CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

public CFD Code development

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   March 2, 1999, 20:54
Default Re: public CFD Code development
  #21
Thomas Huld
Guest
 
Posts: n/a
John C. Chien wrote:

>Since the code has been developed and named, also the
>wishlist has been determined, I don't think there is much
>left to be done. Since it is copy righted, it is up to
>the author to determine what to do with his code. I
>normally think that the most
>important part of CFD is the person not the code. On the
>cfd-online, I can talk to other persons directly, but for
>a code, there is
>not much I can do, except to download it, read it, decode
>it...not really a fun thing to do.

I guess I didn't explain myself well enough. First of all, the wishlist I wrote was not *the* wishlist but just *a* wishlist, examples of what could be done starting from the existing code. If somebody else wants to do something different, we would give support for the implementation of this something, and help on integrating it into the code. You could say, that in this way you would get the code *and* the person. Of course, since we are paid to do development in a specific field, we are best able to give support in that field. Then again, if this thing really took off and many people would be working with the code, there would be a good probability that a newcomer would find somebody with expertise in a given field.

There exist various codes that have been dumped on the net by the author, and which are essentially unsupported. That was not my idea.

The point is: you are not bound by, for example, our choice of compressible solver, as long as you are then willing to write something different. The underlying geometry definition could still be used. Only if you reject the geometry (unstructured, triangular/tetrahedral) would you have to start from scratch or with a different code.

Thomas

  Reply With Quote

Old   March 3, 1999, 02:54
Default Re: public CFD Code development
  #22
Mo Yun
Guest
 
Posts: n/a
I think it may be best to write the solver with Fortran and to write the Pre/Post with Java may be wise.

But how the CFD makers live with the appearance of the public CFD?

  Reply With Quote

Old   March 3, 1999, 03:35
Default Re: public CFD Code development
  #23
Zhong Lei
Guest
 
Posts: n/a
I have ever thought about a similar problem:

Is it possible to develop an enviroment which is something like the software MathLab and GUI but including many intrisical functions and black boxes for CFD development. One can call them into theirown programs but need not to write subroutines which are well development and almost become standard use. It may have interfaces where one can write his code using a special language (CFD language). Also, one can include some self-defined subroutines into it.

Zhong Lei
  Reply With Quote

Old   March 3, 1999, 05:51
Default Re: public CFD Code development
  #24
Gassan Abdoulaev
Guest
 
Posts: n/a
But still, is there anybody, who would say: "I need this code, because I don't want to write my own and I don't have $$ to buy a commercial one. Just a word from a potential user... Up to now I see only developers. Or do you think users appear, as soon as the code will be available?
  Reply With Quote

Old   March 3, 1999, 08:49
Default Re: public CFD Code development
  #25
Roland Bender
Guest
 
Posts: n/a
Hello,

i am a potential user. To myself: I am post-graduate Phd-student and working with a cfd-code as blackbox to couple it with a monte-carlo-pdf-code.

I think the idea of public code is great for scientific purposes. I believe a public code has a great chance on this field. Perhaps if the project is successfull, users in the industry use the code too. Think of linux it is public project too and it gets more important for the industry in the last years.

This leads to the next point. Why not found the developer group as a subproject of GNU. The GNU project has the experience with developing free codes.

Greets
  Reply With Quote

Old   March 3, 1999, 10:36
Default Re: public CFD Code development
  #26
John C. Chien
Guest
 
Posts: n/a
I am pleased to hear that the " person" will be included in the project. The only reason I mentioned that " the person is more important than the code" is that there are hundreds of un-used codes still floating in many companies computer systems, and it takes money to remove the unused codes from the system. This is not a joke. Typically, 5 codes are actually used per group of 5 to 10 persons. So, the total of codes needed for a division of 10 is around 50 codes. But, normally, you will see that about 300 codes are being maintained on the system even if 250 codes are rarely used. If you keep a person for 30 years, all you need is one person. If you keep the codes he developed over 30 years of time, you are going to have a big problem. This is for practical reasons because the language, the operating system and the hardware all change from time to time. The market will force the computing environment to change and pretty soon the code is not going to work ( you don't have to do anything bad to it). And that cycle time is very fast now. Also, the need to have a CFD code is because of the need to solve fluid dynamic related problems. It is totally different from the need of an operating system. The operating system does not solve a fluid dynamic problem, it simply provides mechanical(or actually electronic ) means to operate the machine ( the computer ). It is a one-to-one correspondence device, just like a calculator, a spread sheet program. For a CFD code, it is useless to the user, unless it can solve the fluid dynamic problem for the user. And if you read this forum, you can get a pretty good idea that most people are constantly looking for good answers for his fluid dynamic problem, even with so many CFD codes around and available. The idea is good, but these factors need to factored into the thinking. Can LINUX solve a fluid dynamic problem?
  Reply With Quote

