StrideArraysCore.jl

The core AbstractStrideArray type, separated from StrideArrays.jl to avoid circular dependencies.
Author JuliaSIMD
Popularity
16 Stars
Updated Last
4 Months Ago
Started In
February 2021

StrideArraysCore

Stable Dev Build Status Coverage

Defines the core PtrArray type so that some libraries can make use of it internally without the need for circular dependencies. StrideArrays extends this type with many methods and functionality. It is recommended you depend on and use StrideArrays instead.

Example initialization:

julia> using StrideArraysCore

julia> A = StrideArray(undef, 2, 5) # `Float64` is default eltype
2×5 StrideArray{Tuple{Int64, Int64}, (true, true), Float64, 2, 1, 0, (1, 2), Tuple{StaticInt{8}, Int64}, Tuple{StaticInt{1}, StaticInt{1}}, Vector{Float64}}:
 1.5e-323  -2.0092e-237  4.94e-321   1.0e-322  1.6e-322
 1.6e-322   4.94e-322    1.235e-321  1.5e-323  2.0e-323

julia> B = StrideArray{Float32}(zero, 2, 5) # can specify initialization function; function must have 1-arg method accepting eltype as argument
2×5 StrideArray{Tuple{Int64, Int64}, (true, true), Float32, 2, 1, 0, (1, 2), Tuple{StaticInt{4}, Int64}, Tuple{StaticInt{1}, StaticInt{1}}, Matrix{Float32}}:
 0.0  0.0  0.0  0.0  0.0
 0.0  0.0  0.0  0.0  0.0

julia> C = StrideArray{Float32}(one, static(2), 5) # sizes may optionally be static
2×5 StrideArray{Tuple{StaticInt{2}, Int64}, (true, true), Float32, 2, 1, 0, (1, 2), Tuple{StaticInt{4}, StaticInt{8}}, Tuple{StaticInt{1}, StaticInt{1}}, Vector{Float32}} with indices 1:1:2×Base.OneTo(5):
 1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0

julia> D = StrideArray{Float32}(one, static(2), static(5)) # all sizes being static will allow the compiler to elide the allocation if the array does not escape.
2×5 StrideArraysCore.StaticStrideArray{Tuple{StaticInt{2}, StaticInt{5}}, (true, true), Float32, 2, 1, 0, (1, 2), Tuple{StaticInt{4}, StaticInt{8}}, Tuple{StaticInt{1}, StaticInt{1}}, 10} with indices 1:1:2×1:1:5:
 1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0

julia> using StrideArraysCore, BenchmarkTools

julia> @inline function alloctest()
           D = StrideArray(one, static(2), static(5))
           s = 0.0
           for i in eachindex(D)
               s += D[i]
           end
           s
       end
alloctest (generic function with 1 method)

julia> @btime alloctest()
  1.214 ns (0 allocations: 0 bytes)
10.0

julia> @code_llvm debuginfo=:none alloctest() # compiler compiled-away function
define double @julia_alloctest_1199() #0 {
L67.9:
  ret double 1.000000e+01
}