Simplify specifying benchmark output file

[christiangnrd]

Makes no changes to CI but makes it easier to specify the benchmarks output filename without having to edit the file. Will open an equivalent PR for Metal if this is approved here.

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: 1f9b6e3	Previous: e561e7a	Ratio
latency/precompile	42928882636 ns	43393378645 ns	0.99
latency/ttfp	6986973732 ns	7099882121 ns	0.98
latency/import	3555865382 ns	3463869374 ns	1.03
integration/volumerhs	9611650.5 ns	9623663 ns	1.00
integration/byval/slices=1	146998 ns	146714 ns	1.00
integration/byval/slices=3	426161 ns	425787 ns	1.00
integration/byval/reference	145103 ns	144967 ns	1.00
integration/byval/slices=2	286545 ns	286209 ns	1.00
integration/cudadevrt	103507 ns	103426 ns	1.00
kernel/indexing	14323 ns	14196 ns	1.01
kernel/indexing_checked	14916 ns	14906 ns	1.00
kernel/occupancy	694.0466666666666 ns	759.2189781021898 ns	0.91
kernel/launch	2142.6666666666665 ns	2287.222222222222 ns	0.94
kernel/rand	18371 ns	15792 ns	1.16
array/reverse/1d	19772 ns	19624 ns	1.01
array/reverse/2d	24931 ns	24928.5 ns	1.00
array/reverse/1d_inplace	10538 ns	10448 ns	1.01
array/reverse/2d_inplace	12113 ns	12006 ns	1.01
array/copy	20859 ns	20990 ns	0.99
array/iteration/findall/int	158557 ns	159128.5 ns	1.00
array/iteration/findall/bool	140476 ns	139832 ns	1.00
array/iteration/findfirst/int	163335.5 ns	162546 ns	1.00
array/iteration/findfirst/bool	166096 ns	164393.5 ns	1.01
array/iteration/scalar	72107 ns	72740 ns	0.99
array/iteration/logical	215497.5 ns	216803.5 ns	0.99
array/iteration/findmin/1d	46807 ns	45968 ns	1.02
array/iteration/findmin/2d	96415.5 ns	96433 ns	1.00
array/reductions/reduce/Int64/1d	43110 ns	44555 ns	0.97
array/reductions/reduce/Int64/dims=1	46734 ns	48607 ns	0.96
array/reductions/reduce/Int64/dims=2	62883 ns	63682.5 ns	0.99
array/reductions/reduce/Int64/dims=1L	89091 ns	88842 ns	1.00
array/reductions/reduce/Int64/dims=2L	88266 ns	89417.5 ns	0.99
array/reductions/reduce/Float32/1d	34797 ns	34490 ns	1.01
array/reductions/reduce/Float32/dims=1	51815 ns	50554 ns	1.02
array/reductions/reduce/Float32/dims=2	59786 ns	59726 ns	1.00
array/reductions/reduce/Float32/dims=1L	52383 ns	52852 ns	0.99
array/reductions/reduce/Float32/dims=2L	70338 ns	70052.5 ns	1.00
array/reductions/mapreduce/Int64/1d	43238.5 ns	45547 ns	0.95
array/reductions/mapreduce/Int64/dims=1	51677 ns	48423.5 ns	1.07
array/reductions/mapreduce/Int64/dims=2	62618.5 ns	61443 ns	1.02
array/reductions/mapreduce/Int64/dims=1L	89035 ns	88888 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	87315.5 ns	87908.5 ns	0.99
array/reductions/mapreduce/Float32/1d	34746 ns	34245.5 ns	1.01
array/reductions/mapreduce/Float32/dims=1	41916.5 ns	47287 ns	0.89
array/reductions/mapreduce/Float32/dims=2	59891 ns	59743 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52857.5 ns	53154 ns	0.99
array/reductions/mapreduce/Float32/dims=2L	70310 ns	70503 ns	1.00
array/broadcast	19896 ns	20866 ns	0.95
array/copyto!/gpu_to_gpu	11175 ns	12817 ns	0.87
array/copyto!/cpu_to_gpu	215649.5 ns	213873 ns	1.01
array/copyto!/gpu_to_cpu	283084 ns	284406 ns	1.00
array/accumulate/Int64/1d	125491 ns	125170 ns	1.00
array/accumulate/Int64/dims=1	83763 ns	83519 ns	1.00
array/accumulate/Int64/dims=2	158115 ns	158002 ns	1.00
array/accumulate/Int64/dims=1L	1709771 ns	1709945.5 ns	1.00
array/accumulate/Int64/dims=2L	966596 ns	966571 ns	1.00
array/accumulate/Float32/1d	109531 ns	109737 ns	1.00
array/accumulate/Float32/dims=1	81191 ns	80823.5 ns	1.00
array/accumulate/Float32/dims=2	148152 ns	147778 ns	1.00
array/accumulate/Float32/dims=1L	1619394 ns	1619194 ns	1.00
array/accumulate/Float32/dims=2L	698520 ns	698530 ns	1.00
array/construct	1289.8 ns	1279.85 ns	1.01
array/random/randn/Float32	43954 ns	47253.5 ns	0.93
array/random/randn!/Float32	24863 ns	24573 ns	1.01
array/random/rand!/Int64	27237 ns	27294 ns	1.00
array/random/rand!/Float32	8784 ns	8724.333333333334 ns	1.01
array/random/rand/Int64	38266 ns	29633 ns	1.29
array/random/rand/Float32	13013 ns	12902 ns	1.01
array/permutedims/4d	60387 ns	61250.5 ns	0.99
array/permutedims/2d	54355.5 ns	54865 ns	0.99
array/permutedims/3d	55314 ns	55511 ns	1.00
array/sorting/1d	2758071 ns	2757710 ns	1.00
array/sorting/by	3369468.5 ns	3344132.5 ns	1.01
array/sorting/2d	1088835 ns	1080389 ns	1.01
cuda/synchronization/stream/auto	1044.6666666666667 ns	1015.8333333333334 ns	1.03
cuda/synchronization/stream/nonblocking	8191.4 ns	7618.9 ns	1.08
cuda/synchronization/stream/blocking	843.5106382978723 ns	799.1530612244898 ns	1.06
cuda/synchronization/context/auto	1161 ns	1164.1 ns	1.00
cuda/synchronization/context/nonblocking	8420.4 ns	7651.4 ns	1.10
cuda/synchronization/context/blocking	902.1276595744681 ns	895.8490566037735 ns	1.01

