Should You Use a Single Buffer or Multiple Buffers for Frame Uniform Descriptors in Vulkan?

Explore the benefits and drawbacks of using a `single buffer` versus `multiple buffers` for frame uniform descriptors in Vulkan applications. Find out which method is best suited for your needs!
---
This video is based on the question https://stackoverflow.com/q/72589545/ asked by the user 'CLearner' ( https://stackoverflow.com/u/19035179/ ) and on the answer https://stackoverflow.com/a/72620874/ provided by the user 'Varrak' ( https://stackoverflow.com/u/8519332/ ) at 'Stack Overflow' website. Thanks to these great users and Stackexchange community for their contributions.

Visit these links for original content and any more details, such as alternate solutions, latest updates/developments on topic, comments, revision history etc. For example, the original title of the Question was: Is it better to store uniform descriptors in a single buffer or use seperate buffers for each frame?

Also, Content (except music) licensed under CC BY-SA https://meta.stackexchange.com/help/l...
The original Question post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license, and the original Answer post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license.

If anything seems off to you, please feel free to write me at vlogize [AT] gmail [DOT] com.
---
Understanding Uniform Descriptors in Vulkan

When developing applications with Vulkan, one common question arises: Is it better to store uniform descriptors in a single buffer or use separate buffers for each frame? This query often concerns developers looking to optimize the efficiency and performance of their render pipeline, especially in applications where frames may overlap during rendering. Let's dissect this problem and explore the most effective solution.

The Challenge

In Vulkan, when an application is designed to handle multiple frames simultaneously (in your case, a maximum of 2 frames in flight), the management of memory and resources becomes crucial. You may find yourself debating whether to allocate separate buffers for each frame's uniform data or create a single buffer that encompasses the uniforms of both frames, using offsets for updates. This decision significantly impacts your application's performance and complexity.

Evaluating Your Options

Option 1: Separate Buffers for Each Frame

Allocating individual buffers for each frame ensures that each frame has access to its uniform data without conflicting with the data of other frames. This method can simplify synchronization and data management since each frame operates in isolation.

Pros:

Simpler synchronization: Each buffer is dedicated and does not overlap with others.

Avoids the risk of overwriting data still in use by the GPU.

Cons:

Requires multiple memory allocations, which can increase overhead and complexity in memory management.

May lead to inefficient use of memory if not all frames are in use simultaneously.

Option 2: Single Buffer with Offsets

Creating a single buffer that contains the uniform data for both frames and using offsets to access the respective data can be advantageous in specific scenarios. This approach is deeply rooted in the flexibility that Vulkan provides, allowing developers to maintain tight control over resource usage.

Pros:

Reduces memory allocation overhead by consolidating resources into a single buffer.

Can be more efficient if managed correctly, as it may lead to fewer memory pages being allocated.

Cons:

Requires careful management to prevent overwriting data currently being used by the GPU.

Increases complexity in the implementation as developers need to track offsets and ensure that they are not accessing stale data.

A Practical Approach

The best solution depends on the specific use case of your application. As someone with experience in this field, consider the following strategy for managing uniform descriptors efficiently:

Use of Page Allocations

In my own applications, I leverage a page allocation model similar to a heap.

On-Demand Allocation: Allocate memory as needed for uniform buffers that infrequently change.

Block Recycling: Once the referenced data is no longer active, it is aged out over the number of frames and then freed appropriately.

This method balances memory management and performance, especially if your uniform descriptors are stable across frames.

Exploring Push Constants

For per-draw uniform data which changes frequently, I recommend using push constants. This method allows you to send uniform data to the shader directly without the need for creating separate buffers.

Efficiency: Push constants are lightweight and simple for data that is “used once then discarded”.

Performance Boost: By minimizing buffer overhead and complexity, this method may lead to improved performance.

Conclusion

When faced with the decision of using a single buffer versus multiple, there is no one-size-fits-all answer. It ultimately relies on the specific r