Make CUDA & friends loadable on systems without NVPTX LLVM backend

[giordano]

This makes CUDA.jl loadable for me on macOS, without any warning (as it used to do in v5). PR best reviewed by ignoring whitespace changes

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.42%. Comparing base (d95add6) to head (f59f05f).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #3067   +/-   ##
=======================================
  Coverage   90.42%   90.42%           
=======================================
  Files         141      141           
  Lines       11993    11993           
=======================================
  Hits        10845    10845           
  Misses       1148     1148           

☔ 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.

[github-actions]

CUDA.jl Benchmarks

Details

Benchmark suite	Current: f59f05f	Previous: d95add6	Ratio
latency/precompile	4592146077.5 ns	4591137397 ns	1.00
latency/ttfp	4399155900.5 ns	4388441108 ns	1.00
latency/import	3819111214.5 ns	3806192215 ns	1.00
integration/volumerhs	9439930.5 ns	9441617.5 ns	1.00
integration/byval/slices=1	146076 ns	145778 ns	1.00
integration/byval/slices=3	423209 ns	423134 ns	1.00
integration/byval/reference	144218 ns	143890 ns	1.00
integration/byval/slices=2	284770 ns	284514 ns	1.00
integration/cudadevrt	102812 ns	102578 ns	1.00
kernel/indexing	13471.5 ns	13667 ns	0.99
kernel/indexing_checked	14259 ns	14261 ns	1.00
kernel/occupancy	681.0193548387097 ns	664.75 ns	1.02
kernel/launch	2221.5555555555557 ns	2208.777777777778 ns	1.01
kernel/rand	16362 ns	18252 ns	0.90
array/reverse/1d	18700.5 ns	18706 ns	1.00
array/reverse/2dL_inplace	66103 ns	66090 ns	1.00
array/reverse/1dL	69258 ns	69365 ns	1.00
array/reverse/2d	20940 ns	20874 ns	1.00
array/reverse/1d_inplace	8651.333333333334 ns	10373 ns	0.83
array/reverse/2d_inplace	10268 ns	10322 ns	0.99
array/reverse/2dL	72954 ns	72995 ns	1.00
array/reverse/1dL_inplace	66053.5 ns	65923 ns	1.00
array/copy	18724 ns	18987 ns	0.99
array/iteration/findall/int	149943.5 ns	149956 ns	1.00
array/iteration/findall/bool	132890 ns	132890 ns	1
array/iteration/findfirst/int	84061 ns	84402 ns	1.00
array/iteration/findfirst/bool	81951.5 ns	82118 ns	1.00
array/iteration/scalar	67980.5 ns	68770 ns	0.99
array/iteration/logical	201560.5 ns	204009 ns	0.99
array/iteration/findmin/1d	85760 ns	86819.5 ns	0.99
array/iteration/findmin/2d	117484 ns	118061 ns	1.00
array/reductions/reduce/Int64/1d	43796 ns	43837 ns	1.00
array/reductions/reduce/Int64/dims=1	42901 ns	53019.5 ns	0.81
array/reductions/reduce/Int64/dims=2	59921 ns	60078 ns	1.00
array/reductions/reduce/Int64/dims=1L	87983 ns	88042 ns	1.00
array/reductions/reduce/Int64/dims=2L	85126 ns	85281 ns	1.00
array/reductions/reduce/Float32/1d	35574 ns	35957 ns	0.99
array/reductions/reduce/Float32/dims=1	42129 ns	41971.5 ns	1.00
array/reductions/reduce/Float32/dims=2	57231 ns	57207 ns	1.00
array/reductions/reduce/Float32/dims=1L	52087 ns	52254 ns	1.00
array/reductions/reduce/Float32/dims=2L	70549 ns	70121.5 ns	1.01
array/reductions/mapreduce/Int64/1d	43545 ns	43844 ns	0.99
array/reductions/mapreduce/Int64/dims=1	43249 ns	42506 ns	1.02
array/reductions/mapreduce/Int64/dims=2	59882 ns	59797 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	87920 ns	87969 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	85248 ns	85477 ns	1.00
array/reductions/mapreduce/Float32/1d	35402 ns	35390 ns	1.00
array/reductions/mapreduce/Float32/dims=1	40235.5 ns	39718.5 ns	1.01
array/reductions/mapreduce/Float32/dims=2	57151 ns	56925 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	52079.5 ns	52147 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	70017 ns	69617 ns	1.01
array/broadcast	20624 ns	20791 ns	0.99
array/copyto!/gpu_to_gpu	11393 ns	11363 ns	1.00
array/copyto!/cpu_to_gpu	218446 ns	217254 ns	1.01
array/copyto!/gpu_to_cpu	284049 ns	286261 ns	0.99
array/accumulate/Int64/1d	119013 ns	119453.5 ns	1.00
array/accumulate/Int64/dims=1	79987 ns	80552 ns	0.99
array/accumulate/Int64/dims=2	156049 ns	156859 ns	0.99
array/accumulate/Int64/dims=1L	1694112 ns	1706924 ns	0.99
array/accumulate/Int64/dims=2L	961481 ns	962107.5 ns	1.00
array/accumulate/Float32/1d	102076 ns	101709 ns	1.00
array/accumulate/Float32/dims=1	77112 ns	77739 ns	0.99
array/accumulate/Float32/dims=2	144441 ns	144317 ns	1.00
array/accumulate/Float32/dims=1L	1585891 ns	1586304.5 ns	1.00
array/accumulate/Float32/dims=2L	657924.5 ns	658932 ns	1.00
array/construct	1303.5 ns	1319 ns	0.99
array/random/randn/Float32	39636 ns	44317.5 ns	0.89
array/random/randn!/Float32	31597 ns	27470 ns	1.15
array/random/rand!/Int64	34479 ns	27837 ns	1.24
array/random/rand!/Float32	8529.333333333334 ns	8551.666666666666 ns	1.00
array/random/rand/Int64	37375 ns	30127.5 ns	1.24
array/random/rand/Float32	13215 ns	13103 ns	1.01
array/permutedims/4d	52042 ns	52084 ns	1.00
array/permutedims/2d	52826 ns	52953 ns	1.00
array/permutedims/3d	53261.5 ns	53412.5 ns	1.00
array/sorting/1d	2735692 ns	2736660 ns	1.00
array/sorting/by	3307835 ns	3305151.5 ns	1.00
array/sorting/2d	1076070 ns	1069633 ns	1.01
cuda/synchronization/stream/auto	1018.1 ns	1037.9 ns	0.98
cuda/synchronization/stream/nonblocking	7654.6 ns	7454.700000000001 ns	1.03
cuda/synchronization/stream/blocking	831.952380952381 ns	833.8589743589744 ns	1.00
cuda/synchronization/context/auto	1154.8 ns	1167.8 ns	0.99
cuda/synchronization/context/nonblocking	7605.3 ns	8361.2 ns	0.91
cuda/synchronization/context/blocking	929.5555555555555 ns	926.1538461538462 ns	1.00

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