Old   March 3, 1999, 10:47
Default Re: public CFD Code development
  #27
Alessandro Grossi
Guest
 
Posts: n/a
Hello

I'm a phd student and I work every day with a well know commercial code, but I'm a linux fan, i like very much the idea of a free cfd software. I hope you will demonstrate that the mind are superior to dollars (or euros . You may ask to GNU how to manage all the project. Best whishes!

  Reply With Quote

Old   March 3, 1999, 14:11
Default Re: public CFD Code development
  #28
Aaron
Guest
 
Posts: n/a
Gassan, right now we pay $30k per year for our CFD license (I'll respectfully leave out the company's name) which we could be spending on (1) part of another employee's salary, (2) more lab space, (3) another wind tunnel, (4) more conferences and 'off site' meetings for the employees, (5) more contract work by universities and small companies, (6) etc, (7) etc. So I say, yes I'll definitely make use of, and contribute to such a project. Of the many branches and divisions in our institute, most have researchers who have voiced their desire to add CFD research to their projects, but they can't afford it themselves. $30000 is a whole lot of money to be paying out each, and every year for a license.
  Reply With Quote

Old   March 4, 1999, 04:49
Default Re: public CFD Code development
  #29
Doru Caraeni
Guest
 
Posts: n/a
Hello Heinz,

I thing your idea of a freely available CFD code is excellent and I subscribe to it!

I have "some experience" in this area since I developed my own CFD code (unstructured NS flow solver). I used to work also in a research institute for aerospace engines(I'm doing my PhD now!) Please sign me up for your project!

My e-mail address is: dc@mail.vok.lth.se

My idea, of what a good CFD code should have at basis, is:

- Developed in C++ (at least the part of the code manipulating the data structure for the grid!), and/or Fortran 90, written in a platform independent - "clean" stile (if some graphics for instance will be included).

- Using unstructured grid (but using an octal-tree data structure, to make the multigrid-algorithm easy).

- Modular structured: the "job-doing" parts should be seen as plug-ins into a kernel part of the code (that will ensure the workout of this plug-ins with the data structure). The solver(s) for incompressible and for compressible flows, or any other parts (a turbulence model, etc.) should be connected into the kernel. ( Linux operating system uses this idea! And my code too!!)

This is all! No other restrictions!

Hope this project will start, and it will blow the "competition"! For free!

(at least it will be of big help for students)

Looking forward hearing from you, Doru
  Reply With Quote

Old   March 4, 1999, 12:51
Default Re: public CFD Code development
  #30
Steven Bosscher
Guest
 
Posts: n/a
Zhong Lei suggested a CFD computer language.

But that will involve developing that CFD language and a compiler/interpreter for it. Writing an interpreter is not an easy thing! I had a similair idea once, but it appeared to be much more difficult that developing a traditional CFD application, since writing an interpreter is a VERY difficult problem!

I'm not saying it cannot be done, but it means you'll have in fact two projects: 1. public domain CFD code, and 2. public domain compiler/interpeter with GUI

Now since I don't think many CFD experts are into writing interpreters and GUI's, you're going to need experts in computer sciences too. I'd say you should keep it with the CFD code first, and then perhaps let somebody build his/her CFD language arount that.

