Depth buffer detection modifications

Looks like the latest compiles include quite a number of VLK and other improvements, might help with Surge 2 but Rage 2 and possibly Wolfenstein Youngblood I guess might be a bit problematic still?

...And this might give page 61 here a bit more than just a blank nothing.
(Somehow that just happens occasionally for topics, guess it's a forum software thing.)

EDIT: No go with Surge 2 yet it seems, gets a bit farther but then stops before it can load up the main menu fully.

I'm new here, and I'm sorry for the off-topic post.
It seems like ReShade 4.4 is out already, but only one man claims it's out there, and I can NOT find anyone else other than him talking about 4.4. I'm wondering if that version 4.4 he's claiming is the official one or not.
I'd like to ask you to let me know about ReShade v4.4.
Thank you in advance.

4.4 is NOT out. Official releases are found here on the homepage and ONLY there.

Yeah it's compiled builds from the Discord seen in the above menu for this website, WIP builds though while they have a number of additions since the 4.3.0 release it's still not finished yet.

Been toying around with D3D12 a bit and lately Vulkan seeing if the compatibility improves as additional commits land in the ReShade source code and how it also works with D3D11 and earlier even if these API's were already fairly well supported and so far other than Vulkan it's been quite interesting and if some of the pending additions will also be part of the 4.4 release it's going to be quite a update.
(Although problematic as D3D or OGL can be I expect D3D12 and VLK to be a bit situational from title to title but from what I am reading there's several titles that also work really well already.)

Compiled builds available on the Discord are ONLY experimental builds... There is no official support for them. The builds that i provide in the Discord are always tagged with the current version (4.3). I don't know where you obtained a 4.4 version. It is only on Crosire's side to release a new official Reshade version, and since it is not released, there should not be any 4.4 version of Reshade...

Yea, they aren't even on the official Discord. They are in PGhub #reshade-builds (can get to that Discord from Office Reshade Discord #relevant-discords channel)

It looks like they are compiling builds for people who can't compile dll's on their own. A lot of the new ones are using 4.4.0.xxx as build number.

It's ok. I know where those builds come from, and i think the version number will be correct the next time...

sajittarius wrote:

Aelius Maximus wrote: Does anyone know how to test the new UWP builds on any of the UWP titles? Seems pretty impossible to me considering everything is encrypted and locked down?

I'm also curious how this works. I know the Vulkan support creates extra stuff (json and bat file in the games exe folder), but UWP apps restrict the folder pretty heavily.

In my rig, Forza Horizon 4 files are not encrypted. I use the developper mode, and i disabled bitlocker. The files are stored in "C:\Program Files\WindowsApps". You have to play a little bit with the ACLs to access the content of this folder... As i like FH4, i specifically bought it to check if the folder modifications was possible, and it is...

I've been taking ownership of the folder but that's risky if done for the Program Files paths in general so careful adjustments are needed for accessing these in that way and there's other possibilities such as opening a program with admin rights and browsing into the folder that way though the format UWP and Win32 for a few newer titles would also be a factor as would dependencies and more specifically additions such as Gears 5 using Easy Anti Cheat even if placing ReShade's .dll in place or using a injector with permission to access the process it could still be blocked.

D3D11 is still the most common but a couple of first parties use D3D12 and at least Halo Wars 2 is using purely D3D12 instead of D3D11onD3D12 or how it's called though later game launches might also be going for DirectX 12 only and starting to phase out DirectX 11 for additional benefits of a pure D3D12 environment.

Think Outer Worlds launches in a few days plus the various trials Microsoft is doing for what's it called again GamePass so 2$ or so for a month which is a nice deal so that'd be a interesting experiment but the code is still being worked on and I think there's still a couple of unresolved issues although I also expect it to vary a bit as Microsoft changes up how UWP/Windows 10 Store works like the recent addition of Win32 support over the UWP container type format or sandbox or how to call it.
(Some other issues as well they're still working on but it's improving and them also using Steam for the upcoming release of Halo is a nice sign as well and should simplify things a bit even if it might not be for every released game from Microsoft's first party publishing and development studios.)


And yeah the .DLL you compiled would be 4.3.1.628/629 and the other one on the linked channel was using 4.4.0.854/855 which I guess might confuse things when that gets linked to although also this era of quick "clickbait" articles and being first with material so some video on YouTube or whatever showcasing and linking to a supposed ReShade 4.4 release and that quickly spreads and might be a bit of a confusing mess particularly with links to non verified builds because that's also a problematic development with all sorts of not-entirely-harmless software though I don't think it's gotten to that state yet but I can imagine some confusion from presentations claiming to be using a release build of 4.4 and preset showcasing on this version without having the additional info on how it's a WIP build and where it's really from.


EDIT: Oh and I think the .DLL's are just about identical content wise too containing the latest compile of the ReShade source but excluding the current pending merges for emulator support and UWP so the difference is merely from compiler settings and the resulting final binary or well .DLL files.
(EDIT: No there's one other change from what I'm reading up on seems Matsilagi is taking out the depth buffer check for testing purposes.)

