Expand description
Hodu CPU Kernels
High-performance CPU kernels for tensor operations with support for exotic floating-point formats and comprehensive integer types.
Modules§
- abs
- add
- add_
scalar - all
- any
- argmax
- argmin
- binary
- Binary operations for tensor element-wise computations
- concat
- concat_
split - Concatenation and split operations for tensors
- conv
- Convolution operations for neural network layers
- conv1d
- conv1d_
grad_ weight - conv2d
- conv2d_
grad_ weight - conv3d
- conv3d_
grad_ weight - conv_
transpose1d - conv_
transpose1d_ grad_ weight - conv_
transpose2d - conv_
transpose2d_ grad_ weight - conv_
transpose3d - conv_
transpose3d_ grad_ weight - cos
- div
- div_
scalar - dot
- eq
- eq_
scalar - exp
- exp2
- exp10
- gather
- ge
- ge_
scalar - gelu
- gt
- gt_
scalar - index_
put - index_
select - indexing
- Indexing operations for tensor element selection and manipulation
- le
- le_
scalar - ln
- log2
- log10
- logical_
and - logical_
not - logical_
or - logical_
xor - lt
- lt_
scalar - macros
- matmul
- matrix
- Matrix operations for linear algebra computations
- max
- maximum
- maximum_
scalar - mean
- min
- minimum
- minimum_
scalar - mish
- mul
- mul_
scalar - ne
- ne_
scalar - neg
- norm
- pow
- pow_
scalar - prod
- recip
- reduce
- Reduction operations for aggregating tensor values along dimensions
- reduce_
window_ max - reduce_
window_ mean - reduce_
window_ min - reduce_
window_ sum - relu
- scatter
- scatter_
add - scatter_
max - scatter_
min - sigmoid
- sign
- silu
- sin
- softplus
- split
- sqrt
- square
- std
- sub
- sub_
scalar - sum
- tan
- tanh
- unary
- Unary tensor operations
- var
- windowing
- Windowing operations for tensors
Structs§
Enums§
- CpuKernel
Error - Error type for CPU kernel operations
Functions§
- call_
binary - Execute a binary operation on two tensors
- call_
concat - Execute a concatenation operation on multiple tensors
- call_
conv - Execute a convolution operation
- call_
conv_ grad_ weight - Execute a convolution gradient weight operation
- call_
dot - Execute a 2D matrix multiplication operation
- call_
gather - Execute a gather operation
- call_
index_ put - Execute an index_put operation
- call_
index_ select - Execute an index_select operation
- call_
matmul - Execute a batched matrix multiplication operation with broadcasting
- call_
reduce - Execute a reduction operation
- call_
reduce_ window - Call windowing reduction operation by kernel name
- call_
scatter - Execute a scatter operation (scatter, scatter_add, scatter_max, scatter_min)
- call_
split - Execute a split operation on a tensor
- call_
unary - Call unary operation by kernel name
- call_
unary_ scalar - Call unary operation with scalar by kernel name
Type Aliases§
- Result
- Result type for CPU kernel operations