Benchmarks

MESA - Invert Matrix

DFG 
      GLfloat invDet = 1.0F / det;
      /* Compute rest of inverse. */
      tmp[0] *= invDet;
      tmp[1] *= invDet;
      tmp[2] *= invDet;
      tmp[3] *= invDet;

      tmp[4] = -(m12 * d34 - m13 * d24 + m14 * d23) * invDet;
      tmp[5] =  (m11 * d34 + m13 * d41 + m14 * d13) * invDet;
      tmp[6] = -(m11 * d24 + m12 * d41 + m14 * d12) * invDet;
      tmp[7] =  (m11 * d23 - m12 * d13 + m13 * d12) * invDet;

      /* Pre-compute 2x2 dets for first two rows when computing */
      /* cofactors of last two rows. */
      d12 = m11*m22-m21*m12;
      d13 = m11*m23-m21*m13;
      d23 = m12*m23-m22*m13;
      d24 = m12*m24-m22*m14;
      d34 = m13*m24-m23*m14;
      d41 = m14*m21-m24*m11;

      tmp[8] =  (m42 * d34 - m43 * d24 + m44 * d23) * invDet;
      tmp[9] = -(m41 * d34 + m43 * d41 + m44 * d13) * invDet;
      tmp[10] =  (m41 * d24 + m42 * d41 + m44 * d12) * invDet;
      tmp[11] = -(m41 * d23 - m42 * d13 + m43 * d12) * invDet;
      tmp[12] = -(m32 * d34 - m33 * d24 + m34 * d23) * invDet;
      tmp[13] =  (m31 * d34 + m33 * d41 + m34 * d13) * invDet;
      tmp[14] = -(m31 * d24 + m32 * d41 + m34 * d12) * invDet;
      tmp[15] =  (m31 * d23 - m32 * d13 + m33 * d12) * invDet;

      MEMCPY(out, tmp, 16*sizeof(GLfloat));
Number of nodes 333
Number of edges 354
Avg. edges per node 1.06
Critical Path Length 11
Parallelism (nodes/critical path) 30.27
DOT File
Full C Function
MESA Application

The invert_matrix_general function is a matrix inversion routine. Digital signal processing often involves matricies, which is one reason that it lends itself so nicely to the idea of parallelism. As such, the inverse matrix method is often used in any number of DSP applications.