Use PtrArrays.jl to reduce allocations in bulk sampling

[ameligrana]

to compare with #161

[github-actions]

Benchmark Results

	f72a7bc	71b42cf	f72a7bc / 71b42cf
TTFX excluding time to load	0.0524 ± 0 s	0.0526 ± 0 s	0.994 ± 0,0.98 ± 0,0.997 ± 0
code size in bytes	1.51e+04 ± 0 h	1.52e+04 ± 0 h	0.998 ± 0
code size in lines	564 ± 0 h	569 ± 0 h	0.991 ± 0
code size in syntax nodes	4.08e+03 ± 0 h	4.09e+03 ± 0 h	0.998 ± 0
constructor n=100 σ=0.1	5.78 ± 0.12 μs	5.87 ± 0.14 μs	0.997 ± 0.032,0.971 ± 0.19,0.985 ± 0.031
constructor n=100 σ=1.0	6.13 ± 0.21 μs	6.33 ± 0.25 μs	1.02 ± 0.059,1.04 ± 0.15,0.968 ± 0.051
constructor n=100 σ=10.0	6.78 ± 2.6 μs	6.72 ± 0.16 μs	1.01 ± 0.046,0.963 ± 0.36,1.01 ± 0.39
constructor n=100 σ=100.0	10.3 ± 4.1 μs	10.5 ± 3.9 μs	0.991 ± 0.35,0.982 ± 0.56,0.988 ± 0.54
constructor n=1000 σ=0.1	0.0417 ± 0.001 ms	0.0421 ± 0.0011 ms	1 ± 0.039,1.04 ± 0.086,0.99 ± 0.035
constructor n=1000 σ=1.0	0.0438 ± 0.00096 ms	0.045 ± 0.0015 ms	0.983 ± 0.047,0.968 ± 0.097,0.974 ± 0.039
constructor n=1000 σ=10.0	0.0511 ± 0.00085 ms	0.0516 ± 0.0011 ms	1 ± 0.023,0.97 ± 0.088,0.99 ± 0.026
constructor n=1000 σ=100.0	0.0598 ± 0.001 ms	0.06 ± 0.00099 ms	1.01 ± 0.027,0.998 ± 0.066,0.998 ± 0.024
constructor n=10000 σ=0.1	0.417 ± 0.018 ms	0.424 ± 0.026 ms	1 ± 0.086,0.98 ± 0.09,0.984 ± 0.073
constructor n=10000 σ=1.0	0.426 ± 0.025 ms	0.427 ± 0.017 ms	1 ± 0.086,1.09 ± 0.13,0.998 ± 0.07
constructor n=10000 σ=10.0	0.453 ± 0.032 ms	0.456 ± 0.029 ms	0.938 ± 0.11,1.02 ± 0.12,0.994 ± 0.094
constructor n=10000 σ=100.0	0.585 ± 0.023 ms	0.59 ± 0.031 ms	1 ± 0.071,1.01 ± 0.073,0.993 ± 0.065
delete ∘ rand n=100 σ=0.1	4.46 ± 0.19 μs	4.47 ± 0.2 μs	0.996 ± 0.061,1.01 ± 0.061,0.998 ± 0.062
delete ∘ rand n=100 σ=1.0	4.72 ± 0.2 μs	4.73 ± 0.19 μs	0.998 ± 0.057,1 ± 0.057,0.998 ± 0.058
delete ∘ rand n=100 σ=10.0	4.88 ± 0.18 μs	4.91 ± 0.18 μs	0.998 ± 0.052,0.996 ± 0.053,0.994 ± 0.052
delete ∘ rand n=100 σ=100.0	7.29 ± 0.48 μs	7.32 ± 0.45 μs	1 ± 0.089,0.996 ± 0.088,0.996 ± 0.09
