Metal Performance Shaders

This section lists the package's public functionality that corresponds to the Metal Performance Shaders functions. For more information about these functions, or to see which functions have yet to be implemented in this package, please consult the Metal Performance Shaders Documentation.

Matrices and Vectors

Metal.MPS.MPSMatrix — Type

MPSMatrix(arr::MtlMatrix)

Metal matrix representation used in Performance Shaders.

Note that this results in a transposed view of the input, as Metal stores matrices row-major instead of column-major.

source

MPSMatrix(arr::MtlArray{T,3})

Metal batched matrix representation used in Performance Shaders.

Note that this results in a transposed view of the input, as Metal stores matrices row-major instead of column-major.

source

Metal.MPS.MPSVector — Type

MPSVector(arr::MtlVector)

Metal vector representation used in Performance Shaders.

source

Matrix Arithmetic Operators

Metal.MPS.matmul! — Function

matMulMPS(a::MtlMatrix, b::MtlMatrix, c::MtlMatrix, alpha=1, beta=1,
          transpose_left=false, transpose_right=false)

A MPSMatrixMultiplication kernel thay computes: c = alpha * op(a) * beta * op(b) + beta * C

This function should not typically be used. Rather, use the normal LinearAlgebra interface with any MtlArray and it should be accelerated using Metal Performance Shaders.

source

Metal.MPS.matvecmul! — Function

matVecMulMPS(c::MtlVector, a::MtlMatrix, b::MtlVector, alpha=1, beta=1,
             transpose=false)

A MPSMatrixVectorMultiplication kernel thay computes: c = alpha * op(a) * b + beta * c

This function should not typically be used. Rather, use the normal LinearAlgebra interface with any MtlArray and it should be accelerated using Metal Performance Shaders.

source

Metal.MPS.topk — Function

topk(A::MtlMatrix{T}, k) where {T<:MtlFloat}

Compute the top k values and their corresponding indices column-wise in a matrix A. Return the indices in I and the values in V.

k cannot be greater than 16.

Uses MPSMatrixFindTopK.