ryujinx/Opentk
1
0
Fork 0
mirror of https://github.com/Ryujinx/Opentk.git synced 2025-01-11 09:05:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

Optimized Matrix4(d).Mult for a 7x speedup. Code courtesy of http://www.edais.co.uk/blog/?p=27

Browse source
This commit is contained in:
the_fiddler 2011-09-04 20:52:42 +00:00
parent 9c6a4bb2df
commit d0cd45e6d0
2 changed files with 50 additions and 34 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

42
Source/OpenTK/Math/Matrix4.cs
View file

@ -951,23 +951,31 @@ namespace OpenTK
/// <param name="result">A new instance that is the result of the multiplication</param>
public static void Mult(ref Matrix4 left, ref Matrix4 right, out Matrix4 result)
{
result = new Matrix4(
left.M11 * right.M11 + left.M12 * right.M21 + left.M13 * right.M31 + left.M14 * right.M41,
left.M11 * right.M12 + left.M12 * right.M22 + left.M13 * right.M32 + left.M14 * right.M42,
left.M11 * right.M13 + left.M12 * right.M23 + left.M13 * right.M33 + left.M14 * right.M43,
left.M11 * right.M14 + left.M12 * right.M24 + left.M13 * right.M34 + left.M14 * right.M44,
left.M21 * right.M11 + left.M22 * right.M21 + left.M23 * right.M31 + left.M24 * right.M41,
left.M21 * right.M12 + left.M22 * right.M22 + left.M23 * right.M32 + left.M24 * right.M42,
left.M21 * right.M13 + left.M22 * right.M23 + left.M23 * right.M33 + left.M24 * right.M43,
left.M21 * right.M14 + left.M22 * right.M24 + left.M23 * right.M34 + left.M24 * right.M44,
left.M31 * right.M11 + left.M32 * right.M21 + left.M33 * right.M31 + left.M34 * right.M41,
left.M31 * right.M12 + left.M32 * right.M22 + left.M33 * right.M32 + left.M34 * right.M42,
left.M31 * right.M13 + left.M32 * right.M23 + left.M33 * right.M33 + left.M34 * right.M43,
left.M31 * right.M14 + left.M32 * right.M24 + left.M33 * right.M34 + left.M34 * right.M44,
left.M41 * right.M11 + left.M42 * right.M21 + left.M43 * right.M31 + left.M44 * right.M41,
left.M41 * right.M12 + left.M42 * right.M22 + left.M43 * right.M32 + left.M44 * right.M42,
left.M41 * right.M13 + left.M42 * right.M23 + left.M43 * right.M33 + left.M44 * right.M43,
