Replies: 2 comments
-
Hello, the second question is addressed in the User's Guide. Look for the CHID_cpu.csv file, which is automatically generated for each simulation. It contains the CPU time spent on major FDS processes. As for the first question, let's wait for input from the developers. In my experience, it always depends on the specific case. A general principle is that computational cost should be balanced against the benefit it provides. For example, increasing the computational requirements by 50% to gain only a 2% improvement in accuracy is inadequate. |
Beta Was this translation helpful? Give feedback.
-
No detailed study. The time it takes to run the verification suite is monitored. If there are large jumps in that time, we would look to see if some specific code change caused that. The time FDS spends in the various major components of the code are very dependent on the input file. Typically the largest computational items for a typical fire protection design type simulation are MASS, DIVG, PRES, VELO, and RADI. MAIN, WALL, DUMP, and FIRE are generally small. COMM expense is tied to the number of meshes, how load balanced they are, and the communications fabric. Items like PART, HVAC, GEOM, VEGE, and CHEM depend on if you have specific inputs related to those present or not. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi FDS development team,
I'm curious about the computational cost associated with various FDS entities, such as OBST, SURF, PART, and so on.
I have two questions:
Has the FDS team ever conducted internal studies or published reports analyzing the computational performance breakdown of different model components?
Is there currently a way in FDS to profile or extract the runtime contribution of different elements or modules?
For example, how much simulation time is spent handling particle dynamics, obstruction processing, or surface heat transfer?
If so, I’d be very interested in reading those or understanding the general findings.
I’d greatly appreciate any advice, or even informal insight, into how to measure or estimate the performance cost of these elements.
Beta Was this translation helpful? Give feedback.
All reactions