ruda-tensor-device 0.21.76

Ruda device tensor backend and domain-library dispatch.
docs.rs failed to build ruda-tensor-device-0.21.76
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: ruda-tensor-device-0.21.73

ruda-tensor-device

Runtime-generic device tensor backend and dispatch into the Ruda domain libraries. It connects the framework's backend contracts to kernel-level RudaTensor values.

Interfaces

  • DeviceBackend and DeviceRuntime define the device backend integration.
  • dispatch routes tensor operations to device kernels and domain libraries.
  • cuda::Cuda supplies the NVIDIA tensor adapter when cuda is enabled.

Usage

Cargo package: ruda-tensor-device. Rust import: ruda_tensor_device.

[dependencies]
ruda-tensor-device = "0.21"

Features

Default features: autotune, std, fusion, ruda-kernel/frontend-default, ruda-fusion?/device-default.

Feature Purpose
cuda Enable the CUDA tensor adapter.
cuda-default Enable CUDA with the default fusion and tuning configuration.
fusion Enable operation-fusion integration.
autotune Enable domain-library autotuning.
sparse Enable sparse tensor dispatch.

Links