# Enigma

## Docs

- [Benchmark & Testing](https://klyne-research.mintlify.app/api-reference/benchmark-and-testing.md): Utilities for timing kernels and writing tests.
- [enigma.compile()](https://klyne-research.mintlify.app/api-reference/compile.md): Full reference for the enigma.compile() function and the CompiledKernel it returns.
- [Intrinsics](https://klyne-research.mintlify.app/api-reference/intrinsics.md): Vector geometry, pack/unpack, and type-casting intrinsics callable inside @enigma.kernel bodies.
- [Kernel Decorators & Grid Queries](https://klyne-research.mintlify.app/api-reference/kernel-decorators.md): Reference for @enigma.kernel, @enigma.jit, and the thread/grid query intrinsics available inside kernel bodies.
- [Layout Functions](https://klyne-research.mintlify.app/api-reference/layout-functions.md): Complete reference for all layout creation and transformation functions.
- [Matrix Operations](https://klyne-research.mintlify.app/api-reference/matrix-ops.md): Hardware-accelerated 8x8 simdgroup matrix multiply-accumulate, plus the regular Metal matrix-type operations.
- [MetalRuntime](https://klyne-research.mintlify.app/api-reference/runtime.md): Full reference for MetalRuntime, PreparedKernel, and DeviceCapabilities.
- [Threadgroup Memory & Synchronization](https://klyne-research.mintlify.app/api-reference/threadgroup-and-sync.md): Barriers, threadgroup shared memory allocation, and the buffer/threadgroup API surface.
- [Architecture](https://klyne-research.mintlify.app/concepts/architecture.md): How Enigma transforms Python kernel definitions into Metal GPU binaries.
- [Execution Model](https://klyne-research.mintlify.app/concepts/execution-model.md): How Enigma maps kernel launches to Metal grid geometry, threadgroups, and thread indices.
- [Memory Model](https://klyne-research.mintlify.app/concepts/memory-model.md): Address spaces, shared memory, and device capability checks in Enigma.
- [Common Errors](https://klyne-research.mintlify.app/debugging/common-errors.md): Typical failure modes with concrete diagnostic steps.
- [Inspection and Tracing](https://klyne-research.mintlify.app/debugging/inspection.md): How to inspect every stage of the Enigma pipeline to diagnose compilation and correctness issues.
- [Parallel Reduction](https://klyne-research.mintlify.app/examples/parallel-reduction.md): Sum all elements of a float32 array using SIMD group reductions and shared memory.
- [Tiled GEMM](https://klyne-research.mintlify.app/examples/tiled-gemm.md): Matrix multiply using layout algebra and the @enigma.jit tiling pattern.
- [Vector Add](https://klyne-research.mintlify.app/examples/vector-add.md): The simplest end-to-end Enigma kernel — elementwise addition of two float32 arrays.
- [Compile-Only Workflow](https://klyne-research.mintlify.app/getting-started/compile-only-workflow.md): Iterate on kernel codegen and source inspection without requiring a Metal device.
- [First Kernel](https://klyne-research.mintlify.app/getting-started/first-kernel.md): Write, compile, and dispatch a minimal Enigma kernel end to end.
- [Getting Started](https://klyne-research.mintlify.app/getting-started/index.md): Set up Enigma and run your first GPU kernel on Apple Metal.
- [Installation](https://klyne-research.mintlify.app/getting-started/installation.md): Install Enigma and verify the Metal toolchain before running your first kernel.
- [Enigma](https://klyne-research.mintlify.app/index.md): GPU compute for Apple Metal
- [Async Copy (Experimental)](https://klyne-research.mintlify.app/programming-guide/async-copy.md): AIR-backed async device-threadgroup copy via simdgroup intrinsics.
- [Atomic Operations](https://klyne-research.mintlify.app/programming-guide/atomics.md): Atomic load, store, and RMW operations with memory ordering semantics.
- [Control Flow](https://klyne-research.mintlify.app/programming-guide/control-flow.md): GPU-side loops, conditionals, and predicated memory access.
- [Data Types](https://klyne-research.mintlify.app/programming-guide/data-types.md): All scalar types in Enigma with their Metal and numpy equivalents.
- [High-Level Ops](https://klyne-research.mintlify.app/programming-guide/high-level-ops.md): GEMM, quantization helpers, and reusable compute primitives.
- [Kernels and JIT](https://klyne-research.mintlify.app/programming-guide/kernels-and-jit.md): How @enigma.kernel and @enigma.jit work, and when to use each.
- [Layout Algebra](https://klyne-research.mintlify.app/programming-guide/layout-algebra.md): CuTe-inspired Layout type and tiling operations for writing efficient tiled kernels.
- [Operations](https://klyne-research.mintlify.app/programming-guide/operations.md): Arithmetic, math, vector, comparison, and selection operations available in kernel bodies.
- [Registers, Copy & Pipeline](https://klyne-research.mintlify.app/programming-guide/register-and-pipeline.md): Register tensors, explicit data movement, and software pipelining.
- [SIMD Group Operations](https://klyne-research.mintlify.app/programming-guide/simd-group-ops.md): Reductions, prefix scans, and shuffle operations across a SIMD group (32 threads).
- [Swizzle](https://klyne-research.mintlify.app/programming-guide/swizzle.md): Bank-conflict-free shared memory access via XOR address remapping.
- [Changelog](https://klyne-research.mintlify.app/release-notes/index.md): New features, improvements, and fixes to Enigma.

## OpenAPI Specs

- [openapi](https://klyne-research.mintlify.app/api-reference/openapi.json)