#dev_rc_branch

1 messages · Page 2 of 1

full sonnet
#

can you achieve the effect you want with the non-IR variant? does it work there like you want it to work?

verbal stag
#

(Happy to donate anything that people like, btw. I’m aiming for something more realistic but not everyone may like it)

verbal stag
verbal stag
# full sonnet can you achieve the effect you want with the non-IR variant? does it work there ...

Got a bit more time than expected this evening...

So I think this works as we expect but slightly harder to judge in non IR conditions
Top images are x10 brightness to lower images
https://cdn.discordapp.com/attachments/752836925903077418/1063226013598892112/image.png

I've just realised that looking at this kind of thing through NVG (probably with much lower brightness) might be a better indication of success or not. 🤦‍♂️

#

It's possible that when going up on brightness, we'd want a slightly different texture where the drop off from full saturation to zero is a little tighter but I suspect what we have would be OK.
Although that would rely on there being some control over dot colour/brightness in NVG mode too.

(Oh - and all testing done with 2.12 dev build. Probably obvious that was the case but just to clarify!)

verbal stag
#

It would be super useful to expose the texture of the dot to mod makers as not all devices project just a simple 'dot'.

full sonnet
#

hardcoded quite deeply

verbal stag
unreal arrow
#

SOON (tm)

frosty junco
#

Could we imagine an Arma screen to be able to display on request a truncated map with markers on objectives?
Somewhat a zoom on a chunk of the map for briefings.
Currently interested in with auto generated missions (Liberation, Hearts and Minds, etc.) As there is not really a true briefing feature.

full sonnet
#

Imagination has no bounds

frosty junco
#

In fact, the idea is not so much different from the existing Vanilla GPS, but rendered on a larger screen (BI briefing monitor for example).
Player would be able to scroll and zoom the map on it.

lament crow
unreal arrow
#

Personally I'm not against it but it is not up to me to make this decision. make a ticket just in case

mental fulcrum
#

Aren't custom FOVs breaking quite a bit of things?

wise skiff
#

yeah it has issues with several UI related features

umbral orchid
full sonnet
#

Helo
OnScriptError eventhandler

[error text, filename, fileline, fileoffset (character from start), filecontent (the whole function/file as string), stacktrace (literally output of diag_stacktrace)]
? Enough?

# 
addMissionEventHandler ["ScriptError", {copyToClipboard str _this;}];
[] spawn {
  sleep 5;
  call {
    _x = 50/0;
  };
};

->

[
"Zero divisor",
"",
7,
128,
"addMissionEventHandler [""ScriptError"", {copyToClipboard str _this;}];
[] spawn {
  sleep 5;
  call {
    _x = 50/0;
  };
}",
[
["",6,"",[["_thisscript",<spawn>

sleep 5;
call {
  _x = 50/0;
};

],["_this",[]]]],
["",7,"",[["_x",0]]]
]
]
earnest mural
#

that's great Dedmen! I need to install DEV for this, is there anyplace where I can download the exe only or do install the whole thing?

full sonnet
#

switching from stable to dev isn't that big of a download I think

#

and its not there yet

#

asking for feedback before i finish it

earnest mural
#

looks good to me Dedmen, can say more once starting to work with it

teal snow
#

looks incredibly usefull

scarlet root
#

is variable values feasible? of the error line (if any), or even some/all in the same scope?

unreal arrow
lament crow
#

It seems like most veteran arma players always increase the FOV value to at least 80 so that old school tunnel vision is no longer an issue. It modernizes how A3 looks greatly specially when watching YTB videos and comparing to newer games. Anyway I hope these new sliders can be given the uses they deserve. Keep on rocking guys

full sonnet
lament crow
# wise skiff yeah it has issues with several UI related features

If I make a ticket could you point out the UI related features that have issues and at which FOV values? So that the devs can measure if the FOV slider is feasible. Players are already changing the FOV via config files but that can potentially lead to more issues since it requires external tools to calculate the optimal FOV for each resolution etc etc. Having an official way to do it that in the mean time only allows correct values for any given resolution would reduce the chance of UI getting broken in the first place by those players already modifying their FOVs via the current "hacky" way

wise skiff
lament crow
wise skiff
#

tbh i lost hope on thsoe to be fixed a while ago

#

i accepted the bean formed radar in peoples videos and the many many people asking why it is broken

unreal arrow
lament crow
# wise skiff tbh i lost hope on thsoe to be fixed a while ago

Well yeah, I remember loosing hope on those group event handlers and then we eventually got them so... There is always a chance no matter how small the %... The slider itself is another proof of a feature that many had given up on and it is finally coming 😉

lament crow
wise skiff
mental fulcrum
#

Just get a 21:9 pepe_laugh

shrewd flame
#

ultrawide all the way

lament crow
#

What is sad about the FOV situation is that eventually players could have difficulties finding those external FOV calculators currently required for setting your custom FOV properly... so without an official in game way to do it the external way could become one day inviable

foggy vapor
#

even a script command to change the fov would be nice

lament crow
# foggy vapor even a script command to change the fov would be nice

A script command would pave the way for a config slider anyway. And since the UI issues that custom FOV values cause are minor and players use custom FOVs despite them anyway so even if the UI issues are not fixed the benefit of having a safe way to alter FOV is pretty substantial compared to what we have now

glass talon
full sonnet
#

https://feedback.bistudio.com/T166838
Mh. How did noone notice this yet....

If passive radar sensor detects a active radar. The detection is ignored until the active radar is ALSO in visual range...
Engine ignores all sensor detections, if the target isn't ALSO visible by eye or audible by ear 🤔

So even though ActiveRadar hits your PassiveRadar through fog, you cannot detect it because there is fog 🤯

mental fulcrum
#

fog as in the Fog, or draw distance?

full sonnet
#

both

#

if not visually visible, sensors are ignored (EXCEPT datalink, that actually has a special case which overwrites visibility to always be visible, so if something is also detected via datalink, all your sensors work)

full sonnet
#

A radar nyx sends out radar waves in a 16km range.
So passive radar in ~16km distance should detect it.
But in reality (because of the above) its only detected at 7.8km :sad:
(On my current view distance settings)

#

Actually yes, its view distance dependent. When I up my view distance to 12km, it detects the nyx at about 12.9km (it should be 16)

With my previous settings of 1km view distance, it detected at 7.8km.
So passive radar sensors being view distance dependent will also be fixed by that

lament crow
earnest mural
#

Does DEV always show script errors despite of -showscriptErrors ?

harsh orbit
earnest mural
harsh orbit
#

why don't you fix it? meowsweats

#

also are you running dev or diag?

earnest mural
#

fix it? im on DEV

harsh orbit
#

fix the error

earnest mural
#

oh Im testing the new "ScriptError" EH

#

so made the error on purpose 🙂

harsh orbit
#

showScriptErrors doesn't work in 3den if that's where you're testing it

earnest mural
#

Im running mission from eden

harsh orbit
#

same difference

earnest mural
#

you mean it shouldnt even show the black box?

harsh orbit
#

I mean it always shows the error in 3den regardless of whether you have enabled it or not

earnest mural
#

ooh that explains it

harsh orbit
earnest mural
#

ok thx

#

difficult to make new error dialog when the old one is there XD

harsh orbit
#

it doesn't have an override option? meowsweats
what's the point of making it then?!

earnest mural
#

i mean maybe we can get disable the black error box option?

mental fulcrum
#

return something (bool?) to suppress the error box?

#

Could be abused... thomp

harsh orbit
#

well it would make more sense if you could return true in that EH to override it

earnest mural
#

yes

meager sage
#

When i last tested couple of days ago, dev build did indeed show error box even in SP when not asked for blobdoggoshruggoogly

steep gust
#

Does anyone have any more info on this new dev branch feature, and how it can be utilized?
Added: Engine support for weapon optics with an integrated flashlight / laser

late harbor
wise dragon
#

reminds me of that one terminator sight

steel rune
steep sorrel
#

Is there any reason why when I export a map (as in tactical/strategic map, not terrain) as an .SVG, but convert it to another file format (like .PNG or .JPEG), the roads disappear? How do I prevent that from happening?

#

(I also hope this is an appropriate/applicable place to ask, since I have to opt into the Dev branch to be able to use the commands.)

steel rune
#

sounds like an issue with your conversion from SVG think_turtle

light lintel
# full sonnet Actually yes, its view distance dependent. When I up my view distance to 12km, i...

OMG thank you! I knew the visual sensors were view distance dependent (in fact i think even objectViewDistance dependent, iirc oukej said that was by design) but I didn't know passiveradar was as well.

(Visual/IR sensors being limited by objectviewdistance is annoying because often objectviewdistance needs to be quite small to remain playable for many people still on older systems. As a result, sometimes a ground vehicle should be visible to the sensor but it's not unless it's actually rendered, even if there is LOS and engine is on etc).

light lintel
full sonnet
full sonnet
full sonnet
light lintel
#

exactly 😄

#

can finally make missions involving SEAD elements praise_the_sun

scarlet root
#

how does that radar stuff work with knowsAbout limited by viewdistance? (aka AI forgets about units too far away again)

full sonnet
#

Slightly annoying

devenv_mjS5dhPAEs.png
broken musk
#

are there any known game crashes? cant connect to my MP server. 2.11.xxx

pallid cliff
carmine juniper
#

BI be like

full sonnet
earnest mural
#

any comment from devs about the black error box, if we can get rid of it?

carmine juniper
#

In Eden? I guess not

#

In actual mission, it is simply done by untick Show Script Errors

earnest mural
#

yeah

meager sage
#

except on dev it shows in SP anyways :3

earnest mural
#

but now with the new ScriptError EH it's getting in way

full sonnet
#

I like the black box. We should make it larger

pallid cliff
#

Would be awesome if we had the option to change where on screen it's displayed

#

I think I made a ticket a while back about that.

shrewd flame
#

make it full screen. And add WinXP error sound

pallid cliff
#

Right now it's always on top of the debug console

full sonnet
#

getVideoOptions

[
["aspectRatioName","Auto (Stretch)"],
["brightness",1],
["cloudQualityName","VeryLow"],
["displayModeName","Window"],
["dynLightsCount",4],
["dynLightsQualityName","VeryLow"],
["guiScale",0.55],
["guiScaleName","Small"],
["hdrQuality",16],
["hdrQualityName","Normal"],
["objQualityName","VeryLow"],
["objectVisibility",800],
["overallPresetName","VeryLow"],
["overallVisibility",1000],
["particleQualityName","Low"],
["pipQuality",512],
["pipQualityName","VeryLow"],
["pipVisibility",835],
["ppBloom",0],
["ppBrightness",1],
["ppCausticsQualityName","Disabled"],
["ppColorPresetName","Default"],
["ppContrast",1],
["ppDOF",0],
["ppFSAA",1]
["ppPPAAQualityName","Disabled"],
["ppRadialBlur",0],
["ppRotationBlur",0],
["ppSSAOQualityName","Disabled"],
["ppSaturation",1],
["ppSharpenFilter",0],
["refreshRate",60],
["samplingPercentage",100],
["shadowQualityName","Low"],
["shadowVisibility",100],
["terrainQuality",25],
["terrainQualityName","Low"],
["texQualityName","VeryLow"],
["vsync",false],
["waterReflectionQualityName","Disabled"],
]

Feedback? mainly on names or if something is missing, this is HashMap.
Maybe switch XVisibility to visibilityX
I excluded resolution because getResolution command exists? Is that fine or should it be there anyway?

I'll probably also add some extras that aren't actually in the settings screen (refreshRate already is there)
Could also add some non-option things, like what GPU the user is using? And driver version? Is that useful?

pallid cliff
#

I excluded resolution because getResolution command exists? Is that fine or should it be there anyway?
Maybe to have it all in one command.

full sonnet
#

Annoying thing with resolution is, its 3 values. width, height, string
I guess I could pack it as array though

carmine juniper
#

Are those pp returns 0-1?

full sonnet
#

0-2 actually.
The UI display also goes 0-200 (it displays percent?)