delete ∘ rand n=1000 σ=0.1	0.0441 ± 0.00073 ms	0.0442 ± 0.00084 ms	1 ± 0.027,1.01 ± 0.031,0.998 ± 0.025
delete ∘ rand n=1000 σ=1.0	0.048 ± 0.00097 ms	0.048 ± 0.00086 ms	0.995 ± 0.029,1 ± 0.03,1 ± 0.027
delete ∘ rand n=1000 σ=10.0	0.0492 ± 0.00088 ms	0.0494 ± 0.00081 ms	0.995 ± 0.025,1.01 ± 0.028,0.996 ± 0.024
delete ∘ rand n=1000 σ=100.0	0.0541 ± 0.001 ms	0.0543 ± 0.00097 ms	1 ± 0.025,0.995 ± 0.025,0.996 ± 0.026
delete ∘ rand n=10000 σ=0.1	0.481 ± 0.014 ms	0.479 ± 0.013 ms	1 ± 0.037,1 ± 0.046,1.01 ± 0.039
delete ∘ rand n=10000 σ=1.0	0.517 ± 0.014 ms	0.513 ± 0.014 ms	1 ± 0.029,1 ± 0.041,1.01 ± 0.039
delete ∘ rand n=10000 σ=10.0	0.522 ± 0.013 ms	0.523 ± 0.014 ms	1.01 ± 0.034,1 ± 0.04,0.999 ± 0.036
delete ∘ rand n=10000 σ=100.0	0.518 ± 0.013 ms	0.515 ± 0.015 ms	1 ± 0.035,0.999 ± 0.043,1.01 ± 0.039
empty constructor	1.77 ± 0.28 μs	1.94 ± 0.17 μs	0.936 ± 0.22,1.09 ± 0.45,0.912 ± 0.17
intermixed_h n=100 σ=0.1	11.4 ± 0.86 μs	11.5 ± 0.94 μs	1.02 ± 0.13,1.01 ± 0.17,0.991 ± 0.11
intermixed_h n=100 σ=1.0	11.8 ± 1.2 μs	11.8 ± 1.1 μs	1.02 ± 0.13,1.02 ± 0.17,1 ± 0.14
intermixed_h n=100 σ=10.0	11.6 ± 1.4 μs	11.8 ± 1.9 μs	1.03 ± 0.2,1.03 ± 0.34,0.985 ± 0.2
intermixed_h n=100 σ=100.0	12.5 ± 1.3 μs	12.5 ± 1.3 μs	1.02 ± 0.14,1.04 ± 0.19,1 ± 0.15
intermixed_h n=1000 σ=0.1	0.109 ± 0.009 ms	0.115 ± 0.0084 ms	1.01 ± 0.11,1.01 ± 0.11,0.946 ± 0.1
intermixed_h n=1000 σ=1.0	0.112 ± 0.0087 ms	0.111 ± 0.0087 ms	1.02 ± 0.11,1.02 ± 0.11,1 ± 0.11
intermixed_h n=1000 σ=10.0	0.107 ± 0.0093 ms	0.105 ± 0.0087 ms	1.03 ± 0.13,1.02 ± 0.12,1.01 ± 0.12
intermixed_h n=1000 σ=100.0	0.116 ± 0.013 ms	0.115 ± 0.012 ms	1.03 ± 0.16,1.03 ± 0.15,1.01 ± 0.16
intermixed_h n=10000 σ=0.1	1.2 ± 0.19 ms	1.13 ± 0.16 ms	1.01 ± 0.2,1.04 ± 0.25,1.06 ± 0.23
intermixed_h n=10000 σ=1.0	1.19 ± 0.18 ms	1.17 ± 0.13 ms	1.02 ± 0.19,1.01 ± 0.22,1.02 ± 0.19
intermixed_h n=10000 σ=10.0	1.13 ± 0.18 ms	1.09 ± 0.17 ms	0.961 ± 0.24,1.03 ± 0.24,1.04 ± 0.23
intermixed_h n=10000 σ=100.0	1.16 ± 0.22 ms	1.14 ± 0.21 ms	1.01 ± 0.23,1.02 ± 0.21,1.02 ± 0.27
