Commits
Gil authored and GitHub committed bbc8de404cf
Add a benchmark for query execution (#4731) * Use a valid project id This avoids tripping up over newly added validation in the emulator * Allow passing arguments to the emulator e.g. to pass a non-default host * Add a benchmark for reading documents Baseline results: Run on (12 X 2900 MHz CPU s) CPU Caches: L1 Data 32K (x6) L1 Instruction 32K (x6) L2 Unified 262K (x6) L3 Unified 12582K (x1) Load Average: 2.11, 1.96, 1.92 --------------------------------------------------------------------- Benchmark Time CPU Iterations --------------------------------------------------------------------- BM_QueryMatching/0/1 168359 ns 13750 ns 53878 BM_QueryMatching/1/1 179118 ns 16706 ns 40193 BM_QueryMatching/1/10 721218 ns 35729 ns 10000 BM_QueryMatching/10/10 726533 ns 51864 ns 13587 BM_QueryMatching/1/100 5326040 ns 42820 ns 1000 BM_QueryMatching/100/100 5734101 ns 335289 ns 1000 BM_QueryMatching/1/1000 54110241 ns 41310 ns 100 BM_QueryMatching/10/1000 54584725 ns 95060 ns 100 BM_QueryMatching/100/1000 55083753 ns 527750 ns 100 BM_QueryMatching/1000/1000 58299823 ns 29130 ns 100 BM_QueryAll/1 108892 ns 13355 ns 51197 BM_QueryAll/10 657727 ns 43867 ns 10000 BM_QueryAll/100 5786091 ns 328244 ns 1000 BM_QueryAll/1000 60092767 ns 19380 ns 100 * Add a dedicated XCTest/Google Benchmark adapter Remove FSTLevelDBBenchmarkTests. This benchmark appears to explore some trade-off in document-size vs indexing but doesn't seem useful to re-run.