#

Some possible extras

["xDesktopHeight",1440],
["xGPUName","NVIDIA GeForce GTX 1660"],
["xInstalledVram",6270484480],
["xMonitorH",1440],
["xMonitorW",2304],
["xMonitorX",0],
["xMonitorY",0]
["xSharedVram",8523067392],
["xSystemVram",0],
["xTextureMemory",0],
pallid cliff
#

Great, now we can punish ppl for having bad grafix 😄

carmine juniper
full sonnet
#

😠 First giv feedback on this and get this done

carmine juniper
#

Then push the Dev Branch!!!!!!!!!!!!!!111eleven

pallid cliff
full sonnet
#

Prooobably, but that doesn't fit in here

pallid cliff
#

easy fix

#

new command getHardwareInfo

#

😛

#

Thinking about that. I guess getVideoOptions should indeed only return what's available in A3

full sonnet
#

There are video options that are available in A3, but just have no UI element to configure them

viscid onyx
#

Anything cool?

pallid cliff
#

give

#

I am sure there is a checkbox called "Double FPS"

full sonnet
#

there is a refreshRate one 😄

pallid cliff
#

GIVE

full sonnet
# 
["xStat0","13 MiB"],
["xStat0N","Buffers"],
["xStat1","122 MiB"],
["xStat1N","Textures"],
["xStat2","46 MiB"],
["xStat2N","RenderTargets"]]
["xStat3","2.0 MiB"],
["xStat3N","Other Texs"],
["xStat4","48 MiB"],
["xStat4N","Other"],
["xStat5","231 MiB"],
["xStat5N","Total"],

VRAM Memory usage.
Thats the last potentially useful extras I could find

#

Oki I think thats my final

[
["GPUName","NVIDIA GeForce GTX 1660"],
["aspectRatioName","Auto (Stretch)"],
["brightness",1],
["cloudQualityName","VeryLow"],
["displayModeName","Window"],
["dynLightsCount",4],
["dynLightsQualityName","VeryLow"],
["guiScale",0.55],
["guiScaleName","Small"],
["hdrQuality",16],
["hdrQualityName","Normal"],
["monitorResolution",[0,0,2304,1440]],
["objQualityName","VeryLow"],
["objectVisibility",800],
["overallPresetName","VeryLow"],
["overallVisibility",1000],
["particleQualityName","Low"],
["pipQuality",512],
["pipQualityName","VeryLow"],
["pipVisibility",835],
["ppBloom",0],
["ppBrightness",1],
["ppCausticsQualityName","Disabled"],
["ppColorPresetName","Default"],
["ppContrast",1],
["ppDOF",0],
["ppFSAA",1]
["ppPPAAQualityName","Disabled"],
["ppRadialBlur",0],
["ppRotationBlur",0],
["ppSSAOQualityName","Disabled"],
["ppSaturation",1],
["ppSharpenFilter",0],
["refreshRate",60],
["samplingPercentage",100],
["shadowQualityName","Low"],
["shadowVisibility",100],
["terrainQuality",25],
["terrainQualityName","Low"],
["texQualityName","VeryLow"],
["vramInfo",["5.8 GiB","7.9 GiB","0 B","231 MiB","5.5 GiB"]],
["vsync",false],
["waterReflectionQualityName","Disabled"],
]

yeah vramInfo and gpuName and monitorResolution are not settings, but it somewhat fits in here and I wouldn't want to make a extra command for such.
Will probably document that vramInfo/monitorResolution are optionals that are not guaranteed to be there. So we can remove it IF we add some hardware info command (which I don't think we will)

tepid dock
#

Why would mod makers actually want to access that? Because the first thing that I can think of is game client fingerprinting (which is not a good thing)

full sonnet
#

Vram is useful to detect 3fps bug like problems in script.
monitor blobdoggoshruggoogly

#

If you want fingerprinting you can grab steamId, or store value into profileNamespace

viscid onyx
#

Should those optional elements be moved to the end of the output array, so [properly configured] params on the output don't get broken if they're removed?

full sonnet
#

its not an array, its a hashmap

viscid onyx
#

Oh, okay

tepid dock
full sonnet
#

Yeah. For people who move between steam accounts and also change their profile and their game window resolution everytime they connect
Then you also have audio volume settings.
Which in basically every case is a malicious actor

There are already enough ways to do fingerprinting if someone wanted to

tepid dock
#

Which in basically every case is a malicious actor
Well, I understand your argument and also understand that there are other ways of fingerprinting.

Just wanted to point out that this makes fingerprinting ever easier (or harder to work around). And while I also understand the part about being a "malicious actor" most of the time, it kinda sounds like a "if you have nothing to hide then you shouldn't be concerned" which always surfaces while talking about privacy.

And as a side note and more of a fun fact: the Wasteland server that I've been playing on actually had whole groups of people (at least one that I know of) that were actively switching both player profiles and steam accounts to go undercover so that others don't know who they are facing and what are their usual tactics. They stopped doing that only when officially asked by admins, after numerous complaints from the community. [no, it wasn't me]

lament crow
full sonnet
#

thats not a setting :U

#

but ye might be an idea thanku

steep gust
steep gust
late harbor
#

I know it's not a video thing, but would be cool if we could get ram usage as well.
Especially for server/headless clients. Currently it's only possible for servers via #monitor.

scarlet root
#

The "Name" values - are these internal names, or localized? 😛

hollow basalt
#

Gamma is missing. Which btw is weird when you screenshot/print screen, doesn't affect the result.

full sonnet
full sonnet
#

mh its missing. Probably a copy paste error 😄 Thank you good eyes
Gamma was the main thing they wanted and I didn't see its missing 😄

hollow basalt
#

yw 👍

tranquil sphinx
#

Is it an array of arrays or a hashmap?

full sonnet
#

hashmap

light lintel
broken musk
#

my client crashes whenever i join server

# 
Build: Development
Version: 2.11.150223```
#

okay narrowing it down

#

@full sonnet @unreal arrow game crash addUserActionEventHandler ['turbo','activate',{}];

broken musk
#

@unreal arrow "windyCoef" appears to either not work, or be slaved to "ambientSound" toggle

#

top of tanoa volcano, enableEnvironment [FALSE,TRUE,0];

unreal arrow
#

gimmie repro because idunno what u mean

broken musk
#

just join as client in mp and exec

lament crow
unreal arrow
broken musk
#

@unreal arrow Version: 2.11.150223

broken musk
#

i tested with dedicated server, not MP From editor

unreal arrow
#

there is missing windy input from one of the wind sound shaders

broken musk
#

manage to repro the user action issue?

full sonnet
carmine juniper
#

Lazymen

full sonnet
#

Result: 0.0013 ms
Code: true;true;true;true;true;true;true;true;

Result: 0.0016 ms
Code: var;var;var;var;var;var;var;var;

->

Result: 0.0009 ms
Code: true;true;true;true;true;true;true;true;

Result: 0.0011 ms
Code: var;var;var;var;var;var;var;var;

The world is saved! 💩

#

Result: 0.0007 ms
Code: true;true;true;true;true;true;true;true;

poggers

mental fulcrum
#

Optimized semicolon? kappa

full sonnet
#

split nular commands and variables into separate instructions

#

ah its set to private, sad

mental fulcrum
#

Restricted, yea.

full sonnet
#

Some ticket from some dedmen some 6 years ago

carmine juniper
#

Huh, strange name

earnest mural
#

so do we get the ability to disable the black error box, now that we have ScriptError EH?

full sonnet
#

no

earnest mural
#

oh that's surprising, can I ask why not?

#

u ok man? 👀

carmine juniper
#

Sorry for the ping but is this pushed into Dev-Branch already?

full sonnet
#

yes should be

#

it was in dev changelog on 11th

carmine juniper
#

Oh, was me did wrong

unreal arrow
broken musk
#

I think its good that errors are forced to show for all

#

designers may feel compelled to fix their stuff

#

but options aren't a bad thing

scarlet root
#

repositioning would be quite handy tho as the current position is far from ideal

unreal arrow
#

and yes sometimes options are a bad thing

carmine juniper
# 
import RscPicture;
class test
{
    movingEnable = 0;
    idd = -1;
    class controlsBackground
    {
        class text1: RscPicture
        {
            x = 0;
            y = 0;
            w = pixelW*512;
            h = pixelH*512;
            text = #(rgb,512,512,3)text(0,0,"Caveat",0.3,"#0000ff7f","#ff0000","Hallo\n"Welt"");
        };
    };
};``` @full sonnet sorry for the ping again - do you know why it forces it to be lowercased?
20230124155907_1.jpg
full sonnet
#

Huh...
It was a thing that all proc texture things are forced lowercase, but that shouldn't be the case anymore

carmine juniper
#

Pun detected

full sonnet
#

Ill check that

carmine juniper
#

Thanks

full sonnet
#

Mh, actually i only tested script. Probably smth on the config reading side

carmine juniper
#

Oh and also, this does seem to be happened only in dialogs

#

Either ctrlSetText or text = attribute, both does lowercased text, but setObjectTexture'd object has proper text

earnest mural
full sonnet
#

yep. The control itself has a forced lowercase
so.. previously it would lowercase each texture path twice... :yay:

full sonnet
earnest mural
#

in the editor (launched) that is (error from console)

full sonnet
#

ye inside editor errors are forced on

earnest mural
#

something I already knew..

carmine juniper
#

Display and texture both uses same texture and almost the same script I've ran, but texture doesn't do what I want - which is to sort the width with the text width using ctrlTextWidth, bug?

20230124201338_1.jpg
#

Wait, I've just realized... somehow the text size is different. Maybe I ain't sure how to control it

full sonnet
#

maybe your script needs to run in context of the render, aka inside onDraw event

#

while the thingy is drawn, the engine's screen size and some other stuff changes.
After drawing it changes back

Maybe the textWidth script command output depends on that

carmine juniper
#

Uff so draw3D?

#

No change, sadly

full sonnet
#

no

#

needs to be event fired by the display/control itself

carmine juniper
#

Ah hmm, onLoad'ing did fixed it?

carmine juniper
#

Yeah try to change things in onLoad does fix it, but it is hard to try

full sonnet
#

isnt there some draw or simulate event or smth

carmine juniper
#

Thought map ctrl only has the onDraw?

full sonnet
#

That indeed seems to be the case notlikemeow I guess I forgot that

#

I'll add it then

#

Done, with a bit of luck that makes it into tomorrow's dev

carmine juniper
#

So both control and display will receive it?

full sonnet
#

only display

#

I thought of this problem before and I was sure I had implemented a solution.
Either i forgot to implement it or I forgot what the solution was

#

Ahhhh there is no display onDraw notlikemeow

#

I hate all of this. I guess I'll add a display onDraw event, that only works with UIOnTexture displays

#

Do wiki please, one parameter, the display. onDraw. Only on as above

carmine juniper
#

Done, feel free to correct

full sonnet
#

no

#

i won't even look at it

carmine juniper
#

😭

steep sorrel
#

When I'm in the Dev branch and I start up the arma3diag .exe, SOG: Prairie Fire isn't showing up in the addons menu, yet all other CDLC is. How do I fix this?

steep sorrel
#

Any idea what causes this? I'm trying to use the SVG export command to make an image copy of the strategic map of Cam Lao Nam, but it doesn't work and gives me an error message that makes no sense. It worked for a map from a mod; I don't understand why the command is having issues with this map.

https://imgur.com/yKCwKAy

Imgur
halcyon moth
#

Did you paste code in from elsewhere?

#

A lot of those are caused by hidden characters.

mental fulcrum
#

You're not running the dev exe

halcyon moth
#

oh yeah :P

hot wave
#

dev branch version should be 2.11

steep sorrel
#

@halcyon moth I copied the example text from the BIS wiki and changed it slightly (as in the name of the saved file, where I want to save it, my username, etc. - I'm good with formatting).

#

Just now, I also tried manually typing out the whole code, but it gave me the error message again.

viscid onyx
#

It will always give you that error because the command flat-out does not work unless you're on dev branch with the diagnostics exe. It does not exist on stable branch.

steep sorrel
#

So if that's the case, then how do I activate SOG Prairie Fire with the diagnostics .exe?

halcyon moth
#

You're putting vn in the mod line when you start the diagnostics exe?

steep sorrel
#

Nope, though interestingly, when I launched the diagnostics .exe, this time it actually loaded SPF by itself.

#

I don't know what changed about it, but it works now.

steep sorrel
#

On the subject of the SVG export command, however, is there any way to export it in a lower resolution or convert it to something smaller?

