We have three licences of PHOENICS, and wish run PARALLEL PHOENICS. Our ground.f works well with non-parallelized PHOENICS, but it does not work with PARALLEL PHOENICS. If the way for coding ground.f in PARALLEL PHOENICS is different, why CHAM do not support the document to users for coding PARALLELized ground.f.

(1). It seems to me that the vendor is in the business of developing CFD codes, and you are the user of the codes. So, there is a perfect match. (2). The next step is to draw a specification and work with the vendor so that it can be implemented in the code through a contract or some kind. "business is business", sometimes it is free, sometimes it is not.

Mr. Chien

I dont know if you have understood what I understood from Mr. Zeng message, but I can see a not so easy problem of "matching".

In my oppinion it is a real problem, where a user of a re-compilable PHOENICS version can't use their own codes because he(she) have no way to know how the "new" version handle the F arrays in GROUND, for instance.

As a software vendor you should supply the information which allow the user to handle the software he(she) had bougth.

I don't know if it is a legislation about that or if this feature is clearly remarked on the PHOENICS Lisence Agreement. If so, then you are right with respect to "draw a specification and work with the vendor...", otherwise you don't.

Maybe 10 years ago PHOENICS was the outstanding CFD software (it is a personal oppinion) between a few codes to perform serious CFD calculations, but now you have a more competitive scenario (FLUENT, STAR-CD among others), even with a really well designed free CFD codes (http://www.lmm.jussieu.fr/~zaleski/a...-V3.0.1.tar.gz).

So, now I can see a problem in which I can't say "business is business"; instead that, I should say "business are not like this" or "you can loose a client".

Finally I want to remark that (again according to my own experience) the re-compilation of the PHOENICS executables (earexe or parexe) adding your own code, is the real power of PHOENICS. I grew up working with(on) PHOENICS (profesionally speaking). I had enjoyed and I had used this feature during several PHOENICS versions release and now I am sure that the direction of CHAM will not loose this enormous source of scientific and technical feedback.

Regards

Kike

(1). I don't think the commercial code vendors are interested in solving individual user's problem through the code. (2). If that is the case, they will be very busy answering the questions here. And they are not going to have time to work on the next generation codes. (3). If there are enough users with the similar need, and if they also have made the request, then I think, it is possible to change this. (4). If you think, by not using the code , it will force the vendor to change, then it is also one approach. (5). The fact is the vendors tend to use educational rate to distribute the software to schools to promote the usage. (6). Anyway, if the code is not working as it should be, then one should contact directly the sales department of the vendor company.

Mr. Chien

Too abstract for me. Have you used PHOENICS code? For how long?

Maybe I had assumed you are familiar with the problem we (Mr. Zeng you and me) are talking about, but from your answer I think you don't.

Never mind, I appretiate your comments. Regards

Kike

I am not sure whether you are a maintained PHOENICS User, i.e. whether you are in possession of a valid maintenance contract. If so, then you have access to the support section of CHAM's website where some guidelines are provided in a FAQ in respect of GROUND coding for Parallel PHOENICS.

CHAM's support team have assisted several maintained users in the tracing of problems with cases that work in sequential but not Parallel mode. Such inquiries are sent to support@cham.co.uk for resolution by our support engineers.

In general, the problem does not necessarily reside in the GROUND coding, as the solution methodology is obviously different due to domain decomposition in the z-direction and the use of a different solver. The problem may even be due to an internal PHOENICS problem or sometimes inappropriate user settings for numerical-control parameters when reverting to parallel. In addition some built-in options and features are not as yet designed for parallel use.

If the simulation is performed on a single processor in parallel mode and the run is unsuccessful, this would suggest that the problem is not in the data transfer between sub-domains. I guess you have run successfully on a single processor and so concluded that the problem is in your GROUND coding where data transfers would be important for coding dealing with information involving the z direction.

CHAM can only expend time and resources on investigations for maintained users, but since you request only documentation, I put down below the opening advice to be in the support FAQ section of the website:

"Parallel PHOENICS is restricted to domain decomposition in the z-direction, and the ground coding of IZ loops requires special attention if:

a) summations are made over all IZ slabs; b) references are made to non-neighbour variables in the IZ direction; and c) references are made to non-neighbour IZ indices rather than to Z coordinates.

The reason is that Parallel PHOENICS interprets each IZ loop as referring to the current subdomain, including 'overlap' cells. Such coding is likely to occur in Group 19 or Group 11.