Steven Bosscher (Student LR) Faculty of Aerospace Engineering/Delft University of Technology/ www.lr.tudelft.nl
  Reply With Quote

Old   March 4, 1999, 12:53
Default Re: public CFD Code development
  #31
Steven Bosscher
Guest
 
Posts: n/a
I'll be your potential user, I guess. Besides, it's not only about using the code. Some other people already came up with some other plausible reasons to start sucha project
  Reply With Quote

Old   March 4, 1999, 17:11
Default Re: Why a public CFD Code and how to start
  #32
John C. Chien
Guest
 
Posts: n/a
Since something has been developed, I would strongly suggest that a very detailed handbook be first developed to document every lines of the code ( or something like that). When the book is finished, a copy of the existing code can be attached ( CD ) for sale as a regular book at a regular book price. The users can then send in extensions, modifications for download or in the next version release. Without the document and the code, everything on internet at this stage is something like a guessing game. At that point, if no one understand the code, then you know what to do next. If you receive many extensions and modifications, then you can proceed form that point.
  Reply With Quote

Old   March 4, 1999, 22:18
Default Re: public CFD Code development
  #33
Duane Baker
Guest
 
Posts: n/a
Hi Heinz and all who have contributed,

This certainly has been an interesting discussion to read and think about and I thought that I might do a bit of a summary and analysis here and add some of my own ideas.

First of all I would like to categorize the types of responses.

Type 1: The coder with an unpublished/unused code. This is much like the unpublished artist syndrome. Everyone has their own specific preferences (probably what they coded the first time or for grad-school, etc.) and wants the whole world to adjust to it! Such as statements like: use FORTRAN 90, Java, structured, multiblock, CDS, Petrov-Galerkin, incompressible, all-speed algorithms, explicit, simple, PISO, etc, etc. Much like in any creative art, there are many choices but most of the combinations will not stand the test of time and will remain unpublished and unused. The fact is that just because someone spent 4 years of their life coding DOES NOT MEAN THAT IT IS USEFUL TO OTHERS, PARTICULARLY IF NOT WRITTEN WITH A USER OTHER THAN THEMSELVES IN MIND!

By my analogy, who wants to really read the book that your buddy in an English Doctoral program is writing?? In fact, much like the English student's work, the print-out of the code will sit around in the coder's old desk for years until it is thrown out buy the new occupant! Hopefully, if well supervised, the learning experience will last a lifetime....THIS IS THE PURPOSE OF THE EXERCISE! Sometimes, unfortunatly, the whole thing is a complete waste of time, money, energy, etc.!!

Type 2. The under-resourced CFDer (probably a student) This group of people wrote in saying..."That is a great idea, sign me up!" or "I would love to do CFD, but I don't have 30k/yr, etc. etc". For those at a University and don't have any access, I really feel for you, but just go find a different University, supervisor, etc. YOU DO NOT GO TO A WORLD CLASS INSTITUTION(that is what they all claim to be now isn't it) TO DICK AROUND WITH FREE STUFF THAT YOU CAN DOWNLOAD IN YOUR SPARE TIME! Somebody addressed the issue as use of slave labour grad students: computers and codes expensive and labour cheap! Just get out of this situation, in the long run the Universiy's/department's/supervisor's reputation will not be worth anything!

To draw an analogy (this is one of my favourites!): What would you think of going to a Medical Doctor of the "Home School of Brain Surgery" who was educated by downloading from free medical sites on the net, and came in on weekends to work in the surgical suites with equipment that was either home-made or donated from shut-down hospitals! THEN WHY DO WE THINK THAT WE CAN EDUCATE WORLD CLASS ENGINEERS WITH THE SAME LEVEL OF TECHNOLOGY!

In the practicing engineering firm, the real issue is on of total cost. PROFESSIONAL TIME + CODE + COMPUTER And with professional time at order $50/hr it does not take too many weeks wasted to pay for that liscense BUT the REAL REASON FOR GETTING A COMMERCIAL LISCENCE IS THAT ONE CAN THEN DEMAND A LEVEL OF QUALITY, SUPPORT, TRAINING, ETC.

In fact, I want to pay for everything in a code because then, I know that I have a say in the quality and service!

SOME ANALYSIS 1. Many freeware codes have been developed under Government funding which required that they be distributed. Although some stuff has come out of NASA and some government labs (cfdlib, etc.) most of it takes an extremly high level of knowledge to use and there is no money for support, documentation. In comment, John C. is absolutly right here, without that book of line by line documentation it is pretty useless!

2. Many of the initially free efforts then lead into a higher or later level that costs. Such examples are DIFFpack, PHOENICS, etc. The free level is really only there to allow potential users to learn and "get hooked" on the code.

