Use GPUArrays accumulation implementation

[christiangnrd]

Opened to run benchmarks.

Todo:

• Add compat bound when GPUArrays version released

[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/test/runtests.jl b/test/runtests.jl
index b6c479cce..89bf840c9 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -5,7 +5,7 @@ using Printf: @sprintf
 using Base.Filesystem: path_separator
 
 using Pkg
-Pkg.add(url="https://github.com/christiangnrd/GPUArrays.jl", rev="accumulatetests")
+Pkg.add(url = "https://github.com/christiangnrd/GPUArrays.jl", rev = "accumulatetests")
 
 # parse some command-line arguments
 function extract_flag!(args, flag, default=nothing; typ=typeof(default))

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: f8088b1	Previous: e561e7a	Ratio
latency/precompile	42896033326.5 ns	43393378645 ns	0.99
latency/ttfp	7099587815 ns	7099882121 ns	1.00
latency/import	3574268220 ns	3463869374 ns	1.03
integration/volumerhs	9621277 ns	9623663 ns	1.00
integration/byval/slices=1	146915 ns	146714 ns	1.00
integration/byval/slices=3	425748.5 ns	425787 ns	1.00
integration/byval/reference	144907 ns	144967 ns	1.00
integration/byval/slices=2	286403 ns	286209 ns	1.00
integration/cudadevrt	103446 ns	103426 ns	1.00
kernel/indexing	14264 ns	14196 ns	1.00
kernel/indexing_checked	14951 ns	14906 ns	1.00
kernel/occupancy	672.626582278481 ns	759.2189781021898 ns	0.89
kernel/launch	2152.222222222222 ns	2287.222222222222 ns	0.94
kernel/rand	17637 ns	15792 ns	1.12
array/reverse/1d	20110.5 ns	19624 ns	1.02
array/reverse/2d	24609 ns	24928.5 ns	0.99
array/reverse/1d_inplace	10850 ns	10448 ns	1.04
array/reverse/2d_inplace	13297 ns	12006 ns	1.11
array/copy	20888 ns	20990 ns	1.00
array/iteration/findall/int	116428.5 ns	159128.5 ns	0.73
array/iteration/findall/bool	98530 ns	139832 ns	0.70
array/iteration/findfirst/int	161354.5 ns	162546 ns	0.99
array/iteration/findfirst/bool	163024 ns	164393.5 ns	0.99
array/iteration/scalar	71507 ns	72740 ns	0.98
array/iteration/logical	173144.5 ns	216803.5 ns	0.80
array/iteration/findmin/1d	46910 ns	45968 ns	1.02
array/iteration/findmin/2d	96034 ns	96433 ns	1.00
array/reductions/reduce/Int64/1d	45446 ns	44555 ns	1.02
array/reductions/reduce/Int64/dims=1	51689 ns	48607 ns	1.06
array/reductions/reduce/Int64/dims=2	62807 ns	63682.5 ns	0.99
array/reductions/reduce/Int64/dims=1L	88945 ns	88842 ns	1.00
array/reductions/reduce/Int64/dims=2L	87955 ns	89417.5 ns	0.98
array/reductions/reduce/Float32/1d	34217 ns	34490 ns	0.99
array/reductions/reduce/Float32/dims=1	50563 ns	50554 ns	1.00
array/reductions/reduce/Float32/dims=2	59393 ns	59726 ns	0.99
array/reductions/reduce/Float32/dims=1L	52272 ns	52852 ns	0.99
array/reductions/reduce/Float32/dims=2L	69662 ns	70052.5 ns	0.99
array/reductions/mapreduce/Int64/1d	44348.5 ns	45547 ns	0.97
array/reductions/mapreduce/Int64/dims=1	53384.5 ns	48423.5 ns	1.10
array/reductions/mapreduce/Int64/dims=2	61862 ns	61443 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	88844 ns	88888 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	86736.5 ns	87908.5 ns	0.99
array/reductions/mapreduce/Float32/1d	34025 ns	34245.5 ns	0.99
array/reductions/mapreduce/Float32/dims=1	41472 ns	47287 ns	0.88
array/reductions/mapreduce/Float32/dims=2	59968 ns	59743 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52682 ns	53154 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	70380 ns	70503 ns	1.00
array/broadcast	20047 ns	20866 ns	0.96
array/copyto!/gpu_to_gpu	11206 ns	12817 ns	0.87
array/copyto!/cpu_to_gpu	214939.5 ns	213873 ns	1.00
array/copyto!/gpu_to_cpu	282695.5 ns	284406 ns	0.99
array/accumulate/Int64/1d	80469 ns	125170 ns	0.64
array/accumulate/Int64/dims=1	220590 ns	83519 ns	2.64
array/accumulate/Int64/dims=2	112517 ns	158002 ns	0.71
array/accumulate/Int64/dims=1L	409581.5 ns	1709945.5 ns	0.24
array/accumulate/Int64/dims=2L	5190679 ns	966571 ns	5.37
array/accumulate/Float32/1d	54935 ns	109737 ns	0.50
array/accumulate/Float32/dims=1	201705 ns	80823.5 ns	2.50
array/accumulate/Float32/dims=2	92304 ns	147778 ns	0.62
array/accumulate/Float32/dims=1L	245100 ns	1619194 ns	0.15
array/accumulate/Float32/dims=2L	3737008 ns	698530 ns	5.35
array/construct	1256.7 ns	1279.85 ns	0.98
array/random/randn/Float32	43200 ns	47253.5 ns	0.91
array/random/randn!/Float32	24962 ns	24573 ns	1.02
array/random/rand!/Int64	27381 ns	27294 ns	1.00
array/random/rand!/Float32	8756.333333333334 ns	8724.333333333334 ns	1.00
array/random/rand/Int64	29945 ns	29633 ns	1.01
array/random/rand/Float32	13068 ns	12902 ns	1.01
array/permutedims/4d	60355.5 ns	61250.5 ns	0.99
array/permutedims/2d	54052 ns	54865 ns	0.99
array/permutedims/3d	54989 ns	55511 ns	0.99
array/sorting/1d	2766160 ns	2757710 ns	1.00
array/sorting/by	3354758 ns	3344132.5 ns	1.00
array/sorting/2d	1084688 ns	1080389 ns	1.00
cuda/synchronization/stream/auto	1046 ns	1015.8333333333334 ns	1.03
cuda/synchronization/stream/nonblocking	7903.799999999999 ns	7618.9 ns	1.04
cuda/synchronization/stream/blocking	855.6 ns	799.1530612244898 ns	1.07
cuda/synchronization/context/auto	1161.7 ns	1164.1 ns	1.00
cuda/synchronization/context/nonblocking	8059.799999999999 ns	7651.4 ns	1.05
cuda/synchronization/context/blocking	925.8139534883721 ns	895.8490566037735 ns	1.03

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