docs.rs failed to build vortx-0.2.0
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Visit the last successful build:
vortx-0.0.0
vortx − cross-platform GPU tensor library in Rust
vortx is a cross-platform tensor library exposing linear-algebra operations as GPU compute shaders written in Rust with rust-gpu.
Warning vortx is still very incomplete and under heavy development.
Features
- GPU tensors up to rank 4 with views, strides, transpose, reshape, broadcast, squeeze/unsqueeze
- GEMM (matrix multiplication) — naive and optimized tiled kernels
- Element-wise ops — add, sub, mul, div, copy (in-place)
- Reductions — sum, product, min, max, squared norm
- Multiple backends via khal: WebGPU (default), CUDA. CPU, CPU-parallel, are also supported for debugging.
Development setup
cargo-gpu (required for SPIR-V / WebGPU)
Install cargo-gpu from crates.io:
cargo-cuda (required for CUDA / PTX)
Install cargo-cuda from crates.io:
Important: the cuda backend support requires depending on the github version of khal instead of the version
on crates.io.
This requires the CUDA toolkit to be installed and the CUDA_PATH environment variable to
point to it (e.g. /usr/local/cuda). The install step downloads a pinned Rust nightly, adds the
nvptx64-nvidia-cuda target, and compiles the codegen backend.