8 releases (breaking)
| 0.7.0 | Dec 2, 2025 |
|---|---|
| 0.6.0 | Dec 2, 2025 |
| 0.5.0 | Dec 2, 2025 |
| 0.4.0 | Dec 2, 2025 |
| 0.1.1 | Dec 1, 2025 |
#244 in Data structures
Used in 2 crates
135KB
3K
SLoC
avila-atom
Atomic Computational Structures - High-performance fundamental data structures built from first principles.
Features
- Zero-cost abstractions - No runtime overhead vs manual implementation
- Memory efficiency - Contiguous allocation for optimal cache locality
- Compile-time optimization - Monomorphization enables aggressive inlining
- no_std compatible - Works in embedded and OS development contexts
Data Structures
Core Types
-
Option<T>- Optional value (presence/absence)- Zero-cost abstraction with enum-based representation
- Null pointer optimization for references
-
Result<T, E>- Result type (success/failure)- Tagged union for error handling
- Zero-cost compared to manual error codes
-
DynamicArray<T>- Growable contiguous array- O(1) amortized push with geometric growth
- 24 bytes header (ptr + len + capacity)
-
AssociativeArray<K, V>- Key-value storagestd: HashMap with O(1) average lookupno_std: BTreeMap with O(log n) lookup
-
StringBuffer- UTF-8 encoded string- Guaranteed valid UTF-8 at all times
- Efficient growable capacity
Advanced Types
-
FixedArray<T, N>- Stack-allocated fixed-size array- Zero heap allocation
- Compile-time size known
-
CacheAlignedArray<T, N>- Cache-line aligned array- 64-byte alignment prevents false sharing
- Optimal for multithreaded performance
-
SmallString- Small string optimization- Stores ≤23 bytes inline (no heap)
- Automatic heap promotion for larger strings
Performance Modules
-
perf- Branch prediction hints and alignmentlikely()/unlikely()- Branch prediction hintsassume_aligned()- Pointer alignment assertions
-
simd- SIMD-optimized operations (x86_64)- AVX2/AVX-512 feature detection
- Vectorized memory operations
- 2-4x faster for large buffers
-
DynamicArrayExt- Extended array operationsreserve_exact_fast()- Smart capacity managementextend_from_slice_fast()- Optimized slice extensionclear_and_resize()- Memory-efficient resizing
Memory Management (v0.4.0+)
-
arena::Arena- Bump allocator for batch allocations- O(1) allocation, O(1) bulk deallocation
- Zero per-object overhead
- Perfect for temporary allocations
-
pool::ObjectPool<T>- Object reuse without reallocation- Eliminates allocation overhead
- Per-thread pools for multithreading
- Automatic slot reuse
Concurrency (v0.4.0+)
-
lockfree::LockFreeStack<T>- Wait-free LIFO stack- No mutex overhead
- Compare-and-swap operations
- Thread-safe push/pop
-
lockfree::AtomicCounter- Cache-line padded counter- Prevents false sharing
- Lock-free increment/get/set
- 64-byte alignment
-
lockfree::RingBuffer<T, N>(v0.5.0+) - SPSC circular buffer- Single-producer single-consumer
- Const generic power-of-2 capacity
- Lock-free with Acquire/Release semantics
- Zero-copy inter-thread communication
Advanced Data Structures (v0.5.0+)
-
btree::BPlusTree<K, V>- Cache-optimized B+Tree- 16-way fanout for cache-line alignment
- Database-grade ordered storage
- O(log N) operations
-
robinhood::RobinHoodMap<K, V>- Robin Hood hash table- Linear probing with displacement tracking
- FNV-1a hash for excellent distribution
- Superior cache performance vs chaining
- Bounded probe length variance
Compression (v0.5.0+)
compress- Inline compression algorithmsrle_encode()/rle_decode()- Run-length encoding (2-10x for repetitive data)delta_encode()/delta_decode()- Delta encoding for sequential data- Zero heap allocations, in-place compression
Usage
Add to your Cargo.toml:```toml
[dependencies]
avila-atom = "0.7"
### Complete Data Structure Suite (v0.7.0+)
```rust
// Priority Queues
use avila_atom::heap::{MinHeap, MaxHeap};
let mut heap = MinHeap::new();
heap.push(5);
heap.push(2);
assert_eq!(heap.pop(), Some(2)); // Min element
// Graph Algorithms
use avila_atom::unionfind::UnionFind;
let mut uf = UnionFind::new(100);
uf.union(1, 2);
assert!(uf.connected(1, 2));
// LRU Cache
use avila_atom::lru::LruCache;
let mut cache = LruCache::new(100);
cache.put("key", "value");
assert_eq!(cache.get(&"key"), Some(&"value"));
// BitSet - 64x compression
use avila_atom::bitset::BitSet;
let mut bs = BitSet::new(10000);
bs.insert(42);
assert!(bs.contains(42));
assert_eq!(bs.count(), 1);
// Radix Sort - O(n) for integers
use avila_atom::sort::radix_sort_u32;
let mut arr = vec![170, 45, 75, 90, 802, 24, 2, 66];
radix_sort_u32(&mut arr);
// arr is now [2, 24, 45, 66, 75, 90, 170, 802]
// Segment Tree - Range queries
use avila_atom::segtree::SegmentTree;
let arr = vec![1, 3, 5, 7, 9];
let tree = SegmentTree::new(&arr, 0);
let sum = tree.query(1, 3); // Sum of arr[1..=3]
// Trie - Prefix matching
use avila_atom::trie::Trie;
let mut trie = Trie::new();
trie.insert("hello");
trie.insert("help");
assert!(trie.starts_with("hel"));
// MPMC Queue - Lock-free
use avila_atom::mpmc::MpmcQueue;
let queue = MpmcQueue::new(1024);
queue.push(42).unwrap();
assert_eq!(queue.pop(), Some(42));
// SmallVec - Stack optimization
use avila_atom::smallvec::SmallVec;
let mut sv: SmallVec<i32, 8> = SmallVec::new();
sv.push(1); // On stack (no allocation!)
