Function computeF

Tools/MeshSkinning/main.cpp:687–725

#ifdef USE_AVX

void computeF()
{
	const int numParticles = (int)activeParticles.size();

	#pragma omp parallel default(shared)
	{
		#pragma omp for schedule(static)  
		for (int i = 0; i < (int)numParticles; i++)
		{
			const unsigned int index = activeParticles[i];
			const Vector3r &xi = x[index];
			const Vector3r& xi0 = x0[index];

			const Vector3f8 xi_avx(xi);
			const Vector3f8 xi0_avx(xi0);

			const unsigned int numNeighbors = (unsigned int) initialNeighbors[index].size();

			Matrix3f8 F_avx;
			F_avx.setZero();
			for (unsigned int j = 0; j < numNeighbors; j += 8)
			{
				const unsigned int count = std::min(numNeighbors - j, 8u);

				const Vector3f8 xj_avx = convertVec_zero(&initialNeighbors[index][j], &x[0], count);
				const Vector3f8 xj_xi = xj_avx - xi_avx;
				const Vector3f8& V_gradW = precomp_V_gradW8[precomputed_indices8[i] + j / 8];

				const Scalarf8 restVolume_j_avx = convert_zero(&initialNeighbors[index][j], &restVolumes[0], count);
				const Vector3f8 xj0_avx = convertVec_zero(&initialNeighbors[index][j], &x0[0], count);
				//const Vector3f8& V_gradW = Matrix3f8(L[index]) * CubicKernel_AVX::gradW(xi0_avx - xj0_avx) * restVolume_j_avx;
				Matrix3f8 dyad;
				dyadicProduct(xj_xi, V_gradW, dyad);
				F_avx += dyad;
			}
			F[index] = F_avx.reduce();
		}
	}
}


void deformMesh()