Closed geoah closed 4 years ago
Alternative fix to #30's checkptr validation for go 1.14. Test is from @calmh's in #29.
The fix is based on go-cmp's discussion on a similar issue.
benchmark old ns/op new ns/op delta Benchmark32/1-12 4.41 4.44 +0.68% Benchmark32/2-12 4.79 4.81 +0.42% Benchmark32/4-12 4.73 4.67 -1.27% Benchmark32/8-12 5.62 5.57 -0.89% Benchmark32/16-12 7.59 7.56 -0.40% Benchmark32/32-12 11.5 11.6 +0.87% Benchmark32/64-12 18.8 18.7 -0.53% Benchmark32/128-12 35.5 35.3 -0.56% Benchmark32/256-12 69.0 69.9 +1.30% Benchmark32/512-12 139 139 +0.00% Benchmark32/1024-12 272 273 +0.37% Benchmark32/2048-12 538 542 +0.74% Benchmark32/4096-12 1077 1083 +0.56% Benchmark32/8192-12 2153 2197 +2.04% BenchmarkPartial32/8-12 146 149 +2.05% BenchmarkPartial32/16-12 173 174 +0.58% BenchmarkPartial32/32-12 231 226 -2.16% BenchmarkPartial32/64-12 191 192 +0.52% BenchmarkPartial32/128-12 209 209 +0.00% Benchmark64/1-12 12.1 12.1 +0.00% Benchmark64/2-12 12.3 12.6 +2.44% Benchmark64/4-12 13.5 13.7 +1.48% Benchmark64/8-12 15.3 15.4 +0.65% Benchmark64/16-12 13.4 13.3 -0.75% Benchmark64/32-12 15.4 15.4 +0.00% Benchmark64/64-12 19.8 19.6 -1.01% Benchmark64/128-12 27.9 28.2 +1.08% Benchmark64/256-12 49.3 45.0 -8.72% Benchmark64/512-12 81.4 78.6 -3.44% Benchmark64/1024-12 150 145 -3.33% Benchmark64/2048-12 289 282 -2.42% Benchmark64/4096-12 561 559 -0.36% Benchmark64/8192-12 1093 1098 +0.46% Benchmark128/1-12 17.1 15.4 -9.94% Benchmark128/2-12 15.2 15.0 -1.32% Benchmark128/4-12 16.4 14.9 -9.15% Benchmark128/8-12 17.1 17.2 +0.58% Benchmark128/16-12 15.7 15.3 -2.55% Benchmark128/32-12 16.5 18.9 +14.55% Benchmark128/64-12 22.5 22.7 +0.89% Benchmark128/128-12 30.8 31.0 +0.65% Benchmark128/256-12 47.5 48.7 +2.53% Benchmark128/512-12 80.4 81.3 +1.12% Benchmark128/1024-12 148 147 -0.68% Benchmark128/2048-12 285 288 +1.05% Benchmark128/4096-12 559 553 -1.07% Benchmark128/8192-12 1071 1095 +2.24%
Sorry it's getting late, that is very wrong.
Alternative fix to #30's checkptr validation for go 1.14. Test is from @calmh's in #29.
The fix is based on go-cmp's discussion on a similar issue.