Fill in missing KA functionality (KA.functional + sparse matrices adaption from CUDAbackend)

[Abdelrahman912]

This PR should be able to solve two missing pieces:

julia> KA.functional(CUDABackend())
missing

julia> adapt(CUDABackend(),sprand(10,10,0.2))
10×10 SparseMatrixCSC{Float64, Int64} with 14 stored entries:

I might need help on where I should write the tests, because KA.jl only tests for Array type (https://github.com/JuliaGPU/KernelAbstractions.jl/blob/365bf2cd461b297cf723ee3f819220ecb73c1e1e/test/test.jl#L94), so tests for sparse matrices should be put in test/kernelabstractions.jl , right?

[Abdelrahman912]

Also, one note here, last line will fallback to type Array for matrices other than CuSparseMatrixCSC unless some extension is loaded, for instance #2720. also some alternatives is to fallback to SparseMatrixCSC or just throw not implemented error

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.94%. Comparing base (b24af84) to head (6aa9b1e).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2740      +/-   ##
==========================================
- Coverage   88.99%   88.94%   -0.05%     
==========================================
  Files         153      153              
  Lines       13153    13154       +1     
==========================================
- Hits        11705    11700       -5     
- Misses       1448     1454       +6     

☔ View full report in Codecov by Sentry.
📢 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.

[maleadt]

tests for sparse matrices should be put in test/kernelabstractions.jl

Yeah. test/base/kernelabstractions.jl currently simply launches the KA.jl test suite, but we could test additional functionality there too.

[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 0e18eba61..9e87686a7 100644
--- a/src/CUDAKernels.jl
+++ b/src/CUDAKernels.jl
@@ -32,8 +32,8 @@ KA.synchronize(::CUDABackend) = synchronize()
 KA.functional(::CUDABackend) = CUDA.functional()
 
 Adapt.adapt_storage(::CUDABackend, a::AbstractArray) = Adapt.adapt(CuArray, a)
-Adapt.adapt_storage(::CUDABackend, a::Union{CuArray,CUSPARSE.AbstractCuSparseArray}) = a
-Adapt.adapt_storage(::KA.CPU, a::Union{CuArray,CUSPARSE.AbstractCuSparseArray}) = Adapt.adapt(Array, a)
+Adapt.adapt_storage(::CUDABackend, a::Union{CuArray, CUSPARSE.AbstractCuSparseArray}) = a
+Adapt.adapt_storage(::KA.CPU, a::Union{CuArray, CUSPARSE.AbstractCuSparseArray}) = Adapt.adapt(Array, a)
 
 ## memory operations
 
diff --git a/test/base/kernelabstractions.jl b/test/base/kernelabstractions.jl
index 4f4120032..61519e210 100644
--- a/test/base/kernelabstractions.jl
+++ b/test/base/kernelabstractions.jl
@@ -17,7 +17,7 @@ end
 @testset "CUDA Backend Adapt Tests" begin
     # CPU → GPU
     A = sprand(Float32, 10, 10, 0.5) #CSC
-    A_d = adapt(CUDABackend(), A) 
+    A_d = adapt(CUDABackend(), A)
     @test A_d isa CUSPARSE.CuSparseMatrixCSC
     @test adapt(CUDABackend(), A_d) |> typeof == typeof(A_d)
 
@@ -25,5 +25,5 @@ end
     B_d = A |> cu # CuCSC
     B = adapt(KA.CPU(), A_d)
     @test B isa SparseMatrixCSC
-    @test adapt(KA.CPU(), B) |> typeof == typeof(B) 
+    @test adapt(KA.CPU(), B) |> typeof == typeof(B)
 end

[maleadt]

Test failures related.

