CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLOW-3D (
-   -   single - double precision (

PC October 10, 2008 10:21

single - double precision
Can anyone comment on loss of accuracy (or lack of) using single precision instead of double precision in high pressure castings?

Michael Barkhudarov October 11, 2008 13:01

Re: single - double precision
In my experience the difference in the results between single and double precision versions are typically minute.

There are some cases, when the difference are significant, but not in casting applications.

In fact, the main reason for the existence of the single precision version of FLOW-3D are the die casting applications. Ten years ago, these were, and still often are, the most computationally and memory intensive problems, often reaching the 2 Gb limit of the addressable memory on 32-bit operating systems. It was more important to be able to mesh a part than maintain the doulbe precision accuracy. Of course, if you can comfortably fit the double precision model on your machine, then you should do so.

CK October 14, 2008 04:39

Re: single - double precision
Hi please advice if any significant differences in simulation time between single vs double precison version?

Rgds CK

Michael Barkhudarov October 16, 2008 09:20

Re: single - double precision
Hi Ck,

Not typically. AS far as I understand, the single precision version may run a bit faster (~10%) because it uses less memory, therefore, it has to access memory and refresh cache less often. If this is true, the difference in time will be hardware-dependent.

Of course, if the double precision code uses so much memory that it has to swap some to disk, then the difference in CPU time wil be greater.

CK October 16, 2008 19:28

Re: single - double precision
Hi Michael Thanks a lot for the input.

Rgds CK

All times are GMT -4. The time now is 10:18.