Initial compatibility with CUDA 13

[maleadt]

Closes #2831

NVTX disabled because of NVIDIA/NVTX#125

[github-actions]

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

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: 55b4150	Previous: 205c238	Ratio
latency/precompile	42965722378.5 ns	42934926801 ns	1.00
latency/ttfp	6988682035 ns	7008552789 ns	1.00
latency/import	3569251927 ns	3569139582 ns	1.00
integration/volumerhs	9623821 ns	9606581 ns	1.00
integration/byval/slices=1	147113 ns	147311 ns	1.00
integration/byval/slices=3	426103.5 ns	426127 ns	1.00
integration/byval/reference	144992 ns	145282 ns	1.00
integration/byval/slices=2	286580 ns	286537 ns	1.00
integration/cudadevrt	103446 ns	103674 ns	1.00
kernel/indexing	14148 ns	14638.5 ns	0.97
kernel/indexing_checked	14760 ns	15045 ns	0.98
kernel/occupancy	669.1572327044025 ns	669.9465408805031 ns	1.00
kernel/launch	2136.222222222222 ns	2202.4444444444443 ns	0.97
kernel/rand	18262 ns	17466 ns	1.05
array/reverse/1d	19620 ns	20143 ns	0.97
array/reverse/2d	24878 ns	24692 ns	1.01
array/reverse/1d_inplace	10385 ns	11332 ns	0.92
array/reverse/2d_inplace	12033 ns	13662 ns	0.88
array/copy	21221 ns	21281 ns	1.00
array/iteration/findall/int	157623.5 ns	159966.5 ns	0.99
array/iteration/findall/bool	139804 ns	141602 ns	0.99
array/iteration/findfirst/int	164485 ns	163419 ns	1.01
array/iteration/findfirst/bool	166975 ns	165377 ns	1.01
array/iteration/scalar	72738.5 ns	76152 ns	0.96
array/iteration/logical	215702 ns	219912.5 ns	0.98
array/iteration/findmin/1d	45956 ns	47580 ns	0.97
array/iteration/findmin/2d	96547.5 ns	97060 ns	0.99
array/reductions/reduce/Int64/1d	45046.5 ns	43742.5 ns	1.03
array/reductions/reduce/Int64/dims=1	49930.5 ns	47519.5 ns	1.05
array/reductions/reduce/Int64/dims=2	62791 ns	62503 ns	1.00
array/reductions/reduce/Int64/dims=1L	88804 ns	89134 ns	1.00
array/reductions/reduce/Int64/dims=2L	87453 ns	87634.5 ns	1.00
array/reductions/reduce/Float32/1d	33964 ns	35637 ns	0.95
array/reductions/reduce/Float32/dims=1	49514 ns	51967.5 ns	0.95
array/reductions/reduce/Float32/dims=2	59373 ns	59824 ns	0.99
array/reductions/reduce/Float32/dims=1L	52442 ns	52680 ns	1.00
array/reductions/reduce/Float32/dims=2L	69888 ns	70568 ns	0.99
array/reductions/mapreduce/Int64/1d	43216 ns	43514 ns	0.99
array/reductions/mapreduce/Int64/dims=1	46828 ns	46605.5 ns	1.00
array/reductions/mapreduce/Int64/dims=2	62231 ns	62143.5 ns	1.00
array/reductions/mapreduce/Int64/dims=1L	88831 ns	89174 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	86455 ns	87305.5 ns	0.99
array/reductions/mapreduce/Float32/1d	34051 ns	35464 ns	0.96
array/reductions/mapreduce/Float32/dims=1	51567.5 ns	42505.5 ns	1.21
array/reductions/mapreduce/Float32/dims=2	59608 ns	60252 ns	0.99
array/reductions/mapreduce/Float32/dims=1L	52631 ns	52803 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	69982.5 ns	70795 ns	0.99
array/broadcast	20298 ns	20737 ns	0.98
array/copyto!/gpu_to_gpu	11138.5 ns	13192 ns	0.84
array/copyto!/cpu_to_gpu	214056 ns	217123 ns	0.99
array/copyto!/gpu_to_cpu	283193.5 ns	287100 ns	0.99
array/accumulate/Int64/1d	124851.5 ns	126109 ns	0.99
array/accumulate/Int64/dims=1	83703.5 ns	84201 ns	0.99
array/accumulate/Int64/dims=2	158323 ns	158968 ns	1.00
array/accumulate/Int64/dims=1L	1710623 ns	1710638 ns	1.00
array/accumulate/Int64/dims=2L	966448.5 ns	967410.5 ns	1.00
array/accumulate/Float32/1d	109017.5 ns	109994 ns	0.99
array/accumulate/Float32/dims=1	80827 ns	81343 ns	0.99
array/accumulate/Float32/dims=2	148200 ns	148659 ns	1.00
array/accumulate/Float32/dims=1L	1619508 ns	1619411 ns	1.00
array/accumulate/Float32/dims=2L	698507 ns	699433 ns	1.00
array/construct	1296.6 ns	1288.5 ns	1.01
array/random/randn/Float32	43715 ns	45344 ns	0.96
array/random/randn!/Float32	24865 ns	25330 ns	0.98
array/random/rand!/Int64	27314 ns	27554 ns	0.99
array/random/rand!/Float32	8707 ns	8908.333333333334 ns	0.98
array/random/rand/Int64	29719 ns	30218 ns	0.98
array/random/rand/Float32	12933 ns	13361 ns	0.97
array/permutedims/4d	59903.5 ns	60397 ns	0.99
array/permutedims/2d	54172 ns	54394 ns	1.00
array/permutedims/3d	55007 ns	55362 ns	0.99
array/sorting/1d	2757365 ns	2758561 ns	1.00
array/sorting/by	3343832 ns	3368461 ns	0.99
array/sorting/2d	1080853 ns	1089562 ns	0.99
cuda/synchronization/stream/auto	1031.2 ns	1066.6 ns	0.97
cuda/synchronization/stream/nonblocking	7171.2 ns	7691.3 ns	0.93
cuda/synchronization/stream/blocking	811.1578947368421 ns	844.0121951219512 ns	0.96
cuda/synchronization/context/auto	1164.7 ns	1211.4 ns	0.96
cuda/synchronization/context/nonblocking	7068 ns	6881.1 ns	1.03
cuda/synchronization/context/blocking	890.0363636363636 ns	924.7692307692307 ns	0.96

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

[amontoison]

@maleadt It seems that NVIDIA broke a few headers with the release 13.0.
I wanted to start to update the interface of cusparse, cusolver, cublas, etc... But unable to regenerate the Julia wrappers from the scripts in res/wrap.

If you find a way to update the wrappers, I can take care of the high-level interfaces.

[maleadt]

#2842