Pool sortables used to create attribute sets

[MrAlias]

The metric SDK converts accepted attribute slices into attribute Sets. Currently it creates a temporary slice to sort the attributes each time a measurement is made and it does so for each aggregation. This change optimizes that Set creation in the following ways:

1. Reuse the created attribute Set for each aggregation. Attribute sets are immutable and the only operation the aggregation will perform on them is to filter. Filtering is an operation that returns a new copied Set. Therefore, reusing the attribute set across all aggregations means allocations only need to be made once.
2. Use a pool of Sortable in the attribute package to create Sets that the user does not provide their own Sortable. This amortizes the cost of allocating a new temporary sorting slice during the Set creation across all Set creations.

Benchmarks

$ benchstat old.txt new.txt
name                                   old time/op    new time/op    delta
CounterAddNoAttrs-8                      41.3ns ± 1%    41.7ns ± 4%     ~     (p=0.483 n=9+10)
CounterAddOneAttr-8                       277ns ± 5%     248ns ± 2%  -10.52%  (p=0.000 n=10+10)
CounterAddOneInvalidAttr-8                391ns ± 6%     362ns ± 2%   -7.49%  (p=0.000 n=10+8)
CounterAddSingleUseAttrs-8               1.02µs ± 3%    0.88µs ± 4%  -14.27%  (p=0.000 n=8+9)
CounterAddSingleUseInvalidAttrs-8        1.31µs ± 3%    1.25µs ± 1%   -4.22%  (p=0.000 n=10+10)
CounterAddSingleUseFilteredAttrs-8       1.43µs ± 1%    1.41µs ± 2%   -1.17%  (p=0.016 n=8+10)
CounterCollectOneAttr-8                   327ns ± 7%     302ns ± 5%   -7.63%  (p=0.000 n=9+10)
CounterCollectTenAttrs-8                 3.12µs ± 8%    2.87µs ± 5%   -8.24%  (p=0.000 n=10+9)
CollectHistograms/1-8                     527ns ± 7%     512ns ± 5%     ~     (p=0.236 n=9+8)
CollectHistograms/5-8                    1.95µs ± 4%    1.95µs ± 5%     ~     (p=0.813 n=10+10)
CollectHistograms/10-8                   3.90µs ±11%    3.64µs ± 3%   -6.58%  (p=0.007 n=10+9)
CollectHistograms/25-8                   8.96µs ± 7%    8.93µs ± 6%     ~     (p=0.971 n=10+10)
Instrument/instrumentImpl/aggregate-8    4.22µs ± 4%    3.94µs ±23%     ~     (p=0.481 n=10+10)
Instrument/observable/observe-8          4.01µs ± 6%    3.90µs ±20%     ~     (p=0.661 n=10+9)
ManualReader/Collect-8                   1.22µs ± 3%    1.22µs ± 1%     ~     (p=0.813 n=10+7)
InstrumentCreation-8                     7.93µs ± 7%    7.97µs ± 4%     ~     (p=0.436 n=10+10)
PeriodicReader/Collect-8                 1.24µs ± 1%    1.26µs ± 2%   +1.43%  (p=0.001 n=10+8)