#

When I exported Cam Lao Nam, it's almost 90 MB.

steel rune
#

It’s vector graphics, it does not have a fixed resolution

#

Disable trees or other features to get smaller files

full sonnet
#

About the UIOnTexture crashing on diag binary, thats replaced by skippable assert with todays(maybe) dev update

full sonnet
# carmine juniper 😭

now is happi time meowheart though it might still be a bit annoying that you have to go through draw handler but welp

#

If there is still something wrong with anything now on dev, pong me

steep gust
#

Does the new change mean we have to use onLoad or can we still use the direct method without onLoad just without the added benefits?

full sonnet
#

you dont need onload anymore

carmine juniper
#

Okay my dream's done

20230126155737_1.jpg
#

Is it correct to describe onDraw will fire upon a displayUpdate?

carmine juniper
#

Huh, is TreeFilterUpdated implemented? Can't get it working

# 
class CreateObjectWEST: ctrlTree
{
    onTreeFilterUpdated = "_this call bra_fnc_bra;";
};```correct?
full sonnet
full sonnet
#

nono it does, the config loading code is definitely there
"onTreeFilterUpdated"

carmine juniper
#

Hmm, it is for a tree not an edit right?

full sonnet
#

yes

#

try adding via script ctrlAddEventHandler. Thats what i tried

mossy wyvern
carmine juniper
#

Nu, it is just lightflare on TPP

#

I meant to show my joy after the tilted text

earnest mural
#

in case you guys didn't know yet you can crash arma if you have script error inside of "ScriptError" EH

full sonnet
#

:poutcat:

#

I knew I forgot something

#

Fixed, the second ScriptError will fire but then no more

#

Thanku meowheart

earnest mural
full sonnet
#

max 2 level recursion

earnest mural
#

right, sounds good 🙂

#

ty

carmine juniper
#

Just realized that TreeFilterUpdated does work properly, but not for Eden Editor's right hand tree view

# 
findDisplay 313 displayCtrl 56 ctrlAddEventHandler ["TreeFilterUpdated",{systemChat str _this}]```Repro
#

No wait, just realized that the Eden tree doesn't use idcSearch notlikemeow

#

How come I've never noticed this

full sonnet
#

Eden editor has a custom tree thingy

#

its not just a normal tree view.
i think it also has custom search

unreal arrow
#

For the love of god dont touch it

broken musk
#

weird effect where screen layer disappears when aim down sight

#

is this intended?

#

example (['TAG_my_layer'] call (missionNamespace getVariable 'BIS_fnc_rscLayer')) cutRsc ['RscCBRN_APR','PLAIN'];

#

aim down sight and it disappears

full sonnet
broken musk
#

hmm ill see if i can repro it

#

@full sonnet false alarm, i did not realize cameraView changes to "GUNNER" on ADS

#

my script checks "INTERNAL","EXTERNAL"

gray quarry
#

so release is even numbers and dev are odd numbers?

winter crescent
ocean sail
unreal arrow
#

no you can only enable it if not enabled, cannot disable it if enabled

#

why would you need to disable it, is there any legitimate reason?

ocean sail
unreal arrow
#

could you give me the class i would check

ocean sail
light lintel
#

this is from 2016, i confirmed it was still an issue a year or two ago

#

Land_BagBunker_Tower_F

#

(i can't check again because no gaming PC)

unreal arrow
#

@light lintel bunker enableaudiofeature ["building_exterior", true] works

light lintel
viscid onyx
#

It seems like the environment is determined by what surface you're standing on. That type of bunker doesn't have a floor surface on the lower level, so it may not be practical to do anything about that.

light lintel
#

ah, unfortunate

#

like the fix would be to make it exterior audio on the top floor as well...

#

but how come it works correctly on the cargo towers? does the roof have a different surface from the interior?

viscid onyx
light lintel
#

Oh, i understand what KK said now i guess. That
bunker enableaudiofeature ["building_exterior", true]
would make it sound like the exterior everywhere in/on the bunker?

unreal arrow
#

Fixing the bunker properly would require model editing which at this point may be a bit problematic, so you can use enableaudiofeature as workaround

ocean sail
unreal arrow
#

Yeah, not the worst idea, I will make a ticket

ocean sail
full sonnet
#

Config runs everywhere where the mod is loaded, so no

frosty junco
#

Are you guys working on a system allowing to "write" directly on objects like planes, helicopters, buildings instead of having to create multiple textures?
As a reference, the high watchtower numbered from 1 to 8. It has 8 textures.
As an example, creating a grey texture for an helicopter and labelling it with "Navy" in various locations instead of creating 2 textures for the same goal (a standard one and a specific one).
Asking for a friend

unreal arrow
full sonnet
#

Well with UIOnTexture you can probably do what you want

#

But "we" are not building such a system

frosty junco
#

Thank you.

winter crescent
full sonnet
#

That new update image/thumbnail thing looks really great. much more informative than the previous ones

solid island
#

so about the visible lasers via config

#

how do?

#

nvm, found it

scarlet root
#

Does 2.12 include a volume slider for playSoundUI or what was again the plan to deal with it?

winter crescent
scarlet root
#

thanks a lot for the info 🙇‍♂️

#

we are using playSoundUI for conversations atm

#

this is as usually people mute radio channel when they dislike the radio protocol
the other option is the regular sound channel (fadeSound, soundVolume) - however you want to decouple those from conversations due

  1. people may want to mute conversations
  2. you can make conversations louder to the regular sounds (aka easier to understand conversations in otherwise loud environment)
#

finally playSoundUI also works outside missions (ie main menu without a cutscene running, briefing or debriefing) - this adds a lot of new opportunity and flexibility

calm wasp
#

I hope that helis not force-exploding anymore when falling on the side is also part of 2.12, because it's been a very long time in perf branch. even before 2.10 @full sonnet

full sonnet
#

its not

calm wasp
#

why?

full sonnet
#

because it wasn't enabled

tepid dock
#

For the record, I didn't have any issues on perf branch that could be tied to that fix, so it looks like it doesn't have side effects (at least in my use-cases)

calm wasp
full sonnet
#

Because noone thought of it

calm wasp
#

well. same for patch 2.10, although it's part of perf branch even since before stable 2.10

#

hope it will come in patch 2.14

steep monolith
#

@past moon is there something wrong with arma3diag.exe at the moment?

#

I just updated Dev Branch, using GameUpdater, and cannot run the Diag exe, re-validated and downloaded the binaries, tried with no params and still

versed abyss
#

What does platform support supposed to mean?

carmine juniper
#

In what sense?

versed abyss
#

In the twitter picture it just says that. But I have no idea what it means.

carmine juniper
#

Ah that Platform Support

#

It should just mean it expands the Engine features

meager sage
#

and/or occasional bugfixes

viscid onyx
#

Platform support is the continued maintenance and development of the technical platform - enginestuff, scriptstuff - that content is built on. It means "code updates but no new content".

viscid onyx
#

For people who've tried the new setTurretLimits command, when the gun is outside the new limits when they get set, does it smoothly traverse to the new limits, or instantly snap there?

meager sage
#

iirc it snaps

ocean sail
#

ye instantly to new limits

unreal arrow
#

yeh

#

I can make it zero if limits change if you have concerns someone may exploit it

lone merlin
#

@steep monolith Nothing is wrong, an ASSERT is not an error. In brief, it's a debug message to inform on a value or something. Is it possible to have your log? (will try to have a look)

broken musk
#

Q about that. are the limits player only or apply to AI as well

carmine juniper
#

AI as well as far as I tested

viscid onyx
unreal arrow
#

the command purpose is to set static limits on static weapon so that it won’t clip near objects, so technically you would set it once after you placed your assets

viscid onyx
#

That's definitely the most important use for it.
Personally though, I'm going to use it to restrict the turret traverse on certain vehicles that have exterior infantry seats, so you can't turn the turret backwards while there's infantry sitting behind it. (I also want to do travel locks, but you can't set negative angles, so the gun could only be locked forward, not backward.)
To be clear I was just asking for information, not complaining. If you want to do something cool like an optional bool for smooth/instant, I'm not gonna say no, but it's not strictly necessary for my purpose so I'm not here to pester you about it.

unreal arrow
#

no, animation requires new implementation, so unlikely

teal snow
#

not sure how feasable but you could adjust the gun limits from a to b over 1 second to fake the affect of a slower transition?

scarlet root
#

new RC out?

ocean sail
#

13-02-2023 (notable changes)

Fixed: Crash if a script error occurs in a config file at game start - FT-T170230

weary niche
#

I've recently been experimenting with the new UIonTexture procedural, but I'm finding that when applying the display to my weapon I can't manage to get the alpha to work. Anyone know how to get the controls to display with transparency?

#

I've also been getting error messages upon loading my weapon with the display that state, "Cannot load mipmap file\path\here.paa"

#

This appears to be the same problem TeTeT was having, but when I asked them, they'd not found a workaround.
#arma3_texture message

full sonnet
weary niche
#

I am going to try to set it up where the texture in the p3d is an image with transparency and apply the UIonTexture via hiddenSelections, but if there was some way to tell the model from the p3d that the UIonTexture procedural should consider transparency, that could be useful.

weary niche
full sonnet
#

A simple procedural should work too

void magnet
#

@steep monolith Aye, mine be working. Sorry just now had time to check it.

steep monolith
void magnet
#

gah, that is sub optimal for sure.

dusty scarab
#

Would it be possible to add a missionEventhandler for "TicketsDepleted" as an alternative to having to end the mission via template?

scarlet root
#

its a scripted system. so you need to make a scripted eventhandler with the right logic

unreal arrow
dusty scarab
#

Yes, I can not remove the scripted event

#

When I try to remove the scripted event that you typed it does nothing

harsh orbit
dusty scarab
#

I thini so

#

Will look up my code in a minute

harsh orbit
#

do you remove all of them or do you know the ID?

dusty scarab
#

I tried to remove em all i think

#

Lemme see bootin pc

dusty scarab
harsh orbit
#

you can't actually do anything with it

harsh orbit
#

just end the mission yourself using that event

#

in other words add an EH not remove the EH (plus there was nothing to remove anyway)

dusty scarab
#

Will it overwrite that event?

harsh orbit
#

no. you just end it sooner before it ends the mission itself

dusty scarab
#

I just want the mission to end when there is no respawn tickets, there is still a possibility to get tickets in the mission im building, so I want players to be forced to spectate until a ticket was scored

#

"do not want"

#

But anyway, I accepted it, it will also work this way. I just track in the script that is called on mission ending if there was no tickets left.

harsh orbit
dusty scarab
#

But I will accept the verdict.

scarlet root
#

@dusty scarab you can also duplicate the relevant scripts of the respawn system, modify them to your needs

dusty scarab
timber sorrel
#

Whats coming up for 3.15?

viscid onyx
#

Given that the current stable version is 2.10, it's a little early to be looking as far forward as 3.15

weary niche
#

I'm trying to use UItoTexture for an ammocounter on a weapon similar to how Polpox had shown earlier in this channel, but I've found that the functionality is fairly limited from the sense that all weapons of this particular type would share the same unique ID. Is there any possible chance that we could get a setting for the UItoTexture procedural that randomizes the unique ID on the display's initialization?

carmine juniper
#

I would think that you only see it from First Person Perspective, so did not make any randomization or such

#

But had to admit such limitation is not good since how a weapon behaves

timber sorrel
viscid onyx
#

We really don't know that either. The next update, which is currently on release candidate branch, is 2.12. After that (probably in about 6 months) will be 2.14. 2.15 will be the dev branch for stable update 2.16, and since that's like a year from now we have no idea what's going to be in it.

limpid ledge
#

Perhaps it will be a 10-year Update 🙂

wise dragon
#

i will personally be releasing 3.15

weary niche
# weary niche I'm trying to use UItoTexture for an ammocounter on a weapon similar to how Polp...

To add even more to this, the feature would be extremely useful for workarounds in regards to scripted texture application to vests, headgear, nightvision, you name it. The main thing is that this would be especially useful for objects that cannot have a scripted use case for UItoTexture where instead of application via setObjectTexture, you are forced to use hiddenSelections or application of the texture through the p3d.

full sonnet
#

Well possible for 2.14, to make it generate random

#

that'll be a while though

carmine juniper
#

Nice blobWobble

weary niche
#

Very good news. Thank you!

gray quarry
#

are the rc branch passwords randomized or something known?

#

im asking what's the password of the rc steam branch

#

where is it posted?

wise dragon
#

rc branch for what 2.12 is out

gray quarry
#

says dev hub but im not seeing anything related to RC

#

yeah i know, its late now

#

but im curious anyways

#

was there a rc anyway or not?

wise dragon
#

yes

gray quarry
#

does the password follow a pattern like the other branches?

calm wasp
gray quarry
#

why its not mentioned as such on wiki :/

full sonnet
#

Because noone wrote it there yet 🙂

gray quarry
#

lol

#

it says shared on dev hub during active rc tests

#

was there a dev hub post?

full sonnet
#

Dev hub says this

firefox_2023-02-28_21-03-52.png
#

normally the code would be there

gray quarry
#

ohhh

#

how do i make use of game updater?

#

there's a wiki page for a description of the application but there isn't a tutorial or something i can see

viscid onyx
gray quarry
#

thanks!

full sonnet
# full sonnet Well possible for 2.14, to make it generate random

Actually, i think not.
Each player generates the procedural tex locally.
How to network sync the random unique name, on something that isn't network synced...

