Crate bit_mask_ring_buf

A fast ring buffer implementation with cheap and safe indexing written in Rust. It works by bit-masking an integer index to get the corresponding index in an array/vec whose length is a power of 2. This is best used when indexing the buffer with an isize value. Copies/reads with slices are implemented with memcpy. This is most useful for high performance algorithms such as audio DSP.

This crate has no consumer/producer logic, and is meant to be used as a raw data structure or a base for other data structures.

If your use case needs a buffer with a length that is not a power of 2, and the performance of indexing individual elements one at a time does not matter, then take a look at my crate slice_ring_buf.

Installation

Add bit_mask_ring_buf as a dependency in your Cargo.toml:

bit_mask_ring_buf = 0.5

Example

use bit_mask_ring_buf::{BMRingBuf, BMRingBufRef};

// Create a ring buffer with type u32. The data will be
// initialized with the default value (0 in this case).
// The actual length will be set to the next highest
// power of 2 if the given length is not already
// a power of 2.
let mut rb = BMRingBuf::<u32>::from_len(3);
assert_eq!(rb.len(), 4);

// Read/write to buffer by indexing with an `isize`.
rb[0] = 0;
rb[1] = 1;
rb[2] = 2;
rb[3] = 3;

// Cheaply wrap when reading/writing outside of bounds.
assert_eq!(rb[-1], 3);
assert_eq!(rb[10], 2);

// Memcpy into slices at arbitrary `isize` indexes
// and length.
let mut read_buffer = [0u32; 7];
rb.read_into(&mut read_buffer, 2);
assert_eq!(read_buffer, [2, 3, 0, 1, 2, 3, 0]);

// Memcpy data from a slice into the ring buffer at
// arbitrary `isize` indexes. Earlier data will not be
// copied if it will be overwritten by newer data,
// avoiding unecessary memcpy's. The correct placement
// of the newer data will still be preserved.
rb.write_latest(&[0, 2, 3, 4, 1], 0);
assert_eq!(rb[0], 1);
assert_eq!(rb[1], 2);
assert_eq!(rb[2], 3);
assert_eq!(rb[3], 4);

// Read/write by retrieving slices directly.
let (s1, s2) = rb.as_slices_len(1, 4);
assert_eq!(s1, &[2, 3, 4]);
assert_eq!(s2, &[1]);

// Aligned/stack data may also be used.
let mut stack_data = [0u32, 1, 2, 3];
let mut rb_ref = BMRingBufRef::new(&mut stack_data);
rb_ref[-4] = 5;
assert_eq!(rb_ref[0], 5);
assert_eq!(rb_ref[1], 1);
assert_eq!(rb_ref[2], 2);
assert_eq!(rb_ref[3], 3);

// Get linear interpolation on floating point buffers.
let mut rb = BMRingBuf::<f64>::from_len(4);
rb[0] = 0.0;
rb[1] = 2.0;
rb[2] = 4.0;
rb[3] = 6.0;
assert!((rb.lin_interp_f64(1.0) - 2.0).abs() <= f64::EPSILON);
assert!((rb.lin_interp_f64(1.25) - 2.5).abs() <= f64::EPSILON);
assert!((rb.lin_interp_f64(3.75) - 1.5).abs() <= f64::EPSILON);

Structs

BMRingBuf

A fast ring buffer implementation with cheap and safe indexing. It works by bit-masking an integer index to get the corresponding index in an array/vec whose length is a power of 2. This is best used when indexing the buffer with an isize value. Copies/reads with slices are implemented with memcpy.

BMRingBufRef

A fast ring buffer implementation with cheap and safe indexing. It works by bit-masking an integer index to get the corresponding index in an array/vec whose length is a power of 2. This is best used when indexing the buffer with an isize value. Copies/reads with slices are implemented with memcpy. This works the same as BMRingBuf except it uses a reference as its data source instead of an internal Vec.

Functions

next_pow_of_2

Returns the next highest power of 2 if n is not already a power of 2. This will return 2 if n < 2.