name                                   old alloc/op   new alloc/op   delta
CounterAddNoAttrs-8                       0.00B          0.00B          ~     (all equal)
CounterAddOneAttr-8                        152B ± 0%      128B ± 0%  -15.79%  (p=0.000 n=10+10)
CounterAddOneInvalidAttr-8                 280B ± 0%      256B ± 0%   -8.57%  (p=0.000 n=10+10)
CounterAddSingleUseAttrs-8                 275B ± 0%      228B ± 1%  -16.93%  (p=0.000 n=8+9)
CounterAddSingleUseInvalidAttrs-8          403B ± 0%      379B ± 0%   -5.96%  (p=0.000 n=9+10)
CounterAddSingleUseFilteredAttrs-8         595B ± 0%      571B ± 0%   -4.03%  (p=0.000 n=9+10)
CounterCollectOneAttr-8                    168B ± 0%      144B ± 0%  -14.29%  (p=0.000 n=10+10)
CounterCollectTenAttrs-8                 1.54kB ± 0%    1.30kB ± 0%  -15.62%  (p=0.000 n=10+10)
CollectHistograms/1-8                      592B ± 0%      592B ± 0%     ~     (all equal)
CollectHistograms/5-8                    2.64kB ± 0%    2.64kB ± 0%     ~     (all equal)
CollectHistograms/10-8                   5.20kB ± 0%    5.20kB ± 0%     ~     (all equal)
CollectHistograms/25-8                   13.1kB ± 0%    13.1kB ± 0%     ~     (all equal)
Instrument/instrumentImpl/aggregate-8    1.18kB ± 2%    0.89kB ± 5%  -24.44%  (p=0.000 n=10+8)
Instrument/observable/observe-8          1.17kB ± 3%    0.91kB ± 2%  -21.97%  (p=0.000 n=10+8)
ManualReader/Collect-8                     160B ± 0%      160B ± 0%     ~     (all equal)
InstrumentCreation-8                     2.37kB ± 0%    2.37kB ± 0%     ~     (all equal)
PeriodicReader/Collect-8                   160B ± 0%      160B ± 0%     ~     (all equal)

name                                   old allocs/op  new allocs/op  delta
CounterAddNoAttrs-8                        0.00           0.00          ~     (all equal)
CounterAddOneAttr-8                        3.00 ± 0%      2.00 ± 0%  -33.33%  (p=0.000 n=10+10)
CounterAddOneInvalidAttr-8                 3.00 ± 0%      2.00 ± 0%  -33.33%  (p=0.000 n=10+10)
CounterAddSingleUseAttrs-8                 3.00 ± 0%      2.00 ± 0%  -33.33%  (p=0.000 n=10+10)
CounterAddSingleUseInvalidAttrs-8          3.00 ± 0%      2.00 ± 0%  -33.33%  (p=0.000 n=10+10)
CounterAddSingleUseFilteredAttrs-8         5.00 ± 0%      4.00 ± 0%  -20.00%  (p=0.000 n=10+10)
CounterCollectOneAttr-8                    4.00 ± 0%      3.00 ± 0%  -25.00%  (p=0.000 n=10+10)
CounterCollectTenAttrs-8                   31.0 ± 0%      21.0 ± 0%  -32.26%  (p=0.000 n=10+10)
CollectHistograms/1-8                      6.00 ± 0%      6.00 ± 0%     ~     (all equal)
CollectHistograms/5-8                      22.0 ± 0%      22.0 ± 0%     ~     (all equal)
CollectHistograms/10-8                     42.0 ± 0%      42.0 ± 0%     ~     (all equal)
CollectHistograms/25-8                      102 ± 0%       102 ± 0%     ~     (all equal)
Instrument/instrumentImpl/aggregate-8      7.00 ± 0%      2.00 ± 0%  -71.43%  (p=0.000 n=10+10)
Instrument/observable/observe-8            7.00 ± 0%      2.00 ± 0%  -71.43%  (p=0.000 n=10+10)
ManualReader/Collect-8                     5.00 ± 0%      5.00 ± 0%     ~     (all equal)
InstrumentCreation-8                       70.0 ± 0%      70.0 ± 0%     ~     (all equal)
PeriodicReader/Collect-8                   5.00 ± 0%      5.00 ± 0%     ~     (all equal)

This amortizes the allocation of a Sortable across multiple instrument measurements, meaning the benchmarks show one less allocation per measurement per aggregation.

[codecov]

Codecov Report

Merging #3832 (83c7b62) into main (4af35a9) will increase coverage by 0.0%.
The diff coverage is 100.0%.

Additional details and impacted files

@@          Coverage Diff          @@
##            main   #3832   +/-   ##
=====================================
  Coverage   81.8%   81.8%           
=====================================
  Files        169     169           
  Lines      12570   12581   +11     
=====================================
+ Hits       10283   10294   +11     
  Misses      2072    2072           
  Partials     215     215           

Impacted Files	Coverage Δ
attribute/set.go	62.8% <100.0%> (+0.9%)	⬆️
sdk/metric/instrument.go	96.0% <100.0%> (+0.2%)	⬆️