In conclusion, I think that providing some open access code is a great idea for a few people, including your company if it can get some users, and those who want to get in the guts of a code and try something new. But the majority of the market will still be filled by high $$, high service companies providing commercial code.

My thoughts....................................Duane
  Reply With Quote

Old   March 5, 1999, 05:55
Default Re: public CFD Code development
  #34
Gassan Abdoulaev
Guest
 
Posts: n/a
I do agree with almost every word!

Gassan
  Reply With Quote

Old   March 5, 1999, 09:21
Default Re: public CFD Code development
  #35
Heinz Wilkening
Guest
 
Posts: n/a
Ciao Duane,

thank you very much for the summary you wrote!

As you can expect I can not agree with everything.

You along with Gassan prefer commercial CFD. Your main reason for this is, that you think you will get better support for such a software. But is this support realy so good as you say? Often people had posted specific question about a commercial CFD-Code to this Forum in the past. Why they did not just ask the user support by the company? This is a little bit the strong side, but the truth is alway half way between.

It is correct a public CFD code would also need alot of compromises between thoose who are interested. This might be one major Problem with it. ( There are some programmers who have changed there partners but still using their old line editor! )

Also your arguement about the money is not alway true! The more important is, what is the quality of the product, not the price. You may be right, that there is often is a strong relation between price and quality, but not always, which is often like this for free software. An Example a few years ago I used alot Sun-Sparc 10 Workstation. At least at that time the GNU C compiler produced in many cases better executebles than the SUN C compiler. At least for the programms I wrote. The GNU compiler is for free!

I know there is no final answer like this, except the world also in CFD is big enouth for different ideas and concepts! There is LINUX and at the same time Windows. I think both are usefull, and also doing well.

Ciao Heinz

  Reply With Quote

Old   March 5, 1999, 10:56
Default Re: public CFD Code development
  #36
John C. Chien
Guest
 
Posts: n/a
There are many reasons why software is made available to the public for free. (definitely, it is not a good idea for a student somewhere in a jungle to put it into his thesis.) So, the major advantage of having a piece of free code is to be able to learn the unique idea inside the code. If a code is defined as a tool, then it is up to the author to determine whether he want to collect the fee or not. Some religious books are available to the public for free. It may be possible to define technologies used in CFD as tools, such as certain ways to generate mesh, graphic presentation of results, tri-diagonal matrix solver, etc.. These could, and should be placed inside the tool box. LINUX like any other operating systems is a collection of these tools to handle the file, disk,I/O, screen,mail,editor,network,etc.. Along this line, CFD tools would be of great help to the users when it is available for free. Otherwise, the CFD code is just like a story book, for a user to read it, he must understand the language and the meaning of the code. At this point, it is hard to make any comment without seeing the code itself. Are there published papers related to the code available on internet? In this way, one can learn something about the method, the code, the results from the papers.
  Reply With Quote

Old   March 5, 1999, 11:21
Default Re: public CFD Code development, comments ..
  #37
Doru
Guest
 
Posts: n/a
Dear Mr. Baker, I want to give some comments on the message you posted.

First, as it is normal, peoples opinions can be (and it is better to be!) different. In the subject in concern, there are some, used to run commercial CFD codes (as is your case, I understand), and there are some others willing to run, develop or simply modify a free CFD code, for their own needs. There is nothing wrong in this, if you agree that it is not the code (even a titled commercial CFD code) of main importance but it is the CFD expert who is important, in order to get a valid result from the simulations.

IT IS THE CFD EXPERT NOT THE CODE WHO FINALLY DECIDES THE QUALITY OF THE RESULTS!

You wrote:  "First of all I would like to categorize the types of responses. Type 1: The coder with an unpublished/unused code. This is much like the unpublished artist syndrome. Everyone has their own specific preferences (probably what they coded the first time or for grad-school, etc.) and wants the whole world to adjust to it!"

They have the unpublished artist syndrome?! 

Just because they developed something by them self or because they have their own ideas about a subject?! 

..

" Type 2. The under-resourced CFDer (probably a student) This group of people wrote in saying..."That is a great idea, sign me up!" or "I would love to do CFD, but I don't have 30k/yr, etc.".

