Looks like the DOM operation might also be affected, because when you only add event listeners to window, it is much better and the difference is less severe -
a=performance.now(); for (i=0; i<1000;i++) { addEventListener("touchstart", function () {}); addEventListener("touchmove", function () {}); addEventListener("touchend", function () {}); } b=performance.now(); console.log(a,b,b-a);

a=performance.now(); for (i=0; i<1000;i++) { addEventListener("mousedown", function () {}); addEventListener("mousemove", function () {}); addEventListener("mouseup", function () {}); } b=performance.now(); console.log(a,b,b-a);
// 222926.515 222960.65000000002 34.13500000000931
// 222961.33500000002 222980.545 19.20999999999185

Well, "better" was a strong word. :) It was better for touch, but mouse got significantly worse.

If I remove the lines in the block that log the performance violation all is much faster (https://cs.chromium.org/chromium/src/third_party/WebKit/Source/core/dom/events/EventTarget.cpp?type=cs&q=%22Consider+marking+event+handler+as%22&sq=package:chromium&l=304)

Seems the slowness all comes from logging the performance violation. Over to the performance api team to determine whether this is fast enough and this issue should be closed.

Pavel, does this match the performance characteristics you'd expect from  PerformanceMonitor::ReportGenericViolation?

Unfortunately, it does, but only in case DevTools are open. https://cs.chromium.org/chromium/src/third_party/WebKit/Source/core/frame/PerformanceMonitor.cpp?sq=package:chromium&l=356 is capturing the stack to point at the offender.

@kozy: capturing 100 stacks at the same location takes 11ms, this sounds slow given your recent improvements on the stack capturing front.

100 day ping: Any update on this bug? 

Ping again. Any update?

100 day ping: Has there been any progress here? FWIW, the index.html file from #1 is still showing ~0.5 ms for mouse events and 20-40 ms for touch events on 71.0.3578.21 on my pixelbook...