Linear algebra can of course be dramatically useful for browser applications or embedded applications. For example, vector, matrix, and geometric operations are must-haves for any game running on the browser. Also, you may need matrix decompositions to perform machine-learning related data analysis on an embedded program. Luckily nalgebra supports both compiling for wasm and deactivating its link to the Rust standard library so that it becomes compilable for targets without libstd.
All features of the nalgebra crate, including pure-rust implementations of matrix
decompositions, will work out-of-the-box when compiled for the
target. Recall that compiling to
wasm only amounts to setting-up the target and passing
the right parameters to
On the other hand the nalgebra-lapack crate will not compile for wasm at all since it relies on binding to implementations of Blas and Lapack.
When developing applications targeting embedded platforms, the rust standard library
libstd is not always available. In those situations you typically have to add the
attribute to your project to prevent it from linking to
libstd. Though this
will be useless if nalgebra itself links to the standard library! This section
explains how to compile nalgebra without needing
libstd, and how this affects
the features you can use.
Note that the nalgebra-lapack crate on the other hand does not support being compiled without the Rust standard library yet.
Compiling without the standard library is necessary only for embedded development. You do not need to disable libstd when compiling to wasm!
nalgebra supports the
std feature which is enabled by default. Disabling this
feature will prevent nalgebra itself from linking to
libstd. To achieve this
you have to compile with a nightly version of the Rust compiler and declare your nalgebra
default-features = false, i.e., like the following:
This will affect what feature can be used. In particular:
- All vector and matrix operations on statically-sized matrices and statically-sized vectors will continue work out-of-the box. Of course, this includes all the matrix decompositions!
- Geometric operations will continue to work if the
libmfeature of nalgebra is enabled:
- Creating random matrices or vectors without an user-provided distribution will not be available. Therefore
::new_random()constructor will not exist. You may still use
- Dynamically sized vectors
DVectorand dynamically sized matrices
DMatrixwill not be available unless you activate the
allocfeature as well and provide an allocator for your platform. This will let nalgebra link to the alloc crate and use the
Vecdata structure internally. You may do the following to enable this feature: