Skip to content

Refactor benchmarks #58

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
anicusan merged 2 commits into from
Aug 9, 2025
Merged

Refactor benchmarks #58

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
86fc9e0
Refactor benchmarks
christiangnrd Jun 24, 2025
4ca5dd0
Remove missed debug code
christiangnrd Aug 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions benchmark/Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,10 @@
[deps]
AcceleratedKernels = "6a4ca0a5-0e36-4168-a932-d9be78d558f1"
BenchmarkPlots = "ab8c0f59-4072-4e0d-8f91-a91e1495eb26"
BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
FileIO = "5789e2e9-d7fb-5bc7-8068-2c6fae9b9549"
GPUArrays = "0c68f7d7-f131-5f86-a1c3-88cf8149b2d7"
KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
StableRNGs = "860ef19b-820b-49d6-a774-d7a799459cd3"
StatsPlots = "f3b207a7-027a-5e70-b257-86293d7955fd"
55 changes: 12 additions & 43 deletions benchmark/accumulate_1d.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,53 +1,22 @@
import AcceleratedKernels as AK
using KernelAbstractions
group = addgroup!(SUITE, "accumulate_1d")

using BenchmarkTools
using Random
Random.seed!(0)
acc_f(x, y) = sin(x) + cos(y)


# Choose the Array backend:
#
# using CUDA
# const ArrayType = CuArray
#
# using AMDGPU
# const ArrayType = ROCArray
#
# using oneAPI
# const ArrayType = oneArray
#
# using Metal
# const ArrayType = MtlArray
#
# using OpenCL
# const ArrayType = CLArray
#
const ArrayType = Array


println("Using ArrayType: ", ArrayType)

GPUArrays.neutral_element(::typeof(acc_f), T) = T(0)

n = 1_000_000

for T in [UInt32, Int64, Float32]
local _group = addgroup!(group, "$T")

println("\n===\nBenchmarking accumulate(+) on $n UInt32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=UInt32(0)) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n))))
display(@benchmark AK.accumulate(+, v, init=UInt32(0)) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n))))


println("\n===\nBenchmarking accumulate(+) on $n Int64 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Int64(0)) setup=(v = ArrayType(rand(Int64(1):Int64(100), n))))
display(@benchmark AK.accumulate(+, v, init=Int64(0)) setup=(v = ArrayType(rand(Int64(1):Int64(100), n))))


println("\n===\nBenchmarking accumulate(+) on $n Float32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Float32(0)) setup=(v = ArrayType(rand(Float32, n))))
display(@benchmark AK.accumulate(+, v, init=Float32(0)) setup=(v = ArrayType(rand(Float32, n))))
local randrange = T == Float32 ? T : T(1):T(100)

_group["base_1d"] = @benchmarkable @sb(Base.accumulate(+, v; init=$T(0))) setup=(v = ArrayType(rand(rng, $randrange, n)))
_group["acck_1d"] = @benchmarkable @sb(AK.accumulate(+, v; init=$T(0))) setup=(v = ArrayType(rand(rng, $randrange, n)))

println("\n===\nBenchmarking accumulate((x, y) -> sin(x) + cos(y)) on $n Float32 - Base vs. AK")
display(@benchmark Base.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0)) setup=(v = ArrayType(rand(Float32, n))))
display(@benchmark AK.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0), neutral=Float32(0)) setup=(v = ArrayType(rand(Float32, n))))
T == Float32 || continue

_group["base_1d_sincos"] = @benchmarkable @sb(Base.accumulate(acc_f, v; init=$T(0))) setup=(v = ArrayType(rand(rng, $randrange, n)))
_group["acck_1d_sincos"] = @benchmarkable @sb(AK.accumulate(acc_f, v; init=$T(0), neutral=$T(0))) setup=(v = ArrayType(rand(rng, $randrange, n)))
end
79 changes: 16 additions & 63 deletions benchmark/accumulate_nd.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,75 +1,28 @@
import AcceleratedKernels as AK
using KernelAbstractions
group = addgroup!(SUITE, "accumulate_nd")

using BenchmarkTools
using Random
Random.seed!(0)


# Choose the Array backend:
#
# using CUDA
# const ArrayType = CuArray
#
# using AMDGPU
# const ArrayType = ROCArray
#
# using oneAPI
# const ArrayType = oneArray
#
# using Metal
# const ArrayType = MtlArray
#
# using OpenCL
# const ArrayType = CLArray
#
const ArrayType = Array


println("Using ArrayType: ", ArrayType)
acc_f(x, y) = sin(x) + cos(y)

