Overlays are Nix functions which accept two arguments, conventionally called `self` and `super`, and return a set of packages. For example, the following is a valid overlay.

```nix
self: super:

{
  boost = super.boost.override {
    python = self.python3;
  };
  rr = super.callPackage ./pkgs/rr {
    stdenv = self.stdenv_32bit;
  };
}
```

The first argument (`self`) corresponds to the final package set. You should use this set for the dependencies of all packages specified in your overlay. For example, all the dependencies of `rr` in the example above come from `self`, as well as the overridden dependencies used in the `boost` override.

The second argument (`super`) corresponds to the result of the evaluation of the previous stages of Nixpkgs. It does not contain any of the packages added by the current overlay, nor any of the following overlays. This set should be used either to refer to packages you wish to override, or to access functions defined in Nixpkgs. For example, the original recipe of `boost` in the above example, comes from `super`, as well as the `callPackage` function.

The value returned by this function should be a set similar to `pkgs/top-level/all-packages.nix`, containing overridden and/or new packages.

Overlays are similar to other methods for customizing Nixpkgs, in particular the `packageOverrides` attribute described in [](#sec-modify-via-packageOverrides). Indeed, `packageOverrides` acts as an overlay with only the `super` argument. It is therefore appropriate for basic use, but overlays are more powerful and easier to distribute.

## Using overlays to configure alternatives {#sec-overlays-alternatives}

Certain software packages have different implementations of the same interface. Other distributions have functionality to switch between these. For example, Debian provides [DebianAlternatives](https://wiki.debian.org/DebianAlternatives).  Nixpkgs has what we call `alternatives`, which are configured through overlays.

### BLAS/LAPACK {#sec-overlays-alternatives-blas-lapack}

In Nixpkgs, we have multiple implementations of the BLAS/LAPACK numerical linear algebra interfaces. They are:

-   [OpenBLAS](https://www.openblas.net/)

    The Nixpkgs attribute is `openblas` for ILP64 (integer width = 64 bits) and `openblasCompat` for LP64 (integer width = 32 bits).  `openblasCompat` is the default.

-   [LAPACK reference](https://www.netlib.org/lapack/) (also provides BLAS and CBLAS)

    The Nixpkgs attribute is `lapack-reference`.

-   [Intel MKL](https://software.intel.com/en-us/mkl) (only works on the x86_64 architecture, unfree)

    The Nixpkgs attribute is `mkl`.

-   [BLIS](https://github.com/flame/blis)

    BLIS, available through the attribute `blis`, is a framework for linear algebra kernels. In addition, it implements the BLAS interface.

-   [AMD BLIS/LIBFLAME](https://developer.amd.com/amd-aocl/blas-library/) (optimized for modern AMD x86_64 CPUs)

    The AMD fork of the BLIS library, with attribute `amd-blis`, extends BLIS with optimizations for modern AMD CPUs. The changes are usually submitted to the upstream BLIS project after some time. However, AMD BLIS typically provides some performance improvements on AMD Zen CPUs. The complementary AMD LIBFLAME library, with attribute `amd-libflame`, provides a LAPACK implementation.

Introduced in [PR #83888](https://github.com/NixOS/nixpkgs/pull/83888), we are able to override the `blas` and `lapack` packages to use different implementations, through the `blasProvider` and `lapackProvider` argument. This can be used to select a different provider. BLAS providers will have symlinks in `$out/lib/libblas.so.3` and `$out/lib/libcblas.so.3` to their respective BLAS libraries.  Likewise, LAPACK providers will have symlinks in `$out/lib/liblapack.so.3` and `$out/lib/liblapacke.so.3` to their respective LAPACK libraries. For example, Intel MKL is both a BLAS and LAPACK provider. An overlay can be created to use Intel MKL that looks like: