2 * Copyright (c) 2014 Samsung Electronics Co., Ltd.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include <dali/public-api/math/matrix.h>
26 #include <dali/public-api/common/dali-common.h>
27 #include <dali/public-api/math/vector3.h>
28 #include <dali/public-api/math/vector4.h>
29 #include <dali/public-api/math/quaternion.h>
30 #include <dali/public-api/math/math-utils.h>
31 #include <dali/internal/render/common/performance-monitor.h>
35 const float ROTATION_EPSILON = 0.003f; // Deliberately large
37 const size_t NUM_BYTES_IN_ROW_OF_3( 3 * sizeof( float ) );
38 const size_t NUM_BYTES_IN_ROW( 4 * sizeof( float ) );
39 const size_t NUM_BYTES_IN_MATRIX( 16 * sizeof( float ) );
40 const size_t ROW1_OFFSET( 4 );
41 const size_t ROW2_OFFSET( 8 );
42 const size_t ROW3_OFFSET( 12 );
45 * Helper to convert to Quaternion to float16 array
47 void Convert( float*& m, const Dali::Quaternion& rotation )
49 const float xx = rotation.mVector.x * rotation.mVector.x;
50 const float yy = rotation.mVector.y * rotation.mVector.y;
51 const float zz = rotation.mVector.z * rotation.mVector.z;
52 const float xy = rotation.mVector.x * rotation.mVector.y;
53 const float xz = rotation.mVector.x * rotation.mVector.z;
54 const float wx = rotation.mVector.w * rotation.mVector.x;
55 const float wy = rotation.mVector.w * rotation.mVector.y;
56 const float wz = rotation.mVector.w * rotation.mVector.z;
57 const float yz = rotation.mVector.y * rotation.mVector.z;
59 m[0] = 1.0f - 2.0f * (yy + zz);
60 m[1] = 2.0f * (xy + wz);
61 m[2] = 2.0f * (xz - wy);
64 m[4] = 2.0f * (xy - wz);
65 m[5] = 1.0f - 2.0f * (xx + zz);
66 m[6] = 2.0f * (yz + wx);
69 m[8] = 2.0f * (xz + wy);
70 m[9] = 2.0f * (yz - wx);
71 m[10]= 1.0f - 2.0f * (xx + yy);
84 using Internal::PerformanceMonitor;
86 const float identityArray[] = {1.0f, 0.0f, 0.0f, 0.0f,
87 0.0f, 1.0f, 0.0f, 0.0f,
88 0.0f, 0.0f, 1.0f, 0.0f,
89 0.0f, 0.0f, 0.0f, 1.0f};
91 const Matrix Matrix::IDENTITY(identityArray);
95 memset( mMatrix, 0, NUM_BYTES_IN_MATRIX );
98 Matrix::Matrix( bool initialize )
102 memset( mMatrix, 0, NUM_BYTES_IN_MATRIX );
106 Matrix::Matrix(const float* array)
108 memcpy( mMatrix, array, NUM_BYTES_IN_MATRIX );
111 Matrix::Matrix( const Quaternion& rotation )
113 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,18);
115 float* matrixPtr = &mMatrix[0];
116 Convert( matrixPtr, rotation );
119 Matrix::Matrix( const Matrix& matrix )
121 memcpy( mMatrix, matrix.mMatrix, NUM_BYTES_IN_MATRIX );
124 Matrix& Matrix::operator=( const Matrix& matrix )
126 // no point copying if self assigning
127 if( this != &matrix )
129 memcpy( mMatrix, matrix.mMatrix, NUM_BYTES_IN_MATRIX );
134 void Matrix::InvertTransform(Matrix& result) const
136 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,12);
138 float* m1 = result.AsFloat();
