[Do not merge] Test KernelIntrinsics

[christiangnrd]

[only tests]
[only benchmarks]

[github-actions]

Your PR requires formatting changes to meet the project's style guidelines.
Please consider running Runic (git runic master) to apply these changes.

Click here to view the suggested changes.

diff --git a/src/CUDAKernels.jl b/src/CUDAKernels.jl
index 5e39ab68e..7f8ec7ad4 100644
--- a/src/CUDAKernels.jl
+++ b/src/CUDAKernels.jl
@@ -162,29 +162,29 @@ end
 
 KI.argconvert(::CUDABackend, arg) = cudaconvert(arg)
 
-function KI.kernel_function(::CUDABackend, f::F, tt::TT=Tuple{}; name=nothing, kwargs...) where {F,TT}
+function KI.kernel_function(::CUDABackend, f::F, tt::TT = Tuple{}; name = nothing, kwargs...) where {F, TT}
     kern = cufunction(f, tt; name, kwargs...)
-    KI.Kernel{CUDABackend, typeof(kern)}(CUDABackend(), kern)
+    return KI.Kernel{CUDABackend, typeof(kern)}(CUDABackend(), kern)
 end
 
 function (obj::KI.Kernel{CUDABackend})(args...; numworkgroups = 1, workgroupsize = 1)
     KI.check_launch_args(numworkgroups, workgroupsize)
 
-    obj.kern(args...; threads=workgroupsize, blocks=numworkgroups)
+    obj.kern(args...; threads = workgroupsize, blocks = numworkgroups)
     return nothing
 end
 
 
-function KI.kernel_max_work_group_size(kernel::KI.Kernel{<:CUDABackend}; max_work_items::Int=typemax(Int))::Int
+function KI.kernel_max_work_group_size(kernel::KI.Kernel{<:CUDABackend}; max_work_items::Int = typemax(Int))::Int
     kernel_config = launch_configuration(kernel.kern.fun)
 
-    Int(min(kernel_config.threads, max_work_items))
+    return Int(min(kernel_config.threads, max_work_items))
 end
 function KI.max_work_group_size(::CUDABackend)::Int
-    Int(attribute(device(), CUDA.DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK))
+    return Int(attribute(device(), CUDA.DEVICE_ATTRIBUTE_MAX_THREADS_PER_BLOCK))
 end
 function KI.multiprocessor_count(::CUDABackend)::Int
-    Int(attribute(device(), CUDA.DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT))
+    return Int(attribute(device(), CUDA.DEVICE_ATTRIBUTE_MULTIPROCESSOR_COUNT))
 end
 
 ## indexing
@@ -199,7 +199,7 @@ end
 end
 
 @device_override @inline function KI.get_global_id()
-    return (; x = Int((blockIdx().x-1)*blockDim().x + threadIdx().x), y = Int((blockIdx().y-1)*blockDim().y + threadIdx().y), z = Int((blockIdx().z-1)*blockDim().z + threadIdx().z))
+    return (; x = Int((blockIdx().x - 1) * blockDim().x + threadIdx().x), y = Int((blockIdx().y - 1) * blockDim().y + threadIdx().y), z = Int((blockIdx().z - 1) * blockDim().z + threadIdx().z))
 end
 
 @device_override @inline function KI.get_local_size()
diff --git a/src/accumulate.jl b/src/accumulate.jl
index 54fab2119..e0631d387 100644
--- a/src/accumulate.jl
+++ b/src/accumulate.jl
@@ -22,9 +22,9 @@ function partial_scan(op::Function, output::AbstractArray{T}, input::AbstractArr
     temp = CuDynamicSharedArray(T, (2*threads,))
 
