User:Al Da Best/devlog

Google Summer of Code 2012


Status: Working on fluids system, starting with drag caused from falling object.

Gain Commit access ASAP
Work out how density is handled during simulation and add it to the mass/volume calculations in a more accurate fashion. WOrking out how to best calculate the coefficient of drag for objects.

Coding Environments Status:
VS2010x64 : Success, no bullet.
Fedora 17 KDE-Spin : Success, Bullet running.


April 2012

BRLCAD compiled and run with Windows Visual Studio 10 x64 Debug

May 2012

8th : Updated Wiki pages

9th-10th : Tried to compile BRLCAD with bullet on VS2010, consistently recieved 63/70+ errors for debug/release versions. Using linux as development environment until I have a chance to fix issues.

11th-12th : Setup a Fedora 16 environment with BRLCAD compiled and running simulations.

13th-16th : Testing various simulations, spent time looking through the simulation code to learn more about how it currently works and how to best implement new features.

24-25th : Reinstalling Windows on my SSD (From HDD), computer unavailable for a day 26th : Everything back to normal with PC, build times using VS2010 decreased from 3+ hours to 9 minutes (from clean)

27th-29th : Fixed issues with SA/Centroid functions for Ellipsoid Primitive patch

29th : Patch for Surface Area for Ellipsoid Primitive accepted
29th : Downloaded Fedora 17 and began to set it up for brlcad development

30th : Successfully built and installed brlcad on Fedora 17, as well as ran some small test simulations without issue.
30th : Working on a function to calculate the gravitational force between every object in the simulation, in order to be able to calculate the acceleration between them.

31st : Continued work on aforementioned function, testing tomorrow for patch submission.

June 2012

2nd : Submitted initial patch for implementation of gravity forces between rigid bodies.

4th : Added to patch with a small update for calculating mass using the bounding box of an object, and using that to find the volume, assuming a density of one. Density to be integrated better soon.

5th : Working on the resolving of acceleration due to gravitational force, mainly how to store the information since there will be n(n-1) number of forces which will then be condensed to n number of accelerations, where n is the number of rigid bodies.

6th-8th : Finished work on resolving the acceleration and begun working on how to integrate this into the current simulation, in terms of running the calculations every time and altering the linear_velocity values of the rigid bodies.

11-15th : Continue work on gravitational functions.

16th-17th : Looking at how better to implement mass and center of mass. Trying to learn how to utilise gqa.c and also whether it would be best to add to the rigid_body struct with volume, mass and center of mass (as opposed to center of BB) for use in gravity functions to avoid repeatedly recalculating it.

18th - 22nd : Continued work on gravitational function. Working out how to intergrate the changing angular and linear velocities into the bullet simulation, since the brlcad system would ideally recalculate this every step.

23rd - 26th : Finalising the gravity calculations and working on the fluid system. Need to work out suitable intervals for these calculations. 60 times per second would be a massive performance hit, even with the calculations themselves being relatively simple.

27th - 29th : Finishing off the gravity, videos coming soon (with highly exaggerated values to make it obvious, movement is negligible with small mass objects, or large distances).

30th : Continue working on fluid. Namely how to best teach the system to use the best coefficient of drag for a shape.

July 2012

1st - 2nd : Integrating into system with a similar method to gravity, alter the acceleration for the rigid body then use that to change the velocity every step.

3rd - 5th : Learning about how to go about creating video footage. Will need to save a render of every step of the simulation, I think. Also worked on testing the drag functions.

6th - 7th : Changed how mass is handled in the simulation, added new parameter to rigid_body structure to handle this. Temp using bb_vol assuming density is 1 to calculate mass, will be changed to something more precise soon. Also working on making sure the drag function gets the right area, and started to try and calculate any rotational forces around a rigid body

8th : Tested basic drag and gravity functions with rest of simulation and values output are correct with my calculations. Made progress with video. Still working on how to calculate the surface area for drag function, as well as how to get a reasonable coefficient of drag.