So well yes i can make it random, but it wouldn't be same random value per weapon in MP. Which is what you'd actually want...

#

What you'd need is a way to get the parent entity.
Had that idea before but oof. And especially when texture is reused in different places it becomes impossible (unless array of parents)

winter crescent
# gray quarry im asking what's the password of the rc steam branch

Just to have all info in one, we share intel about RC tests via:

unreal arrow
solid crown
#

So is there any hope to get the number of available custom chat channels dramatically increased? Like from 10 to 100 or something like that. There was chat about it on scripting channel already, but just posting it here as well just in case because I feel it would be an important upgrade. The current number of channels is understandable due to it being apparently hardcoded pretty well, but if the workload to change it is not extreme, I think many of the community devs would appreciate increasing the number drastically a lot. Requesting this explicitly here because the concept of customizable chat channels itself has so much potential but the implementation breaks it with such a low number of channels. Gib plz meowheart meowawww

unreal arrow
#

you have a link to the discussion?

late harbor
unreal arrow
#

is there a ft ticket I can have a look if there is a less painful way to up it

late harbor
#

Don't think so, but that can be arranged I'm sure 🤔

solid crown
#

Thanks for looking into it in any case!

unreal arrow
#

2021 though, some code have changed since

unreal arrow
#

200 no 101 maybe

rich pond
earnest mural
#

never got findDisplay "dispNameHere" working. my display is defined in description.ext

full sonnet
#

only works for UIOnTexture

steep gust
#

Just to clarify at the moment, you can only have a single UI on texture on a spawned weapon/scope per MP server right?

#

Like due to the unique name/number at spawn

earnest mural
#

oh that explains it. not sure if that is mentioned somewhere in the wiki

full sonnet
steady sundial
#

Noticed recent bug, in zeus, place a group and click-drag over them to select the lot of them, the audio bugs out at everone of them doesn't stop talking until you quit.

#

Anyone able to confirm this? I'll then hit up a tracker issue

frail wharf
#

Is FSR for Arma 3 dreaming too much?

steel rune
# frail wharf Is FSR for Arma 3 dreaming too much?

There’s https://store.steampowered.com/app/993090/Lossless_Scaling/ as well as many other to enable FSR or NIS for any game

Steam
Lossless Scaling

Lossless Scaling lets you upscale windowed games to full screen using state-of-the-art spatial scaling algorithms, sharpening algorithms and machine learning.Scaling algorithms are currently presented: LS1 AMD FidelityFX Super Resolution (AMD FSR) NVIDIA Image Scaling (NIS) Integer Scaling Nearest Neighbor xBR Anime4K Sharp Bilinear Bicubic CASL...

Price

$4.99

Recommendations

1025

sleek cliff
frail wharf
thick zephyr
calm wasp
past moon
#

@steady sundial wait till Thursday DEV build, if it still keeps doing hit the tracker

full sonnet
wise dragon
#

I max out my gpu usage on a3 🤷

light lintel
#

Which GPU? What resolution?

#

Is it possible that the GPU is down clocking itself to save power, and as a result showing higher GPU utilisation?

#

I know it can become GPU bottlenecked on certain CUP maps and Tanoa when using 8x SSAA even at 1080p

wise dragon
#

1050 ti at 1080p

#

if it was down clocking itself i wouldnt expect its temp to get into late 90s

void harbor
#

I'm curious to see what people think about the fatigue/stamina changes.

charred kite
#

fatigue feedback thread on the BI forums

#

mostly negative ATM

void harbor
#

I mean people here, but yeah I've seen that

#

I tend to agree. I wish they wouldn't step away from a system that could have used some refinement but was generally good as it was

#

But I'm curious to see if people have more suggestions as to how to make the new system better.

light lintel
#

@wise dragon modern cards can down clock to lower power consumption independent of temperature (so do verify clocks), but yeah a 1050 Ti is more likely to get GPU bottlenecked

wise dragon
tepid dock
#

Why don't you use the overlay that comes with Nvidia experience? I seem to remember it shows enough information (and will be more accurate than task manager)

wise dragon
#

theres one with nvidia experience? thonk

true peak
#

dosnt show memory utilization though.

wise dragon
#

Bruh moment ill look later

tepid dock
#

Just remember to set your performance HUD settings to advanced to see more stats

light lintel
#

There's also stuff like HWiNFO and Riva Tuner Statistics Server if you (like me) don't use GeForce Experience

primal wharf
#

i was happy with the system as it is. staminabar is a nice extension to the hud, tho

frail wharf
#

Can we have an option to not log the Server: Network message XXXXXX is pending T_YYY logs on the server RPT like we have for Server: Object X:YYYY not found (message Type_ZZZ) logs with the LogObjectNotFound=false setting?

full sonnet
full sonnet
full sonnet
#

Theoretically possible to render different UI for different levels lol. But probably better to leave that stone burried

#

I mean... Its quite simple to do, for uiEx that could be an option. But it will be much more expensive. It has to process all mips at once even if it doesn't currently need them

trim crane
full sonnet
tepid dock
#

Maybe you can choose another scaling function? (can you even?)

full sonnet
#

I can ye

tepid dock
#

Lanczos was quite good at scaling, AFAIR, but there were sinuses involved, so I guess that's not optimal for something like game 😆

full sonnet
#

Well I can choose the ones DirectX11 offers me
I don't really want to write a new scaling shader

tepid dock
#

What is it that's currently being used, out of curiosity?

full sonnet
#

bottom row D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT, top D3D11_FILTER_MIN_MAG_MIP_POINT

Arma3NoOptInt_DX11_x64_2023-04-05_13-11-42.png
#

I could do MSAA too but ugh no

#

Now just gotta fix game crashing when you hide the Eden UI thonk

tepid dock
#

I wonder how it would look with D3D11_FILTER_ANISOTROPIC 🤔

full sonnet
#

Ignore right column.
D3D11_FILTER_ANISOTROPIC
vs D3D11_FILTER_MIN_MAG_MIP_POINT

looks basically same as D3D11_FILTER_MIN_MAG_LINEAR_MIP_POINT looked in that its just darker

Arma3NoOptInt_DX11_x64_2023-04-05_13-25-26.png
#

mh okey a bit smootherish

Arma3NoOptInt_DX11_x64_2023-04-05_13-26-40.png
tepid dock
#

I guess it could make an impact if you're looking at it at an angle

trim crane
#

Can make you a repro that ui2displays all vehicle textures and sets them on a vehicle next to a normal textures

trim crane
#

Normal vs Crispy

image.png
#

Aniso Filtering: Ultra va Disabled vs Crispy

image.png
carmine juniper
#

Fixed: Changes of slider Control positions caused by the engine or scripting did not trigger the "SliderPosChanged" Event Handler
Was even an issue?!

full sonnet
#

There aren't many places where engine updates slider position

#

mainly everywhere where you have text boxes that also update the slider

pallid cliff
#

I guess that fixes the Eden issue where edit boxes stayed empty?

full sonnet
#

that was caused by that yea, but thats #perf_prof_branch, dev branch has that thing and the fix for it. prof is missing the fix for it

steep gust
ocean sail
#

In audio settings all sound previews (effects, radio and music) are start playing automatically
in new Dev and Prof

trim crane
#

Thought about how great it would be for PIP scopes and there it is

full sonnet
trim crane
#

Do I have to pay to play my own missions now? meowsweats

image.png
carmine juniper
#

Yes of course!!!!11

harsh orbit
trim crane
#

Game HAS to verify I paid for the custom mission or it won't run

trim crane
#

2.13.150442

full sonnet
trim crane
#

Can try it again

formal cipher
#

Did you try Dev branch before yesterday?

trim crane
#

Version: 2.13.150408
been on this build for like a week or two

formal cipher
#

That one did not have the issue?

trim crane
#

Nope, it was fine

#

Updated, then no longer was able to run any missions, also none of DLCs or mods were shown in the main menu

