Benchmark reverse on bigger arrays

[christiangnrd]

Currently part of #2832 but this can (and should) be merged seperately beforehand.

[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/perf/array.jl b/perf/array.jl
index 30348a512..7adb375db 100644
--- a/perf/array.jl
+++ b/perf/array.jl
@@ -55,11 +55,11 @@ let group = addgroup!(group, "reverse")
     group["1d"] = @async_benchmarkable reverse($gpu_vec)
     group["1dL"] = @async_benchmarkable reverse($gpu_vec_long)
     group["2d"] = @async_benchmarkable reverse($gpu_mat; dims=1)
-    group["2dL"] = @async_benchmarkable reverse($gpu_mat_long; dims=1)
+    group["2dL"] = @async_benchmarkable reverse($gpu_mat_long; dims = 1)
     group["1d_inplace"] = @async_benchmarkable reverse!($gpu_vec)
     group["1dL_inplace"] = @async_benchmarkable reverse!($gpu_vec_long)
     group["2d_inplace"] = @async_benchmarkable reverse!($gpu_mat; dims=1)
-    group["2dL_inplace"] = @async_benchmarkable reverse!($gpu_mat_long; dims=2)
+    group["2dL_inplace"] = @async_benchmarkable reverse!($gpu_mat_long; dims = 2)
 end
 
