docs.rs failed to build amx-rs-0.0.3
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Visit the last successful build:
amx-rs-0.0.1
amx-rs
High-level ergonomic API for AMX (Apple Matrix eXtensions)
This crate provides type-safe, ergonomic abstractions over the low-level amx-sys instruction set. It includes:
- [
Matrix<T>] - Generic matrix type with operations - [
Vector<T>] - Generic vector type with iterators - [
MatMulBuilder] - Fluent API for matrix multiplication - [
ConvBuilder] - Fluent API for convolution operations
Features
- ✅ Generic over data types
- ✅ Type-safe bounds checking
- ✅ Iterator support
- ✅ Fluent builder API
- ✅ Zero-copy operations where possible
- ✅
no_stdcompatible (withalloc)
Quick Start
use Matrix;
// Create matrices
let a = zeros?;
let b = zeros?;
// Matrix operations
let c = a.transpose?;
println!;
Vectors
use Vector;
let mut v = zeros?;
v.set?;
v.set?;
// Iterate
for val in v.iter
Algorithms
Matrix Multiplication
use MatMulBuilder;
let builder = new
.transpose_a
.transpose_b
.alpha
.beta;
let c = builder.execute?;
Convolution
use ConvBuilder;
let conv = new // 3x3 kernel
.stride
.padding;
let = conv.output_dims;
Examples
Run examples:
Architecture
- amx-sys - Low-level instruction emulation (23 instructions)
- amx-rs - High-level API (this crate)
The split enables:
- Low-level research and education (amx-sys)
- Production usage with ergonomic API (amx-rs)
- Independent versioning and updates
Testing
All algorithms include unit tests:
Performance
- Dimension checking at construction time (when possible)
- Runtime bounds checking for safety
- Zero-copy operations via references
- Minimal memory overhead
License
MIT OR Apache-2.0