# User guide

```{toctree}
:caption: Getting started
:maxdepth: 1
:titlesonly:

getting_started
supported_systems
```

```{toctree}
:caption: Core concepts
:maxdepth: 1
:titlesonly:

tensor_types
scalar_tensors
matrix_vector
matrix_vector_per_thread
linalg_per_thread
tensor
compound_types
buffer_view
static
sub_functions
parallelization
```

```{toctree}
:caption: Integration
:maxdepth: 1
:titlesonly:

interop
metal_shared_queue
```

```{toctree}
:caption: Autodiff
:maxdepth: 1
:titlesonly:

autodiff
```

```{toctree}
:caption: SIMT primitives
:maxdepth: 1
:titlesonly:

gpu_execution_model
atomics
block
grid
math
subgroup
tile
```

```{toctree}
:caption: Algorithms
:maxdepth: 1
:titlesonly:

algorithms
```

```{toctree}
:caption: Performance
:maxdepth: 1
:titlesonly:

performance
fastcache
graph
streams
perf_dispatch
init_options
```

```{toctree}
:caption: Testing
:maxdepth: 1
:titlesonly:

unit_testing
kernel_coverage
```

```{toctree}
:caption: Reference
:maxdepth: 1
:titlesonly:

unsupported_python
python_backend
debug
quirks
troubleshooting
```

```{toctree}
:caption: Advanced
:maxdepth: 1
:titlesonly:

optimization_passes
```

```{toctree}
:caption: Contributing
:maxdepth: 1
:titlesonly:

contributing
```

```{toctree}
:caption: Internal
:maxdepth: 1
:titlesonly:

building_cudagraph_conditional_fatbin
```