5] + z * m[9] + w * m[13],
|
|
|
x * m[2] + y * m[6] + z * m[10] + w * m[14]);
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void Matrix::transformVector(Vector4* vector) const
|
|
|
@@ -957,11 +1070,33 @@ void Matrix::transformVector(const Vector4& vector, Vector4* dst) const
|
|
|
{
|
|
|
GP_ASSERT(dst);
|
|
|
|
|
|
+#ifdef USE_NEON
|
|
|
+
|
|
|
+ asm volatile
|
|
|
+ (
|
|
|
+ "vld1.32 {d0, d1}, [%1] \n\t" //Q0 = v (x, y, z, w)
|
|
|
+ "vld1.32 {d18 - d21}, [%0]! \n\t" //Q1 = M (m0-m7)
|
|
|
+ "vld1.32 {d22 - d25}, [%0]! \n\t" //Q2 = M (m8-m15)
|
|
|
+
|
|
|
+ "vmul.f32 q13, q9, d0[0] \n\t" //Q5 = Q0*Q0[0]
|
|
|
+ "vmla.f32 q13, q10, d0[1] \n\t" //Q5 += Q1*Q0[1]
|
|
|
+ "vmla.f32 q13, q11, d1[0] \n\t" //Q5 += Q2*Q0[2]
|
|
|
+ "vmla.f32 q13, q12, d1[1] \n\t" //Q5 += Q3*Q0[3]
|
|
|
+ "vst1.32 {d26, d27}, [%2] \n\t" //Q4 = m+12
|
|
|
+ :
|
|
|
+ : "r"(m), "r"(&vector), "r"(dst)
|
|
|
+ : "q0", "q9", "q10","q11", "q12", "q13", "memory"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
+
|
|
|
dst->set(
|
|
|
vector.x * m[0] + vector.y * m[4] + vector.z * m[8] + vector.w * m[12],
|
|
|
vector.x * m[1] + vector.y * m[5] + vector.z * m[9] + vector.w * m[13],
|
|
|
- vector.x * m[2] + vector.y * m[6] + vector.z * m[10] + vector.w * m[14],
|
|
|
+ vector.x * m[2] + vector.y * m[6] + vector.z * m[10] + vector.w * m[14],
|
|
|
vector.x * m[3] + vector.y * m[7] + vector.z * m[11] + vector.w * m[15]);
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
void Matrix::translate(float x, float y, float z)
|
|
|
@@ -994,7 +1129,24 @@ void Matrix::transpose()
|
|
|
void Matrix::transpose(Matrix* dst) const
|
|
|
{
|
|
|
GP_ASSERT(dst);
|
|
|
+
|
|
|
+#ifdef USE_NEON
|
|
|
|
|
|
+ asm volatile(
|
|
|
+ "vld4.32 {d0[0], d2[0], d4[0], d6[0]}, [%0]! \n\t"
|
|
|
+ "vld4.32 {d0[1], d2[1], d4[1], d6[1]}, [%0]! \n\t"
|
|
|
+ "vld4.32 {d1[0], d3[0], d5[0], d7[0]}, [%0]! \n\t"
|
|
|
+ "vld4.32 {d1[1], d3[1], d5[1], d7[1]}, [%0]! \n\t"
|
|
|
+
|
|
|
+ "vst1.32 {q0-q1}, [%1]! \n\t"
|
|
|
+ "vst1.32 {q2-q3}, [%1]! \n\t"
|
|
|
+ :
|
|
|
+ : "r"(this->m), "r"(dst->m)
|
|
|
+ : "q0", "q1", "q2", "q3", "memory"
|
|
|
+ );
|
|
|
+
|
|
|
+#else
|
|
|
+
|
|
|
float t[16] = {
|
|
|
m[0], m[4], m[8], m[12],
|
|
|
m[1], m[5], m[9], m[13],
|
|
|
@@ -1002,6 +1154,8 @@ void Matrix::transpose(Matrix* dst) const
|
|
|
m[3], m[7], m[11], m[15]
|
|
|
};
|
|
|
memcpy(dst->m, t, MATRIX_SIZE);
|
|
|
+
|
|
|
+#endif
|
|
|
}
|
|
|
|
|
|
}
|
setaylor