140 DALI_ASSERT_ALWAYS( EqualsZero( mMatrix[3] ) && EqualsZero( mMatrix[7] ) && EqualsZero( mMatrix[11] ) && Equals( mMatrix[15], 1.0f ) && "Must be a transform matrix" );
154 m1[10] = mMatrix[10];
157 m1[12] = -( ( mMatrix[0] * mMatrix[12] ) + ( mMatrix[1] * mMatrix[13] ) + ( mMatrix[2] * mMatrix[14] ) + ( mMatrix[3] * mMatrix[15] ) );
158 m1[13] = -( ( mMatrix[4] * mMatrix[12] ) + ( mMatrix[5] * mMatrix[13] ) + ( mMatrix[6] * mMatrix[14] ) + ( mMatrix[7] * mMatrix[15] ) );
159 m1[14] = -( ( mMatrix[8] * mMatrix[12] ) + ( mMatrix[9] * mMatrix[13] ) + ( mMatrix[10] * mMatrix[14] ) + ( mMatrix[11] * mMatrix[15] ) );
163 static bool InvertMatrix(const float* m, float* out)
167 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,192); // 12 x 16 multiples
169 inv[0] = m[5]*m[10]*m[15] - m[5]*m[11]*m[14] - m[9]*m[6]*m[15] + m[9]*m[7]*m[14] + m[13]*m[6]*m[11] - m[13]*m[7]*m[10];
170 inv[4] = -m[4]*m[10]*m[15] + m[4]*m[11]*m[14] + m[8]*m[6]*m[15] - m[8]*m[7]*m[14] - m[12]*m[6]*m[11] + m[12]*m[7]*m[10];
171 inv[8] = m[4]*m[9]*m[15] - m[4]*m[11]*m[13] - m[8]*m[5]*m[15] + m[8]*m[7]*m[13] + m[12]*m[5]*m[11] - m[12]*m[7]*m[9];
172 inv[12] = -m[4]*m[9]*m[14] + m[4]*m[10]*m[13] + m[8]*m[5]*m[14] - m[8]*m[6]*m[13] - m[12]*m[5]*m[10] + m[12]*m[6]*m[9];
173 inv[1] = -m[1]*m[10]*m[15] + m[1]*m[11]*m[14] + m[9]*m[2]*m[15] - m[9]*m[3]*m[14] - m[13]*m[2]*m[11] + m[13]*m[3]*m[10];
174 inv[5] = m[0]*m[10]*m[15] - m[0]*m[11]*m[14] - m[8]*m[2]*m[15] + m[8]*m[3]*m[14] + m[12]*m[2]*m[11] - m[12]*m[3]*m[10];
175 inv[9] = -m[0]*m[9]*m[15] + m[0]*m[11]*m[13] + m[8]*m[1]*m[15] - m[8]*m[3]*m[13] - m[12]*m[1]*m[11] + m[12]*m[3]*m[9];
176 inv[13] = m[0]*m[9]*m[14] - m[0]*m[10]*m[13] - m[8]*m[1]*m[14] + m[8]*m[2]*m[13] + m[12]*m[1]*m[10] - m[12]*m[2]*m[9];
177 inv[2] = m[1]*m[6]*m[15] - m[1]*m[7]*m[14] - m[5]*m[2]*m[15] + m[5]*m[3]*m[14] + m[13]*m[2]*m[7] - m[13]*m[3]*m[6];
178 inv[6] = -m[0]*m[6]*m[15] + m[0]*m[7]*m[14] + m[4]*m[2]*m[15] - m[4]*m[3]*m[14] - m[12]*m[2]*m[7] + m[12]*m[3]*m[6];
179 inv[10] = m[0]*m[5]*m[15] - m[0]*m[7]*m[13] - m[4]*m[1]*m[15] + m[4]*m[3]*m[13] + m[12]*m[1]*m[7] - m[12]*m[3]*m[5];
180 inv[14] = -m[0]*m[5]*m[14] + m[0]*m[6]*m[13] + m[4]*m[1]*m[14] - m[4]*m[2]*m[13] - m[12]*m[1]*m[6] + m[12]*m[2]*m[5];
181 inv[3] = -m[1]*m[6]*m[11] + m[1]*m[7]*m[10] + m[5]*m[2]*m[11] - m[5]*m[3]*m[10] - m[9]*m[2]*m[7] + m[9]*m[3]*m[6];
182 inv[7] = m[0]*m[6]*m[11] - m[0]*m[7]*m[10] - m[4]*m[2]*m[11] + m[4]*m[3]*m[10] + m[8]*m[2]*m[7] - m[8]*m[3]*m[6];
183 inv[11] = -m[0]*m[5]*m[11] + m[0]*m[7]*m[9] + m[4]*m[1]*m[11] - m[4]*m[3]*m[9] - m[8]*m[1]*m[7] + m[8]*m[3]*m[5];
184 inv[15] = m[0]*m[5]*m[10] - m[0]*m[6]*m[9] - m[4]*m[1]*m[10] + m[4]*m[2]*m[9] + m[8]*m[1]*m[6] - m[8]*m[2]*m[5];
186 float det = m[0]*inv[0] + m[1]*inv[4] + m[2]*inv[8] + m[3]*inv[12];
188 // In the case where the determinant is exactly zero, the matrix is non-invertible
189 if ( EqualsZero( det ) )
196 for (int i = 0; i < 16; i++)
