search

QF-Error-Tracking / QFVD5

0 stars 0 forks source link

TEST_FOR_DAVID-FDRS_QF_Simulation_Time #23

Closed zacharycope0 closed 1 year ago

zacharycope0 commented 1 year ago

TEST:FDRS_QF_Simulation_Time David, As we discussed in the FDRS meeting today, I would like to test the following simulation to see if there are noticeable differences in simulation times across machines/why the QF sims are taking so long on Matt Jolly's Server. I cut the simulation times 600s and have them running on 1 thread. Please run with QF_523.

I'm currently running the tests on MJs server. I will report the run times in the comments.

To Reproduce Test 1 (Low Wind & High Fuel): https://drive.google.com/file/d/12dbuPQ84XLi_O2ujBN9ISIlF6oy6tibD/view?usp=drive_link Test 2 (High Wind & Low Fuel): https://drive.google.com/file/d/1jXwZfJ7ypoBTlGJuwAZ275zVK8rmXONl/view?usp=drive_link

zacharycope0 commented 1 year ago

As David and I discussed in the QF meeting, it seems like these simulations are taking a long time because of the long fireline and QF's plume protocol. I do not believe the long run times have to do with the server I'm running them on.

@drobinson6045. Can you confirm that your runtimes are similar to mine?

  | Scenarios |   -- | -- | --   | Low Wind & High Fuel | High Wind & Low Fuel |   USFS Workstation (1-Thread) |   |   |   Simulation time (s) | 75345 | 89778 |   Sub time fire (s) | 408 | 622 |   Sub time plumes (s) | 62812 | 77081 |   Sub time mass-consistency (s) | 11805 | 11745 |   Bigfire Server (1-Thread) |   |   |   Simulation time (s) | 64300 | 79972 |   Sub time fire (s) | 459 | 608 |   Sub time plumes (s) | 53852 | 69419 |   Sub time mass-consistency (s) | 9755 | 9681 |   Bigfire Server (8-Thread) |   |   |   Simulation time (s) | 20380 | 20080 |   Sub time fire (s) | 1630 | 1520 |   Sub time plumes (s) | 16383 | 16234 |   Sub time mass-consistency (s) | 2133 | 2069 |  
drobinson6045 commented 1 year ago

Definitely the plumes are the slowdown in this case. We have some ideas to improve the scaling of the algorithm and I hope that it will get addressed sooner than later. We have very similar timings:

<html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40">

  | Scenarios |   -- | -- | --   | Low Wind & High Fuel | High Wind & Low Fuel MacbookPro (1-Thread) |   |   Simulation time (s) | 64038 | 63185 Sub time fire (s) | 280 | 381 Sub time plumes (s) | 59237 | 58621 Sub time mass-consistency (s) | 4376 | 4016