Evidence that Macros Do NOT Cause GCD Delay or Lost Casts

[LilimoLimomo]

If you're referring to the video in the original post: I took your macro exactly as-is, and I tried the same thing. I hit Blizzard over and over again against a striking dummy. And I hit the Macro over and over again against a striking dummy. Cast and recast of 2.26s.

Can you clarify whether by "over and over again" you mean once per cast, or mashing repeatedly?

Hitting Blizzard directly was fine and comfy. Every cast went off, exactly as expected.

Hitting the macro was a complete disaster. Only 50% of the casts went off. The other 50% of the time, nothing happened.

How could we have such different experiences?

I can hit Blizzard when the "dinner plate" is a mere 5/6th full and have it go off reliably. Doing the same with the macro is a 100% failure rate. This means that /ac doesn't queue actions the same way as hitting the action directly.

This is covered in my first post. When using programmatic queuing, macros will have a shorter queue window than normal actions. Because of this, the timing of when you press them is different. That needs to be taken into account when performing this experiment. Naturally if you press any button outside of the window it's capable of queuing in, you're not going to get the results you're after.

Furthermore:

5 / 6 * 2.26s = 1.88s.

15 frames at 60fps = 0.25s.

1.88s + 0.25s = 2.13s < 2.26s.

I'm not sure what you're trying to say here. Can you perhaps add some labels to clarify?

Which means even the Swiftcast action didn't queue, despite being an oGCD in the common parlance, which normally can queue 1+ seconds in advance of actually being usable.

Earlier you stated that you took my macro "exactly as-is". If that's the case, then the macro wouldn't have tried to cast Swiftcast, because that action wasn't in the macro. Instead, it would have tried to cast Sharpcast, which is an action that is no longer in the game. That's why you're not getting an oGCD cast.

Hopefully these explanations can provide the guidance you need to give the test another go by performing it the way I did! If it helps, there's a visual effect that happens on the crosshotbar when a button is pressed, so you can watch my video to get a better idea of when I'm pressing the button.

But the real point of contention is the following:

That latter point is a quirk of the few jobs and actions that feature actions with "cast time >= recast time." If you want to make a convincing point about macros and performance, you need to demonstrate the value for the more common case where "cast time < recast time."

This is a genuinely good point; it's entirely possible that the whole "caster tax" thing might mean that such actions aren't impacted as meaningfully by macros as actions that more strictly depend on the GCD. That's well worth doing another round of tests for! Thank you for pointing this out.