# 
16:28:18 SteamAPI initialization failed. Steam features won't be accessible!
```first RPT line
#

Steam was running fine, restarted the game few times including disabling all mods, had same picture

#

Redownloading dev branch again, gonna see if it happens after reboot

steel rune
#

bad steam dll?

trim crane
#

At this point this is likely my connection issues

formal cipher
#

Did you try a PC restart yet?

trim crane
#

Going to soon

earnest mural
#

steam is down for me

formal cipher
trim crane
#

Restarting Steam was enough. It was totally on, everything working but Arma refused to work for some reason 🤔

#

Didn't check if stable was affected too or not, so its likely not an issue of dev branch but my local connection\steam issue

#

Had it working fine just few hours ago, then went to update and it didn't anymore, restarting the Steam helped

#

Sorry for false alarm

steel rune
#

bottom of steam window usually says "NO CONNECTION" when this happens

#

one of the causes can be creating another session with same id, for example when running steamcmd

trim crane
full sonnet
#

That's how you want it right?

Arma3NoOptInt_DX11_x64_2023-04-06_13-59-29.png
#

or with offset's/viewport changes

Arma3NoOptInt_DX11_x64_2023-04-06_14-08-00.png
#

Also a thing
#(rgb,2048,4096,8)ui("RscDisplayFieldManual","uniqueName")
#(rgb,2048,2048,8)ui("RscDisplayFieldManual","uniqueName")

Will render two different display instances, even though they have same uniqueName, because they are different "texture sources" due to the different resolution.
I'm thinking about changing it such that only display class and uniqueName are combined to specify the instance, and that multiple textures can share the same display.
if I do that, the above will render the display twice on every change, into the two textures. But there is only one instance of the display in scripting.

I think I'll not worry about backwards compat on that, the feature is new enough that people probably aren't doing that. And.. unique name is supposed to be unique anyway. If it breaks your code that you expect two textures with same uniqueName to be different then.. well bad luck.

This can then be combined with the cropping settings, and the changed viewport, maybe different background color

#

He be so focused in his work that he doesn't realize I'm talking to him 🍿 I don't wanna ping :u ping is annoy

trim crane
#

I'm fine with being pinged

trim crane
#

As for as different display instances go, not an issue for me either, just have a note to differentiate with different unique names for different resolutions or setups

#

I am already using same display config to create different displays for different objects

#

Maybe store texture name in the display along with its unique name?

#

STRING = displayTextureSource DISPLAY

#

Gonna return #(rgb,2048,2048,8)ui("RscDisplayFieldManual","uniqueName")

#

Though not really needed, if you know what you're doing its not an issue at all

full sonnet
#

Well that won't work if multiple texture sources can refer to same display

trim crane
#

Didn't you mean that having different resolution but same unique name will make a new display?

#

How do you even have ui2texture refer same display? 🤔

full sonnet
#

Currently its that way.
I want to change it so that its not that anymore

trim crane
#

I'll be fine with the change as well

full sonnet
#

Example above, under last image.
Same display class, same unique name, just different resolution.
Should just spawn one display instance and render it twice, instead of spawning two instances

#

also findDisplay only finds the first instance, if you have multiple with same name

trim crane
#

Yeah, I get what you want to implement. I'd be fine with this as well, re-using same display for different textures sounds like a potentially useful feature

#

Say somebody makes some cool billboard with a display, they'll want to have it with different resolutions when displayed on a huge billboard or a small info stand somewhere

#

Cool animated billboard even, processing one display and then rendering it to different textures sounds more proper than having to same operations with copies of the display

#

(That somebody might be me, I want to display scoreboard on some objects at base in game)

full sonnet
#

Yeah. And also for last screenshot.
The one on the right is same display, but split into two halves. That surely can be useful for smth too

Example syntax for that
#(rgb,2048,4096,8)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportX:-0.5, viewportH:0.5)

#

You can display different parts of display, even zoom in, different resolutions, different background colors. But all have it be the same display that you only need to update once in script, and they will all be in sync

trim crane
#

Hmm, if there was scaling then we wouldn't need any of these fit or cut options

#

Fit or cut exactly like you want it

full sonnet
#

Thats what I'll do with the viewport

#

The fit version shown above is viewportH:0.5 (only render half height), the second is viewportW:2 (render double width, excess will be cut off)

trim crane
#

If its non-1:1 texture, just scale it by width or height (or shrink) and cut off as you need with offset

#

Yup, I get your idea, that would be even more flexible

full sonnet
#

Mh..
for uiEx
Do I do for test " test
uniqueName: "test "" test"
or
uniqueName: "test \" test"
I guess the "" is more in line with rest of Arma, but I hate it

#

omg notlikemeow No I just realized I won't allow it at all :harold:
You gotta live with no quotes in your uniqueName 😄

trim crane
#

I see no problem with that

full sonnet
#

ah background color isn't even a thing.
What else could be a parameter next to the viewport stuff
Could make a flag to bypass the same name == same display, and crate new display regardless if it already exists with same name. But I don't like people doing that so won't until I get a good usecase that needs it.

Ah the filtering type for mipmaps!

trim crane
#

Name already gives you all the options you might want for using same or a new display

full sonnet
#

Also.. I now have code to render any post-processing effect over top of UI 🤔
But thats an idea thats probably also better left burried in its grave

trim crane
#

Not sure why this could be useful 🤔

#

Animate vehicle being shiny through texture?

bleak canopy
#

when it rains, or force aircraft de-ice xD

full sonnet
#

For like in-vehicle displays, having like ChromAberration ontop of your UI to make it look brokey-ish, or put blur or filmgrain on it

trim crane
#

Actually, this can be useful for PIP scopes 🤔

#

Though r2t already have something like that?

full sonnet
#

Yeah I think it does

full sonnet
#

And proper error messages meowheart
#(rgb,2048,4096,8)uiEx(display:"RscDisplayFieldManual, uniqueName:"testName", viewportX:-0.5, viewportH:0.5)
-> uiEx procedural texture syntax error: 'Unexpected character, should be ',' or end' at -> 'testName", viewportX'

#(rgb,2048,4096,8)uiEx(displaiy:"RscDisplayFieldManual")
-> uiEx procedural texture syntax error: 'Unknown property' at -> 'displaiy:"RscDisplay'

#(rgb,2048,4096,8)uiEx(display"RscDisplayFieldManual")
-> uiEx procedural texture syntax error: 'Invalid character after property name' at -> ', uniqueName:"testNa'

#(rgb,2048,4096,8)uiEx(display:"RscDisplayFieldManual)
-> uiEx procedural texture syntax error: 'String missing closing quotation mark' at -> '"RscDisplayFieldManu'

#(rgb,2048,4096,8)uiEx(display:, uniqueName : "testo")
-> uiEx procedural texture syntax error: 'Property has no value' at -> ', uniqueName : "test'

Also if you need quotes in string, you can like
#(rgb,2048,4096,8)uiEx(uniqueName : 'testo " test ')

trim crane
#

Nice!

full sonnet
#

display: string
uniqueName: string
viewportX: number
viewportY: number
viewportW: number
viewportH: number
texType: string

quotes are optional so texType:co is possible
Still need to add that mipmap filter type. Anything else?

trim crane
#

I guess that's it

#

Did you try my repro with the vehicles, how did mipmap options compared to stock Arma filtering and mipmapping?

full sonnet
#

nu, you can try that yourself when I'm done

full sonnet
#

Mh viewport in pixels, or percent? 🤔
#(rgb,2048,4096,1)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportW:2)
#(rgb,2048,4096,1)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportH:0.5)
#(rgb,2048,4096,1)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportX:-1024, viewportW:2)

Weird mix out of pixels and percent now.
I think percent is probably easier to use

Arma3NoOptInt_DX11_x64_2023-04-06_17-46-31.png
bleak canopy
#

Virtual multi-monitor bezels Dx

full sonnet
#

Its basically UV mapping by hand XD
You could have many different displays handled by making one huge UI, but with different zones on it. I'm not sure if rendering is efficient... Optimally it would not try to render stuff thats offscreen but not sure if that can be trusted

#

#(rgb,2048,4096,1)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportX:-0.5, viewportW:2, bgColor:#ff000000)
#(rgb,2048,4096,1)uiEx(display:"RscDisplayFieldManual", uniqueName:"testName", viewportX:-0.5, viewportW:2, bgColor:#00ff0080)
Setting background alpha to zero... still makes it red 😠 what the hell, alpha is zero, it should have no color at all 😢

Arma3NoOptInt_DX11_x64_2023-04-06_18-03-07.png
#

Blame ClearRenderTargetView 😠
Welp that'll stay.

Now.. It would be really nice if you could adjust the viewport and stuff, on the fly by script think_turtle
You can re-set the texture but that's very bad please don't do
Could give it a second uniqueName (for the texture, vs the other one being for the display) so you can refer to the texture and change properties on it notlikemeow (Though these texture changes would then not be MP synced, unlike if changing texture string)

trim crane
#

Maybe its an issue of texType?

#

Also why black BG by default and not transparency? 🤔

full sonnet
#

default is transparent I was just testing

trim crane
#

As in -0.5 = half display to left

full sonnet
#

UI is in 0-1 so. yes percent then

full sonnet
#

Mh no live changing via script will also be shitty, because due to memory usage, your texture may randomly disappear and re-appear which would reset your changed values 🤔
Though that also happens with the display where it would already be bad/annoying 🤔

trim crane
#

Live changing viewport values with script?

full sonnet
#

ye

#

and background color, and mip interpolation

trim crane
#

So texture name remains the same, with some viewport numbers in yet they they can still be different since changed by script?

trim crane
#

Thought it was just RscPicture scaling or something

#

But its the whole thing scaling

full sonnet
#

They are just a value stored in the texture, that is re-read on every render. Nothing really stopping from changing these values (besides the complication of having to find where they are at)
Basically scripted UV Animations 😄

trim crane
#

Off for today, can't wait to test it out

full sonnet
#

We have a couple holidays till tuesday
I'll have to do the one display multiple textures thing still.
And that scripted animation thing, I guess I'll think about it

lament crow
#

Happy Holidays!

vague saddle
#

I'm in favor of minimizing the hud. I don't like being able to see when I'm going to be out of breath (feels mainstream-fps-like). I'd rather like right now, base it off of assumption compared to how much weight I'm carrying, but that's just me.

#

I'd recommend they instead focus on their scrolling action menu. Maybe color code some options? Like make the eject option on the menu red to make it more visible.

#

Or maybe require a shift+click for more important options

void harbor
#

Or just get rid of the god-awful thing...

vocal zealot
#

😄
Use a collapsible 3D menu.

mossy wyvern
#

Was their any final decission on wether the amount of custom von channels will be increased from 10 to more channels?

trim crane
#

I'd find a use for it too

unreal arrow
#

it’s on back burner. the problem is that the engine is simulating all custom channels regardless of whether they are used on not. Having 10 is acceptable, more it can start to interfere with overall performance. Some kind of dynamic system needs to be implemented to allow sensible extension which means significant rewrite.

full sonnet
#

This is my blacklist for now for commands that cannot appear in short-circuit

Unary;
for, Sleep, UISleep, call
createHashMapObject (If in scheduled environment, because constructor)

  • (IF in scheduled env with non-resolvable arg or hashmap, hashmap copy may call copy constructor)
    waitUntil
    exitWith
    break/breakWith/continue will not work (technically there is no loop now that they can refer to)
    try/catch
    && CODE (maybe all of && if right side is non-resolvable variable)

Binary:
call
do (switch, while, with)
forEach, forEachReversed
then
apply, select (unless right side is number constant constant), findIf, count
(These loops (and "then") could work, if they themselves can also short-circuit, or if in unscheduled, could force these loops to also run unscheduled (but then we loose break/breakWith/... so probably not))

I think that's managable. For the option to get 2x or more perf improvement if you don't use any of these blacklisted ones

#

Most simple select/apply loops will end up optimized

#

Also extra idea.

private _array = ...;
_stuff select {_x in _array;}

usually the in command has to check every iteration if the right side value is an array or object or location or string or hashmap, in order to select the correct implementation
But in short-circuit I know what type we have before running the loop, so I can move that check out

#

And another
groups apply {leader _x}
Currently creates one array of groups, passes that to apply, which returns another array, then the groups one is deleted.
With the new stuff, it could re-use the groups array and not have to allocate a new one 🤔
But performance difference of that probably wont be high

halcyon moth
#

_stuff select {_x in _array;}

This one reminds me. Could arrayIntersect be re-specced to preserve the order of the left input? Currently it seems to preserve the order of the largest input, which wrecks a lot of optimisation opportunities. It's not documented so it's probably a safe change :P

trim crane
#

Oh yeah it does, very strange

unreal arrow
#

Not strange but yes, why not, could be optimised as well a bit

full sonnet
#

Hashmaps for the win 😢

full sonnet
#

Another new optimization idea
x apply {leader group _x}

usually engine must check everytime leader is executed, if the right side argument is Group or Object or TeamMember, because depending on that, different implementation of leader command must run.
But with short-circuiting, it knows that the group command can only return a Group, so it just completely gets rid of that check on leader command.

_bigArray = [];
_bigArray resize 10000;
_bigArray = _bigArray apply {player};

[
diag_codePerformance [{_this apply {leader group _x}}, _bigArray],
diag_codePerformance [{_this apply {;;leader group _x}}, _bigArray]
]

Before: [[7.74615,130],[2.29062,437]]
After: [[7.67176,131],[2.32251,431]]

Welp, the new type check is already so much faster that this optimization doesn't make sense anymore 😢 Neat!

#

Some benchmarking because we all love data!

With short circuiting:
45.4% of time is spent inside the Leader command (including allocating the result value)
19.7% of time is spent inside Group command (including allocating the result value)
9.5% inside diag code performance, getting rid of that result of the apply
5.2% getting rid of the temporary group value that was returned by group command
2% getting type for the type check on group command
1.7% inside the apply, pushing the result of Leader command, to the end of the array that will be returned.

Without short circuiting:
13.9% inside leader command
8.3% in group command
16.7% getting rid of the scope/callstack level that gets created every iteration of the apply
10% creating memory to store the _x variable inside the apply code (It has to re-allocate space every iteration, because it always starts with a empty local namespace. The short circuit doesn't have a namespace at all it just passes a value directly)
5.1% allocating memory to store the string "_x" (frankly we already have ways to get rid of that, must've been forgotten to use it here, I'll just fix that, same for select. hashmap apply doesn't have it because that was done right 😄 )
3.3% inside diag code performance, getting rid of that result of the apply
2.7% for every group/leader execution, re-check if its disabled via CfgDisabledCommands.
2.4% apply setting the _x variable (short circuit doesn't have any _x variable, its resolved before execution)
2.1% adding the scope/callstack level for next the apply iteration
2.1% getting the type of the argument passed to group/leader command, to select correct version or throw type error if none found (short circuiting skips this for leader command)
1.1% getting rid of the temporary group value that was returned by group command

The time that the Leader/Group commands themselves need didn't change, just the surrounding fluff became so much bigger relative to it

full sonnet
#

I don't think I can optimize it much further. The rest of the stuff that remains must be there.

Actually.. out of the total 2% for that type check, half of it is checking if the type is a number and if yes it checks if its NaN to turn it into NaN type.
And it does that check even on my objects and groups 😢
So I can shave a tiiiiiny bit more time off. If I get rid of it you wouldn't get "expected number, got number not a number" errors inside short circuited stuff.. I think that's fine. Or I can just reimplement that type more efficiently 🤔 but not much

#

jippie I saved 50% on the type check, which is like, immeasurably tiny bit faster

#

75% is now inside the commands itself, vs 22% before.
I can get one or two percent more but ugh. I guess supporting binary commands is next 😢 then I need a two element stack

#

small difference, big difference.
shaved off another percent

8nuWiHuAss.png
harsh orbit
full sonnet
#

Intel vtune

harsh orbit
#

never seen that interface meowsweats

full sonnet
#

big gud

harsh orbit
#

so what kind of analysis gives that? meowsweats

#

hotspots?

full sonnet
#

yes

#

under process set finalization to full.
then hotspots usermode

full sonnet
#

Now the fun game of shrinking 32 bytes of array (dataptr, size, allocatorptr, capacity) down to 8 bytes so I can fit more instructions in a cache line meowheart

harsh orbit
#

you meant down to 16 bytes? think_turtle

full sonnet
#

no :3 Thats the fun (I don't actually need allocator, its constant)

tepid dock
#

Just winzip them 🙃

harsh orbit
full sonnet
#

capacity is not needed as that is only needed to check if it needs to allocate when adding values.
The size I can fit into the pointer

harsh orbit
#

size I can fit into the pointer
really? how? think_turtle

full sonnet
#

Well.. I don't want to bother with this crap. working with 64bit pointers is fun. Making this stuff work for 32bit is ugh. So.. You'll not get short circuiting on 32bit 🔨

full sonnet
#

Also pointers are usually 8byte aligned, so there are free bits on both ends. Though I would then need to 16byte align it meh

harsh orbit
#

well it's just too hacky meowsweats

full sonnet
#

yes :3

harsh orbit
#

plz don't 😅

full sonnet
#

But its the good kind of hacky

tepid dock
# full sonnet capacity is not needed as that is only needed to check if it needs to allocate w...

I was about to say that the software I was working on has a cool optimization for short strings: they use a union of size and ptr and (ab)use the fact that malloc will always align allocations on a 16-bit boundary (actually 32-bit, but they use just the last nibble anyway).
So if (ptr & 0xF) > 0, then size == ptr & 0xF and the rest of the ptr is the string.
Otherwise it's a pointer to a regular allocated struct with size and the string itself.

full sonnet
#

Yes I'll do that too to store single-value directly in-line

#

first bit is inline flag, then 5 bits for size if its pointer to array

tepid dock
#

Actually, correction, I messed up, it's as you're saying, regarding alignment

full sonnet
#

But I need atleast 5 bits, and I don't want to align that much so I'm using the other end of the pointers.
That may potentially be unsafe if in the next decade or so we get full 64bit virtual addresses, but I'll just check for it

tepid dock
#

Too bad you have to account for both linux and windows if you're writing such hacks

harsh orbit
#

well imo just do the 16 byte thing. at least it's 2x as before so it's still a win and it won't break meowsweats

tepid dock
#

For what it's worth, from my code that seems to be working 😛

    // https://stackoverflow.com/a/48081399/6543759
    // Windows does not allow allocating the bottom 64k
    // Vtables will be aligned to 64-bit
    // https://docs.microsoft.com/en-us/windows-hardware/drivers/gettingstarted/virtual-address-spaces
    // 64bit userspace ends at 0x7FFFFFFFFFF

Not saying that you should use it, but it may be good to be aware of that (see: links)

#

Although that's just for win-x64

full sonnet
#

Don't need to worry about linux. It uses 63rd bit to identify kernel stuff, but I'm working with a static buffer on stack so that doesn't interfere.
And even if anything goes wrong, I'm checking if the bits are free in my buffer area

unreal arrow
halcyon moth
#

I just realised that findAny is actually better for at least one of my use cases, so good job adding that :P

#

Now I just want command replacement for BIS_fnc_nearest

wise hill
#

Roman, i think we will see just that in the next year ;)

crisp wyvern
#

@vague saddle There's an unspecified interaction overhaul planned in the lead-up to the Apex DLC (previously aka the expansion), hence what @wise hill mentioned

#

it was mentioned in the 2015-2016 Roadmap

vague saddle
#

That's great.

grave viper
full sonnet
full sonnet
full sonnet
#

Quickly need to apply {[_x,2]}
sure thing
[[3.715,1000],[0.644,1000]]

I will probably only support a 2 element stack, which is required for binary commands. Meaning only 2 long arrays in short-circuited code
The stack size is kinda arbitrary, actually don't think it has much influence on performance so I can probably do like 8

full sonnet
# 
_bigArray = []; 
_bigArray resize 10000; 
_bigArray = _bigArray apply {[1,2]}; 

[ 
diag_codePerformance [{_this apply {_x select 1}}, _bigArray, 1000], 
diag_codePerformance [{_this apply {;;_x select 1}}, _bigArray, 1000]
];

First test with binary commands, a simple _x select 1
[[2.565,1000],[0.286,1000]]

10x improvement

trim crane
meager sage
#

;; - can be called a "vampire operator" 🙃

full sonnet
#

The more I work on this, the more I think this should be how SQF should originally have been written 😄
I'm basically implementing a proper cache-friendly VM notlikemeow

full sonnet
full sonnet
#

Now that we can run binary commands, lets see what it does to configClasses.
Old string matching based optimization
String changed just eeever so slightly to stop the old string matching optimization
remove the checks from configClasses, and do them in a separate select, without and with short circuiting

[ 
diag_codePerformance [{ "getNumber (_x >> 'scope') == 2" configClasses (configFile >> "CfgMagazines") }, nil, 1000], 
diag_codePerformance [{ "getNumber (_x >> 'scope') == 2 " configClasses (configFile >> "CfgMagazines") }, nil, 1000], 
diag_codePerformance [{ ("true" configClasses (configFile >> "CfgMagazines")) select {getNumber (_x >> 'scope') == 2} }, nil, 1000], 
diag_codePerformance [{ ("true" configClasses (configFile >> "CfgMagazines")) select {;;getNumber (_x >> 'scope') == 2} }, nil, 1000]
];

[[0.104,1000],[0.526,1000],[0.539,1000],[0.221,1000]]

First and last are not same speed because the last one still creates two arrays and iterates over the stuff twice. So once its actually integrated into configClasses itself it should be close enough to the string matching optimization.
Will do benchmarks when I get to that.

steel rune
#

My in-game config validation tool is gonna love this ❤️

full sonnet
#

And a way for you to check if short circuiting works, and if not, why not.

For example, && code is blacklisted, but if it can detect that right side is bool (including resolving local variables from parent scope) then it'll let it pass

This will land in dev branch this week, for now requiring the ;; prefix. I'll remove that the week after, after we get some testing

Arma3Retail_DX11_x64_2023-04-17_21-37-28.png
#

for now for
array select code, array apply code, code count array, array findIf code
configClasses some time later, hashmap apply

What else?

harsh orbit
#

configProperties?

#

speaking of which, is there a quick way to tell how many classes a config entry has?

#

there is count but it counts props + classes meowsweats

full sonnet
#

{isClass _x} count ...
😄

harsh orbit
#

in that case I would just do count ("true" configClasses _cfg) 😅

full sonnet
#

now that that is very fast, yeah 😄

distant halo
full sonnet
# distant halo While you're looking at `configClasses`, could support be added to the command f...

"performance benefit"
Well easy to test..

[ 
diag_codePerformance [{ "true" configClasses (configFile >> "CfgMagazines") }, nil, 1000], 
diag_codePerformance [{ configProperties [configFile >> "CfgMagazines", "isClass _x", true]; }, nil, 1000], 

("true" configClasses (configFile >> "CfgMagazines")) isEqualTo (configProperties [configFile >> "CfgMagazines", "isClass _x", true])
];

[[0.086,1000],[0.169,1000],true]
mhhhh Interesting I didn't expect such a big difference

"isClass _x" configClasses changes to
[[0.288,1000],[0.174,1000],true]

#

The iteration code that configProperties uses is a bit more complex (which it needs to be for base-class support). Adding that to configClasses would also slow it down

full sonnet
#

Mh the most expensive part is that config properties checks if the subclass overwrote a class in its parent.
So that the class is not listed twice. That's 30% of the time spent. Cannot fix that

#

Or can I...
[[0.083,1000],[0.099,1000],true]
There ya go, good nuff

full sonnet
#

Please help me with testing the short-circuiting, I've tried to find edge cases but I can't find all of them.
On #perf_prof_branch I already know the command blacklist is broken, exitWith/breakWith/continueWith/uiSleep/forEach all commands that have a uppercase letter are not detected correctly. And they'll cause chaos when used inside short-circuited. That'll be fixed on dev branch (coming out this week) though

diag_testScriptShortCircuiting CODE command is there to test and see why it may be rejected
Just figured out I can easily support continue/break commands so that's already something to tweak later.
I'll post the command blacklists tomorrow

earnest mural
#

did i understand right, is the prof broken right now?

full sonnet
#

no

# 
    BlacklistedCommandsNular[] = {
      ("break"), // 
      ("continue") // 
    };

    BlacklistedCommandsUnary[] = {
      ("call"), // 
      ("sleep"), // 
      ("uisleep"), // 
      ("for"), // 
      ("waituntil"), /
      ("breawith"), //
      ("continuewith")
      ("try"), // 
      ("breakout"), //
      ("breakto") // 
    };

    BlacklistedCommandsBinary[] = {
      ("call"), // 
      ("try"), // 
      ("foreach"), // 
      ("foreachreversed")
      ("apply"), // 
      ("count"), // 
      ("findif"), // 
      ("breakout"), // 
      ("exitwith"), // 
      ("catch"), // 
      ("do"), // 
      ("then"), // 
      ("setvariable"), // we don't allow modifying variables (because we resolve them before execution)
"isNil"  // could set variables //#TODO can run if we check its code for blacklisted commands and var assignment
    };

    BlacklistedCommandsBinaryConditional[] = {
      ("select"), // unless right arg is scalar constant
      ("&&"), // unless right arg is non-code
    };

These blacklists are not working right, but short circuiting is still locked behind ;; prefix anyway
Just don't try to use these

scarlet root
#

Does it matter if simple statements/code vs more complex code is tested for this testing?

earnest mural
#

ok good , I havent followed the discussion too closely to understand what you guys are doing here but it sounds good 😄 wish I could help somehow

full sonnet
#

Doesn't matter. Do anything to break it 😄
There is a limit on max stack size, and maximum number of instructions, and the command blacklist.

cedar frost
full sonnet
#

👀 Thank you

meager sage
#

on that point, is "isNil" not being in () intended?

full sonnet
#

ye I got that idea while I was cleaning it up and just plopped it at the end. (it isn't actually in the current list but should be)
formatting is irrelevant

full sonnet
#

test applying short-circuiting to configClasses
Hardcoded string matching optimization.
No optimization because the string matching fails.
Short-circuiting.

[  
diag_codePerformance [{ "true" configClasses (configFile >> "CfgMagazines") }, nil, 10000],
diag_codePerformance [{ "true " configClasses (configFile >> "CfgMagazines") }, nil, 10000],
diag_codePerformance [{ ";;true" configClasses (configFile >> "CfgMagazines") }, nil, 10000],

("true" configClasses (configFile >> "CfgMagazines")) isEqualTo (";;true" configClasses (configFile >> "CfgMagazines")) 
];
``` -> `[[0.1727,10000],[0.4267,10000],[0.1822,10000],true]`
short circuiting is only a tiiiny bit slower than the hardcoded strings (because hardcoded true literally does no extra checking, for the fact that its actually executing script, it feels quite amazing).
And... Its universal
```sqf
[  
diag_codePerformance [{ "configName _x == 'test'" configClasses (configFile >> "CfgMagazines") }, nil, 10000],
diag_codePerformance [{ ";;configName _x == 'test'" configClasses (configFile >> "CfgMagazines") }, nil, 10000]
];
``` -> `[[0.5327,10000],[0.1978,10000]]`
#