 group["broadcast"] = @async_benchmarkable $gpu_mat .= 0f0

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 7c005df	Previous: d670186	Ratio
latency/precompile	43572403994.5 ns	43647228816 ns	1.00
latency/ttfp	7274860573 ns	7277435679 ns	1.00
latency/import	3835703622.5 ns	3852618873.5 ns	1.00
integration/volumerhs	9609940 ns	9627931.5 ns	1.00
integration/byval/slices=1	147028 ns	147110 ns	1.00
integration/byval/slices=3	426147 ns	426020 ns	1.00
integration/byval/reference	145137 ns	145041 ns	1.00
integration/byval/slices=2	286450 ns	286551.5 ns	1.00
integration/cudadevrt	103552 ns	103546 ns	1.00
kernel/indexing	14270 ns	14322 ns	1.00
kernel/indexing_checked	14989 ns	14915 ns	1.00
kernel/occupancy	688.0526315789474 ns	667.2075471698113 ns	1.03
kernel/launch	2132.95 ns	2223.222222222222 ns	0.96
kernel/rand	14752 ns	18438 ns	0.80
array/reverse/1d	19648 ns	19972 ns	0.98
array/reverse/2dL_inplace	66921 ns
array/reverse/1dL	69924 ns
array/reverse/2d	22043 ns	24087.5 ns	0.92
array/reverse/1d_inplace	9561 ns	10265 ns	0.93
array/reverse/2d_inplace	11078 ns	11666 ns	0.95
array/reverse/2dL	74018 ns
array/reverse/1dL_inplace	66771 ns
array/copy	20334 ns	21022 ns	0.97
array/iteration/findall/int	157708 ns	157451 ns	1.00
array/iteration/findall/bool	139787 ns	138964 ns	1.01
array/iteration/findfirst/int	2161061.5 ns	2145633 ns	1.01
array/iteration/findfirst/bool	2143779 ns	2125537.5 ns	1.01
array/iteration/scalar	72525 ns	72116 ns	1.01
array/iteration/logical	236269.5 ns	235859 ns	1.00
array/iteration/findmin/1d	258429 ns	258194 ns	1.00
array/iteration/findmin/2d	96303 ns	96186 ns	1.00
array/reductions/reduce/Int64/1d	147342 ns	147739.5 ns	1.00
array/reductions/reduce/Int64/dims=1	43938 ns	44174 ns	0.99
array/reductions/reduce/Int64/dims=2	61319 ns	61550 ns	1.00
array/reductions/reduce/Int64/dims=1L	88919 ns	89011.5 ns	1.00
array/reductions/reduce/Int64/dims=2L	654799 ns	657807 ns	1.00
array/reductions/reduce/Float32/1d	103721 ns	104011 ns	1.00
array/reductions/reduce/Float32/dims=1	40824 ns	40950 ns	1.00
array/reductions/reduce/Float32/dims=2	59404 ns	59626 ns	1.00
array/reductions/reduce/Float32/dims=1L	52297 ns	52408 ns	1.00
array/reductions/reduce/Float32/dims=2L	544052 ns	547881 ns	0.99
array/reductions/mapreduce/Int64/1d	148541 ns	149756 ns	0.99
array/reductions/mapreduce/Int64/dims=1	43984 ns	44269 ns	0.99
array/reductions/mapreduce/Int64/dims=2	61397 ns	61933 ns	0.99
array/reductions/mapreduce/Int64/dims=1L	88787 ns	89034 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	680736 ns	685908.5 ns	0.99
array/reductions/mapreduce/Float32/1d	104121 ns	105015.5 ns	0.99
array/reductions/mapreduce/Float32/dims=1	40907 ns	40993 ns	1.00
array/reductions/mapreduce/Float32/dims=2	59419 ns	59573 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52511 ns	52874 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	546452 ns	550366 ns	0.99
array/broadcast	20216 ns	20326 ns	0.99
array/copyto!/gpu_to_gpu	12817 ns	12862 ns	1.00
array/copyto!/cpu_to_gpu	216578 ns	214253 ns	1.01
array/copyto!/gpu_to_cpu	286120.5 ns	286370.5 ns	1.00
array/accumulate/Int64/1d	124945 ns	124932 ns	1.00
array/accumulate/Int64/dims=1	83680 ns	83506 ns	1.00
array/accumulate/Int64/dims=2	157660.5 ns	157937 ns	1.00
array/accumulate/Int64/dims=1L	1710260 ns	1720004 ns	0.99
array/accumulate/Int64/dims=2L	966715 ns	968081 ns	1.00
array/accumulate/Float32/1d	109842 ns	109339 ns	1.00
array/accumulate/Float32/dims=1	80526 ns	80455 ns	1.00
array/accumulate/Float32/dims=2	147358 ns	147609.5 ns	1.00
array/accumulate/Float32/dims=1L	1618698.5 ns	1618340 ns	1.00
array/accumulate/Float32/dims=2L	698592 ns	700504 ns	1.00
array/construct	1616.4 ns	1612.7 ns	1.00
array/random/randn/Float32	44402 ns	44195.5 ns	1.00
array/random/randn!/Float32	24940 ns	24926 ns	1.00
array/random/rand!/Int64	27447 ns	27742 ns	0.99
array/random/rand!/Float32	8834.333333333334 ns	8671.666666666666 ns	1.02
array/random/rand/Int64	30044 ns	30303 ns	0.99
array/random/rand/Float32	12934.5 ns	13171 ns	0.98
array/permutedims/4d	59971 ns	60857.5 ns	0.99
array/permutedims/2d	53804 ns	54404 ns	0.99
array/permutedims/3d	54872 ns	55128 ns	1.00
array/sorting/1d	2757614 ns	2756220.5 ns	1.00
array/sorting/by	3344832.5 ns	3354879 ns	1.00
array/sorting/2d	1081338 ns	1084703.5 ns	1.00
cuda/synchronization/stream/auto	1004.6923076923077 ns	1064.1 ns	0.94
cuda/synchronization/stream/nonblocking	7875.1 ns	7654.4 ns	1.03
cuda/synchronization/stream/blocking	798.6210526315789 ns	827.0243902439024 ns	0.97
cuda/synchronization/context/auto	1170 ns	1151.2 ns	1.02
cuda/synchronization/context/nonblocking	7330.2 ns	7325.5 ns	1.00
cuda/synchronization/context/blocking	899.3617021276596 ns	916.8333333333334 ns	0.98

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

[christiangnrd]

Test failures seem unrelated?

[maleadt]

Yeah, that's #2885