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

Solver Precision in CFX-5.

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

Reply
 
LinkBack Thread Tools Display Modes
Old   August 14, 2001, 16:55
Default Solver Precision in CFX-5.
  #1
cfd guy
Guest
 
Posts: n/a
Hi people,
I know it sounds a weird question, but how do I know if I'm running in single or double precision in the CFX-5? I've already searched the manual, but I found nothing it could help me. Where I can change it? In the Solver Manager? Build 5?
Regards,
cfd guy
  Reply With Quote

Old   August 14, 2001, 20:24
Default Re: Solver Precision in CFX-5.
  #2
Dan Williams
Guest
 
Posts: n/a
The problem is that no double recision version of the CFX-5 solver has been released yet. There will be when CFX-5.5 is released this fall.

What sort of problem are you doing that you need double precision? Certainly the coefficient matrix does not benifit much from double precision. Geometric calculations may be another story depending on the grid that you are using.

Dan.
  Reply With Quote

Old   August 14, 2001, 20:39
Default Re: Solver Precision in CFX-5.
  #3
John C. Chien
Guest
 
Posts: n/a
(1). I had spent a lot of time trying to identify the source of error when using CFX-TASCflow in 3-D turbine applications some times ago. (2). I was convinced that the error was related to the single precision math used in the code. The double-precision version was not available then. My application required fine mesh and low Re turbulence model (two-layer model was available and was used). (3). This limited the Y+ values and the Reynolds number one can compute. (4). My feeling is: in the professional field, the double-precision math is required, for the solution to be possible or even useful. ( I had similar bad experience with an in-house cfd code, which uses single precision math in I/O and each time the calculation do a restart read, it change the accuracy of the computed variables. To be in the CFD field, you must be a professional programmer and mathematician first. It is just a waste of time when you are creating garbages all the time.)
  Reply With Quote

Old   August 15, 2001, 00:50
Default Re: Solver Precision in CFX-5.
  #4
Dan Williams
Guest
 
Posts: n/a
As you say, double precision math may be necessary, but it really only depends on the grid. With hexahedral meshes used in CFX-TASCflow, when the elements have large aspect ratios then double precision may be necessary. It depends on the application. Double precision math only matters when accuracy to the 5th or 6th significant figure becomes important.

It is not clear that the coefficient matrix benefits from double precision as the coeffients all come from a linearisation of the various non-linear terms anyways. And yes, you are correct, a double precision code should really restart in double precision.

Dan.

  Reply With Quote

Old   August 15, 2001, 10:15
Default Re: Solver Precision in CFX-5.
  #5
cfd guy
Guest
 
Posts: n/a
Thanks Dan,
Actually, I was just curious about it. Right now, I'm using hexa meshes and I didn't have any convergence problems yet.
Regards,
cfd guy
  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
thobois class engineTopoChangerMesh error Peter_600 OpenFOAM 4 August 2, 2014 09:52
How to start CFX Solver with an existing parfile lentschi CFX 3 February 20, 2011 14:31
CFX 5.7.1 PRE and solver won't start daniel CFX 1 January 20, 2006 11:09
amd athlon64 cpu and cfx 5.7.1 solver cmete CFX 3 October 20, 2005 22:16
compressible two phase flow in CFX4.4 youngan CFX 0 July 1, 2003 23:32


All times are GMT -4. The time now is 23:38.