And sometimes the original can still be tweaked a bit when you find bad code in code
[[0.174,10000],[0.4117,10000],[0.1821,10000],true] though.. not that much

#

I'll add a way to run diag_codePerformance bypassing the 1 second max duration.
That's friggin annoying, if I want 10k runs then I want 10k runs please, thank you

full sonnet
#

The other hardcoded optimization of getNumber (_x >> 'scope') > 0
[[0.237,10000],[0.8096,10000],[0.388,10000],true]
yeah.. somewhat worse.
Maybe it makes sense to still leave the string comparison optimizations in there
For the extra edge on commonly used code

on CfgPatches (where the lookup for "scope" will always fail because that entry doesn't exist) its actually terribad
[[0.098,1000],[5.895,1000],[4.965,1000],true]
Any script variant suffers hard, the engine sided value reading is not affected though.

Actually, the performance impact isn't even where I would expect it to be. Its not inside the script looking up non existing entry 🤔
It seems to be in creating the script value to pass as _x, which the string hardcoded one doesn't have
But then it should also be the same on the other case, I probably fucked up that benchmark ⏳

halcyon moth
#

Oh, someone else already picked it up.

harsh orbit
#

bruhwith

trim crane
#

Wondering if short circuiting will work with drawTriangle? 🤔

#

I was about to request drawTriangles to feed it list of triangles instead of doing them one by one

full sonnet
# full sonnet The other hardcoded optimization of `getNumber (_x >> 'scope') > 0` `[[0.237,100...

Turns out, there was a decade old inefficiency in how config values were handled...

Entry finding either uses a hashmap, or it goes O(N) through every entry comparing their names until it finds a match.
CfgVehicles,CfgAmmo,CfgNonAIVehicles are hardcoded to have a hashmap
And it can be defined in PreloadConfig class if thing is set to "fastFind". If a thing is set in PreloadConfig, streaming is also turned off (otherwise default on).

Streaming lets Arma unload configs from memory to save space, and re-load them on demand.
And if streaming is on, you run into a shitty situation.

CfgPatches is not listed in PreloadConfig, thus it neither has Hashmap lookup, nor does it have streaming turned off.

Say you had a config entry (from something that has streaming enabled), and wanted to store that config entry inside a script value.

Instead of just... taking the entry and passing it through.
The game grabbed the parent class of the config entry, grabbed the name of the config entry, then did a lookup by name inside the parent class, to re-find the entry (that it already had to begin with) to then pass that newly re-found entry along.