// Sparse Set - O(1) operations
use avila_atom::sparseset::SparseSet;
let mut ss = SparseSet::new(1000);
ss.insert(42);
assert!(ss.contains(42));
// Matrix operations
use avila_atom::matrix::Matrix;
let mut m = Matrix::new(3, 3);
m.set(0, 0, 1.0);
let result = m.multiply(&m); // Matrix multiplication
// Slab Allocator
use avila_atom::slab::SlabAllocator;
let mut slab = SlabAllocator::new(1000);
let id = slab.allocate(42).unwrap();
assert_eq!(slab.get(id), Some(&42));
Revolutionary Features (v0.6.0+)
use avila_atom::skiplist::SkipList;
use avila_atom::bloom::BloomFilter;
use avila_atom::radix::RadixTree;
use avila_atom::cow::CowArray;
use avila_atom::numa::{NumaPool, NumaNode};
// Lock-free concurrent skip list
let mut skiplist: SkipList<i32, String> = SkipList::new();
skiplist.insert(42, "answer".to_string());
let exists = skiplist.contains(&42); // O(log N) concurrent
// Bloom filter - probabilistic set membership
let mut bloom: BloomFilter<&str> = BloomFilter::new(10000, 0.01); // 1% FPR
bloom.insert(&"user123");
if bloom.contains(&"user123") { // Fast negative lookups
// Might be present (or false positive)
}
// Radix tree for prefix matching
let mut radix = RadixTree::new();
radix.insert(b"192.168.1.0", "subnet_a");
radix.insert(b"192.168.2.0", "subnet_b");
let route = radix.get(b"192.168.1.0"); // O(k) where k = key length
// Copy-on-Write array for immutable sharing
let mut arr1: CowArray<i32> = CowArray::new();
arr1.push(1);
arr1.push(2);
let arr2 = arr1.clone(); // O(1) - shared until write
// NUMA-aware memory pool
let mut pool = NumaPool::new(2); // 2-socket system
pool.set_node(NumaNode(0)); // Bind to socket 0
pool.push(data); // Allocate on local NUMA node
Examples
use avila_atom::{DynamicArray, AssociativeArray, StringBuffer};
// Dynamic array with type inference
let mut numbers = DynamicArray::new();
numbers.push(1);
numbers.push(2);
numbers.push(3);
// Map with convenient macro
use avila_atom::map;
let config = map! {
"host" => "localhost",
"port" => "8080",
};
// UTF-8 string buffer
let mut text = StringBuffer::from("Hello, ");
text.push_str("Ávila!");
assert_eq!(text.len(), 14); // UTF-8 byte count
Macros
use avila_atom::{map, list, array};
// Map with capacity pre-allocation
let m = map! {
"key1" => "value1",
"key2" => "value2",
};
// Dynamic array
let v = list![1, 2, 3, 4, 5];
// Fixed-size stack array
let arr = array![10, 20, 30];
assert_eq!(arr.len(), 3);
Performance Optimizations
use avila_atom::{DynamicArray, DynamicArrayExt};
use avila_atom::fixed::CacheAlignedArray;
use avila_atom::perf::{likely, unlikely};
// Cache-aligned array for multithreading
let aligned = CacheAlignedArray::new([0u64; 8]);
// Smart capacity management
let mut v = DynamicArray::new();
v.reserve_exact_fast(1000); // Only allocates if needed
v.extend_from_slice_fast(&[1, 2, 3]);
// Branch prediction hints
fn process(value: i32) -> i32 {
if likely(value > 0) {
value * 2
} else {
0
}
}
SIMD Operations (x86_64)
#[cfg(target_arch = "x86_64")]
use avila_atom::simd::{has_avx2, has_avx512f, fast_copy};
#[cfg(target_arch = "x86_64")]
fn optimized_copy(dst: &mut [u8], src: &[u8]) {
if has_avx512f() {
unsafe {
fast_copy(
dst.as_mut_ptr(),
src.as_ptr(),
src.len()
);
}
} else {
dst.copy_from_slice(src);
}
}
Arena Allocator
use avila_atom::arena::Arena;
let mut arena = Arena::with_capacity(4096);
// Allocate multiple values
let val1 = arena.alloc_value(42u64).unwrap();
let val2 = arena.alloc_value(String::from("hello")).unwrap();
// Use values...