     # iterate the main dimension using threads and the first block dimension
-    i = (KI.get_group_id().x-1i32) * KI.get_local_size().x + KI.get_local_id().x
+    i = (KI.get_group_id().x - 1i32) * KI.get_local_size().x + KI.get_local_id().x
     # iterate the other dimensions using the remaining block dimensions
-    j = (KI.get_group_id().z-1i32) * KI.get_num_groups().y + KI.get_group_id().y
+    j = (KI.get_group_id().z - 1i32) * KI.get_num_groups().y + KI.get_group_id().y
 
     if j > length(Rother)
         return
@@ -105,9 +105,9 @@ function aggregate_partial_scan(op::Function, output::AbstractArray,
     block = KI.get_group_id().x
 
     # iterate the main dimension using threads and the first block dimension
-    i = (KI.get_group_id().x-1i32) * KI.get_local_size().x + KI.get_local_id().x
+    i = (KI.get_group_id().x - 1i32) * KI.get_local_size().x + KI.get_local_id().x
     # iterate the other dimensions using the remaining block dimensions
-    j = (KI.get_group_id().z-1i32) * KI.get_num_groups().y + KI.get_group_id().y
+    j = (KI.get_group_id().z - 1i32) * KI.get_num_groups().y + KI.get_group_id().y
 
     @inbounds if i <= length(Rdim) && j <= length(Rother)
         I = Rother[j]
diff --git a/src/device/random.jl b/src/device/random.jl
index 7d72d90a1..063c736ed 100644
--- a/src/device/random.jl
+++ b/src/device/random.jl
@@ -73,8 +73,8 @@ end
         @inbounds global_random_counters()[warpId]
     elseif field === :ctr2
         globalId = KI.get_global_id().x +
-                   (KI.get_global_id().y - 1i32) * KI.get_global_size().x +
-                   (KI.get_global_id().z - 1i32) * KI.get_global_size().x * KI.get_global_size().y
+            (KI.get_global_id().y - 1i32) * KI.get_global_size().x +
+            (KI.get_global_id().z - 1i32) * KI.get_global_size().x * KI.get_global_size().y
         globalId%UInt32
     end::UInt32
 end
diff --git a/src/mapreduce.jl b/src/mapreduce.jl
index 97a4176b4..6fccff91e 100644
--- a/src/mapreduce.jl
+++ b/src/mapreduce.jl
@@ -294,8 +294,9 @@ function GPUArrays.mapreducedim!(f::F, op::OP, R::AnyCuArray{T},
         end
 
         partial_kernel(f, op, init, Rreduce, Rother, Val(shuffle), partial, A;
-                    threads=partial_threads, blocks=partial_blocks, shmem=partial_shmem)
-                    # workgroupsize=partial_threads, numworkgroups=partial_blocks, shmem=partial_shmem)
+            threads = partial_threads, blocks = partial_blocks, shmem = partial_shmem
+        )
+        # workgroupsize=partial_threads, numworkgroups=partial_blocks, shmem=partial_shmem)
 
         GPUArrays.mapreducedim!(identity, op, R, partial; init)
     end
diff --git a/test/base/kernelabstractions.jl b/test/base/kernelabstractions.jl
index 2f2c4300b..1e674d3be 100644
--- a/test/base/kernelabstractions.jl
+++ b/test/base/kernelabstractions.jl
@@ -4,9 +4,14 @@ using SparseArrays
 
 include(joinpath(dirname(pathof(KernelAbstractions)), "..", "test", "testsuite.jl"))
 