Arma needs to do this IF the config is unloaded but it did this everytime when a config entry CAN be unloaded, but not checking if it actually IS unloaded.
And because everything that CAN be unloaded, also doesn't have hashmap lookup. You not only pay for a needless lookup, you also pay for a needless extra slow lookup.

Before [[0.098,1000],[5.895,1000],[4.965,1000],true]
After [[0.102,1000],[1.537,1000],[0.679,1000],true]

The extreme performance difference for the first one, is because it does the scope check before creating the script value.
Whereas the others create the script value, so they can then pass it to the condition script.

full sonnet
full sonnet
trim crane
#

Thought I'll check if it works through select/apply

full sonnet
#

I thought forEach may be not too useful, because you often use alot more complex code in there. I need to measure the short circuiting overhead, as it will fail slot more often on foreach, but should be fast enough if the early checks fail (code too large) but if the later checks (command blacklist) may not be so nice

scarlet root
#

wouldnt it make sense to also define other root classes in PreloadConfig with todays RAM/hardware? like animations, FX

full sonnet
#

Could add some logging for which classes are looked up often, that are not in preload. And use that to see.
In practice they are probably not unloaded (because there's enough ram) but the hash index will provide speed

#

I can even apply short-circuiting to call and && {} code. But the overhead of checking if it can short-circuit on every usage would probably be too high
though for &&{} its probably generally worth it

scarlet root
#

Like with FX and dialogs we notice a short lag. However could be also from textures - saw BI added quite some textures to PreloadTextures

full sonnet
#

I think you can always add your own stuff to preload if you want

scarlet root
#

with a mod yep. however i think sa-matra mentioned recently the workaround he uses for missions/scripts (dummy dialog).

way back one also used to put down all vehicles/entities/created them briefly at mission start for those only created by script lag to also preload the data. not sure if thats still needed in A3

full sonnet
#

preloading the models and stuff will still be needed in A3

full sonnet
# 
[  
diag_codePerformance [{ true && {isNull player} }, nil, 10000, true],
diag_codePerformance [{ true && {;;isNull player} }, nil, 10000, true]
];

[[0.0007,10000],[0.001,10000]] well yeah nah.
The script does execute faster, but the overhead of building the short-circuiting is higher than the time saved, for probably most simpl-ish && CODE's

full sonnet
#

wow 🤡
diag_codePerformance uses https://learn.microsoft.com/en-us/cpp/c-runtime-library/reference/clock?view=msvc-170 to measure ticks.
And the resolution is 1000 ticks per second, aka one tick per millisecond
We are measuring sub-millisecond timings, with a timing function that only counts in milliseconds? 🤔

LOL yes indeed.
profiling configFile command.
on one run
start time 181712
end time 181712

total runtime, zero.

diag_codePerformance profiling depends on.. luck...
If clock ticks over during run, it measures higher time than what is really happening, if it ticks between runs it measures zero.
You getting a "real" timing, depends on the ratio of how often it ticks inside and between. The longer the code runs the higher the chance it ticks between.
But its all random chance based

full sonnet
#

well I'm changing it to 100ns precision blobdoggoshruggoogly

toFixed 9;
diag_codePerformance [{true}, nil, 10000, true]

First value is with 100ns precision, second value is with millisecond precision
[0.000208420,0.000100000,10000.000000000]
[0.000207250,0.000200000,10000.000000000]
[0.000208140,0.000300000,10000.000000000]
timing is basically exactly the same, but the old code jumps between 1 and 3

lol.
And if we only do 1000 total runs
[0.000207900,0.000000000,1000.000000000]
suddenly it takes zero time 🤣

full sonnet
#

config properties with short circuiting, with the new profiling thing. Code,time,percentage

[
[{configProperties [configFile, "true", true];},"0.205679","87.6"],
[{configProperties [configFile, ";;true", true];},"0.234793","100.0"]
]
``` ```sqf
[
[{configProperties [configFile, "isText _x", true];},"0.560549","100.0"],
[{configProperties [configFile, ";;isText _x", true];},"0.208435","37.2"]
]

even a simple isText does alot, neat
I'll remove the "isClass _x && (...)" string optimization from configClasses because.. meh. At the point where any script needs to execute, just short-circuiting the whole thing is better

[
[{configProperties [configFile, "isClass _x && true", true];},"0.769286","100.0"],
[{configProperties [configFile, ";;isClass _x && true", true];},"0.293197","38.1"]
]
harsh orbit
full sonnet
#

Because that's the older hardcoded optimization if condition == "true" That one runs no script at all

harsh orbit
#

Oh I thought you said you removed it 😅

#

btw which expressions can be short circuited?
do you detect if the expression doesn't modify anything outside or...?

full sonnet
#

current limits are

  • less than 16 instructions
  • doesn't need more than 8 elements on stack
  • cannot contain variable assignment (aka _x =..)
  • command blacklist

It also depends on memory usage of the instructions (binary commands need more), it's all stuffed into a static sized buffer, currently only 512 bytes
instruction limit really only depends on that buffer size, I didn't measure how many I can actually fit
binary instruction is 32 bytes so.. ... 16 😄

harsh orbit
#

that 32 bytes is: 4x8 = command+2 args+return ptrs?

full sonnet
#

32 bytes
8 bytes commandptr
8 bytes argtypes table (8 bytes for single entry, 8+8*count if multiple entries)
8 bytes arg2types table
8 bytes implementation per argtype table

nular command for example is
8 bytes commandptr

doesn't need any types or different implementations.
And everything also has a 8 byte type (plus padding :/)

{;;true] ["Success",16]
{;;isPlayer _x} ["Success",80]
{;;[1,2,3,4]} ["Success",104]

#

I can test with larger buffer on heap, but spatial locality is quite important.
Right now they are also right next to the stack. Should probably be fine but I didn't get to it yet

harsh orbit
#

why not make it an arbitrary size on the heap? meowsweats
it should still be faster than raw SQF...think_turtle

full sonnet
#

I want the fastest though 😄
I'll have to test, I'll get to it
First making the small things work, then expand it

ocean sail
#

Eden checkbox "Place vehicles with crew" lost a few Dev updates ago and still not fixed
only on Dev branch, all ok on Prof

formal cipher
#

Is there a ticket for it?

ocean sail
#

no

ocean sail
trim crane
#

@full sonnet No mipmapping on uiEx?

full sonnet
#

it has mip

#

or should, same as ui

trim crane
#

Comparing ui and uiEx mipmap-wise and it seems to be identical

full sonnet
#

selecting a filter, no is not in, is also not listed on wiki

trim crane
#

I expected this feature the most, to solve issue of aliasing at distance and at at angles 🤔

#

Normal textures VS uiEx\ui textures

image.png
#

Not selecting in particular, but having multiple mipmaps\filtering, or option to enable that in general

full sonnet
#

multiple mipmaps was implemented

#

and is on dev and perf prof

trim crane
#

current sharp by default, smooth sharp depending on filtering in video settings?

trim crane
#

Otherwise uiEx is looking great so far, no more ui2texture inside ui2texture mess needed

ocean tundra
#

Can we please get this fixed? It's very stressful having to deal with this stutter every time especially when you're doing mod development where you have to open and close a3 a lot and you have to go through all that lag for hundreds of times sadtrollfacewhy

full sonnet
#

I heard of this and investigated it, but couldn't reproduce and from guessing around also couldn't find any reason for it

#

maybe you can grab #perf_prof_branch the profiling exe, and let it profile to show where the lag is at

ocean tundra
#

For me, stuttering disappear completely once that steam download is stopped

full sonnet
#

Yes I know, I've seen all the reports, but that got me nowhere.

#

I either need to repro, or I need to have someone show me what's going on (for example by running profiling)

#

Might be that arma integrated profiling is not sufficient, may need more advanced profiling.
But if I cannot reproduce it (which I couldn't last time), then I cannot do that on my own

ocean tundra
#

Ok, i can try to help provide data

#

i have around 2k subscribed workshop items so im guessing certain items in that list causes this

#

maybe deleted/hidden workshop items that are still subscribed?

#

cause i had missions that cannot be played cause they're no longer on the workshop but they still show up in my MP mission list

#

Can you try subscribing to a whole lot of workshop items and see if the issue pops up?

full sonnet
#

"to a whole lot" that's relative.
We tried with over 200 items, and neither me nor QA could repro

ocean tundra
#

probably to 2k items if possible? as i have 2087

#

and maybe try subscribing to an item then hide/un-public it

ocean tundra
full sonnet
#

download profiling branch on steam.
Go to arma folder, and launch the profiling exe.
Go into eden, debug console, run diag_captureSlowFrame ["total", 0.05];

#

it will then open a popup, select copy, it copies text to clipboard, put in text file, send to me

#

preferably make multiple of these captures during the lagging

harsh orbit
cedar frost
#

i also had the problem with the 0 byte download for a while, also with a game not downloading workshop content when it got added. it turned out to have to do with family sharing

ocean tundra
harsh oyster
#

I have had this bug for a while, I heard it was having too many compositions subbed so I unsubbed from a bunch and it stopped. Resubbed to a bunch recently and the bug is back

#

It only starts when I load into editor/a game, doesn't happen in the main menu

#

It also continues after I close the game

full sonnet
gray quarry
#

workshop folder gets bigger because it has to maintain the other user's different workshop items

cedar frost
#

Exactly that

#

My problem disappeared once I removed the additional accounts

tardy silo
#

why on earth is me running the profiling exe from cmd trying to load mods at all? It complains about @ace having missing deps

#

(oops, sorry for the ping!)

full sonnet
#

maybe you got mod pbo files in basegame addons folders

tardy silo
#

I found the culprit: stuff had puked into my Arma3.cfg at some point 🙄

ocean tundra
#

it was running a bunch of mods too for me for some reason, had to disable them in-game

tardy silo
#

I am currently trying to provide that profiling stuff you and exocet discussed

#

Thing is, I don't have any mods enabled.

#

aaand the cruft is back in the cfg file. where does this come from!?

#

Ok, moved all the profile stuff away, that seems to have helped

#

Ok, got profiling output, will send dm to dedmen

gray quarry
#

its a feature

tardy silo
#

ah, thanks, good to know

full sonnet
#

The profiling data for the lags helped me narrow it down to about 20 lines of code.
Problem.. steam callbacks are in there. Which widens it again to thousands of lines of code that could be the culprit, and we actually cannot use the game's profiling stuff in our steam callback code to find out which callback is bad notlikemeow

full sonnet
#

[
[{[1,2,3,4,5,6] apply {continueWith 8;}},"0.002697","100.0"],
[{[1,2,3,4,5,6] apply {;;continueWith 8;}},"0.001065","39.5"]
]
🤔

continue/break support maybe isn't all too useful if you cannot do if/then :/
I can probably implement my own then but uff. Though might be fun, that'd be a real if with a jump instruction 🤔

full sonnet
#

1k element array in _this

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"0.374090","100.0"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.063156","16.9"]
]
``` :pog:
And  that even is with setting _forEachIndex every iteration (which I don't need to do)

```sqf
[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"0.358731","100.0"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.056011","15.6"]
]
``` though might not even be worth to remove forEachIndex
harsh orbit
#

I wonder how fast the first one is compared to arrayIntersect 😅

#

I mean _this arrayIntersect _this

full sonnet
#

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.599603","5.6"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.546255","0.8"],
[{_result = _this arrayIntersect _this;},"64.700760","100.0"]
]

#

huh, am I doing something wrong there?

harsh orbit
#

Yes it's too fast 😠

full sonnet
#

1000 element array intersect can't take 64ms

harsh orbit
#

Oh right 😅

full sonnet
#

Result:
65.1871 ms

Cycles:
16/10000

Code:

_bigArray = []; 
_bigArray resize 10000; 
_bigArray = _bigArray apply {;;1}; 

_bigArray arrayIntersect _bigArray

wuht

harsh orbit
#

I wonder if it's broken after the change?

full sonnet
#

Result:
3.80391 ms

Cycles:
263/10000

Code:

_bigArray = []; 
_bigArray resize 10000; 
_bigArray = _bigArray apply {;;1}; 
 
_result = [];
{_result pushBackUnique _x} forEach _bigArray;
_result
#

something is way wonky here

harsh orbit
#

Maybe arrayIntersect checks uniqueness 10000 times 😂

full sonnet
#

current profiling branch is 43ms vs 4.1ms

#

but that has the arrayIntersect update I think, I'll check on stable

harsh orbit
#

Btw we have _arr resize [10000, 1] now blobcloseenjoy

#

No need for apply

full sonnet
#

2.12 stable

Result:
46.8636 ms

Cycles:
22/10000

Code:

_bigArray = []; 
_bigArray resize 10000; 
_bigArray = _bigArray apply {;;1}; 

_bigArray arrayIntersect _bigArray
#

It has always been so bad, wtf

#

Added a special case if both arrayIntersect args are the same array

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"0.368161","100.0"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.056166","15.3"],
[{_result = _this arrayIntersect _this;},"0.022143","6.0"]
]

harsh orbit
#

Why is it so slow tho? 😓

full sonnet
#

0.02ms is big fast tho

harsh orbit
#

I mean the old one

full sonnet
#

It always does O(N²) iterating through full second array (which has 10k elements), for every element in first array
Whereas the new one iterates through full output array (which only has one element) for every element in first array

full sonnet
harsh orbit
#

Can you test this plz?