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: 6aa9b1e	Previous: b24af84	Ratio
latency/precompile	47227094040 ns	47408635362.5 ns	1.00
latency/ttfp	6824720867 ns	6756583810 ns	1.01
latency/import	3219728914.5 ns	3201988386.5 ns	1.01
integration/volumerhs	9610143 ns	9608856 ns	1.00
integration/byval/slices=1	147044 ns	146935 ns	1.00
integration/byval/slices=3	425640.5 ns	425262 ns	1.00
integration/byval/reference	145233 ns	144977 ns	1.00
integration/byval/slices=2	286533 ns	286173 ns	1.00
integration/cudadevrt	103559 ns	103373 ns	1.00
kernel/indexing	14361.5 ns	14266 ns	1.01
kernel/indexing_checked	15137 ns	14874.5 ns	1.02
kernel/occupancy	691.7651006711409 ns	689.8741721854304 ns	1.00
kernel/launch	2506.5 ns	2164.3 ns	1.16
kernel/rand	16558 ns	18183.5 ns	0.91
array/reverse/1d	19851 ns	20047 ns	0.99
array/reverse/2d	23852 ns	25366 ns	0.94
array/reverse/1d_inplace	10978.5 ns	11030.666666666666 ns	1.00
array/reverse/2d_inplace	12652 ns	11508 ns	1.10
array/copy	21244.5 ns	21080 ns	1.01
array/iteration/findall/int	160140 ns	158388 ns	1.01
array/iteration/findall/bool	140519 ns	139415 ns	1.01
array/iteration/findfirst/int	165650 ns	154381 ns	1.07
array/iteration/findfirst/bool	168634.5 ns	155031 ns	1.09
array/iteration/scalar	73635 ns	71204.5 ns	1.03
array/iteration/logical	219015.5 ns	210201.5 ns	1.04
array/iteration/findmin/1d	42078 ns	41765 ns	1.01
array/iteration/findmin/2d	94351 ns	95061.5 ns	0.99
array/reductions/reduce/1d	40507.5 ns	35804 ns	1.13
array/reductions/reduce/2d	51737 ns	42000.5 ns	1.23
array/reductions/mapreduce/1d	39119.5 ns	33568 ns	1.17
array/reductions/mapreduce/2d	50881 ns	41606.5 ns	1.22
array/broadcast	21490 ns	21172 ns	1.02
array/copyto!/gpu_to_gpu	11970 ns	13803 ns	0.87
array/copyto!/cpu_to_gpu	210872 ns	210795 ns	1.00
array/copyto!/gpu_to_cpu	245606.5 ns	245433 ns	1.00
array/accumulate/1d	110045.5 ns	109840.5 ns	1.00
array/accumulate/2d	80742 ns	80884 ns	1.00
array/construct	1290.9 ns	1282.8 ns	1.01
array/random/randn/Float32	49916.5 ns	47662 ns	1.05
array/random/randn!/Float32	26896 ns	26788 ns	1.00
array/random/rand!/Int64	27380 ns	27194 ns	1.01
array/random/rand!/Float32	8841.333333333334 ns	8682.333333333334 ns	1.02
array/random/rand/Int64	30085 ns	30217 ns	1.00
array/random/rand/Float32	13174 ns	13197 ns	1.00
array/permutedims/4d	61281 ns	62953 ns	0.97
array/permutedims/2d	55930 ns	55536.5 ns	1.01
array/permutedims/3d	56595 ns	56437 ns	1.00
array/sorting/1d	2778451 ns	2776763 ns	1.00
array/sorting/by	3369102 ns	3368018 ns	1.00
array/sorting/2d	1086558 ns	1085390 ns	1.00
cuda/synchronization/stream/auto	1086.5 ns	1028.5 ns	1.06
cuda/synchronization/stream/nonblocking	6519.2 ns	6435 ns	1.01
cuda/synchronization/stream/blocking	835.8333333333334 ns	810.6022727272727 ns	1.03
cuda/synchronization/context/auto	1213.4 ns	1167.4 ns	1.04
cuda/synchronization/context/nonblocking	6796.8 ns	6638.299999999999 ns	1.02
cuda/synchronization/context/blocking	961.5675675675676 ns	898.7291666666666 ns	1.07

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

[Abdelrahman912]

I believe CI failure is not from my end. is it? if not, then this should be ready