pathological 1	0.0457 ± 0.00031 μs	0.0456 ± 0.00025 μs	1 ± 0.0085,0.998 ± 0.01,1 ± 0.0087
pathological 1′	0.178 ± 0.0018 μs	0.178 ± 0.0016 μs	0.965 ± 0.012,1.01 ± 0.015,1 ± 0.014
pathological 2	0.0636 ± 0.00039 μs	0.0636 ± 0.00026 μs	0.921 ± 0.0055,1 ± 0.0067,1 ± 0.0074
pathological 2′	0.206 ± 0.0013 μs	0.194 ± 0.0016 μs	1.06 ± 0.011,1.03 ± 0.011,1.06 ± 0.011
pathological 2′′	0.229 ± 0.0045 μs	0.221 ± 0.023 μs	0.999 ± 0.21,1.02 ± 0.078,1.04 ± 0.11
pathological 3	16.9 ± 0.23 ns	16.9 ± 0.23 ns	1 ± 0.019,1 ± 0.019,0.998 ± 0.019
pathological 4	0.063 ± 0.00024 μs	0.063 ± 0.00038 μs	1 ± 0.007,1 ± 0.0061,1 ± 0.0072
pathological 4′	0.197 ± 0.0021 μs	0.2 ± 0.0016 μs	1 ± 0.014,0.989 ± 0.013,0.986 ± 0.013
pathological 4′′	0.225 ± 0.015 μs	0.225 ± 0.0065 μs	1.01 ± 0.089,1.05 ± 0.12,1 ± 0.071
pathological 5a	0.0453 ± 0.00028 μs	0.0453 ± 0.0002 μs	0.999 ± 0.0078,0.999 ± 0.009,1 ± 0.0076
pathological 5b	0.0464 ± 0.00043 μs	0.0453 ± 0.00022 μs	1.12 ± 0.029,1.03 ± 0.096,1.02 ± 0.011
pathological 5b′	0.341 ± 0.0036 μs	0.341 ± 0.0035 μs	0.967 ± 0.014,1.02 ± 0.014,0.999 ± 0.015
pathological 5b′′	0.344 ± 0.0033 μs	0.345 ± 0.012 μs	1 ± 0.045,1.01 ± 0.062,0.997 ± 0.037
pathological large compaction (133380-op)	16.4 ± 1.3 ms	15 ± 0.23 ms	0.988 ± 0.021,1.01 ± 0.027,1.1 ± 0.091
pathological medium compaction (1254-op)	0.104 ± 0.013 ms	0.104 ± 0.022 ms	0.996 ± 0.14,1 ± 0.24,0.992 ± 0.25
pathological old compaction (6-op)	0.224 μs	0.212 μs	1.29,1.04,1.06
pathological small compaction (18-op)	0.924 ± 0.012 μs	0.918 ± 0.011 μs	0.993 ± 0.023,1.02 ± 0.017,1.01 ± 0.017
pathological tiny compaction (6-op)	0.287 ± 0.0029 μs	0.282 ± 0.0043 μs	0.99 ± 0.036,1 ± 0.024,1.02 ± 0.019
sample (bulk) n=1000 k=10000 σ=1	0.166 ± 0.02 ms	0.181 ± 0.02 ms	0.91 ± 0.15,0.909 ± 0.16,0.918 ± 0.15
sample (bulk) n=1000 k=10000 σ=100	0.1 ± 0.037 ms	0.113 ± 0.035 ms	0.885 ± 0.43,0.93 ± 0.44,0.891 ± 0.43
sample (bulk) n=1000 k=1000000 σ=1	16.1 ± 2.1 ms	17.2 ± 2.3 ms	0.926 ± 0.19,0.934 ± 0.17,0.934 ± 0.17