println!("val1: {}", *val1);
// Bulk deallocation
arena.reset(); // O(1) - frees all at once
Object Pool
use avila_atom::pool::ObjectPool;
let mut pool = ObjectPool::with_capacity(100);
// Acquire object
let id = pool.acquire(|| vec![0u8; 1024]);
// Use object
if let Some(buffer) = pool.get_mut(id) {
buffer.extend_from_slice(&[1, 2, 3]);
}
// Release back to pool (for reuse)
pool.release(id);
Lock-Free Concurrency
use avila_atom::lockfree::{LockFreeStack, AtomicCounter};
use std::thread;
// Lock-free stack
let stack = LockFreeStack::new();
let handles: Vec<_> = (0..4).map(|i| {
let stack_ref = &stack;
thread::spawn(move || {
stack_ref.push(i);
})
}).collect();
for handle in handles {
handle.join().unwrap();
}
// Atomic counter (cache-line padded)
let counter = AtomicCounter::new(0);
let handles: Vec<_> = (0..1000).map(|_| {
let counter_ref = &counter;
thread::spawn(move || {
counter_ref.increment();
})
}).collect();
for handle in handles {
handle.join().unwrap();
}
assert_eq!(counter.get(), 1000);
Advanced Structures (v0.5.0+)
use avila_atom::robinhood::RobinHoodMap;
use avila_atom::lockfree::RingBuffer;
use avila_atom::compress::{rle_encode, rle_decode};
use avila_atom::DynamicArray;
// Robin Hood hash map
let mut map = RobinHoodMap::new();
map.insert("key1", 100);
map.insert("key2", 200);
assert_eq!(map.get(&"key1"), Some(&100));
// Lock-free SPSC ring buffer (power-of-2 capacity)
let buffer: RingBuffer<u64, 1024> = RingBuffer::new();
buffer.push(42);
buffer.push(99);
assert_eq!(buffer.pop(), Some(42));
// Run-length encoding compression
let data = [1u8, 1, 1, 2, 2, 3];
let mut compressed = DynamicArray::new();
rle_encode(&data, &mut compressed);
// Result: [3, 1, 2, 2, 1, 3] = 3 ones, 2 twos, 1 three
Performance Characteristics
| Operation | Complexity | Notes |
|---|---|---|
DynamicArray::push |
O(1) amortized | Geometric growth (2x) |
DynamicArray::get |
O(1) | Direct index access |
AssociativeArray::get |
O(1) avg / O(log n) | HashMap / BTree |
StringBuffer::push_str |
O(n) | UTF-8 validation required |
FixedArray operations |
O(1) | Stack-allocated, inlined |
Arena::alloc |
O(1) | Bump pointer increment |
Arena::reset |
O(1) | Bulk deallocation |
ObjectPool::acquire |
O(1) | Slot reuse from free list |
LockFreeStack::push/pop |
O(1) | Wait-free CAS operations |
RobinHoodMap::insert/get |
O(1) avg | Bounded probe length |
RingBuffer::push/pop |
O(1) | Lock-free SPSC |
BPlusTree::insert |
O(log N) | 16-way fanout |
rle_encode/decode |
O(n) | 2-10x compression |
SkipList::insert/search |
O(log N) | Probabilistic, concurrent |
RadixTree::insert/get |
O(k) | k = key length |
BloomFilter::insert/contains |
O(h) | h = hash count |
CowArray::clone |
O(1) | Lazy copy, refcounted |
IntrusiveList::push/pop |
O(1) | Zero allocations |
NumaPool::push/pop |
O(1) | NUMA-local access |
Compile-Time Guarantees
use avila_atom::static_assert_size;
// Verify structure sizes for ABI compatibility
static_assert_size!(Option<&str>, 16); // Two pointers on 64-bit
no_std Support
Disable default features and enable alloc:
[dependencies]
avila-atom = { version = "0.2", default-features = false }
Note: AssociativeArray falls back to BTreeMap in no_std mode.
Architecture
Built following the Ávila Engineering Philosophy:
- Stack-preferred - Minimize heap allocations
- Zero dependencies - Built from Rust core types
- Performance-first - Optimized for modern CPU architectures
- Type-safe - Compile-time guarantees prevent runtime errors
License
Licensed under either of:
- Apache License, Version 2.0 (LICENSE-APACHE)
- MIT license (LICENSE-MIT)
at your option.
Contributing
Contributions are welcome! Please ensure:
- All tests pass (
cargo test) - Code is formatted (
cargo fmt) - No clippy warnings (
cargo clippy) - Documentation is updated
Part of the Ávila Computational Stack - Building high-performance systems from first principles.