_new =[];
{
   If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;
#

If possible apply your optimizations too 😅

full sonnet
#

then isn't supported in short-circuiting because it executes code

harsh orbit
#

Well still how does it compare with arrayIntersect?

full sonnet
#

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.661153","52.0"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.553641","7.9"],
[{_result = _this arrayIntersect _this;},"0.210790","3.0"],
[{_new =[];
{
If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;},"7.038748","100.0"]
]

harsh orbit
#

Is this the new arrayIntersect?

full sonnet
#

ye

harsh orbit
#

Well I mean old one 😛

full sonnet
#

Is your code intentionally bad?
the in check checks against _this not _new

harsh orbit
#

That's what arrayIntersect does

full sonnet
#

did you mean
_new =[];
{
If !(_x in _new) then {_new pushBack _x}
} forEach _this;

harsh orbit
#

No

full sonnet
#

ah oki

#

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.589542","5.5"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.583446","0.9"],
[{_result = _this arrayIntersect _this;},"64.792931","100.0"],
[{_new =[];
{
If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;},"7.123693","11.0"]
]

harsh orbit
#

But both return the same thing right?

full sonnet
#

yeah, all of them

harsh orbit
#

Well in my code forEach _this was _arg1, and _x in _this is _arg2

#

So if my code is faster you can write it in engine code to make arrayIntersect fast

# 
_new =[];
{
   If (_x in _arg2) then {_new pushBackUnique _x}
} forEach _arg1;
full sonnet
#

That is essentially what it does

harsh orbit
#

Then why is it slower? thonk

full sonnet
#

except, that it doesn't abort at the first find, like in does
it goes till the end of the arg2 array, always

harsh orbit
#

What for?! meowfacepalm

full sonnet
#

Dunno, there was some reason.
I thought I can get rid of it but I don't have the headspace to think of all the test cases to verify that doesn't break anything

harsh orbit
#

Isn't it meant to return what's in both arrays? Except unique? If so you only need to check the first occurance

full sonnet
#

if I just add a if(match)break
so it stops at first match

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.580807","16.2"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.565247","2.6"],
[{_result = _this arrayIntersect _this;},"22.046898","100.0"],
[{_new =[];
{
If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;},"6.964725","31.6"]
]

#

its still worse, wow

harsh orbit
#

Just call compile my SQF code instead of the engine code for arrayIntersect 🤣

full sonnet
#

Ah actually we have a "optimization" in there to try to skip iterations on the second array

#

[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.650128","49.6"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.711296","9.7"],
[{_result = _this arrayIntersect _this;},"0.228172","3.1"],
[{_new =[];
{
If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;},"7.358353","100.0"]]

with that optimization removed

harsh orbit
#

I think not think_turtle

full sonnet
#

Well if the arrays still isEqualTo compare the same

harsh orbit
#

Yeah then I guess no reason 😅

full sonnet
#

Ah. yes that broke it

Arma3Retail_DX11_x64_2023-04-25_13-20-51.png
#

Ah because I'm not doing a pushBackUnique on the second 😄

#

But that would become slower
Say
[0,1,2,3,4] arrayIntersect [0,1,2,3,4,4,4,4]
when we get to the 4 in right side
we iterate over 0,1,2,3,4 to find that it is present
and then we iterate through the output array 0,1,2,3,4 to check and find its already in there
so we do two big iterations per element in that case

harsh orbit
#

You do _new pushBackUnique _x tho think_turtle

#

It's not slower as far as I see

#

_new is output

full sonnet
#

another idea
[
[{_result = []; {_result pushBackUnique _x} forEach _this;},"3.579499","49.6"],
[{_result = []; {;;_result pushBackUnique _x} forEach _this;},"0.681426","9.4"],
[{_result = _this arrayIntersect _this;},"0.238816","3.3"], // broken, ignore
[{_new =[];
{
If (_x in _this) then {_new pushBackUnique _x}
} forEach _this;},"7.222832","100.0"],
[{_new =[];
_new insert [-1, _this, true]; // arg1
_new insert [-1, _this, true]; // arg2
},"0.389969","5.4"]
]

harsh orbit
#

That's wrong tho meowsweats

full sonnet
#

yes

#

ah

harsh orbit
#

You're not supposed to add _arg2

full sonnet
#

I give up we'll just keep it as is and maybe potentially fix someday or not

harsh orbit
#

Well mine was good 😢

full sonnet
#

nu 😠

#

pushBackUnique iterates the whole output.
[0,1,2,3,4] arrayIntersect [0,1,2,3,4]

when you get to the 4 in first array
it does 5 isEqualTo's on second array, to find the 4
then it has to do another 4 isEqualTo's on the result array to find if 4 is not already in it.

the current stable branch arrayIntersect would do 1 isEqualTo on second array, and none on the result
it keeps a extra skip list to mark elements in second array that were already matched

harsh orbit
#

oki...meowsweats
I guess your isEqualTo check is good enough for now

#

Or I guess insert one is faster for uniqueness now

full sonnet
#

We have the most common use case optimized by alot now, good enough for today

harsh orbit
#

nvm that's wrong too meowfacepalm

full sonnet
#

The code really should be using a hashMap
But I don't know how the overhead of creating that map would compare to plainly bruteforcing isEqualTo's
Also out hashmap would be somewhat inefficient for that due to the intra-bucket lookups being by value and not by hash, so I'd want to write a more efficient hashmap implementation and bleeeeergh
Another alternative is also to replace the second array isEqualTo's by a hash check in a continuous array of hashes, so alot more cache friendly. But for that we would need to rely on different values always having different hash, or we'll have to do another isEqualTo on match..
All doable but I don't want to spend the time on that

harsh orbit
#

nvm then... meowsweats

#

But either way wasn't the "all the way thru _arg2 isEqualTo" check redundant?

full sonnet
#

it is done that way so we don't need to do a unique check when inserting into output

harsh orbit
#

Ah that. Well then mine is either faster or same perf?

#

Oh right not when at end nvm

full sonnet
#

in some cases

#

In my test, the in always did only one isEqualTo check
If the arrays actually have different/unique values inside them, it behaves differently

We can make the skip list thing work better but.. so much stuff that can be done it all just takes time
I'm pretty sure 90% of uses just use it as a way to get unique elements, so they'll hit the new optimization.
So making the rest faster is even less worth investing the time now

harsh orbit
#

Maybe put a ref equals check first 😅

full sonnet
#

that is the new optimization yeah

#

Short-Circuiting is starting to become a bad name for what it is.
What it really is, is a more efficient script VM that just formats normal instructions into instructions for the new VM...
I kinda don't really want to rewrite the whole script engine notlikemeow

#

Now the kinda next step is to implement a jump instruction so I can implement if

current sqf

boolean on stack
if (stack) call trueCode else call elseCode
... instructions after if/then/else block

new thing

boolean on stack
if (!stack) jump ElseCode
... instructions of then code block
Jump PostIfBlock;
ElseCode:
 ... instructions of else code block
PostIfBlock:
 ... instructions after if/then/else block

If I do that we also get && {} support
And theoretically call too if I can inline the instructions

#

ohh right I need to blacklist params because that sets local vars, aha!
Or I'll just have to implement local var scope, but then I only want one level of scope, the more it creates the worse performance gets again

tepid dock
#

for the new VM
Do you mind elaborating on that? I must have missed it earlier, when reading.

What new VM are you taking about? Are there actually two VMs interpreting code right now or do you just mean new code for some of the commands that you plug instead of the old one?

mental fulcrum
#

What is this gibberish. <@&105621371547045888>

lucid totem
#

!purgeban 617279509888106498 60d cross-posting spam, random rambling, you're better out than inside, ciao! 👋

cloud solarBOT
lucid totem
umbral orchid
#

RE: the mipmaps that were adding for UItoTexture is a similar thing available for the procedural text?

full sonnet
tepid dock
full sonnet
#

new written from scratch. And it converts the old instructions into new format

tepid dock
#

Woah... only a crazy person would want to... oh wait! 😁

frail wharf
#

can it be used within FSMs?

halcyon moth
#

I feel like the lower-hanging fruit for FSM optimisation is making a way to not run them every damned frame :/

light lintel
#

Short-Circuiting is starting to become a bad name for what it is.
What it really is, is a more efficient script VM that just formats normal instructions into instructions for the new VM...
Just invent a german word for it 😛

full sonnet
limber forum
#

ever though of adding more options for how the lightning interacts with the skyObject?
it would be nice to have the option to not have it affect texture saturation
allowing the usage of less neutral skyboxes

#

even if at the cost of sky quality

crisp wyvern
ocean sail
#

Project Drive 'Extract Game Data' official tool are not extracting _contact .pbo files
_tacops extracts ok, _contact no, is this normal for contact?
fixed

carmine juniper
#

Mikero's ones does finely

ocean sail
#

I'm talking about official tools so on release 2.14 there would be no problem that was in the Tank dlc content extraction

scarlet root
#

in regards to testing the ShortCircuiting - is it "only" to report broken/non working things, or also performance stats or what other things?

ocean tundra
tepid dock
formal cipher
#

I don't think we adjusted the tools yet to account for that

ocean tundra
#

where are the contact pbos

#

oh its in enoch

ocean tundra
#

so both contact ebo and pbo are kept in arma3 game folder?

harsh orbit
ocean tundra
#

Mine is thonotlikemeow

#

Should i just delete the ebo

harsh orbit
ocean tundra
#

No ebos in contact
Pbo in enoch

harsh orbit
#

well that's the way it currently is yes

#

I thought by in "same folder" you mean pbos and ebos in the addons folder

formal cipher
#

Contact folder has the PBO's on dev branch, as expected:

image.png
#

The Enoch folder has been decrypted for a while now, I think that was done with 2.10?

full sonnet
scarlet root
#

ok so A-B testing/stats

the general idea is you outsource specific operations to ShortCircuiting with the specific syntax/tag (;;) to the new VM to make use of the faster computation, correct?
should you try to keep the executions to a minimum within a SC statement, or rather get as many into it (with the said limit)? in other words is there some "hand-over" cost between the two VMs o between SC to SC?

full sonnet
#

There is cost to translate the normal instructions into SC instructions yes.
But in general its already faster to execute even if you have just a single item in the array

specific operations, well the code part of forEach,select,apply,findIf

I don't know what you mean by executions within a SC statement

scarlet root
#

my bad. not that solid anymore on the correct terminology
;; is SC till the end of the scope I would assume. so each sqf command is probably an operation or instruction - in general terms should you try to maximize those within the SC scope, or rather separate to multiple SC statements - or is this more case by case, or doesnt matter really

full sonnet
#

A whole piece of code gets short circuited. Code starts with { and ends with }
Currently during experiment it needs to have ;; right at the start.
It just makes everything faster, doesn't matter how much code you have. Having more code going faster is ofc better than having less code go faster.

In general in SQF, the less code you have the better. If you can do the same work in one loop, or split it into 4 loops. Then adding 3 more loops is adding more code and more work to do for the game

full sonnet
trim crane
weary niche
#

Quick question regarding UI On Texture, when applied directly in the model as opposed to via hiddenSelections or setObjectTexture, any transparency becomes blacked out. Is there any possible fix for this? I know previously when discussing it, the face that gets the UI On Texture applied to it has to have transparency before hiddenSelections or setObjectTexture is used to apply said UI display, but since the bistudio wiki states the default UI On Texture texType should be CA one would think that that should apply transparency to the face similar to a Color procedural using CA with alpha.

full sonnet
#

You need a binarize.exe version that supports UiToTexture.. and i think it's currently plain disabled in binarize..
It checks at model pack time if the texture you supplied has transparency (it would just check what you specify as background color)

#

If binarize doesn't know the texture type, it probably just falls back to black

#

I haven't thought about that we'll have to test that. And that'll need a tools dev branch update

weary niche
trim crane
#

I did some playing around with materials and you can get transparency with some renderFlags in .rvmat, but unfortunately there is no simple flag to force allow transparency.

# 
renderFlags[]={"AlphaTest32"};
20230313215254_1.jpg