-Testsuite.testsuite(()->CUDABackend(false, false), "CUDA", CUDA, CuArray, CuDeviceArray; skip_tests=Set([
-    "CPU synchronization",
-    "fallback test: callable types",]))
+Testsuite.testsuite(
+    () -> CUDABackend(false, false), "CUDA", CUDA, CuArray, CuDeviceArray; skip_tests = Set(
+        [
+            "CPU synchronization",
+            "fallback test: callable types",
+        ]
+    )
+)
 for (PreferBlocks, AlwaysInline) in Iterators.product((true, false), (true, false))
     Testsuite.unittest_testsuite(()->CUDABackend(PreferBlocks, AlwaysInline), "CUDA", CUDA, CuDeviceArray)
 end
diff --git a/test/runtests.jl b/test/runtests.jl
index 802c832e5..11584ca92 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -1,6 +1,6 @@
 @static if VERSION < v"1.11" && get(ENV, "BUILDKITE_PIPELINE_NAME", "CUDA.jl") == "CUDA.jl"
     using Pkg
-    Pkg.add(url="https://github.com/JuliaGPU/KernelAbstractions.jl", rev="main")
+    Pkg.add(url = "https://github.com/JuliaGPU/KernelAbstractions.jl", rev = "main")
 end
 
 using Distributed

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 2e7c994	Previous: fbb9098	Ratio
array/accumulate/Float32/1d	100136 ns	98020 ns	1.02
array/accumulate/Float32/dims=1	74958 ns	73968 ns	1.01
array/accumulate/Float32/dims=1L	1631537 ns	1598577 ns	1.02
array/accumulate/Float32/dims=2	139754 ns	139829 ns	1.00
array/accumulate/Float32/dims=2L	664739 ns	660479 ns	1.01
array/accumulate/Int64/1d	118892 ns	118236 ns	1.01
array/accumulate/Int64/dims=1	79762 ns	78818 ns	1.01
array/accumulate/Int64/dims=1L	1745959 ns	1717142 ns	1.02
array/accumulate/Int64/dims=2	155075 ns	151913 ns	1.02
array/accumulate/Int64/dims=2L	987375 ns	987639 ns	1.00
array/broadcast	19008 ns	18185 ns	1.05
array/construct	897.9 ns	900.1904761904761 ns	1.00
array/copy	16063 ns	16081 ns	1.00
array/copyto!/cpu_to_gpu	208640 ns	210623 ns	0.99
array/copyto!/gpu_to_cpu	241080 ns	242406 ns	0.99
array/copyto!/gpu_to_gpu	9847 ns	8883.666666666666 ns	1.11
array/iteration/findall/bool	132942 ns	132352 ns	1.00
array/iteration/findall/int	146162 ns	146447 ns	1.00
array/iteration/findfirst/bool	68025 ns	67784 ns	1.00
array/iteration/findfirst/int	69669 ns	69220 ns	1.01
array/iteration/findmin/1d	63892 ns	62612 ns	1.02
array/iteration/findmin/2d	100230 ns	99990 ns	1.00
array/iteration/logical	189337 ns	186900 ns	1.01
array/iteration/scalar	63169 ns	62246 ns	1.01
array/permutedims/2d	49271 ns	49143 ns	1.00
array/permutedims/3d	51982 ns	49912 ns	1.04
array/permutedims/4d	54559 ns	50252 ns	1.09
array/random/rand/Float32	11779 ns	10722 ns	1.10
array/random/rand/Int64	22985 ns	20792 ns	1.11
array/random/rand!/Float32	7710.25 ns	7728 ns	1.00
array/random/rand!/Int64	18105 ns	20553 ns	0.88
array/random/randn/Float32	32991 ns	33301 ns	0.99
array/random/randn!/Float32	23238 ns	24939 ns	0.93
array/reductions/mapreduce/Float32/1d	32051 ns	31290 ns	1.02
array/reductions/mapreduce/Float32/dims=1	40230 ns	37312 ns	1.08
array/reductions/mapreduce/Float32/dims=1L	50153 ns	50036 ns	1.00
array/reductions/mapreduce/Float32/dims=2	59204 ns	54292 ns	1.09
array/reductions/mapreduce/Float32/dims=2L	68355 ns	65880 ns	1.04
array/reductions/mapreduce/Int64/1d	39690 ns	39323 ns	1.01
