Work around NVML issue on Jetson Orin.

[maleadt]

Works around #2580 until we have a proper issue.

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 1 line in your changes missing coverage. Please review.

Project coverage is 73.50%. Comparing base (6bccad6) to head (6b7319e).
Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
lib/nvml/NVML.jl	50.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master    #2620       +/-   ##
===========================================
+ Coverage   11.24%   73.50%   +62.26%     
===========================================
  Files         152      157        +5     
  Lines       14923    15228      +305     
===========================================
+ Hits         1678    11194     +9516     
+ Misses      13245     4034     -9211     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[github-actions]

CUDA.jl Benchmarks

Benchmark suite	Current: 6b7319e	Previous: 774abc6	Ratio
latency/precompile	45436554103.5 ns	45532671418 ns	1.00
latency/ttfp	6431238243.5 ns	6382276443.5 ns	1.01
latency/import	3058519609.5 ns	3039078540.5 ns	1.01
integration/volumerhs	9568375 ns	9567627 ns	1.00
integration/byval/slices=1	147044 ns	146713 ns	1.00
integration/byval/slices=3	426020 ns	425286 ns	1.00
integration/byval/reference	145116 ns	144622 ns	1.00
integration/byval/slices=2	286382 ns	286077 ns	1.00
integration/cudadevrt	103646 ns	103283 ns	1.00
kernel/indexing	14223 ns	14073 ns	1.01
kernel/indexing_checked	15450.5 ns	15126 ns	1.02
kernel/occupancy	686.1895424836601 ns	710.5460992907801 ns	0.97
kernel/launch	2057.6 ns	2120.3 ns	0.97
kernel/rand	15051 ns	14743 ns	1.02
array/reverse/1d	19755.5 ns	19325.5 ns	1.02
array/reverse/2d	25051 ns	24669 ns	1.02
array/reverse/1d_inplace	11034 ns	10913.666666666666 ns	1.01
array/reverse/2d_inplace	12765 ns	11253 ns	1.13
array/copy	21133 ns	20229 ns	1.04
array/iteration/findall/int	158802 ns	157863.5 ns	1.01
array/iteration/findall/bool	139144.5 ns	138404.5 ns	1.01
array/iteration/findfirst/int	162357.5 ns	153375 ns	1.06
array/iteration/findfirst/bool	165076.5 ns	154273 ns	1.07
array/iteration/scalar	76458.5 ns	75697 ns	1.01
array/iteration/logical	210618 ns	212853.5 ns	0.99
array/iteration/findmin/1d	41806 ns	41543 ns	1.01
array/iteration/findmin/2d	94268 ns	93933.5 ns	1.00
array/reductions/reduce/1d	40676 ns	35999 ns	1.13
array/reductions/reduce/2d	47706 ns	41907.5 ns	1.14
array/reductions/mapreduce/1d	38099.5 ns	33891.5 ns	1.12
array/reductions/mapreduce/2d	52201.5 ns	41528 ns	1.26
array/broadcast	21688.5 ns	21376 ns	1.01
array/copyto!/gpu_to_gpu	11721 ns	11516 ns	1.02
array/copyto!/cpu_to_gpu	211559 ns	210665 ns	1.00
array/copyto!/gpu_to_cpu	244510 ns	243223.5 ns	1.01
array/accumulate/1d	108964 ns	108164 ns	1.01
array/accumulate/2d	80430 ns	79823.5 ns	1.01
array/construct	1223.35 ns	1284.3 ns	0.95
array/random/randn/Float32	44283.5 ns	49740 ns	0.89
array/random/randn!/Float32	26478 ns	26117 ns	1.01
array/random/rand!/Int64	27293 ns	27030 ns	1.01
array/random/rand!/Float32	8848.666666666666 ns	8836.333333333334 ns	1.00
array/random/rand/Int64	30030 ns	37762.5 ns	0.80
array/random/rand/Float32	13023 ns	13046 ns	1.00
array/permutedims/4d	67646 ns	66810 ns	1.01
array/permutedims/2d	57500 ns	56518 ns	1.02
array/permutedims/3d	59592.5 ns	59273.5 ns	1.01
array/sorting/1d	2933746 ns	2933200.5 ns	1.00
array/sorting/by	3500000.5 ns	3500043 ns	1.00
array/sorting/2d	1085592 ns	1084935 ns	1.00
cuda/synchronization/stream/auto	1090.4 ns	1035.9 ns	1.05
cuda/synchronization/stream/nonblocking	6436.6 ns	6536.8 ns	0.98
cuda/synchronization/stream/blocking	804.5 ns	791.2244897959183 ns	1.02
cuda/synchronization/context/auto	1183.5 ns	1182.9 ns	1.00
cuda/synchronization/context/nonblocking	6649.6 ns	6769.6 ns	0.98
cuda/synchronization/context/blocking	920.390243902439 ns	915.2666666666667 ns	1.01

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

[maleadt]

With this, tests pass on a Jetson Orin Nano devkit.