Vulkan Grass Rendering

University of Pennsylvania, CIS 5650: GPU Programming and Architecture, Project 5

• Michael Rabbitz
  • LinkedIn
• Tested on: Windows 10, i7-9750H @ 2.60GHz 32GB, RTX 2060 6GB (Personal)

Part 1: Introduction

This project is an implementation of techniques described in Responsive Real-Time Grass Rendering for General 3D Scenes, using Vulkan to build an efficient grass simulator and renderer. Grass blades are represented as Bezier curves, with a compute shader handling physics and culling processes, while graphics shaders manage rendering. The goal is to achieve realistic, performance-efficient grass rendering suitable for real-time applications.

The base code includes a basic Vulkan setup with a compute pipeline and two graphics pipelines. This implementation focuses on developing shaders for the grass compute and graphics pipelines, along with custom descriptor bindings necessary to manage data between these pipelines.

Part 2: Simulating Forces

To create realistic movement, we simulate environmental forces on each grass blade using a compute shader.

Before we get into the simulated forces, here is an image showing the static, initial setup of the grass blades with no forces applied.

Gravity

The first simulated force is gravity, which is an application of the Earth's downward gravitational force at 9.81 m/s². As we can see, without a counter force, the blades are flattened to the ground plane.

Gravity

Recovery

Next, we add the second simulated force, recovery, which counteracts gravity and returns the blades to equilibrium, following Hooke's law as derived in the paper.

Gravity + Recovery

Wind

Finally, we add the third simulated force, wind, by implementing a custom wind function to influence the grass, considering the alignment of the blades with the wind direction.

The arbitrary wind function we use is: WIND_INTENSITY * vec3(cos(totalTime), 0.0, sin(totalTime)) * directional_alignment * height_ratio

Gravity + Recovery + Wind

Part 3: Culling Tests

To optimize performance, various culling methods are implemented to avoid rendering blades that don’t contribute to the final image:

Orientation Culling

In this technique, blades near-perpendicular or perpendicular to the camera are culled to prevent rendering artifacts.

View-Frustum Culling

In this technique, blades outside the view-frustum are excluded based on visibility tests for each Bezier curve.

Distance Culling

In this technique, blades beyond a certain distance are culled in buckets, with more distant blades culled more aggressively.

Part 4: Performance Analysis

• Frames Per Second (FPS) is the measurment of performance in this section. FPS is measured using a GLFW timer within the main loop.
• The Test Scene is positioned to render many grass blades as we increase the count, and to apply all three culling options.

Test Scene

Runtime vs Blade Count

• Culling ON refers to when all three culling options are enabled.
• Culling OFF refers to when all three culling options are disabled.

Blade Count	Culling OFF (FPS)	Culling ON (FPS)
210	1235	1245
212	1145	1230
214	565	1130
216	195	545
218	62	180
220	17	60
222	4	16
224	1	4

Observations

• Trend: As the blade count increases, FPS decreases significantly in both cases, but with culling enabled, FPS remains higher.
• Culling Efficiency: At higher blade counts, culling's impact on performance becomes much more noticeable, maintaining playable FPS even as the blade count reaches 2²⁰.
• Culling Overhead: There’s minimal overhead for culling at lower blade counts, as FPS differences between Culling ON and OFF remain small.

Runtime vs Culling Options

• Blade Count is 2¹⁶ for the following tests.

Culling Option(s)	FPS
Culling OFF	195
Orientation	225
View Frustrum	200
Distance	500
Orientation + View Frustrum	215
Orientation + Distance	530
View Frustrum + Distance	505
All Culling	545

Observations

• Distance Culling Effectiveness: Distance culling alone boosts FPS by over 2x, suggesting that rendering fewer distant blades is the most impactful optimization.
• All Culling: Using all three techniques provides a slight additional FPS boost over distance culling alone, though the improvement is minor, indicating diminishing returns with combined techniques.