GPUArrays.neutral_element(::typeof(acc_f), T) = T(0)

n1 = 3
n2 = 1_000_000

for T in [UInt32, Int64, Float32]
local _group = addgroup!(group, "$T")

println("\n===\nBenchmarking accumulate(+, dims=1) on $n1 × $n2 UInt32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=UInt32(0), dims=1) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n1, n2))))
display(@benchmark AK.accumulate(+, v, init=UInt32(0), dims=1) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n1, n2))))

println("\n===\nBenchmarking accumulate(+, dims=2) on $n1 × $n2 UInt32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=UInt32(0), dims=2) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n1, n2))))
display(@benchmark AK.accumulate(+, v, init=UInt32(0), dims=2) setup=(v = ArrayType(rand(UInt32(1):UInt32(100), n1, n2))))




println("\n===\nBenchmarking accumulate(+, dims=1) on $n1 × $n2 Int64 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Int64(0), dims=1) setup=(v = ArrayType(rand(Int64(1):Int64(100), n1, n2))))
display(@benchmark AK.accumulate(+, v, init=Int64(0), dims=1) setup=(v = ArrayType(rand(Int64(1):Int64(100), n1, n2))))

println("\n===\nBenchmarking accumulate(+, dims=2) on $n1 × $n2 Int64 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Int64(0), dims=2) setup=(v = ArrayType(rand(Int64(1):Int64(100), n1, n2))))
display(@benchmark AK.accumulate(+, v, init=Int64(0), dims=2) setup=(v = ArrayType(rand(Int64(1):Int64(100), n1, n2))))




println("\n===\nBenchmarking accumulate(+, dims=1) on $n1 × $n2 Float32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Float32(0), dims=1) setup=(v = ArrayType(rand(Float32, n1, n2))))
display(@benchmark AK.accumulate(+, v, init=Float32(0), dims=1) setup=(v = ArrayType(rand(Float32, n1, n2))))

println("\n===\nBenchmarking accumulate(+, dims=2) on $n1 × $n2 Float32 - Base vs. AK")
display(@benchmark Base.accumulate(+, v, init=Float32(0), dims=2) setup=(v = ArrayType(rand(Float32, n1, n2))))
display(@benchmark AK.accumulate(+, v, init=Float32(0), dims=2) setup=(v = ArrayType(rand(Float32, n1, n2))))
local randrange = T == Float32 ? T : T(1):T(100)

_group["base_dims=1"] = @benchmarkable @sb(Base.accumulate(+, v, init=$T(0), dims=1)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))
_group["acck_dims=1"] = @benchmarkable @sb(AK.accumulate(+, v, init=$T(0), dims=1)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))

_group["base_dims=2"] = @benchmarkable @sb(Base.accumulate(+, v, init=$T(0), dims=2)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))
_group["acck_dims=2"] = @benchmarkable @sb(AK.accumulate(+, v, init=$T(0), dims=2)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))

T == Float32 || continue

println("\n===\nBenchmarking accumulate((x, y) -> sin(x) + cos(y)), dims=1) on $n1 × $n2 Float32 - Base vs. AK")
display(@benchmark Base.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0), dims=1) setup=(v = ArrayType(rand(Float32, n1, n2))))
display(@benchmark AK.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0), neutral=Float32(0), dims=1) setup=(v = ArrayType(rand(Float32, n1, n2))))
_group["base_sincos_dims=1"] = @benchmarkable @sb(Base.accumulate(acc_f, v, init=$T(0), dims=1)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))
_group["acck_sincos_dims=1"] = @benchmarkable @sb(AK.accumulate(acc_f, v, init=$T(0), neutral=$T(0), dims=1)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))

println("\n===\nBenchmarking accumulate((x, y) -> sin(x) + cos(y)), dims=2) on $n1 × $n2 Float32 - Base vs. AK")
display(@benchmark Base.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0), dims=2) setup=(v = ArrayType(rand(Float32, n1, n2))))
display(@benchmark AK.accumulate((x, y) -> sin(x) + cos(y), v, init=Float32(0), neutral=Float32(0), dims=2) setup=(v = ArrayType(rand(Float32, n1, n2))))
_group["base_sincos_dims=2"] = @benchmarkable @sb(Base.accumulate(acc_f, v, init=$T(0), dims=2)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))
_group["acck_sincos_dims=2"] = @benchmarkable @sb(AK.accumulate(acc_f, v, init=$T(0), neutral=$T(0), dims=2)) setup=(v = ArrayType(rand(rng, $randrange, n1, n2)))
end
Loading
Loading