198 out[i] = inv[i] * det;
204 bool Matrix::Invert()
208 return InvertMatrix(temp.AsFloat(), mMatrix);
211 void Matrix::Transpose()
213 float temp = mMatrix[1];
214 mMatrix[1] = mMatrix[4];
218 mMatrix[2] = mMatrix[8];
222 mMatrix[3] = mMatrix[12];
226 mMatrix[6] = mMatrix[9];
230 mMatrix[7] = mMatrix[13];
234 mMatrix[11] = mMatrix[14];
238 void Matrix::SetIdentity()
240 memcpy( mMatrix, identityArray, NUM_BYTES_IN_MATRIX );
243 void Matrix::SetIdentityAndScale( const Vector3& scale )
245 // initialize to zeros
246 memset( mMatrix, 0, NUM_BYTES_IN_MATRIX );
248 // just apply scale on the diagonal
249 mMatrix[0] = scale.x;
250 mMatrix[5] = scale.y;
251 mMatrix[10] = scale.z;
255 void Matrix::SetTranslation(const Vector4& translation)
257 memcpy( mMatrix + ROW3_OFFSET, &translation, NUM_BYTES_IN_ROW );
259 void Matrix::SetTranslation(const Vector3& other)
261 memcpy( mMatrix + ROW3_OFFSET, &other, NUM_BYTES_IN_ROW_OF_3 );
265 void Matrix::Multiply( Matrix& result, const Matrix& lhs, const Matrix& rhs )
267 MATH_INCREASE_COUNTER(PerformanceMonitor::MATRIX_MULTIPLYS);
268 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,64); // 64 = 16*4
270 float* temp = result.AsFloat();
271 const float* rhsPtr = rhs.AsFloat();
272 const float* lhsPtr = lhs.AsFloat();
276 for( int i=0; i < 4; i++ )
278 // i<<2 gives the first vector / column
283 float value0 = lhsPtr[loc];
284 float value1 = lhsPtr[loc1];
285 float value2 = lhsPtr[loc2];
286 float value3 = lhsPtr[loc3];
287 temp[loc] = (value0 * rhsPtr[0]) +
288 (value1 * rhsPtr[4]) +
289 (value2 * rhsPtr[8]) +
290 (value3 * rhsPtr[12]);
292 temp[loc1] = (value0 * rhsPtr[1]) +
293 (value1 * rhsPtr[5]) +
294 (value2 * rhsPtr[9]) +
295 (value3 * rhsPtr[13]);
297 temp[loc2] = (value0 * rhsPtr[2]) +
298 (value1 * rhsPtr[6]) +
299 (value2 * rhsPtr[10])+
300 (value3 * rhsPtr[14]);
302 temp[loc3] = (value0 * rhsPtr[3]) +
303 (value1 * rhsPtr[7]) +
304 (value2 * rhsPtr[11])+
305 (value3 * rhsPtr[15]);
310 // 64 32bit registers,
312 // d = 64 bit double-word d0 -d31
313 // q =128 bit quad-word q0 -q15 (enough to handle a column of 4 floats in a matrix)
314 // e.g. q0 = d0 and d1
316 // load and stores interleaved as NEON can load and store while calculating
317 asm volatile ( "VLDM %1, {q0-q3} \n\t" // load matrix 1 (lhsPtr) q[0..q3]
318 "VLDM %0, {q8-q11} \n\t" // load matrix 2 (rhsPtr) q[q8-q11]
319 "VMUL.F32 q12, q8, d0[0] \n\t" // column 0 = rhsPtr[0..3] * lhsPtr[0..3]
320 "VMUL.F32 q13, q8, d2[0] \n\t" // column 1 = rhsPtr[0..3] * lhsPtr[4..7]
321 "VMUL.F32 q14, q8, d4[0] \n\t" // column 2 = rhsPtr[0..3] * lhsPtr[8..11]
322 "VMUL.F32 q15, q8, d6[0] \n\t" // column 3 = rhsPtr[0..3] * lhsPtr[12..15]
324 "VMLA.F32 q12, q9, d0[1] \n\t" // column 0 += rhsPtr[4..7] * lhsPtr[0..3]
325 "VMLA.F32 q13, q9, d2[1] \n\t" // column 1 += rhsPtr[4..7] * lhsPtr[4..7]
326 "VMLA.F32 q14, q9, d4[1] \n\t" // column 2 += rhsPtr[4..7] * lhsPtr[8..11]
327 "VMLA.F32 q15, q9, d6[1] \n\t" // column 3 += rhsPtr[4..7] * lhsPtr[12..15]
329 "VMLA.F32 q12, q10, d1[0] \n\t" // column 0 += rhsPtr[8..11] * lhsPtr[0..3]
330 "VMLA.F32 q13, q10, d3[0] \n\t" // column 1 += rhsPtr[8..11] * lhsPtr[4..7]
331 "VMLA.F32 q14, q10, d5[0] \n\t" // column 2 += rhsPtr[8..11] * lhsPtr[8..11]
332 "VMLA.F32 q15, q10, d7[0] \n\t" // column 3 += rhsPtr[8..11] * lhsPtr[12..15]
334 "VMLA.F32 q12, q11, d1[1] \n\t" // column 0 += rhsPtr[12..15] * lhsPtr[0..3]
335 "VMLA.F32 q13, q11, d3[1] \n\t" // column 1 += rhsPtr[12..15] * lhsPtr[4..7]
336 "VMLA.F32 q14, q11, d5[1] \n\t" // column 2 += rhsPtr[12..15] * lhsPtr[8..11]
337 "VMLA.F32 q15, q11, d7[1] \n\t" // column 3 += rhsPtr[12..15] * lhsPtr[12..15]
338 "VSTM %2, {q12-q15} \n\t" // store entire output matrix.
339 : "+r"(rhsPtr), "+r"(lhsPtr), "+r"(temp)
341 : "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11", "q12", "q13", "q14", "q15", "memory" );
346 void Matrix::Multiply( Matrix& result, const Matrix& lhs, const Quaternion& rhs )
348 MATH_INCREASE_COUNTER(PerformanceMonitor::MATRIX_MULTIPLYS);
349 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,54); // 54 = 36+18
352 float* rhsPtr = &matrix[0];
353 Convert( rhsPtr, rhs );
355 // quaternion contains just rotation so it really only needs 3x3 matrix
357 float* temp = result.AsFloat();
358 const float* lhsPtr = lhs.AsFloat();
362 for( int i=0; i < 4; i++ )
364 // i<<2 gives the first vector / column
369 float value0 = lhsPtr[loc];
370 float value1 = lhsPtr[loc1];
371 float value2 = lhsPtr[loc2];
372 float value3 = lhsPtr[loc3];
373 temp[loc] = (value0 * rhsPtr[0]) +
374 (value1 * rhsPtr[4]) +
375 (value2 * rhsPtr[8]) +
376 (0.0f); //value3 * rhsPtr[12] is 0.0f
378 temp[loc1] = (value0 * rhsPtr[1]) +
379 (value1 * rhsPtr[5]) +
380 (value2 * rhsPtr[9]) +
381 (0.0f); //value3 * rhsPtr[13] is 0.0f
383 temp[loc2] = (value0 * rhsPtr[2]) +
384 (value1 * rhsPtr[6]) +
385 (value2 * rhsPtr[10])+
386 (0.0f); //value3 * rhsPtr[14] is 0.0f
388 temp[loc3] = (0.0f) + //value0 * rhsPtr[3] is 0.0f
389 (0.0f) + //value1 * rhsPtr[7] is 0.0f
390 (0.0f) + //value2 * rhsPtr[11] is 0.0f
391 (value3); // rhsPtr[15] is 1.0f
396 // 64 32bit registers,
398 // d = 64 bit double-word d0 -d31
399 // q =128 bit quad-word q0 -q15 (enough to handle a column of 4 floats in a matrix)
400 // e.g. q0 = d0 and d1
401 // load and stores interleaved as NEON can load and store while calculating
402 asm volatile ( "VLDM %1, {q4-q6} \n\t" // load matrix 1 (lhsPtr)
403 "VLD1.F32 {q7}, [%2]! \n\t" // load matrix 2 (rhsPtr) [0..3]
404 "VMUL.F32 q0, q7, d8[0] \n\t" // column 0 = rhsPtr[0..3] * lhsPtr[0..3]
405 "VMUL.F32 q1, q7, d10[0] \n\t" // column 1 = rhsPtr[0..3] * lhsPtr[4..7]
406 "VMUL.F32 q2, q7, d12[0] \n\t" // column 2 = rhsPtr[0..3] * lhsPtr[8..11]
