jklae / gridliquid2d Goto Github PK

Description
Currently, the drawing coordinates and the simulation coordinates are the same. It is not good because the smaller the grid scale, the larger the calculation error. Also, some coordinates are negative to center the grid position. These coordinates are very cumbersome in simulation calculations, especially when interpolating particle variables. Therefore, the simulation calculations will be more efficient by having the drawing coordinates and the simulation coordinates separately.

Todo

• Do not equate simulation coordinates with drawing coordinates.
• Normalize the simulation coordinates [0, N+2]. These coordinates make simulation calculations convenient.
• Remove variables like stride and offset. As long as the coordinates are normalized, these variables are not needed.
• Modify the interpolation function. In the interpolation function, the offset and stride of the particle are calculated. Since the coordinate system has changed, these must also be corrected.

Note
A method for compensating for the difference between the two coordinate systems is presented in jklae/DXViewer#25.
This issue is derived from #3.

Description
Hybrid methods PIC and FLIP are implemented.

Todo

• Implement PIC and FLIP.
• Modify iWMCreate() to support PIC/FLIP.

Note
PIC is already implemented in commit ab9ea96.

Description
Further implementations are underway to support the DXViewer GUI.

Todo

• Create a parent fluid simulation class.
• Respond to various functions added to DXViewer GUI.

Note
This issue is related to jklae/DXViewer/issues/27.

Description
Currently, center-type velocity storage is being used. Although this method is easy to implement, it is outdated and causes problems with boundary conditions. Therefore, MAC grid is adopted.

Todo

• Implement MAC grid. It should provide various functions using classes. (ex. getHalfVelocity())

Describe the bug
I just found a terrible bug. The particle creation location was wrong. (Rows and columns are reversed)
This will probably cause problems everywhere index function is used.

Screenshots

Description
Liquid simulation doesn't seem plausible, especially when falling due to gravity. It moves like gas.
The cause is probably because the gas method is being used as it is. Additional work is required for liquid simulation.

Comments
A few doubts are, first, that additional methods such as volume conservation or surface tension need to be implemented.
Second, I think that it is not necessary to update the entire velocity vector field, but only the vector field near the liquid.

The image on the right is the result of https://github.com/kbladin/Fluid_Simulation.

Description
A marker particles technique is implemented for Eulerian simulations.

Todo

• Define marker particle-related variables.
• Implement abstract methods so that DXViewer can use the above variables.

Note
This issue is related to jklae/DXViewer#21.

Describe the bug
A problem occurred while implementing subWndProc in ISimulation. ISimulation object was nullptr during initialization with subWndProc.

Screenshots

Comments on the bug
This problem occurs because the initialization of subWndProc precedes the initialization of the ISimulation object.
The initialization order of Win32App needs to be changed.

Description
The ultimate goal, Eulerian fluid, is realized.

Todo

• Implement the Eulerian steps, which are force, advect, diffuse, and project.
• Integrate eulerian fluid with marker particles.

Note
This implementation refers to https://github.com/frostsim/StableFluids.