Page 1 of 1

msmac3D 3D Simulation Speed Test Results

Posted: Sun Feb 16, 2014 5:03 pm
by MSI
Feb 2014: As we prepare for our release of our msmac3D program in the near future, we have recently begun actively running speed benchmark tests on msmac3D and the results are amazing!
The msmac3D simulations of vehicles colliding is approaching real-time with simple optimization of the code (our 2D simulation code has always been faster than realtime).
Next up:
  • Use of INTEL software tools to further improve the optimization and then onto parallelization! (parallelization should also dramatically increase the speed of our SMAC Automatic Iteration program which we are also in the process of preparing for release, both for the msmac2D and msmac3D programs)

    Special discounts and exclusive Beta participation will be offered in the next 30 to 60 days ONLY to CURRENT medit32/msmac clients.
    • If you are current, Thank you! We will be sending out emails in the next month.
    • If you aren't current and/or never updated to medit32, it's OK. We want you back!
      We have a very reasonable pathway to re-activating your McHenry Software license
Join us as we continue to advance the science of 3D accident reconstruction and 3D simulation.
We expect to publish preliminary benchmark msmac3D results and timings for some full-scale tests including the RICSAC tests.

Re: msmac3D 3D Simulation Speed Test Results

Posted: Wed Feb 19, 2014 6:34 pm
by MSI
Feb 19, 2014: Below is a chart of the Speed Factor comparison between msmac, our currently distributed 2D smac program with our soon to be released msmac3D.
  • We used the inputs for msmac validation runs from our RICSAC97 paper
  • For msmac3D we first ran it in '2D' mode to insure the results are the same as the 'gold standard' msmac 2D program results
  • We then ran msmac3D in 3D mode to quantify the increase in execution time to simulate the vehicle responses in 3D.
    • To run and msmac3D Simulation in 3D only requires 3 additional inputs for each simulated vehicle: CG height, Bumper height and Overall height.
    • The program automatically approximates all the required 3D simulation inputs for the vehicles.
    • Any and all 3D simulation inputs (suspension properties, tire properties, etc) can be user input to provide an override of the defaults
For the speed test we calculated a Speed Factor:
  • Factor is calculated by dividing the Simulation Time by the time required to execute a simulation (calculation time).
    So for example, a 5 second collision simulation run:
    • if it takes 5 second actual time to run msmac or msmac3D the Factor would be 5/5 = 1 which represents REAL TIME
      • if it takes 1 second actual time to run msmac or msmac3D the Factor would be 5/1 = 5 which represents 5 times faster than REAL TIME
      • if it takes 0.25 second actual time to run msmac or msmac3D the Factor would be 5/0.25 = 20 which represents 20 times faster than REAL TIME
    • Note that the range for the McHenry programs msmac and msmac3D are between 0.50 and 34.6!
    • All tests include a collision simulation.
    • the first 12 simulations are the RICSAC full-scale tests numbers 1 through 12, run 13 is a simulation of a NYSTARS full scale crash test
    • We'll add more tests soon
    benchmark1.jpg (46.27 KiB) Viewed 1908 times
    From the chart you'll notice the following:
    • msmac as distributed runs at greater than 5.0 to 30.0 times real time which means a 5 sec collision simulation takes less than 1 sec.
    • msmac3D running in 2D mode runs at comparably fast calculation times and with the same results as msmac
    • msmac3D in 3D mode, which includes the concurrent running of 2 mHVOSM 3D simulations for the colliding vehicles, currently runs at a factor of approximately 0.50 or 50% of real time, meaning a 5 second collision simulation takes approximately 10 seconds
In the next few weeks we will be working on making the execution speed of msmac3D faster through a 'tune up' us using optimizations and tuning tools so that we will get our msmac3D 3D collision simulation program running at real time or better.
  • Our Current McHenry Software licensees are used to fast executions times for our 2D msmac program
  • Speed will also help in our implementation of the automatic iteration of msmac3D which is to be an extension of our SMACITER program (both of these programs we also expect to release in the next 30 to 60 days)
This benchmark is our "starting point".
Our simulation programs run faster than any other 3D simulation program in the reconstruction community
We also are insuring that ALL our program results are accurate and valid and will not change with optimization