that's great, you've written a very fast driver. My benchmarks are comparing psycopg2 to itself, under threads / gevent vs. asyncio, to show that asyncio is in

that's great, you've written a very fast driver. My benchmarks are comparing psycopg2 to itself, under threads / gevent vs. asyncio, to show that asyncio is inherently slower. It follows that if you wrote your driver and all of its protocol improvements into a traditional blocking model, it would also be much faster. You've written something so fast that it overcomes the latency of asyncio (this whole thread is a huge plug for your driver, btw). It's not that the latency of asyncio is really that big of a deal. The point I try to make is, asyncio != speed. For databases, 95% of the time, it means, a little less speed. Not a big deal, but if you aren't into writing explicit context switching directives throughout 100% of your code when the OS does it for you just fine, it isn't worth it, unless you are in the zone of thousands of arbitrarily performing connections, which is really more of an incoming client connection thing, not usually a database thing.