Alex Riasanovsky

Add `heapless` to `Common > General > Stack-allocated arrays` (and other data structures)

1

Alongside `arrayvec` and `tinyvec`, [heapless](https://docs.rs/heapless/latest/heapless/) has stack-allocated arrays, but also includes: - [Arc](https://docs.rs/heapless/latest/heapless/pool/arc/struct.Arc.html) – like `std::sync::Arc` but backed by a lock-free memory pool rather than `#[global_allocator]` - [Box](https://docs.rs/heapless/latest/heapless/pool/boxed/struct.Box.html) – like...

Add `faer` to `math-scientific`

1

[faer](https://github.com/sarah-ek/faer-rs) is > a pure Rust implementation of low level linear algebra algorithms that leverages SIMD instructions, modern cache friendly algorithms and thread parallelism to achieve high performance. > --...

use cudaOccupancyMaxPotentialBlockSize in LaunchConfig::for_num_elems

2

for_num_elements defaults to block size 1024 but this is often suboptimal for performance. See [NVIDIA's article](https://developer.nvidia.com/blog/cuda-pro-tip-occupancy-api-simplifies-launch-configuration/) on optimal number of blocks and threads. > cudaOccupancyMaxPotentialBlockSize makes it possible to compute...