left.M41 * right.M14 + left.M42 * right.M24 + left.M43 * right.M34 + left.M44 * right.M44);
float lM11 = left.Row0.X, lM12 = left.Row0.Y, lM13 = left.Row0.Z, lM14 = left.Row0.W,
lM21 = left.Row1.X, lM22 = left.Row1.Y, lM23 = left.Row1.Z, lM24 = left.Row1.W,
lM31 = left.Row2.X, lM32 = left.Row2.Y, lM33 = left.Row2.Z, lM34 = left.Row2.W,
lM41 = left.Row3.X, lM42 = left.Row3.Y, lM43 = left.Row3.Z, lM44 = left.Row3.W,
rM11 = right.Row0.X, rM12 = right.Row0.Y, rM13 = right.Row0.Z, rM14 = right.Row0.W,
rM21 = right.Row1.X, rM22 = right.Row1.Y, rM23 = right.Row1.Z, rM24 = right.Row1.W,
rM31 = right.Row2.X, rM32 = right.Row2.Y, rM33 = right.Row2.Z, rM34 = right.Row2.W,
rM41 = right.Row3.X, rM42 = right.Row3.Y, rM43 = right.Row3.Z, rM44 = right.Row3.W;
result.Row0.X = (((lM11 * rM11) + (lM12 * rM21)) + (lM13 * rM31)) + (lM14 * rM41);
result.Row0.Y = (((lM11 * rM12) + (lM12 * rM22)) + (lM13 * rM32)) + (lM14 * rM42);
result.Row0.Z = (((lM11 * rM13) + (lM12 * rM23)) + (lM13 * rM33)) + (lM14 * rM43);
result.Row0.W = (((lM11 * rM14) + (lM12 * rM24)) + (lM13 * rM34)) + (lM14 * rM44);
result.Row1.X = (((lM21 * rM11) + (lM22 * rM21)) + (lM23 * rM31)) + (lM24 * rM41);
result.Row1.Y = (((lM21 * rM12) + (lM22 * rM22)) + (lM23 * rM32)) + (lM24 * rM42);
result.Row1.Z = (((lM21 * rM13) + (lM22 * rM23)) + (lM23 * rM33)) + (lM24 * rM43);
result.Row1.W = (((lM21 * rM14) + (lM22 * rM24)) + (lM23 * rM34)) + (lM24 * rM44);
result.Row2.X = (((lM31 * rM11) + (lM32 * rM21)) + (lM33 * rM31)) + (lM34 * rM41);
result.Row2.Y = (((lM31 * rM12) + (lM32 * rM22)) + (lM33 * rM32)) + (lM34 * rM42);
result.Row2.Z = (((lM31 * rM13) + (lM32 * rM23)) + (lM33 * rM33)) + (lM34 * rM43);
result.Row2.W = (((lM31 * rM14) + (lM32 * rM24)) + (lM33 * rM34)) + (lM34 * rM44);
result.Row3.X = (((lM41 * rM11) + (lM42 * rM21)) + (lM43 * rM31)) + (lM44 * rM41);
result.Row3.Y = (((lM41 * rM12) + (lM42 * rM22)) + (lM43 * rM32)) + (lM44 * rM42);
result.Row3.Z = (((lM41 * rM13) + (lM42 * rM23)) + (lM43 * rM33)) + (lM44 * rM43);
result.Row3.W = (((lM41 * rM14) + (lM42 * rM24)) + (lM43 * rM34)) + (lM44 * rM44);
}
#endregion

