Get rid of workspace buffers when requiring a new one.

[maleadt]

This drops the libraries/cusolver/dense_generic which has been failing recently from 3710 MB RSS to 2110 MB.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.95%. Comparing base (b4d4284) to head (8b1a02b).
⚠️ Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3062      +/-   ##
==========================================
+ Coverage   89.74%   89.95%   +0.21%     
==========================================
  Files         159      159              
  Lines       13374    13374              
==========================================
+ Hits        12002    12031      +29     
+ Misses       1372     1343      -29     

☔ 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: 8b1a02b	Previous: b4d4284	Ratio
latency/precompile	4058602565 ns	4061473970 ns	1.00
latency/ttfp	14087237272 ns	14133764963 ns	1.00
latency/import	3542171230 ns	3547097268 ns	1.00
integration/volumerhs	9446785 ns	9426493 ns	1.00
integration/byval/slices=1	146076 ns	145913 ns	1.00
integration/byval/slices=3	423162 ns	423161 ns	1.00
integration/byval/reference	144173 ns	143947 ns	1.00
integration/byval/slices=2	284748 ns	284734 ns	1.00
integration/cudadevrt	102772 ns	102560 ns	1.00
kernel/indexing	13565 ns	13222 ns	1.03
kernel/indexing_checked	14271 ns	13991 ns	1.02
kernel/occupancy	659.5465838509317 ns	673.639751552795 ns	0.98
kernel/launch	2162.5 ns	2213.3333333333335 ns	0.98
kernel/rand	18279 ns	16428 ns	1.11
array/reverse/1d	18306 ns	18516 ns	0.99
array/reverse/2dL_inplace	66102.5 ns	66029 ns	1.00
array/reverse/1dL	68897 ns	68986 ns	1.00
array/reverse/2d	20632 ns	21144 ns	0.98
array/reverse/1d_inplace	10466.833333333332 ns	10316.333333333334 ns	1.01
array/reverse/2d_inplace	10670 ns	10547 ns	1.01
array/reverse/2dL	72742 ns	73179 ns	0.99
array/reverse/1dL_inplace	65960 ns	65885 ns	1.00
array/copy	18593 ns	18715 ns	0.99
array/iteration/findall/int	148897.5 ns	149711.5 ns	0.99
array/iteration/findall/bool	131473 ns	132274 ns	0.99
array/iteration/findfirst/int	82767.5 ns	83298 ns	0.99
array/iteration/findfirst/bool	80893 ns	81740 ns	0.99
array/iteration/scalar	68739 ns	68091 ns	1.01
array/iteration/logical	199703.5 ns	203632 ns	0.98
array/iteration/findmin/1d	84759 ns	87161.5 ns	0.97
array/iteration/findmin/2d	116619 ns	117314 ns	0.99
array/reductions/reduce/Int64/1d	42709 ns	43093.5 ns	0.99
array/reductions/reduce/Int64/dims=1	51699 ns	43339.5 ns	1.19
array/reductions/reduce/Int64/dims=2	59821 ns	59714.5 ns	1.00
array/reductions/reduce/Int64/dims=1L	87629 ns	87713 ns	1.00
array/reductions/reduce/Int64/dims=2L	84896 ns	84933 ns	1.00
array/reductions/reduce/Float32/1d	34824 ns	35099.5 ns	0.99
array/reductions/reduce/Float32/dims=1	40249.5 ns	49465.5 ns	0.81
array/reductions/reduce/Float32/dims=2	57452.5 ns	56884 ns	1.01
array/reductions/reduce/Float32/dims=1L	51898 ns	51935 ns	1.00
array/reductions/reduce/Float32/dims=2L	70145 ns	69990 ns	1.00
array/reductions/mapreduce/Int64/1d	42594 ns	43043 ns	0.99
array/reductions/mapreduce/Int64/dims=1	42456 ns	44383.5 ns	0.96
array/reductions/mapreduce/Int64/dims=2	59973 ns	59605 ns	1.01
array/reductions/mapreduce/Int64/dims=1L	87657 ns	87872 ns	1.00
array/reductions/mapreduce/Int64/dims=2L	85167.5 ns	85316 ns	1.00
array/reductions/mapreduce/Float32/1d	34151.5 ns	35025 ns	0.98
array/reductions/mapreduce/Float32/dims=1	39911 ns	41001.5 ns	0.97
array/reductions/mapreduce/Float32/dims=2	57040 ns	56849 ns	1.00
array/reductions/mapreduce/Float32/dims=1L	51996 ns	51771 ns	1.00
array/reductions/mapreduce/Float32/dims=2L	69528 ns	69231 ns	1.00
array/broadcast	20572 ns	20483 ns	1.00
array/copyto!/gpu_to_gpu	11131 ns	11330 ns	0.98
array/copyto!/cpu_to_gpu	213964 ns	214913 ns	1.00
array/copyto!/gpu_to_cpu	283062 ns	282237 ns	1.00
array/accumulate/Int64/1d	118404 ns	118707 ns	1.00
array/accumulate/Int64/dims=1	79481 ns	79832 ns	1.00
array/accumulate/Int64/dims=2	155850 ns	155666 ns	1.00
array/accumulate/Int64/dims=1L	1693815.5 ns	1695124.5 ns	1.00
array/accumulate/Int64/dims=2L	960659 ns	961014 ns	1.00
array/accumulate/Float32/1d	101080 ns	101589.5 ns	0.99
array/accumulate/Float32/dims=1	76385 ns	76781 ns	0.99
array/accumulate/Float32/dims=2	143233.5 ns	143784 ns	1.00
array/accumulate/Float32/dims=1L	1584876 ns	1585667 ns	1.00
array/accumulate/Float32/dims=2L	656755 ns	657563 ns	1.00
array/construct	1280.8 ns	1327.7 ns	0.96
array/random/randn/Float32	44009 ns	38841 ns	1.13
array/random/randn!/Float32	27321.5 ns	31414 ns	0.87
array/random/rand!/Int64	34378 ns	33583 ns	1.02
array/random/rand!/Float32	8588.333333333334 ns	8570 ns	1.00
array/random/rand/Int64	32498.5 ns	37488 ns	0.87
array/random/rand/Float32	13223 ns	13104 ns	1.01
array/permutedims/4d	51659 ns	52907.5 ns	0.98
array/permutedims/2d	52304 ns	52478 ns	1.00
array/permutedims/3d	52716 ns	53034 ns	0.99
array/sorting/1d	2734941 ns	2735432.5 ns	1.00
array/sorting/by	3303943 ns	3304825.5 ns	1.00
array/sorting/2d	1068616 ns	1068239 ns	1.00
cuda/synchronization/stream/auto	1088 ns	1031.3 ns	1.05
cuda/synchronization/stream/nonblocking	7461.1 ns	7742.8 ns	0.96
cuda/synchronization/stream/blocking	860.6739130434783 ns	831.5466666666666 ns	1.04
cuda/synchronization/context/auto	1241.4 ns	1181 ns	1.05
cuda/synchronization/context/nonblocking	6884.8 ns	8005.7 ns	0.86
cuda/synchronization/context/blocking	947.741935483871 ns	938.6410256410256 ns	1.01

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