Because they like this idea, you call them "under-resourced CFDer (probably a student)"?! 

Yes, a student!

" In fact, I want to pay for everything in a code because then, I know that I have a say in the quality and service! "

Paying a lot for the license of a CFD code isn't a guaranty of the quality of the simulations. Or for the service or the support they provided! Please see the numerous questions, posted here by peoples running commercial codes!

Second,

Yes, peoples have their own preferences and it is good to discuss about it. This is a discussion forum! Maybe not always the questions or the answers have been of the highest quality. I agree! Not everyone is an expert! But, it is not wrong in not knowing something, it is wrong in not trying to learn about it! And peoples posting here are willing to learn from the others experience! Or to help the others! Is there anyone here claiming to know everything in the CFD field? If it is, this place must be too boring for him!

The fact of running a commercial CFD code, or getting support for it, didn't made anyone and expert!

ANYWAY, IT IS DANGEROUS TO USE A CFD CODE (COMMERCIAL OR NOT) NOT KNOWING ANYTHING ABOUT CFD!

Instead of conclusion:

What is wrong to have also commercial codes and a free CFD code?

And, why should be somebody angry about it?!

With all my respect!

Doru
  Reply With Quote

Old   March 5, 1999, 12:31
Default Re: public CFD Code development, comments ..
  #38
Roland Bender
Guest
 
Posts: n/a
Hello

i ask myself why some people argue against the idea of this project?

Perhabs because they are from commercial developers and are frightend about their jobs?
  Reply With Quote

Old   March 5, 1999, 12:44
Default Re: public CFD Code development, comments ..
  #39
John C. Chien
Guest
 
Posts: n/a
I like your comments! At least, I know that I am not talking to the computer screen. And that is very important. I just want to add some back ground information. The cut back in defense and aerospace industries in 90's has a tremendous impact on the emerging CFD technology. The trend created in 70's and 80's for the CFD was suddenly disrupted. This provided opportunities for some experienced engineers to start or expand their CFD business. Companies without permanent staff of CFD experts still need their problems solved. The use of commercial CFD codes fit this market needs. The user-friendly, the automatic mesh generation, the push button GUI features are all very attractive to companies without CFD experts. So, they hired the CFD codes ( which is much cheaper than the salary of an expert ) instead of the CFD expert. And the strong service support is one key to the successful renew of license agreement. In reality, the quality of support varies from company to company, from branch to branch, just like any car dealers. At least in this cost cutting approach, on the surface, the company can always have a working code for their non-CFD engineer to use. so it becomes a common practice for a company to hire an engineer with one year ( or even zero year) of experience in using a commercial code. ( sure, this is o.k. because it would reduce the training period) This approach will reduce the cost, but the quality of the solution is normally unknown ( at least , not consistent.) At the same time, if the commercial CFD company is not getting enough money from their sale, they are not going to invest time to improve the quality and reliability of the product. If every company is using the same commercial code, then in principle, the software company should have enough resources to improve the quality of the software. The problem is every one will be producing the identical car model.(by using the identical commercial code.) This is a story related to the commercial CFD code and the cost reduction approach of a company----the use of a commercial code will cut the cost. So the first thing a company does is to cut the CFD expert position and hire the CFD code. The story is just a background information. And I don't think it is related to the public domain CFD code issue. Sometimes, it maybe a good approach for the software company to give away a software for free. "free commercial CFD codes on internet?" At this stage of CFD development, I think most people would agree that " CFD code is not equal to CFD solution." But it is hard to rule out the thinking that " CFD code has the potential to be equal to CFD solution." or "CFD code sometimes is equal to CFD solution." The only sure way to success is to know all the negative aspect of the problem. Looking from this point of view, negative comments are always wellcome.
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Searching team members for CFD Code Development krikicha Main CFD Forum 6 August 20, 2006 11:24
Activities of national CFD code development? Dong-Kyu, Chung Main CFD Forum 1 July 6, 2000 09:12
CFD Design...The CFD Future John C. Chien Main CFD Forum 19 October 6, 1999 11:57
CFD Salary CFD Main CFD Forum 15 September 4, 1999 14:04
Since Last June John C. Chien Main CFD Forum 3 July 12, 1999 09:38


All times are GMT -4. The time now is 06:55.