BabelStream/JuliaStream.jl/src/ThreadedStream.jl

112 lines
3.0 KiB
Julia

include("Stream.jl")
function devices()::Vector{DeviceWithRepr}
return [(undef, "$(Sys.cpu_info()[1].model) ($(Threads.nthreads())T)", "Threaded")]
end
function make_stream(
arraysize::Int,
scalar::T,
_::DeviceWithRepr,
silent::Bool,
)::Tuple{VectorData{T},Nothing} where {T}
if !silent
println("Using max $(Threads.nthreads()) threads")
end
return (
VectorData{T}(
Vector{T}(undef, arraysize),
Vector{T}(undef, arraysize),
Vector{T}(undef, arraysize),
scalar,
arraysize,
),
nothing
)
end
function init_arrays!(data::VectorData{T}, _, init::Tuple{T,T,T}) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.a[i] = init[1]
@inbounds data.b[i] = init[2]
@inbounds data.c[i] = init[3]
end
end
function copy!(data::VectorData{T}, _) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.c[i] = data.a[i]
end
end
function mul!(data::VectorData{T}, _) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.b[i] = data.scalar * data.c[i]
end
end
function add!(data::VectorData{T}, _) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.c[i] = data.a[i] + data.b[i]
end
end
function triad!(data::VectorData{T}, _) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.a[i] = data.b[i] + (data.scalar * data.c[i])
end
end
function nstream!(data::VectorData{T}, _) where {T}
Threads.@threads for i = 1:data.size
@inbounds data.a[i] += data.b[i] + data.scalar * data.c[i]
end
end
# Threads.@threads/Threads.@spawn doesn't support OpenMP's firstprivate, etc
function static_par_ranged(f::Function, range::Int, n::Int)
stride = range ÷ n
rem = range % n
strides = map(0:n) do i
width = stride + (i < rem ? 1 : 0)
offset = i < rem ? (stride + 1) * i : ((stride + 1) * rem) + (stride * (i - rem))
(offset, width)
end
ccall(:jl_enter_threaded_region, Cvoid, ())
try
foreach(wait, map(1:n) do group
(offset, size) = strides[group]
task = Task(() -> f(group, offset+1, offset+size))
task.sticky = true
ccall(:jl_set_task_tid, Cvoid, (Any, Cint), task, group-1) # ccall, so 0-based for group
schedule(task)
end)
finally
ccall(:jl_exit_threaded_region, Cvoid, ())
end
end
function dot(data::VectorData{T}, _) where {T}
partial = Vector{T}(undef, Threads.nthreads())
static_par_ranged(data.size, Threads.nthreads()) do group, startidx, endidx
acc = zero(T)
@simd for i = startidx:endidx
@inbounds acc += data.a[i] * data.b[i]
end
@inbounds partial[group] = acc
end
return sum(partial)
# This doesn't do well on aarch64 because of the excessive Threads.threadid() ccall
# and inhibited vectorisation from the lack of @simd
# partial = zeros(T, Threads.nthreads())
# Threads.@threads for i = 1:data.size
# @inbounds partial[Threads.threadid()] += (data.a[i] * data.b[i])
# end
# return sum(partial)
end
function read_data(data::VectorData{T}, _)::VectorData{T} where {T}
return data
end
main()