available_parallelism=18; 15 batches x 10000 calls; us/call median [min,max] direct (0.0005125, 0.0003833, 0.0005458) threads=1 outside=(3.0368333, 2.9182040999999996, 4.872579099999999) inside=(0.0015375000000000002, 0.0015166000000000001, 0.0020417) threads=2 outside=(4.120825, 3.9938209000000002, 4.3873541000000005) inside=(0.0013625, 0.0013, 0.0018167) threads=8 outside=(14.398083300000001, 14.1847875, 14.869708300000001) inside=(0.0013083, 0.0012958, 0.0014290999999999998) threads=18 outside=(26.3802875, 26.301795799999997, 26.703279099999996) inside=(0.0015209, 0.0015208, 0.0023791)