407 "VLD1.F32 {q7}, [%2]! \n\t" // load matrix 2 (rhsPtr) [4..7]
408 "VMLA.F32 q0, q7, d8[1] \n\t" // column 0+= rhsPtr[4..7] * lhsPtr[0..3]
409 "VMLA.F32 q1, q7, d10[1] \n\t" // column 1+= rhsPtr[4..7] * lhsPtr[4..7]
410 "VMLA.F32 q2, q7, d12[1] \n\t" // column 2+= rhsPtr[4..7] * lhsPtr[8..11]
411 "VLD1.F32 {q7}, [%2]! \n\t" // load matrix 2 (rhsPtr) [8..11]
412 "VMLA.F32 q0, q7, d9[0] \n\t" // column 0+= rhsPtr[8..11] * lhsPtr[0..3]
413 "VMLA.F32 q1, q7, d11[0] \n\t" // column 1+= rhsPtr[8..11] * lhsPtr[4..7]
414 "VMLA.F32 q2, q7, d13[0] \n\t" // column 2+= rhsPtr[8..11] * lhsPtr[8..11]
415 "VSTM %0, {q0-q2} \n\t" // store entire output matrix.
417 : "r"(temp), "r"(lhsPtr), "r" (rhsPtr)
418 : "%r0", "%q0", "%q1", "%q2", "%q4", "%q5", "%q6", "%q7", "memory" );
427 Vector4 Matrix::operator*(const Vector4& rhs) const
429 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,16);
435 temp.x = rhs.x * mMatrix[0] + rhs.y * mMatrix[4] + rhs.z * mMatrix[8] + rhs.w * mMatrix[12];
436 temp.y = rhs.x * mMatrix[1] + rhs.y * mMatrix[5] + rhs.z * mMatrix[9] + rhs.w * mMatrix[13];
437 temp.z = rhs.x * mMatrix[2] + rhs.y * mMatrix[6] + rhs.z * mMatrix[10] + rhs.w * mMatrix[14];
438 temp.w = rhs.x * mMatrix[3] + rhs.y * mMatrix[7] + rhs.z * mMatrix[11] + rhs.w * mMatrix[15];
442 // 64 32bit registers,
444 // d = 64 bit double-word d0 -d31
445 // q =128 bit quad-word q0 -q15 (enough to handle a column of 4 floats in a matrix)
446 // e.g. q0 = d0 and d1
447 // load and stores interleaved as NEON can load and store while calculating
448 asm volatile ( "VLD1.F32 {q0}, [%1] \n\t" //q0 = rhs
449 "VLD1.F32 {q9}, [%0]! \n\t"
450 "VMUL.F32 q10, q9, d0[0] \n\t"
451 "VLD1.F32 {q9}, [%0]! \n\t"
452 "VMLA.F32 q10, q9, d0[1] \n\t" //q10 = mMatrix[0..3] * rhs + mMatrix[4..7] * rhs
453 "VLD1.F32 {q9}, [%0]! \n\t"
454 "VMUL.F32 q11, q9, d1[0] \n\t"
455 "VLD1.F32 {q9}, [%0]! \n\t"
456 "VMLA.F32 q11, q9, d1[1] \n\t" //q11 = mMatrix[8..11] * rhs + mMatrix[12..15] * rhs
457 "VADD.F32 q10, q10, q11 \n\t"
458 "VST1.F32 {q10},[%2] \n\t" //temp = q10 + q11
460 : "r"(mMatrix), "r"(&rhs), "r"(&temp)
461 : "q0", "q9", "q10", "q11", "memory" );
466 bool Matrix::operator==(const Matrix& rhs) const
469 ( fabsf( mMatrix[0] - rhs.mMatrix[0] ) <= GetRangedEpsilon( mMatrix[0], rhs.mMatrix[0] ) ) &&
470 ( fabsf( mMatrix[1] - rhs.mMatrix[1] ) <= GetRangedEpsilon( mMatrix[1], rhs.mMatrix[1] ) ) &&
471 ( fabsf( mMatrix[2] - rhs.mMatrix[2] ) <= GetRangedEpsilon( mMatrix[2], rhs.mMatrix[2] ) ) &&
472 ( fabsf( mMatrix[3] - rhs.mMatrix[3] ) <= GetRangedEpsilon( mMatrix[3], rhs.mMatrix[3] ) ) &&
473 ( fabsf( mMatrix[4] - rhs.mMatrix[4] ) <= GetRangedEpsilon( mMatrix[4], rhs.mMatrix[4] ) ) &&
474 ( fabsf( mMatrix[5] - rhs.mMatrix[5] ) <= GetRangedEpsilon( mMatrix[5], rhs.mMatrix[5] ) ) &&
475 ( fabsf( mMatrix[6] - rhs.mMatrix[6] ) <= GetRangedEpsilon( mMatrix[6], rhs.mMatrix[6] ) ) &&
476 ( fabsf( mMatrix[7] - rhs.mMatrix[7] ) <= GetRangedEpsilon( mMatrix[7], rhs.mMatrix[7] ) ) &&
477 ( fabsf( mMatrix[8] - rhs.mMatrix[8] ) <= GetRangedEpsilon( mMatrix[8], rhs.mMatrix[8] ) ) &&
478 ( fabsf( mMatrix[9] - rhs.mMatrix[9] ) <= GetRangedEpsilon( mMatrix[9], rhs.mMatrix[9] ) ) &&
479 ( fabsf( mMatrix[10] - rhs.mMatrix[10] ) <= GetRangedEpsilon( mMatrix[10], rhs.mMatrix[10] ) ) &&
480 ( fabsf( mMatrix[11] - rhs.mMatrix[11] ) <= GetRangedEpsilon( mMatrix[11], rhs.mMatrix[11] ) ) &&
481 ( fabsf( mMatrix[12] - rhs.mMatrix[12] ) <= GetRangedEpsilon( mMatrix[12], rhs.mMatrix[12] ) ) &&
482 ( fabsf( mMatrix[13] - rhs.mMatrix[13] ) <= GetRangedEpsilon( mMatrix[13], rhs.mMatrix[13] ) ) &&
483 ( fabsf( mMatrix[14] - rhs.mMatrix[14] ) <= GetRangedEpsilon( mMatrix[14], rhs.mMatrix[14] ) ) &&
484 ( fabsf( mMatrix[15] - rhs.mMatrix[15] ) <= GetRangedEpsilon( mMatrix[15], rhs.mMatrix[15] ) ) );
487 bool Matrix::operator!=(const Matrix& rhs) const
497 void Matrix::OrthoNormalize()
499 Vector4 vector0(GetXAxis());
500 Vector4 vector1(GetYAxis());
501 Vector4 vector2(GetZAxis());
505 vector2 = vector0.Cross( vector1 );
506 vector1 = vector2.Cross( vector0 );
508 memcpy( mMatrix, &vector0, NUM_BYTES_IN_ROW );
509 memcpy( mMatrix + ROW1_OFFSET, &vector1, NUM_BYTES_IN_ROW );
510 memcpy( mMatrix + ROW2_OFFSET, &vector2, NUM_BYTES_IN_ROW );
513 Vector3 Matrix::GetXAxis() const
515 return Vector3(mMatrix[0], mMatrix[1], mMatrix[2]);
518 Vector3 Matrix::GetYAxis() const
520 return Vector3(mMatrix[4], mMatrix[5], mMatrix[6]);
523 Vector3 Matrix::GetZAxis() const
525 return Vector3(mMatrix[8], mMatrix[9], mMatrix[10]);
528 void Matrix::SetXAxis(const Vector3& axis)
535 void Matrix::SetYAxis(const Vector3& axis)
542 void Matrix::SetZAxis(const Vector3& axis)
546 mMatrix[10] = axis.z;
549 void Matrix::SetTransformComponents(const Vector3& scale,
550 const Quaternion& rotation,
551 const Vector3& translation )
553 if( rotation.IsIdentity() )
555 mMatrix[0] = scale.x;
561 mMatrix[5] = scale.y;
567 mMatrix[10]= scale.z;
572 MATH_INCREASE_COUNTER(PerformanceMonitor::MATRIX_MULTIPLYS);
573 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,27); // 27 = 9+18
575 const float xx = rotation.mVector.x * rotation.mVector.x;
576 const float yy = rotation.mVector.y * rotation.mVector.y;
577 const float zz = rotation.mVector.z * rotation.mVector.z;
578 const float xy = rotation.mVector.x * rotation.mVector.y;
579 const float xz = rotation.mVector.x * rotation.mVector.z;
580 const float wx = rotation.mVector.w * rotation.mVector.x;
581 const float wy = rotation.mVector.w * rotation.mVector.y;
582 const float wz = rotation.mVector.w * rotation.mVector.z;
583 const float yz = rotation.mVector.y * rotation.mVector.z;
585 mMatrix[0] = (scale.x * (1.0f - 2.0f * (yy + zz)));
586 mMatrix[1] = (scale.x * ( 2.0f * (xy + wz)));
587 mMatrix[2] = (scale.x * ( 2.0f * (xz - wy)));
