|
|
Computes the product of a 3-component vector and a 3x3 matrix.
| m3x3 dst, src0, src1 |
|---|
where
| Vertex shader versions | 1_1 | 2_0 | 2_x | 2_sw | 3_0 | 3_sw |
|---|---|---|---|---|---|---|
| m3x3 | x | x | x | x | x | x |
The xyz mask is required for the destination register. Negate and swizzle modifiers are allowed for src0 but not for src1.
The following code fragment shows the operations performed.
dest.x = (src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z); dest.y = (src0.x * src2.x) + (src0.y * src2.y) + (src0.z * src2.z); dest.z = (src0.x * src3.x) + (src0.y * src3.y) + (src0.z * src3.z);
The input vector is in register src0. The input 3x3 matrix is in register src1, and the next two higher registers, as shown in the expansion below. A 3D result is produced, leaving the other element of the destination register (dest.w) unaffected.
This operation is commonly used for transforming normal vectors during lighting computations. This instruction is implemented as a pair of dot products as shown below.
m3x3 r0.xyz, r1, c0 which will be expanded to: dp3 r0.x, r1, c0 dp3 r0.y, r1, c1 dp3 r0.z, r1, c2