Expand description
Compute server module.
Structs§
- Allocation
- An allocation with associated strides. Strides depend on tensor layout.
- Allocation
Descriptor - Descriptor for a new tensor allocation
- Binding
- Binding of a tensor handle to execute a kernel.
- Bindings
- Bindings to execute a kernel.
- Copy
Descriptor - A binding with shape and stride info for non-contiguous reading
- Handle
- Server handle containing the memory handle.
- Metadata
Binding - Binding of a set of scalars of the same type to execute a kernel.
- Profiling
Token - Profiling identification so that the server can support recursive and overlapping profilings.
- Scalar
Binding - Binding of a set of scalars of the same type to execute a kernel.
- Server
Utilities - Contains many different types that are useful for server implementations and compute clients.
- Tensor
MapBinding - A tensor map used with TMA ops
- Tensor
MapMeta - TensorMap metadata for the opaque proxy used in TMA copies
Enums§
- Allocation
Kind - Type of allocation, either contiguous or optimized (row-aligned when possible)
- Cube
Count - Specifieds the number of cubes to be dispatched for a kernel.
- IoError
- Error returned from
create/read/writefunctions. Due to async execution not all errors are able to be caught, so some IO errors will still panic. - Profile
Error - An error during profiling.
Traits§
- Compute
Server - The compute server is responsible for handling resources and computations over resources.
- Server
Communication - Defines functions for optimized data transfer between servers, supporting custom communication mechanisms such as peer-to-peer communication or specialized implementations.