array/reductions/mapreduce/Int64/dims=1	44307 ns	40145 ns	1.10
array/reductions/mapreduce/Int64/dims=1L	88124 ns	87779 ns	1.00
array/reductions/mapreduce/Int64/dims=2	64626 ns	57093 ns	1.13
array/reductions/mapreduce/Int64/dims=2L	85416 ns	82500 ns	1.04
array/reductions/reduce/Float32/1d	32037 ns	31207 ns	1.03
array/reductions/reduce/Float32/dims=1	40461 ns	37284 ns	1.09
array/reductions/reduce/Float32/dims=1L	49846 ns	50034 ns	1.00
array/reductions/reduce/Float32/dims=2	58804 ns	54677 ns	1.08
array/reductions/reduce/Float32/dims=2L	69093 ns	67287 ns	1.03
array/reductions/reduce/Int64/1d	39093 ns	38695 ns	1.01
array/reductions/reduce/Int64/dims=1	44297 ns	39938 ns	1.11
array/reductions/reduce/Int64/dims=1L	88231 ns	87933 ns	1.00
array/reductions/reduce/Int64/dims=2	64577 ns	56801 ns	1.14
array/reductions/reduce/Int64/dims=2L	84719 ns	82417 ns	1.03
array/reverse/1d	16309 ns	16212 ns	1.01
array/reverse/1dL	69042 ns	69012 ns	1.00
array/reverse/1dL_inplace	66983 ns	66891 ns	1.00
array/reverse/1d_inplace	9577.333333333334 ns	9703.666666666666 ns	0.99
array/reverse/2d	19619 ns	19393 ns	1.01
array/reverse/2dL	72912 ns	72842 ns	1.00
array/reverse/2dL_inplace	66835 ns	66766 ns	1.00
array/reverse/2d_inplace	10363 ns	10134 ns	1.02
array/sorting/1d	2658417 ns	2658542 ns	1.00
array/sorting/2d	1039100 ns	1039211 ns	1.00
array/sorting/by	3194272 ns	3193897 ns	1.00
cuda/synchronization/context/auto	1075.3 ns	1040 ns	1.03
cuda/synchronization/context/blocking	813.2222222222222 ns	795.5444444444445 ns	1.02
cuda/synchronization/context/nonblocking	5741 ns	5738.666666666667 ns	1.00
cuda/synchronization/stream/auto	927.483870967742 ns	889.2083333333334 ns	1.04
cuda/synchronization/stream/blocking	695.6095890410959 ns	676.8387096774194 ns	1.03
cuda/synchronization/stream/nonblocking	5409 ns	5562.285714285715 ns	0.97
integration/byval/reference	147391 ns	147385 ns	1.00
integration/byval/slices=1	149295 ns	149634 ns	1.00
integration/byval/slices=2	292146 ns	292176 ns	1.00
integration/byval/slices=3	434924 ns	435096 ns	1.00
integration/cudadevrt	104399 ns	104391 ns	1.00
integration/volumerhs	9311191 ns	9304645 ns	1.00
kernel/indexing	12529 ns	12267 ns	1.02
kernel/indexing_checked	13393 ns	13063 ns	1.03
kernel/launch	2014.888888888889 ns	1980.4 ns	1.02
kernel/occupancy	650.9285714285714 ns	638.797619047619 ns	1.02
kernel/rand	13685 ns	14927 ns	0.92
latency/import	3976654766 ns	3890391684 ns	1.02
latency/precompile	4745040524 ns	4668709973 ns	1.02
latency/ttfp	4939834474 ns	4883527794 ns	1.01

This comment was automatically generated by workflow using github-action-benchmark.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 17.80%. Comparing base (f36a98c) to head (f9ec011).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2944      +/-   ##
==========================================
+ Coverage   16.88%   17.80%   +0.92%     
==========================================
  Files         124      124              
  Lines        9886     9886              
==========================================
+ Hits         1669     1760      +91     
+ Misses       8217     8126      -91     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.