9th : Accidentally broke my fedora VM by running a simulation with 1 million steps as an experiment. Files were backed up, currently setting up another one, trying out the KDE distribution of fedora. Issue with building that I will fix tomorrow (fb.h, unknown types window, XVisualInfo, and a couple of others on line 136) and then I will debug the error. VM broke due to a full harddrive of error messages as far as I can tell. Error was somehow related to collision, from the little I could make out from the mass of error messages. Situation was a simple sphere dropping onto a surface.

10th : Wasted several hours getting a new VM working, had compile issues. Now working again, however. Came across an error in one of my functions whilst debugging, which involved NaN results. Working on finding the cause and will fix asap.

11th : Fixed NaN results, had a 0/0 division which I changed. Also refined code to better handle ground plane with respect to gravity from it (i.e ignore it). Gravity is still counted as 10 down, with the gravitational attraction seperate. Should be in position to finally make video tomorrow morning.

12th : Ran simulation and took raytraces at 50 step intervals. Looks like the groundplane moves, which isn't meant to happen. Will look into this and check it's not moving, and if it is stop it. Also changed use of gqa slightly, and implemented a, for now, debug usage rt, which I intend to make a user option later on, if they wish to automatically have the simulation saved as several (hundred/thousand) images.

Development Schedule

(Unless otherwise specified, 40+ hours of work to be assumed):

23rd April – 5th May

  • Flesh out any last details in project, have all objectives clear and the schedule complete.
  • Gain commit access before or during this point
  • Before the official start of coding, create basic, general (and some specific) unit tests for functions that will be implemented, such as testing the variable gravity, the resolving of drag forces, the moments and couples calculated, and so forth.

5th May – 21st May

  • Learn as much about any relevant libraries such as Bullet, and BRL-CAD libraries such as librt and libged as I can, to enable me to be able to get straight into coding with the knowledge of what is already there and how to use it.
  • Start writing function outlines, header files to make it clearer as to what goes where and how to reuse code efficiently.
  • Iron out any current bugs in the simulation files, to ensure a fully working simulation environment.

21st May – 28th May

  • Coding begins. Start by working on the gravitational system as this will be the easiest to implement and so will take the shortest amount of time, and will make sure I am completely familiar with the code.

28th May – 4th June

  • Milestone 1: Gravitational system complete and fully bug tested.
  • Once milestone 1 is reached, begin work on the fluid system, starting with functions to calculate drag on surfaces

4th June – 11th June

  • Exam period starts at University, limited time available.
  • In available time, work on drag functions and start to implement functions to resolve the forces from the drag (In x,y and z directions)

11th June – 18th June

  • Exam period ends
  • Carry on implementing functions to resolve forces and start on functions to resolve moment forces to show rotation in the simulation

18th June – 25th June

  • Continue working on the fluid system and begin bug testing
  • After 22nd June assume 50+ hours working, to make up for lost time during University term

25th June – 2nd July

  • Milestone 2: Fluid system complete and fully bug tested.
  • Start work on the collision system

2nd July – 9th July

  • Preparing for mid-term evaluations, fixing all current code up to this point which will be around two thirds of the whole project
  • Continue work on the collision system

9th July - 16th July

  • Continue work on the collision system

16th July – 23rd July

  • Milestone 3: Collision system completed
  • From this point forward, start black box testing to ensure everything works

23rd July – 30th July

  • Milestone 4: Black box testing complete
  • Start white box testing to ensure code is of a high standard

30th July – 3rd August

  • Milestone 5: White box testing complete
  • All following time is a buffer zone. If everything complete, begin work on adding more functionality to the fluid system, namely the underwater physics

3rd August – 17th August

  • I will be in Germany during this period on holiday. I will however, still have a laptop as well as internet access so although this is a critical time with respect to final evaluations and so on, there is no reason for me to not be able to work
  • Time spent cleaning up code if needed, write documentation to enable users to easily utilise any of the functionality of the simulation system.

17th August – 24th August

  • Any last minute fixes are made, final evaluations sorted and project finished.
  • Milestone 6: Project finish, submit evaluations.