AudioKit能否设置非2的幂的bufferLength？限制来源咨询

Great question! Let's break this down into two clear parts to answer your concerns fully.

1. Yes, you can use non-power-of-two buffer lengths (with important caveats) #

The AKSettings.BufferLength enum is just a convenience wrapper AudioKit provides—it doesn’t lock you into only power-of-two sizes. To use a custom buffer length that isn’t a power of two, you can bypass the enum and directly interact with the underlying AVAudioSession API before starting AudioKit.

Here’s a quick example of how to do this:

• First, calculate your desired buffer duration using your target frame count and the audio session’s sample rate. For instance, if you want 768 frames at 44.1kHz, the duration would be 768 / 44100 ≈ 0.01741 seconds.
• Then, call the system’s buffer duration method and handle potential errors:

do {
    let sampleRate = 44100.0
    let desiredFrameCount = 768
    let targetDuration = Double(desiredFrameCount) / sampleRate
    
    try AVAudioSession.sharedInstance().setPreferredIOBufferDuration(targetDuration)
    try AudioKit.start()
} catch {
    print("Failed to configure buffer: \(error.localizedDescription)")
}

A few things to keep in mind:

• Not all devices support arbitrary buffer lengths. The system will automatically adjust your requested duration to the closest supported value, so you might not get exactly the frame count you asked for.
• Some AudioKit nodes (especially those relying on FFT or block-based processing) are optimized for power-of-two buffers. Using non-power-of-two sizes could lead to reduced performance or unexpected behavior in these cases—always test thoroughly if you go this route.

2. The power-of-two restriction is an AudioKit choice, not a system requirement #

AVAudioSession does not force buffer lengths to be powers of two. It only requires that the buffer duration falls within a device-specific range (usually between ~0.000277s and 0.5s, depending on hardware capabilities).

AudioKit’s AKSettings.BufferLength enum only includes power-of-two options for two key reasons:

• Performance & Compatibility: Most audio processing algorithms (like FFT effects, reverbs, or analyzers) run most efficiently with power-of-two buffer sizes. Providing these as default options ensures smooth performance for the majority of use cases.
• API Simplicity: Limiting pre-defined options to common power-of-two sizes makes the API easier to navigate for developers who don’t need custom buffer configurations, avoiding confusion around valid duration values.

内容的提问来源于stack exchange，提问作者jubl