
Computational Core of HyperFLOW2D now open source, meet OpenHyperFLOW2D !
Well, first of all it is necessary to agree on the terms.
I do not know which solver you use. So let's start with their classification.
At first:
Solvers are dimensional and dimensionless. Most solvers dimensionless. In contrast OpenHyperFLOW2D  originally the dimensional solver as its core code was created for the engineering design. That is, all the differential equation in this case are written in dimensional form.
Secondly:
Solvers are use the concept of a gauge pressure (p_g) and using the absolute pressure directly. In case of gauge pressure concept, to obtain the absolute pressure we should add gauge pressure to the pressure, which operates solver.
OpenHyperFLOW2D initially uses the concept of absolute pressure directly
Next point, the physical aspect of your case.
When you say "Oulet pressure BC  0 P" it means outflow in the vacuum ?
In my case outflow occurs in the atmosphere,
that is equal to the ambient pressure 1.0e5 Pa
If your solver uses the concept of a gauge pressure then your BC can mean both.
It depends on the value of gauge pressure:
p_g = 0 Pa > vacuum outflow
p_g = 1.e 5 Pa > atmosphere outflow
In the case of outflow in vacuum instability is normal. In this case, usually zero pressure replace a very small nonzero,
but not everyone solver can be considered this (OpenHyperFLOW2D can)
Now concerning the the initial and BC of the above case:
Since the problem of axially symmetric, simulated only half the nozzle.
Image obtained with a full nozzle in the postprocessor.
BC:
 X axi  symmetry BC (radial velocity and gradients of all parameters is zero)
 left boundary in chamber  Dirichlet BC with total pressure 7.2e6 Pa and total temperature 3338.5 K.
concentration of combustion products  100%, air  0%
 left boundary in ambient area  Dirichlet BC with total pressure 1.0e5 Pa and total temperature 300 K.
concentration of combustion products  0%, air  100%
 top boundary  Dirichlet BC with total pressure 1.0e5 Pa and total temperature 300 K.
concentration of combustion products  0%, air  100%
 right boundary  Neumann BC with zero gradient in axial direction
 nozzle walls  "noslip" BC
Initial conditions:
 chamber from left boundary to throat filled gas with total pressure 7.2e6 Pa and total temperature 3338.5 K.
concentration of combustion products  100%, air  0%
 another domain part is filled gas with total pressure 1.0e5 Pa and total temperature 300 K.
concentration of combustion products  0%, air  100%
The boundary between the domains simulates a frangible disc in throat of nozzle.
So as not to attract attention spambots I do not give here my email, but you can easily find it in any files header of OpenHyperFLOW2D projectPosted July 13, 2015 at 03:52 by SergeAS 
Computational Core of HyperFLOW2D now open source, meet OpenHyperFLOW2D !
Boundary conditions
Yes i saw your supersonic nozzle picture.
I am also trying to design a supersonic nozzle with a Exit Mach of 3.0
but the results we diverging .
I used BC as (InletPressure of 300000 P , Oulet pressure BC  0 P , walls as stationary )
so wanted to know your BC's and model setup details of how did you specify inlet , outlet and walls and did you use freestream conditions in the surrounding?
hope you can help and this clearifies your question.
and if you can share me your Email, i can send you picture showing geometry i am working on. (tried attaching picture over here but failed)
Regards,
Mehlam.Posted July 12, 2015 at 14:02 by Mehlam 
Computational Core of HyperFLOW2D now open source, meet OpenHyperFLOW2D !
Quote:Posted July 12, 2015 at 05:47 by SergeAS 
Computational Core of HyperFLOW2D now open source, meet OpenHyperFLOW2D !
i am doing simillar work.
can you tell me what initial & boundary conditions did you use?
Also plz just which flow model you used??Posted July 11, 2015 at 23:06 by Mehlam 
Dark side of Amdahl's law
This is due to the fact that the one cluster was used Infiniband DDR and has nodes on NUMA architecture and a smaller cache (Opteron 285) against the other cluster which used SDR on the SMP (Xeon) and large cache.
The addition for each node we has two data exchanges (except first and last node) on Infiniband and 3 exchanges of internal memory (SMP or NUMA).
PS: If I use one process per node (exchange only for InfiniBand) that scalability will be even better, it's a paradox but a factPosted March 28, 2012 at 10:20 by SergeAS 
Posted March 28, 2012 at 09:20 by lakeat