Referee 1 Report ************************************************************************

The manuscript describes a methodology to simulate static or quasi-static frictional contact
between multi-elasto-plastic bodies using a general nonlinear friction law. The approach uses
a "node-to-point" contact element strategy in order to overcome convergence problems existing
in other methodologies. Interesting results are obtained with the approach which shows that
the model can be successfully applied to active faults.

However, the manuscript fails to demonstrate the efficiency and stability of the approach.
Therefore I recommand the manuscript for publication providing the suggested changes are
made. I believe that the paper should be revised in various ways before re-submission.
Minor problems with the manuscript are:

1) "in our laboratory" should be replaced by the institution name.

2) The introduction states that the model aims to "further predict earthquake occurence". I
suggest removing this statement or rephrasing it into a more general statement such as
"the approach may permit study of earthquake prediction".

3) References to other methodologies are required for the statement "it is also well known
that it is quite time consuming and also difficult for dynamic-explicit FEM to predict the
stress distribution with a high accuracy".

4) Eq. (1) states that when contact occurs the normal contact velocity is null (ie. d(g sub n)/dt = 0 )
this would require explanation since it would appear that for a contact to occur the normal
contact velocity must not be zero.

5) Eq. (2) states that (f sup alpha). (n sup alpha) < 0 when a contact occurs, 
   whereas Figure 1 shows otherwise (  (f sup alpha). (n sup alpha)  > 0 ).
   
6) Eq. (3) assumes that g sub n < 0 because the two bodies are in contact, 
   this should be explicitly stated in the text.

7) Eq. (5) f sub m is not defined in the text, and the term sqrt( (f sub m)(f sub m) )
   is used in the equation. Do the author(s) mean abs(f sub m) ?
   
8) Eq. (7) needs more explanation.

9) Figure 6 shows units in mm. Do the author(s) mean m? The dimensions of model also
appears to be wrong ( 20038 X 51 X 600 does not appear to correspond to the dimensions
of the model shown in Figure 6).

Major problems with the manuscript are:
10) Section 5: as stated in the manuscript, one of the goals of this paper is to show the efficiency,
stability and usefulness of the proposed algorithm. However the manuscript failed
to demonstrate the efficiency and stability of the algorithm. The algorithm utilizes a parallel
sparse solver which is the only parallel part of the algorithm (cf. section 5, the contact
search and stiffness matrix assembly computations are serial). The only performance estimate
is given by Figure 5 which shows computational time vs. numbers of contact nodes.

To demonstrate efficiency, a speed-up plot with number of processors, estimate of flops
vs. model size, and/or comparison with other algorithms would be required. The stability
of the algorithm is not demonstrated in the manuscript. The rate of convergence (number
of iterations required) for various problems should be stated and eventually compared
with other methodologies to demonstrate stability and efficiency.

11) Section 6: References to the actual analysis and description of the numerical experiment
or a more detail analysis is required. Only a few input parameters are given in the text.
Enough details should be given so that the experiment may be reproduced.

Presentation Changes
12) Figure 6b showing a 3D view of the mesh is not necessary and should be removed.

Referee 2 Report ************************************************************************
This is an interesting paper which should be published with some enhancement.
Sections 4 and 5 need extension with a deeper discussion of parallel
computational structure -- what architecture and programming model do you use.
Does method scale to large parallel machines?
The presentation -- including english style -- is good.