590 mMatrix[4] = (scale.y * ( 2.0f * (xy - wz)));
591 mMatrix[5] = (scale.y * (1.0f - 2.0f * (xx + zz)));
592 mMatrix[6] = (scale.y * ( 2.0f * (yz + wx)));
595 mMatrix[8] = (scale.z * ( 2.0f * (xz + wy)));
596 mMatrix[9] = (scale.z * ( 2.0f * (yz - wx)));
597 mMatrix[10]= (scale.z * (1.0f - 2.0f * (xx + yy)));
601 mMatrix[12] = translation.x;
602 mMatrix[13] = translation.y;
603 mMatrix[14] = translation.z;
607 void Matrix::SetInverseTransformComponents(const Vector3& scale,
608 const Quaternion& rotation,
609 const Vector3& translation )
611 Vector3 inverseTranslation = -translation;
612 Vector3 inverseScale( 1.0f/scale.x, 1.0f/scale.y, 1.0f/scale.z);
613 Quaternion inverseRotation(rotation);
614 bool isRotated = ! inverseRotation.IsIdentity();
616 // Order of application is translation, rotation, scale.
617 // Ensure translation is relative to scale & rotation:
621 inverseRotation.Invert();
622 inverseTranslation = inverseRotation.Rotate(inverseTranslation);
625 inverseTranslation *= inverseScale;
629 MATH_INCREASE_COUNTER(PerformanceMonitor::MATRIX_MULTIPLYS);
630 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,27); // 27 = 9+18
632 const float xx = inverseRotation.mVector.x * inverseRotation.mVector.x;
633 const float yy = inverseRotation.mVector.y * inverseRotation.mVector.y;
634 const float zz = inverseRotation.mVector.z * inverseRotation.mVector.z;
635 const float xy = inverseRotation.mVector.x * inverseRotation.mVector.y;
636 const float xz = inverseRotation.mVector.x * inverseRotation.mVector.z;
637 const float wx = inverseRotation.mVector.w * inverseRotation.mVector.x;
638 const float wy = inverseRotation.mVector.w * inverseRotation.mVector.y;
639 const float wz = inverseRotation.mVector.w * inverseRotation.mVector.z;
640 const float yz = inverseRotation.mVector.y * inverseRotation.mVector.z;
642 mMatrix[0] = (inverseScale.x * (1.0f - 2.0f * (yy + zz)));
643 mMatrix[1] = (inverseScale.y * (2.0f * (xy + wz)));
644 mMatrix[2] = (inverseScale.z * (2.0f * (xz - wy)));
647 mMatrix[4] = (inverseScale.x * (2.0f * (xy - wz)));
648 mMatrix[5] = (inverseScale.y * (1.0f - 2.0f * (xx + zz)));
649 mMatrix[6] = (inverseScale.z * (2.0f * (yz + wx)));
652 mMatrix[8] = (inverseScale.x * (2.0f * (xz + wy)));
653 mMatrix[9] = (inverseScale.y * (2.0f * (yz - wx)));
654 mMatrix[10]= (inverseScale.z * (1.0f - 2.0f * (xx + yy)));
659 mMatrix[0] = inverseScale.x;
665 mMatrix[5] = inverseScale.y;
671 mMatrix[10]= inverseScale.z;
676 mMatrix[12] = inverseTranslation.x;
677 mMatrix[13] = inverseTranslation.y;
678 mMatrix[14] = inverseTranslation.z;
682 void Matrix::SetInverseTransformComponents(const Vector3& xAxis,
683 const Vector3& yAxis,
684 const Vector3& zAxis,
685 const Vector3& translation )
687 // x, y, z axis parameters represent a orthonormal basis with no scaling, i.e. a rotation matrix.
