1: \begin{abstract}

2:  Implicit integration of the viscous term can significantly improve

3:  performance in computational fluid dynamics for highly viscous fluids

4:  such as lava. We show improvements over our previous proposal for

5:  semi-implicit viscous integration in Smoothed Particle Hydrodynamics,

6:  extending it to support a wider range of boundary models. Due to the

7:  resulting loss of matrix symmetry, a key advancement is a more robust

8:  version of the biconjugate gradient stabilized method to solve the

9:  linear systems, that is also better suited for parallelization in both

10:  shared-memory and distributed-memory systems. The advantages of the new

11:  solver are demostrated in applications with both Newtonian and

12:  non-Newtonian fluids, covering both the numerical aspect (improved

13:  convergence thanks to the possibility to use more accurate boundary

14:  model) and the computing aspect (with excellent strong scaling and

15:  satisfactory weak scaling).

16: \end{abstract}

17: