Performance Measurements and Comparisons

[itteerde]

Establish performance indicators, and make them useable, regression testing and testing if improvements are improvements indeeed.

[itteerde]

• Pragmatic step 1, reproduce the behavior on fresh installs, is there a general problem in the first place, or is there something instance, if not scene specific (sadly those are hard to profile [changing that would be the ultimate goal, short of fixing if there is something to fix that is]).
• Did setup two empty fresh 11.308 instances for 1:1 comparisons between dnd5e and cpr

1. Install system (and only that).
2. Create scene "start", all defaults accepted (hope those are the same for both systems). Remove Token Vision.

dnd5e

1. Create new Actor "pc1", Player Character, all defaults.
2. Create new Actor "npc1", Non-Player Character, all defaults.
3. Puth 1 Token for "pc1" and 100 Tokens for "npc1" on "start".
4. Move "npc1" with computer game like expectation of responsiveness.
   1. No decernable lag at all.

cpr

1. Create new Actor "pc1", Player Character, all defaults.
2. Create new Actor "npc1", Mook, all defaults.
3. Puth 1 Token for "pc1" and 100 Tokens for "npc1" on "start".
4. Move "npc1" with computer game like expectation of responsiveness.
   1. No decernable lag at all.

[itteerde]

One difference that is there already with the fresh 101 is that in dnd5e the block of 100 can be moved without any perceivable lag, too, while in cpr moving that block of 100 Tokens has a clear lag -- though not at all an alarming one in itself (probably below 1s, that should be measureable easy enough with using a Macro to move them [@itteerde Write that Macro!]). The comparison however might be relevant.

so what to call for the movement in a measured context (Macro)?: https://discord.com/channels/170995199584108546/699750150674972743/1148278914230276199

[itteerde]

the simple highest level macro

const start = performance.now();
await canvas.tokens.moveMany({ dx: 1, dy: 1 });
console.log(`time measured: ${performance.now() - start}ms`);

[itteerde]

first line each: PC selected, second line each: all Mooks/NPCs selected CPR 101 tokens (1 PC, 100 Mooks) at otherwise empty scene DND 101 (1 PC, 100 NPC) tokens at otherwise empty scene the explosion here is only in moving the many, but there is a significant hit to moving just the one PC, too. Next step should be to demonstrate a condition where that moving one explodes, too -- as it did in the game blocking/ session aborting scene.

[itteerde]

looking into any kinds of warming up or building up blockage scenarios

CPR, moving the 1 PC repeatedly

DND, moving the 1 PC repeatedly

oh!

[itteerde]

pf2e

[itteerde]

Measurements of ~0ms are hitting the macro to quickly. Other than that I think the numbers are reliable enough.

[itteerde]

[itteerde]

full text search indicates that the registering hooks section is complete, and no other registering is hidden anywhere

in cpr.js

registerHooks();

calling hooks.js

export default function registerHooks() {
  actorHooks.default();
  actorSheetHooks.default();
  chatHooks.default();
  folderHooks.default();
  itemHooks.default();
  tokenHudHooks.default();
  uiHooks.default();
  hotbarHooks.default();
  tokenHooks.default();
  externalHooks.default();
}

[itteerde]

therefore WTF?! indeed. The answer probably being polymorphism.

[itteerde]

What is happening here? Is it really destroying the old one and making a new one? And it is doing it for every Token, despite only one requiring to be mutated?

[itteerde]

how can it be 1/2 time today? I mean that is still to slow, but it is way faster than ever before.

[itteerde]

Not really getting anywhere, next step destructive testing trying to get some idea what at all is making an impact, an in parallel reading the whole Foundry code.