Expand description
Multi-stream kernel launch support.
Launches the same kernel across multiple CUDA streams simultaneously, enabling concurrent execution on the GPU when streams have no inter-dependencies. This is useful for data-parallel workloads where independent chunks can be processed in parallel.
§Example
// Assuming you have a kernel, streams, params, and args set up:
// multi_stream_launch(&kernel, &streams, ¶ms, &args)?;Functions§
- multi_
stream_ launch - Launches the same kernel across multiple streams with per-stream parameters and arguments.
- multi_
stream_ launch_ uniform - Launches the same kernel across multiple streams with uniform parameters and arguments.