Less allocations in CombineLatest

[danielcweber]

BenchmarkDotNet-results:

Original:

                  Method |      Mean |     Error |    StdDev |   Gen 0 | Allocated |
------------------------ |----------:|----------:|----------:|--------:|----------:|
  CombineLatest_Typical2 |  7.782 us | 0.1516 us | 0.1418 us | 18.1732 |  37.49 KB |
  CombineLatest_Typical3 | 10.117 us | 0.1421 us | 0.1259 us | 19.4092 |  39.94 KB |
  CombineLatest_Typical4 | 12.379 us | 0.1666 us | 0.1558 us | 20.4010 |  41.96 KB |
  CombineLatest_Typical5 | 14.565 us | 0.1809 us | 0.1692 us | 21.2708 |  43.91 KB |
  CombineLatest_Typical6 | 17.205 us | 0.3376 us | 0.4019 us | 22.3083 |  45.91 KB |
  CombineLatest_Typical7 | 19.341 us | 0.3281 us | 0.3069 us | 23.4375 |  48.17 KB |
  CombineLatest_Typical8 | 21.923 us | 0.3786 us | 0.3541 us | 24.3835 |  50.16 KB |
  CombineLatest_Typical9 | 24.930 us | 0.3194 us | 0.2988 us | 25.3601 |  52.18 KB |
 CombineLatest_Typical10 | 26.821 us | 0.2250 us | 0.2105 us | 26.3062 |  54.16 KB |
 CombineLatest_Typical11 | 29.256 us | 0.6369 us | 0.5646 us | 27.3438 |  56.15 KB |
 CombineLatest_Typical12 | 30.725 us | 0.6072 us | 0.5679 us | 28.3203 |  58.15 KB |
 CombineLatest_Typical13 | 34.625 us | 0.5876 us | 0.5209 us | 29.2969 |  60.14 KB |
 CombineLatest_Typical14 | 37.387 us | 0.9611 us | 0.8520 us | 30.2734 |  62.13 KB |
 CombineLatest_Typical15 | 39.565 us | 0.7561 us | 0.7072 us | 31.4941 |  64.66 KB |
 CombineLatest_Typical16 | 42.625 us | 0.4401 us | 0.3902 us | 32.4097 |  66.64 KB |


Less allocations:

                  Method |      Mean |     Error |    StdDev |   Gen 0 | Allocated |
------------------------ |----------:|----------:|----------:|--------:|----------:|
  CombineLatest_Typical2 |  7.994 us | 0.1915 us | 0.4284 us | 18.1732 |  37.49 KB |
  CombineLatest_Typical3 | 10.390 us | 0.2061 us | 0.2205 us | 19.4092 |  39.86 KB |
  CombineLatest_Typical4 | 12.876 us | 0.1537 us | 0.1438 us | 20.4010 |  41.84 KB |
  CombineLatest_Typical5 | 15.253 us | 0.1810 us | 0.1693 us | 21.2708 |  43.79 KB |
  CombineLatest_Typical6 | 17.808 us | 0.4094 us | 0.5324 us | 22.2168 |  45.76 KB |
  CombineLatest_Typical7 | 20.226 us | 0.3981 us | 0.4088 us | 23.2544 |     48 KB |
  CombineLatest_Typical8 | 22.777 us | 0.3718 us | 0.3478 us | 24.3835 |  49.97 KB |
  CombineLatest_Typical9 | 25.220 us | 0.1248 us | 0.0974 us | 25.2991 |  51.96 KB |
 CombineLatest_Typical10 | 28.183 us | 0.4242 us | 0.3968 us | 26.3062 |  53.95 KB |
 CombineLatest_Typical11 | 30.336 us | 0.3151 us | 0.2793 us | 27.1912 |  55.92 KB |
 CombineLatest_Typical12 | 32.899 us | 0.6381 us | 0.7349 us | 28.1372 |   57.9 KB |
 CombineLatest_Typical13 | 36.574 us | 0.6383 us | 0.5971 us | 29.1138 |  59.84 KB |
 CombineLatest_Typical14 | 38.209 us | 0.5974 us | 0.5588 us | 30.0903 |   61.8 KB |
 CombineLatest_Typical15 | 40.917 us | 0.5025 us | 0.4196 us | 31.2500 |  64.29 KB |
 CombineLatest_Typical16 | 43.664 us | 0.8243 us | 0.7711 us | 32.2266 |  66.32 KB |

[glennawatson]

Be curious if it'd be worthwhile having tests running over a much larger sample. Like 10000 instead of like 16. I see the trend though, slight decrease in speed, decrease in allocs.

[danielcweber]

The overload takes a max of 16 observables. You can combine more observables but that will be handled by some other code, not in the scope of the optimizations of this PR.

[danielcweber]

Do you have an idea what would make speed decrease slightly? There wasn't any additional workload introduced.