sample (bulk) n=1000 k=1000000 σ=100	9.04 ± 3.7 ms	8.63 ± 4.3 ms	0.834 ± 0.55,0.883 ± 0.53,1.05 ± 0.67
sample (bulk) n=1000000 k=10000 σ=1	0.342 ± 0.018 ms	0.479 ± 0.033 ms	0.865 ± 0.088,0.871 ± 0.15,0.713 ± 0.062
sample (bulk) n=1000000 k=10000 σ=100	0.169 ± 0.042 ms	0.175 ± 0.017 ms	0.991 ± 0.26,0.95 ± 0.32,0.966 ± 0.26
sample (bulk) n=1000000 k=1000000 σ=1	21.6 ± 1.3 ms	23.6 ± 0.5 ms	0.864 ± 0.063,0.87 ± 0.068,0.913 ± 0.06
sample (bulk) n=1000000 k=1000000 σ=100	9.83 ± 2.9 ms	9.6 ± 2.3 ms	0.853 ± 0.56,0.931 ± 0.34,1.02 ± 0.39
sample n=100 σ=0.1	25.4 ± 0.66 ns	25.5 ± 0.63 ns	1 ± 0.035,0.979 ± 0.035,0.997 ± 0.036
sample n=100 σ=1.0	29.4 ± 2.3 ns	29.6 ± 2.2 ns	0.998 ± 0.11,0.979 ± 0.11,0.994 ± 0.11
sample n=100 σ=10.0	19.4 ± 5 ns	19.4 ± 4.5 ns	1 ± 0.34,0.978 ± 0.34,0.997 ± 0.35
sample n=100 σ=100.0	16.9 ± 3.8 ns	17 ± 3.8 ns	0.998 ± 0.33,0.985 ± 0.33,0.996 ± 0.32
sample n=1000 σ=0.1	23.3 ± 6.4 ns	23.2 ± 4.6 ns	1.01 ± 0.31,0.989 ± 0.34,1 ± 0.34
sample n=1000 σ=1.0	0.0322 ± 0.0022 μs	0.0322 ± 0.0024 μs	1 ± 0.1,0.979 ± 0.097,0.998 ± 0.1
sample n=1000 σ=10.0	20.4 ± 5.2 ns	20.5 ± 5.2 ns	0.989 ± 0.35,0.969 ± 0.39,0.997 ± 0.36
sample n=1000 σ=100.0	17 ± 3.9 ns	17.1 ± 4.1 ns	1 ± 0.32,0.992 ± 0.36,0.99 ± 0.33
sample n=10000 σ=0.1	0.0318 ± 0.0012 μs	0.0318 ± 0.0013 μs	1.01 ± 0.054,0.984 ± 0.059,0.999 ± 0.055
sample n=10000 σ=1.0	0.0344 ± 0.0014 μs	0.0344 ± 0.0014 μs	1.01 ± 0.06,1 ± 0.15,0.999 ± 0.058
sample n=10000 σ=10.0	21.8 ± 6 ns	21.2 ± 4.9 ns	1.03 ± 0.38,0.96 ± 0.4,1.03 ± 0.37
sample n=10000 σ=100.0	17.1 ± 4.7 ns	17.6 ± 4.2 ns	1.01 ± 0.31,0.963 ± 0.35,0.973 ± 0.35
summarysize n=100 σ=0.1	1.19e+05 ± 0 h	1.19e+05 ± 0 h	1 ± 0
summarysize n=100 σ=1.0	1.19e+05 ± 0 h	1.19e+05 ± 0 h	1 ± 0
summarysize n=100 σ=10.0	1.19e+05 ± 0 h	1.19e+05 ± 0 h	1 ± 0
summarysize n=100 σ=100.0	1.19e+05 ± 0 h	1.19e+05 ± 0 h	1 ± 0
summarysize n=1000 σ=0.1	1.52e+05 ± 0 h	1.52e+05 ± 0 h	1 ± 0
summarysize n=1000 σ=1.0	1.52e+05 ± 0 h	1.52e+05 ± 0 h	1 ± 0
summarysize n=1000 σ=10.0	1.52e+05 ± 0 h	1.52e+05 ± 0 h	1 ± 0
summarysize n=1000 σ=100.0	1.52e+05 ± 0 h	1.52e+05 ± 0 h	1 ± 0
summarysize n=10000 σ=0.1	1.13e+06 ± 0 h	1.13e+06 ± 0 h	1 ± 0
summarysize n=10000 σ=1.0	1.13e+06 ± 0 h	1.13e+06 ± 0 h	1 ± 0
summarysize n=10000 σ=10.0	1.13e+06 ± 0 h	1.13e+06 ± 0 h	1 ± 0
summarysize n=10000 σ=100.0	1.13e+06 ± 0 h	1.13e+06 ± 0 h	1 ± 0
update ∘ rand n=100 σ=0.1	0.0827 ± 0.0022 μs	0.082 ± 0.0021 μs	1 ± 0.037,1 ± 0.037,1.01 ± 0.037
update ∘ rand n=100 σ=1.0	0.0889 ± 0.0026 μs	0.0884 ± 0.0025 μs	1 ± 0.043,1 ± 0.043,1.01 ± 0.041
update ∘ rand n=100 σ=10.0	0.0988 ± 0.0037 μs	0.0987 ± 0.0032 μs	1 ± 0.048,1 ± 0.047,1 ± 0.05
update ∘ rand n=100 σ=100.0	0.175 ± 0.015 μs	0.175 ± 0.015 μs	0.993 ± 0.12,1.01 ± 0.12,1 ± 0.12
update ∘ rand n=1000 σ=0.1	0.0827 ± 0.0022 μs	0.0824 ± 0.0021 μs	1 ± 0.036,1 ± 0.035,1 ± 0.037
update ∘ rand n=1000 σ=1.0	0.0885 ± 0.0019 μs	0.0881 ± 0.0019 μs	1 ± 0.034,1 ± 0.03,1 ± 0.03
update ∘ rand n=1000 σ=10.0	0.0961 ± 0.0018 μs	0.096 ± 0.0018 μs	1.01 ± 0.027,0.998 ± 0.029,1 ± 0.027
update ∘ rand n=1000 σ=100.0	0.171 ± 0.0065 μs	0.17 ± 0.0063 μs	1 ± 0.055,1 ± 0.056,1 ± 0.053
update ∘ rand n=10000 σ=0.1	0.0905 ± 0.0018 μs	0.0899 ± 0.0012 μs	1 ± 0.021,1 ± 0.017,1.01 ± 0.025
update ∘ rand n=10000 σ=1.0	0.0946 ± 0.0018 μs	0.0932 ± 0.001 μs	1 ± 0.018,0.996 ± 0.017,1.02 ± 0.022
update ∘ rand n=10000 σ=10.0	0.0954 ± 0.0018 μs	0.0942 ± 0.0019 μs	1 ± 0.017,0.998 ± 0.016,1.01 ± 0.028
update ∘ rand n=10000 σ=100.0	0.165 ± 0.0032 μs	0.164 ± 0.0022 μs	0.996 ± 0.02,0.999 ± 0.024,1.01 ± 0.024
time_to_load	0.0836 ± 0.001 s	0.083 ± 0.00081 s	0.998 ± 0.019,0.993 ± 0.035,1.01 ± 0.016

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[ameligrana]

Seems slower than the current implementation actually

[ameligrana]

An interesting thing is that this PR got this bad message in Julia Pre (https://github.com/LilithHafner/WeightVectors.jl/actions/runs/15939048226/job/44964315259):

double free or corruption (!prev)
[2192] signal 6 (-6): Aborted

not sure why, but maybe could be interesting to you @LilithHafner