PARALLEL PHOENICS operates on a user-specified number of subdomains. These are created automatically by splitting the solution domain in the z-direction. Each subdomain is assigned to a different parallel processor, and each subdomain is enlarged at each end by the additional of two slabs of cells. These additional 'overlap' slabs are used to store field values from the adjacent subdomains so as to facilitate data exchange between subdomains during the parallel computation."

(1). Currently, I am not using any commercial CFD codes at all. (2). For me, I think, these codes are useless, and all the users of commercial codes can do is to ask some questions instead of providing answers. (3). There are people interested in spending time, money , asking questions and still are not getting the right answers. (4). I am not using commercial cfd codes, because for me these codes do not provide me answers, they only create more questions. (5). The right way to do is to become smart enough to write your own software. In that case, you can answer your own questions. (6). My goal is to keep this forum alive, active. I don't have the answer to your questions. (but someone might.)

> CHAM can only expend time and resources on investigations for maintained users

(1) Our Lab. has bought PHOENICS in about three years ago, and bought three licences this year after renew all the computer system. There are two licences (PHOENICS3.2) in DEC server and HITACHI supercomputer are claimed to be possible run PARALLEL PHOENICS. I believe we should be maintained PHOENICS Users.

>Such inquiries are sent to support@cham.co.uk

(2) According to our experiences support@cham.co.uk seldom give supports directly.

(3)We have used neighbour variables in the IZ direction, is it be possible reason?

>(1) Our Lab. has bought PHOENICS in about three years >ago, and bought three licences this year after renew >all the computer system. There are two licences
PHOENICS3.2) in DEC server and HITACHI supercomputer >are claimed to be possible run PARALLEL PHOENICS. I >believe we should be maintained PHOENICS Users.

If you require assistance from CHAM then I suggest that you send your inquiry through the proper channels to support@cham.co.uk, where it will be processed by the support team provided that your are in possession of a valid maintenance contract. Please note that some PHOENICS users choose to buy licences but not technical support.

(3)We have used neighbour variables in the IZ direction, is it be possible reason?

If you are in an active IZ slab on a processor then you will be able to access up to two slabs away without problem, i.e. IZ+1, IZ+2, IZ-1, IZ-2. This is OK because there are two overlap slabs between adjacent sub domains facilitating data transfer. However, accessing IZ+3 with the function ANYZ say, will be a problem because then that information is beyond reach, as it is in the next processor.

We will need to know the name of your organisation and department and if possible, your site reference number. This will allow us to determine the status of your maintenance contract.

>According to our experiences support@cham.co.uk
: seldom give supports directly.

I am sorry to hear of this, and it is of some concern to us. Therefore, please send details to support@cham.co.uk of your earlier unsuccessful attempts to obtain support.

There are over 50 inquiries per month and users may have to enter a queue, but as far as we know every incoming inquiry is logged and assigned to a support engineer. If the user is not maintained, then a response is issued stating this fact plus an invitation to renew maintenance.

Rarely, some incoming inquiries have been lost because of server problems either at CHAM or the user's end.

>We will need to know the name of your organisation and department and if possible, your site reference number. This will allow us to determine the status of your maintenance contract.

Institute for Materials Research Tohoku University Japan

I do not know the means of site reference number.

It seems you are maintained by CHAM Japan rather than CHAM UK, and that CHAM Japan have referred your problem to CHAM UK for resolution by Dr Jeremy Wu, who is our Technical Director and ulltimately responsible for Far East users.

Jeremy says that your model runs successfully on the Linux platform, but as yet he has been unable to test it on the Compaq (DEC server) due to a hardware fault, which he hopes will be remedied shortly so that he can report to you on his progress by the end of the week.

Since your experience is that the case does not run on the DEC server, Jeremy's test under Linux would suggest that your ground file is probably OK.

Thanks. We are responded very quickly by CHAM Japan when we ask them questions after we bought PHOENICS from them in this year.

I do not know we should sent questions directly to your agency rather than support@cham.co.uk before, because it is said to sent questions directly to support@cham.co.uk based on the document of PHOENICS. For this reason, I would like to appologize to Dr Jeremy Wu.

>>you have access to the support section of CHAM's
:>website where some guidelines are provided in a FAQ
:>in respect of GROUND coding for Parallel PHOENICS.

I have check the FAQ, I failed to find any guideline in FAQ in respect of GROUND coding for Parallel PHOENICS.

Are you really sure your mentioned guideline for parallelization of PHOENICS exists in FAQ? Could you please tell me where it is?

It appears that our software development team have yet to update the faq section of our website with the latest material. I am told this will happen on Wednesday, so I suggest you try again on Thursday 12 April. In the FAQ home page you should be able to see the hyperlink to Parallel PHOENICS, as this is how you can tell you are looking at the latest web material.