@nospecialize testf/_compare forwarders in CUDA tests.

[maleadt]

The main test/helpers.jl `testf` and the private clones in lib/cublas, lib/cufft, lib/cusolver are all thin forwarders that call out to deepcopy / adapt / ≈. Each unique (f, xs...) call site was getting its own compiled method even though the body does no per-type compute.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: 85552c4	Previous: 22b2689	Ratio
array/accumulate/Float32/1d	101042 ns	102014.5 ns	0.99
array/accumulate/Float32/dims=1	76713 ns	77428 ns	0.99
array/accumulate/Float32/dims=1L	1586552 ns	1587073 ns	1.00
array/accumulate/Float32/dims=2	143947.5 ns	144782 ns	0.99
array/accumulate/Float32/dims=2L	658019 ns	661233 ns	1.00
array/accumulate/Int64/1d	118225 ns	118923 ns	0.99
array/accumulate/Int64/dims=1	80053.5 ns	80472.5 ns	0.99
array/accumulate/Int64/dims=1L	1695730 ns	1706560 ns	0.99
array/accumulate/Int64/dims=2	156555 ns	156801 ns	1.00
array/accumulate/Int64/dims=2L	962158 ns	962520 ns	1.00
array/broadcast	20369 ns	20669 ns	0.99
array/construct	1256.9 ns	1256.8 ns	1.00
array/copy	18009 ns	18092 ns	1.00
array/copyto!/cpu_to_gpu	216616 ns	217195 ns	1.00
array/copyto!/gpu_to_cpu	280853 ns	284274 ns	0.99
array/copyto!/gpu_to_gpu	10742 ns	10873 ns	0.99
array/iteration/findall/bool	134816 ns	134986 ns	1.00
array/iteration/findall/int	149572 ns	151080 ns	0.99
array/iteration/findfirst/bool	81437 ns	81781 ns	1.00
array/iteration/findfirst/int	83808 ns	84023 ns	1.00
array/iteration/findmin/1d	86422.5 ns	87902.5 ns	0.98
array/iteration/findmin/2d	117166 ns	117417 ns	1.00
array/iteration/logical	198169.5 ns	201691.5 ns	0.98
array/iteration/scalar	66685 ns	66086 ns	1.01
array/permutedims/2d	51987.5 ns	52480 ns	0.99
array/permutedims/3d	52434 ns	53206 ns	0.99
array/permutedims/4d	51399.5 ns	52056 ns	0.99
array/random/rand/Float32	13113 ns	13513 ns	0.97
array/random/rand/Int64	25032 ns	25225 ns	0.99
array/random/rand!/Float32	8447.666666666666 ns	9480 ns	0.89
array/random/rand!/Int64	21770 ns	21874.5 ns	1.00
array/random/randn/Float32	43447 ns	39930.5 ns	1.09
array/random/randn!/Float32	30894 ns	30795 ns	1.00
array/reductions/mapreduce/Float32/1d	34244 ns	35149.5 ns	0.97
array/reductions/mapreduce/Float32/dims=1	49219.5 ns	40034.5 ns	1.23
array/reductions/mapreduce/Float32/dims=1L	51349 ns	51335 ns	1.00
array/reductions/mapreduce/Float32/dims=2	56457 ns	56606 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	69468.5 ns	69556 ns	1.00
array/reductions/mapreduce/Int64/1d	42976.5 ns	42597 ns	1.01
array/reductions/mapreduce/Int64/dims=1	43024.5 ns	50815 ns	0.85
array/reductions/mapreduce/Int64/dims=1L	87461 ns	87295 ns	1.00
array/reductions/mapreduce/Int64/dims=2	59307 ns	59460 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	84710 ns	84825 ns	1.00
array/reductions/reduce/Float32/1d	34755 ns	35215.5 ns	0.99
array/reductions/reduce/Float32/dims=1	41015.5 ns	40067 ns	1.02
array/reductions/reduce/Float32/dims=1L	51446 ns	51370 ns	1.00
array/reductions/reduce/Float32/dims=2	56728.5 ns	56646 ns	1.00
array/reductions/reduce/Float32/dims=2L	70065 ns	69814 ns	1.00
array/reductions/reduce/Int64/1d	42968 ns	42572 ns	1.01
array/reductions/reduce/Int64/dims=1	42021.5 ns	50859 ns	0.83
array/reductions/reduce/Int64/dims=1L	87380 ns	87191 ns	1.00
array/reductions/reduce/Int64/dims=2	59393 ns	59743 ns	0.99
array/reductions/reduce/Int64/dims=2L	84703.5 ns	84667 ns	1.00
array/reverse/1d	17939 ns	18121 ns	0.99
array/reverse/1dL	68538 ns	68646 ns	1.00
array/reverse/1dL_inplace	65727 ns	65864 ns	1.00
array/reverse/1d_inplace	10232.833333333332 ns	10292.666666666668 ns	0.99
array/reverse/2d	20679 ns	21138 ns	0.98
array/reverse/2dL	72699 ns	73135 ns	0.99
array/reverse/2dL_inplace	65688 ns	65713 ns	1.00
array/reverse/2d_inplace	11080 ns	11182 ns	0.99
array/sorting/1d	2735481 ns	2735147 ns	1.00
array/sorting/2d	1068965 ns	1073189 ns	1.00
array/sorting/by	3304941 ns	3303865.5 ns	1.00
cuda/synchronization/context/auto	1182.1 ns	1112 ns	1.06
cuda/synchronization/context/blocking	938.6515151515151 ns	888.829268292683 ns	1.06
cuda/synchronization/context/nonblocking	7634.4 ns	6852.5 ns	1.11
cuda/synchronization/stream/auto	996 ns	970.8 ns	1.03
cuda/synchronization/stream/blocking	796.6421052631579 ns	800.5102040816327 ns	1.00
cuda/synchronization/stream/nonblocking	8034.1 ns	7101.299999999999 ns	1.13
integration/byval/reference	143711 ns	143926 ns	1.00
integration/byval/slices=1	145657 ns	145976 ns	1.00
integration/byval/slices=2	284549 ns	284590 ns	1.00
integration/byval/slices=3	423087 ns	423197 ns	1.00
integration/cudadevrt	102345 ns	102511 ns	1.00
integration/volumerhs	23471812 ns	23505751 ns	1.00
kernel/indexing	13153 ns	13253 ns	0.99
kernel/indexing_checked	13896 ns	13903 ns	1.00
kernel/launch	2147.222222222222 ns	2079.8888888888887 ns	1.03
kernel/occupancy	664.1314102564102 ns	693.52 ns	0.96
kernel/rand	17278.5 ns	15424 ns	1.12
latency/import	3821736397.5 ns	3843588033 ns	0.99
latency/precompile	4587847485 ns	4581549814.5 ns	1.00
latency/ttfp	4396109132 ns	4430341918 ns	0.99

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

[codecov]

Codecov Report

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 16.58%. Comparing base (22a3b2c) to head (85552c4).
⚠️ Report is 5 commits behind head on master.

Files with missing lines	Patch %	Lines
lib/cublas/test/setup.jl	0.00%	4 Missing ⚠️
lib/cusolver/test/setup.jl	0.00%	4 Missing ⚠️
lib/cufft/test/setup.jl	0.00%	3 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3117   +/-   ##
=======================================
  Coverage   16.58%   16.58%           
=======================================
  Files         120      120           
  Lines        9586     9586           
=======================================
  Hits         1590     1590           
  Misses       7996     7996           

☔ 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.