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

[christiangnrd]

Error unrelated

[github-actions]

Your PR no longer requires formatting changes. Thank you for your contribution!

[maleadt]

Can you elaborate why this is needed? It seems like a very hacky way of feeding this into the script, why not parse an actual (but optional) CLI argument?

[christiangnrd]

Can you elaborate why this is needed? It seems like a very hacky way of feeding this into the script, why not parse an actual (but optional) CLI argument?

Fair enough. l'll do that

[christiangnrd]

Benchmark output file is now an optional argument

[maleadt]

FYI, it's good to use [ci skip] or [only benchmarks] or so to avoid kicking off a whole CI run.

[christiangnrd]

Thank you!

FYI, it's good to use [ci skip] or [only benchmarks] or so to avoid kicking off a whole CI run.

I didn't bother because my latest push to #2815 kicked off a whole CI run anyway...

[maleadt]

I didn't bother because my latest push to #2815 kicked off a whole CI run anyway...

I'm not sure how that's relevant? It's not because CI runs on another PR anyway, that it needs to run here needlessly.

[christiangnrd]

I should have been more clear in my last response. The linked PR has an [only benchmarks] in the commit description but full CI ran anyway, so I assumed that the CI config would also ignore it here. In the future I'll make sure to properly tag my commits for CI regardless.

[maleadt]

I see. That's surprising, I'm not sure what regressed there.