neb0lle / slimesimulation Goto Github PK

Add support for multiple species with interaction capabilities b/w them.

• Add support for multiple species with different SpeciesSettings for each.
• Needs converging and diverging interaction methods for different species.
• Interaction based on SpeciesSettings.color.

keeps crashing on linux

Issue

• The FPS in the current setup is really low ~ (10 FPS/ 10000 Agents).
• Even for smaller sample sizes its still low.

Causes

• Most of the math is done on the cpu.
• The whole simulation process is single-threaded.
• DDTexture() produces a huge performance hit ie; RES * RES iterations.

Possible fixes

• Use Compute shaders for agent and *pixels array member calculations. (recommended)
• Optimizing DDTexture() by using a shader.
• Faster math?

Problem

The *pixels array allocated in render.cpp never gets freed.

Using freeParray() produces a double free error.

Maybe the UnloadTexture frees the array bcz it loads texture using *pixels data. (Not sure)

Steps to reproduce

1. Use freeParray() at the end of the main simulation loop.
2. Close the simulation window.
3. Leak?

INFO: TEXTURE: [ID 2] Texture loaded successfully (128x128 | GRAY_ALPHA | 1 mipmaps)
INFO: TEXTURE: [ID 3] Texture loaded successfully (1000x1000 | R8G8B8A8 | 1 mipmaps)
INFO: TEXTURE: [ID 3] Unloaded texture data from VRAM (GPU)
slimesimulation(58977,0x10457c580) malloc: *** error for object 0x140008000: pointer being freed was not allocated
slimesimulation(58977,0x10457c580) malloc: *** set a breakpoint in malloc_error_break to debug

Expected behaviour

No memory leak or double free error.

Specifications

raylib version: 4.5.0
OS: macOS 12.6

Add Z axis index in SpeciesSettings

• Each species should be rendered on the screen based on Z index.
• Z index needs to be added as a new member in SpeciesSettings.

Add better bounds detection in Clamp().

• The current clamp function has issues with agents being rendered out of bounds.
• At the bounds the texture gets saturated and agents accumulate.
• agent.angle recalculation using InvResolveAngle() is not accurate at bounds collision.
• Add feature in Clamp() for random angle agent redirection at bounds collision.

Option to generate agents based on config file.

• Agent generation on simulation start with parameters and gen-method.
• Periodic agent generation based on intervals and other params.
• Periodic agent decay (Not to be confused with decayRate in DDTexture).

Problem

The texture produces an artifact right on initialization but not always.

Randomly comes and goes but never occurs when no agents are initialized.

Bug only exists on macOS. (only M1 tested)

The artifact may be generated from either the texture or pixelarray. (not sure)

Havnt tested on other x86 maybe issue wont persist there ig.

Steps to reproduce

Run the simulation a few times till u get the artifact.

Specifications

raylib version: 4.5.0
OS: macOS 12.6