42
Source/OpenTK/Math/Matrix4d.cs
View file

@ -945,23 +945,31 @@ namespace OpenTK
/// <param name="result">A new instance that is the result of the multiplication</param>
public static void Mult(ref Matrix4d left, ref Matrix4d right, out Matrix4d result)
{
result = new Matrix4d();
result.M11 = left.M11 * right.M11 + left.M12 * right.M21 + left.M13 * right.M31 + left.M14 * right.M41;
result.M12 = left.M11 * right.M12 + left.M12 * right.M22 + left.M13 * right.M32 + left.M14 * right.M42;
result.M13 = left.M11 * right.M13 + left.M12 * right.M23 + left.M13 * right.M33 + left.M14 * right.M43;
result.M14 = left.M11 * right.M14 + left.M12 * right.M24 + left.M13 * right.M34 + left.M14 * right.M44;
result.M21 = left.M21 * right.M11 + left.M22 * right.M21 + left.M23 * right.M31 + left.M24 * right.M41;
result.M22 = left.M21 * right.M12 + left.M22 * right.M22 + left.M23 * right.M32 + left.M24 * right.M42;
result.M23 = left.M21 * right.M13 + left.M22 * right.M23 + left.M23 * right.M33 + left.M24 * right.M43;
result.M24 = left.M21 * right.M14 + left.M22 * right.M24 + left.M23 * right.M34 + left.M24 * right.M44;
result.M31 = left.M31 * right.M11 + left.M32 * right.M21 + left.M33 * right.M31 + left.M34 * right.M41;
result.M32 = left.M31 * right.M12 + left.M32 * right.M22 + left.M33 * right.M32 + left.M34 * right.M42;
result.M33 = left.M31 * right.M13 + left.M32 * right.M23 + left.M33 * right.M33 + left.M34 * right.M43;
result.M34 = left.M31 * right.M14 + left.M32 * right.M24 + left.M33 * right.M34 + left.M34 * right.M44;
result.M41 = left.M41 * right.M11 + left.M42 * right.M21 + left.M43 * right.M31 + left.M44 * right.M41;
result.M42 = left.M41 * right.M12 + left.M42 * right.M22 + left.M43 * right.M32 + left.M44 * right.M42;
result.M43 = left.M41 * right.M13 + left.M42 * right.M23 + left.M43 * right.M33 + left.M44 * right.M43;
result.M44 = left.M41 * right.M14 + left.M42 * right.M24 + left.M43 * right.M34 + left.M44 * right.M44;
double lM11 = left.Row0.X, lM12 = left.Row0.Y, lM13 = left.Row0.Z, lM14 = left.Row0.W,
lM21 = left.Row1.X, lM22 = left.Row1.Y, lM23 = left.Row1.Z, lM24 = left.Row1.W,
lM31 = left.Row2.X, lM32 = left.Row2.Y, lM33 = left.Row2.Z, lM34 = left.Row2.W,
lM41 = left.Row3.X, lM42 = left.Row3.Y, lM43 = left.Row3.Z, lM44 = left.Row3.W,
rM11 = right.Row0.X, rM12 = right.Row0.Y, rM13 = right.Row0.Z, rM14 = right.Row0.W,
rM21 = right.Row1.X, rM22 = right.Row1.Y, rM23 = right.Row1.Z, rM24 = right.Row1.W,
rM31 = right.Row2.X, rM32 = right.Row2.Y, rM33 = right.Row2.Z, rM34 = right.Row2.W,
rM41 = right.Row3.X, rM42 = right.Row3.Y, rM43 = right.Row3.Z, rM44 = right.Row3.W;
result.Row0.X = (((lM11 * rM11) + (lM12 * rM21)) + (lM13 * rM31)) + (lM14 * rM41);
result.Row0.Y = (((lM11 * rM12) + (lM12 * rM22)) + (lM13 * rM32)) + (lM14 * rM42);
result.Row0.Z = (((lM11 * rM13) + (lM12 * rM23)) + (lM13 * rM33)) + (lM14 * rM43);
result.Row0.W = (((lM11 * rM14) + (lM12 * rM24)) + (lM13 * rM34)) + (lM14 * rM44);
result.Row1.X = (((lM21 * rM11) + (lM22 * rM21)) + (lM23 * rM31)) + (lM24 * rM41);
result.Row1.Y = (((lM21 * rM12) + (lM22 * rM22)) + (lM23 * rM32)) + (lM24 * rM42);
result.Row1.Z = (((lM21 * rM13) + (lM22 * rM23)) + (lM23 * rM33)) + (lM24 * rM43);
result.Row1.W = (((lM21 * rM14) + (lM22 * rM24)) + (lM23 * rM34)) + (lM24 * rM44);
result.Row2.X = (((lM31 * rM11) + (lM32 * rM21)) + (lM33 * rM31)) + (lM34 * rM41);
result.Row2.Y = (((lM31 * rM12) + (lM32 * rM22)) + (lM33 * rM32)) + (lM34 * rM42);
result.Row2.Z = (((lM31 * rM13) + (lM32 * rM23)) + (lM33 * rM33)) + (lM34 * rM43);
result.Row2.W = (((lM31 * rM14) + (lM32 * rM24)) + (lM33 * rM34)) + (lM34 * rM44);
result.Row3.X = (((lM41 * rM11) + (lM42 * rM21)) + (lM43 * rM31)) + (lM44 * rM41);
result.Row3.Y = (((lM41 * rM12) + (lM42 * rM22)) + (lM43 * rM32)) + (lM44 * rM42);
result.Row3.Z = (((lM41 * rM13) + (lM42 * rM23)) + (lM43 * rM33)) + (lM44 * rM43);
result.Row3.W = (((lM41 * rM14) + (lM42 * rM24)) + (lM43 * rM34)) + (lM44 * rM44);
}
#endregion