Probe rocSPARSE with Libdl only

[luraess]

Relates to #1040. Follow-up to #1001.

#1001 moved the rocSPARSE version query into a subprocess so a segfault on a broken ROCm install (#920) degrades to err instead of killing the session. But the child ran using AMDGPU under the active project which is expensive.

The child now uses Base.Libc.Libdl only: it dlopens librocsparse by the absolute path the parent already knows and calls rocsparse_create_handle / rocsparse_get_version directly. The result is memoized for the session.

cc @simeonschaub

[github-actions]

AMDGPU.jl Benchmarks

Details

Benchmark suite	Current: b05f351	Previous: 16f5974	Ratio
amdgpu/synchronization/context/device	550 ns	755 ns	0.73
amdgpu/synchronization/stream/blocking	230 ns	317.5 ns	0.72
amdgpu/synchronization/stream/nonblocking	310 ns	432.25 ns	0.72
array/accumulate/Float32/1d	73244 ns	97092.25 ns	0.75
array/accumulate/Float32/dims=1	248444.5 ns	299718.5 ns	0.83
array/accumulate/Float32/dims=1L	79688.75 ns	121446.75 ns	0.66
array/accumulate/Float32/dims=2	71968.75 ns	130896.75 ns	0.55
array/accumulate/Float32/dims=2L	2617124.75 ns	2806321.5 ns	0.93
array/accumulate/Int64/1d	77849 ns	113217 ns	0.69
array/accumulate/Int64/dims=1	242724.5 ns	269376.25 ns	0.90
array/accumulate/Int64/dims=1L	84306.5 ns	144501.75 ns	0.58
array/accumulate/Int64/dims=2	83984 ns	127879.25 ns	0.66
array/accumulate/Int64/dims=2L	2892025.5 ns	2948585.75 ns	0.98
array/broadcast	73151.25 ns	64054.75 ns	1.14
array/construct	2027.5 ns	3220 ns	0.63
array/copy	36905.5 ns	39815 ns	0.93
array/copyto!/cpu_to_gpu	111524.75 ns	94539.5 ns	1.18
array/copyto!/gpu_to_cpu	110772.25 ns	95039.5 ns	1.17
array/copyto!/gpu_to_gpu	48313.25 ns	42115 ns	1.15
array/iteration/findall/bool	138179.25 ns	198036.5 ns	0.70
array/iteration/findall/int	150259.5 ns	207411.5 ns	0.72
array/iteration/findfirst/bool	146234.5 ns	167311.75 ns	0.87
array/iteration/findfirst/int	145174.5 ns	155156.75 ns	0.94
array/iteration/findmin/1d	112614 ns	179444.25 ns	0.63
array/iteration/findmin/2d	108961.5 ns	161089.25 ns	0.68
array/iteration/logical	247388.25 ns	317088.25 ns	0.78
array/iteration/scalar	296551.5 ns	343653.25 ns	0.86
array/permutedims/2d	72169 ns	82694.75 ns	0.87
array/permutedims/3d	71301.25 ns	82232.25 ns	0.87
array/permutedims/4d	73846.25 ns	84369.75 ns	0.88
array/random/rand/Float32	45208.5 ns	53677.5 ns	0.84
array/random/rand/Int64	54011 ns	66274.75 ns	0.81
array/random/rand!/Float32	65121.25 ns	49389.75 ns	1.32
array/random/rand!/Int64	53471 ns	59654.75 ns	0.90
array/random/randn/Float32	79321.5 ns	93112.25 ns	0.85
array/random/randn!/Float32	80596.5 ns	69269.75 ns	1.16
array/reductions/mapreduce/Float32/1d	97961.25 ns	141249.25 ns	0.69
array/reductions/mapreduce/Float32/dims=1	84738.75 ns	105462 ns	0.80
array/reductions/mapreduce/Float32/dims=1L	830031 ns	862533.5 ns	0.96
array/reductions/mapreduce/Float32/dims=2	84221.25 ns	108059.5 ns	0.78
array/reductions/mapreduce/Float32/dims=2L	142407.75 ns	155446.75 ns	0.92
array/reductions/mapreduce/Int64/1d	97658.75 ns	140932 ns	0.69
array/reductions/mapreduce/Int64/dims=1	83453.5 ns	104764.5 ns	0.80
array/reductions/mapreduce/Int64/dims=1L	831688.25 ns	860696 ns	0.97
array/reductions/mapreduce/Int64/dims=2	84576 ns	107757 ns	0.78
array/reductions/mapreduce/Int64/dims=2L	145124.25 ns	163496.75 ns	0.89
array/reductions/reduce/Float32/1d	98161.25 ns	145184.25 ns	0.68
array/reductions/reduce/Float32/dims=1	86813.75 ns	105042 ns	0.83
array/reductions/reduce/Float32/dims=1L	829623.5 ns	856898.25 ns	0.97
array/reductions/reduce/Float32/dims=2	84368.75 ns	106659.5 ns	0.79
array/reductions/reduce/Float32/dims=2L	144459.5 ns	161129.25 ns	0.90
array/reductions/reduce/Int64/1d	98753.75 ns	141302 ns	0.70
array/reductions/reduce/Int64/dims=1	84486 ns	105144.5 ns	0.80
array/reductions/reduce/Int64/dims=1L	834881 ns	851721 ns	0.98
array/reductions/reduce/Int64/dims=2	84691.25 ns	107839.5 ns	0.79
array/reductions/reduce/Int64/dims=2L	144876.75 ns	162989.25 ns	0.89
array/reverse/1d	45693 ns	53687.25 ns	0.85
array/reverse/1dL	70586 ns	78502.25 ns	0.90
array/reverse/1dL_inplace	80313.5 ns	62952.25 ns	1.28
array/reverse/1d_inplace	60795.75 ns	45130 ns	1.35
array/reverse/2d	50175.75 ns	55499.75 ns	0.90
array/reverse/2dL	82723.5 ns	111822 ns	0.74
array/reverse/2dL_inplace	75833.5 ns	74057 ns	1.02
array/reverse/2d_inplace	55493.25 ns	49312.25 ns	1.13
array/sorting/1d	336523.25 ns	359238.75 ns	0.94
integration/byval/reference	38900 ns	41860 ns	0.93
integration/byval/slices=1	39881 ns	42700 ns	0.93
integration/byval/slices=2	155333 ns	133939 ns	1.16
integration/byval/slices=3	238334 ns	239880 ns	0.99
integration/volumerhs	5013417 ns	4916200 ns	1.02
kernel/indexing	49345.5 ns	44594.75 ns	1.11
kernel/indexing_checked	59145.75 ns	47782.25 ns	1.24
kernel/launch	1422.5 ns	1785 ns	0.80
kernel/rand	98486.25 ns	110472 ns	0.89
latency/import	1706991022 ns	2242598207 ns	0.76
latency/precompile	40075227986 ns	52102215931 ns	0.77
latency/ttfp	2333714860 ns	3070234127 ns	0.76

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