688 // Invert rotation by transposing in place
690 // Order of application is translation, rotation
692 mMatrix[0] = xAxis.x;
693 mMatrix[1] = yAxis.x;
694 mMatrix[2] = zAxis.x;
697 mMatrix[4] = xAxis.y;
698 mMatrix[5] = yAxis.y;
699 mMatrix[6] = zAxis.y;
702 mMatrix[8] = xAxis.z;
703 mMatrix[9] = yAxis.z;
704 mMatrix[10] = zAxis.z;
711 // Ensure translation is relative to scale & rotation:
713 Vector4 inverseTranslation( -translation.x, -translation.y, -translation.z, 1.0f);
714 inverseTranslation = *this * inverseTranslation; // Rotate inverse translation
715 inverseTranslation.w = 1.0f;
716 SetTranslation(inverseTranslation);
720 void Matrix::GetTransformComponents(Vector3& position,
721 Quaternion& rotation,
722 Vector3& scale) const
724 position = GetTranslation3();
726 // Derive scale from axis lengths.
727 Vector3 theScale(GetXAxis().Length(), GetYAxis().Length(), GetZAxis().Length());
730 if( ! ( fabs(theScale.x - Vector3::ONE.x) < ROTATION_EPSILON &&
731 fabs(theScale.y - Vector3::ONE.y) < ROTATION_EPSILON &&
732 fabs(theScale.z - Vector3::ONE.z) < ROTATION_EPSILON ) )
734 MATH_INCREASE_COUNTER(PerformanceMonitor::MATRIX_MULTIPLYS);
735 MATH_INCREASE_BY(PerformanceMonitor::FLOAT_POINT_MULTIPLY,9);
737 // Non-identity scale is embedded into rotation matrix. Remove it first:
739 Vector3 inverseScale(1.0f/theScale.x, 1.0f/theScale.y, 1.0f/theScale.z);
740 m.mMatrix[0] *= inverseScale.x;
741 m.mMatrix[1] *= inverseScale.x;
742 m.mMatrix[2] *= inverseScale.x;
743 m.mMatrix[4] *= inverseScale.y;
744 m.mMatrix[5] *= inverseScale.y;
745 m.mMatrix[6] *= inverseScale.y;
746 m.mMatrix[8] *= inverseScale.z;
747 m.mMatrix[9] *= inverseScale.z;
748 m.mMatrix[10] *= inverseScale.z;
750 Quaternion theRotation(m);
752 // If the imaginary components are close to zero, then use null quaternion instead.
753 if( fabs(theRotation.mVector.x) < ROTATION_EPSILON &&
754 fabs(theRotation.mVector.y) < ROTATION_EPSILON &&
755 fabs(theRotation.mVector.z) < ROTATION_EPSILON )
757 theRotation = Quaternion();
759 rotation = theRotation;
763 Quaternion theRotation(*this);
765 // If the imaginary components are close to zero, then use null quaternion instead.
766 if( fabs(theRotation.mVector.x) < ROTATION_EPSILON &&
767 fabs(theRotation.mVector.y) < ROTATION_EPSILON &&
768 fabs(theRotation.mVector.z) < ROTATION_EPSILON )
770 theRotation = Quaternion();
772 rotation = theRotation;
778 std::ostream& operator<< (std::ostream& o, const Matrix& matrix)
780 return o << "[ [" << matrix.mMatrix[0] << ", " << matrix.mMatrix[1] << ", " << matrix.mMatrix[2] << ", " << matrix.mMatrix[3] << "], "
781 << "[" << matrix.mMatrix[4] << ", " << matrix.mMatrix[5] << ", " << matrix.mMatrix[6] << ", " << matrix.mMatrix[7] << "], "
782 << "[" << matrix.mMatrix[8] << ", " << matrix.mMatrix[9] << ", " << matrix.mMatrix[10] << ", " << matrix.mMatrix[11] << "], "
783 << "[" << matrix.mMatrix[12] << ", " << matrix.mMatrix[13] << ", " << matrix.mMatrix[14] << ", " << matrix.mMatrix[15] << "] ]";