Array programming

The Metal array type, MtlArray, generally implements the Base array interface and all of its expected methods.

However, there is the special function mtl for transferring an array over to the gpu. For compatibility reasons, it will automatically convert arrays of Float64 to Float32.

Metal.mtl — Function

mtl(A; storage=Metal.PrivateStorage)

storage can be Metal.PrivateStorage (default), Metal.SharedStorage, or Metal.ManagedStorage.

Opinionated GPU array adaptor, which may alter the element type T of arrays:

For T<:AbstractFloat, it makes a MtlArray{Float32} for performance and compatibility reasons (except for Float16).
For T<:Complex{<:AbstractFloat} it makes a MtlArray{ComplexF32}.
For other isbitstype(T), it makes a MtlArray{T}.

By contrast, MtlArray(A) never changes the element type.

Uses Adapt.jl to act inside some wrapper structs.

Examples

julia> mtl(ones(3)')
1×3 adjoint(::MtlVector{Float32, Metal.PrivateStorage}) with eltype Float32:
 1.0  1.0  1.0

julia> mtl(zeros(1,3); storage=Metal.SharedStorage)
1×3 MtlMatrix{Float32, Metal.SharedStorage}:
 0.0  0.0  0.0

julia> mtl(1:3)
1:3

julia> MtlArray(1:3)
3-element MtlVector{Int64, Metal.PrivateStorage}:
 1
 2
 3

source

Metal.MtlArray — Type

MtlArray{T,N,S} <: AbstractGPUArray{T,N}

N-dimensional Metal array with storage mode S and elements of type T.

S can be Metal.PrivateStorage (default), Metal.SharedStorage, or Metal.ManagedStorage.

See the Array Programming section of the Metal.jl docs for more details.

source

Metal.MtlVector — Type

MtlVector{T,S} <: AbstractGPUVector{T}

One-dimensional array with elements of type T for use with Apple Metal-compatible GPUs. Alias for MtlArray{T,1,S}.

See also Vector(@ref), and the Array Programming section of the Metal.jl docs for more details.

source

Metal.MtlMatrix — Type

MtlMatrix{T,S} <: AbstractGPUMatrix{T}

Two-dimensional array with elements of type T for use with Apple Metal-compatible GPUs. Alias for MtlArray{T,2,S}.

See also Matrix(@ref), and the Array Programming section of the Metal.jl docs for more details.

source

Metal.MtlVecOrMat — Type

MtlVecOrMat{T,S}

Union type of MtlVector{T,S} and MtlMatrix{T,S} which allows functions to accept either an MtlMatrix or an MtlVector.

Storage modes

The Metal API has various storage modes that dictate how a resource can be accessed. MtlArrays are Metal.PrivateStorage by default, but they can also be Metal.SharedStorage or Metal.ManagedStorage. For more information on storage modes, see the official Metal documentation.

Metal.MTL.PrivateStorage — Type

struct Metal.PrivateStorage <: MTL.StorageMode

Used to indicate that the resource is stored using MTLStorageModePrivate in memory.

For more information on Metal storage modes, refer to the official Metal documentation.

source

Metal.MTL.SharedStorage — Type

struct Metal.SharedStorage <: MTL.StorageMode

Used to indicate that the resource is stored using MTLStorageModeShared in memory.

For more information on Metal storage modes, refer to the official Metal documentation.

source

Metal.MTL.ManagedStorage — Type

struct Metal.ManagedStorage <: MTL.StorageMode

Used to indicate that the resource is stored using MTLStorageModeManaged in memory.

For more information on Metal storage modes, refer to the official Metal documentation.

source

There also exist the following convenience functions to check if an MtlArray is using a specific storage mode:

Metal.is_private — Function

is_private(A::MtlArray) -> Bool

Returns true if A has storage mode Metal.PrivateStorage.