diff options
author | Oz Linden <oz@lindenlab.com> | 2016-12-05 15:20:27 -0500 |
---|---|---|
committer | Oz Linden <oz@lindenlab.com> | 2016-12-05 15:20:27 -0500 |
commit | c21571474c0022f83a4efc28d59e97fc020fd0e7 (patch) | |
tree | 15586cbd2055a1c84dad4ce2f74efe1a885d997d /indra/llmath/llmatrix4a.h | |
parent | fd2ccb16068dfd21307b17e78e384d8ae19fef12 (diff) | |
parent | 68413474c4479eee9bdbeb34ea131475ba1d646e (diff) |
merge changes for 5.0.0-release
Diffstat (limited to 'indra/llmath/llmatrix4a.h')
-rw-r--r-- | indra/llmath/llmatrix4a.h | 39 |
1 files changed, 38 insertions, 1 deletions
diff --git a/indra/llmath/llmatrix4a.h b/indra/llmath/llmatrix4a.h index d141298f69..216334752a 100644 --- a/indra/llmath/llmatrix4a.h +++ b/indra/llmath/llmatrix4a.h @@ -121,7 +121,7 @@ public: res.add(z); } - inline void affineTransform(const LLVector4a& v, LLVector4a& res) + inline void affineTransformSSE(const LLVector4a& v, LLVector4a& res) { LLVector4a x,y,z; @@ -137,6 +137,43 @@ public: z.add(mMatrix[3]); res.setAdd(x,z); } + + inline void affineTransformNonSSE(const LLVector4a& v, LLVector4a& res) + { + F32 x = v[0] * mMatrix[0][0] + v[1] * mMatrix[1][0] + v[2] * mMatrix[2][0] + mMatrix[3][0]; + F32 y = v[0] * mMatrix[0][1] + v[1] * mMatrix[1][1] + v[2] * mMatrix[2][1] + mMatrix[3][1]; + F32 z = v[0] * mMatrix[0][2] + v[1] * mMatrix[1][2] + v[2] * mMatrix[2][2] + mMatrix[3][2]; + F32 w = 1.0f; + res.set(x,y,z,w); + } + + inline void affineTransform(const LLVector4a& v, LLVector4a& res) + { + affineTransformSSE(v,res); + } }; +inline LLVector4a rowMul(const LLVector4a &row, const LLMatrix4a &mat) +{ + LLVector4a result; + result = _mm_mul_ps(_mm_shuffle_ps(row, row, _MM_SHUFFLE(0, 0, 0, 0)), mat.mMatrix[0]); + result = _mm_add_ps(result, _mm_mul_ps(_mm_shuffle_ps(row, row, _MM_SHUFFLE(1, 1, 1, 1)), mat.mMatrix[1])); + result = _mm_add_ps(result, _mm_mul_ps(_mm_shuffle_ps(row, row, _MM_SHUFFLE(2, 2, 2, 2)), mat.mMatrix[2])); + result = _mm_add_ps(result, _mm_mul_ps(_mm_shuffle_ps(row, row, _MM_SHUFFLE(3, 3, 3, 3)), mat.mMatrix[3])); + return result; +} + +inline void matMul(const LLMatrix4a &a, const LLMatrix4a &b, LLMatrix4a &res) +{ + LLVector4a row0 = rowMul(a.mMatrix[0], b); + LLVector4a row1 = rowMul(a.mMatrix[1], b); + LLVector4a row2 = rowMul(a.mMatrix[2], b); + LLVector4a row3 = rowMul(a.mMatrix[3], b); + + res.mMatrix[0] = row0; + res.mMatrix[1] = row1; + res.mMatrix[2] = row2; + res.mMatrix[3] = row3; +} + #endif |