Thank you all. I searched for ReShade 4.4 on twitter to check if there's any info available, and the guy has shared the link to the beta version of dlls. The guy said who doesn't know that 4.4 is out already is out-dated, so I got it wrong:(

thalixte wrote:

sajittarius wrote:

Aelius Maximus wrote: Does anyone know how to test the new UWP builds on any of the UWP titles? Seems pretty impossible to me considering everything is encrypted and locked down?

I'm also curious how this works. I know the Vulkan support creates extra stuff (json and bat file in the games exe folder), but UWP apps restrict the folder pretty heavily.

In my rig, Forza Horizon 4 files are not encrypted. I use the developper mode, and i disabled bitlocker. The files are stored in "C:\Program Files\WindowsApps". You have to play a little bit with the ACLs to access the content of this folder... As i like FH4, i specifically bought it to check if the folder modifications was possible, and it is...

Yea, i'm pretty sure i have had bitlocker disabled for a long time but the ACL's were still a problem. I even had windows install my games to a different drive so that it was not inside "C:\Program Files" and i have had varying levels of success touching files in these folders (its like some of the apps work ok and others wont start if you edit or add any files to the folder). Also I have Xbox Gamepass for PC and i think it does some very strange stuff like mounting the folder as a drive. I have seen in the event log many mount/dismount of drives, like "Drive 8 mounted" and then "Drive 8 dismounted" which correspond to the game being installed/uninstalled, may have been when bitlocker was still enabled though.

I have done all 3, bitlocker, dev mode and messing around with the ACL, and i still get a CTD with Forza 4..

Nvidia Drivers Support Reshade?
www.tomshardware.com/news/nvidia-reshade...-experience-graphics

I've tested Doom with the latest build from discord. And i found a problem there similar to the one i described in my thread here with ACO:
reshade.me/forum/troubleshooting/5826-ac...e-performance-impact

Pattern is the same, in a demanding scene if your game is running with a Reshade hooked in, it will introduce frame drops. I.e. let's say on my system R5 1600@4ghz+1660Ti i'am able to run Doom at max 1440p@75fps locked. When i introduce Reshade into that i'am no longer able to do that, i get major frame dips during movement, no shaders used, Reshade toggled off via GUI, just hooked into the game. The similarities i see between Doom and ACO is that both games are heavily threaded. The problem here is that if this issue is universal then it compromises testing, you toggle Reshade off and expect the game's perfomance to be as is, but it's not and u can only find out about this basically by chance, because you need to specifically exit the game and uninstall Reshade and then launch a game clean to see the difference. I thought this issue was exclusive to AC anvil engine, but i see now similar pattern in Doom. I thought i'd post this here to let ppl who develop latest builds know, plus according to Kaldaien this might be related to depth buffer access.

Yeah something about it locking it but I forgot the exact explanation since that time (Been a while since 3.0.8 came out.), there's some newer commits on Github as well but I think most of these are in the 4.3.1 / 4.4.0 compiles from Discord now since the latest versions went up minus the October 23th and on code updates.
github.com/crosire/reshade/commits/master

Even more VLK improvements and a couple of fixes and what looks like a possible crash(?) fix for D3D12 I think possibly also newer versions of 11 like 11.4 maybe if anything actually uses that instead of 11.1 or 11.2 maybe because backwards compatibility with older OS's is still important.

EDIT: Would be interesting to see if it's just that or doing a compile without the depth buffer hook or having it disabled for comparison which I think I saw some flag for though it might be other factors to the performance regression too.
(Simulating network activity so ReShade disables the depth buffer might also be a possibility but perhaps that's not enough.)


Going by ReShade's Discord discussion there might also be changes to how ReShade is setting up the VLK layer compared to the current batch file and adding it via registry instead much like how Steam and the display drivers add their own layers into Vulkan.
(Though off by default so not everything using the Vulkan API tries loading ReShade .JSON and ReShade .DLL and then enabled via a .bat file I assume per-game much like the current means of dropping in and renaming the .dll for hooking into D3D or OGL.)

Well, this is a pattern at this point. Tested Shadow of the Tomb Raider -- same thing. There are places in Hidden City that can cause frame dips and as soon as Reshade is eliminated from the equation the dips are gone. It feels like Reshade somehow messes with the workflow which creates bottlenecks much earlier when system is pushed compared to the clean state without Reshade.

pyewacket wrote: Well, this is a pattern at this point. Tested Shadow of the Tomb Raider -- same thing. There are places in Hidden City that can cause frame dips and as soon as Reshade is eliminated from the equation the dips are gone. It feels like Reshade somehow messes with the workflow which creates bottlenecks much earlier when system is pushed compared to the clean state without Reshade.

Have you checked the "Statistics" tab ? Which shader has the heaviest impact ? If you disable all the shaders, does it has an impact ?

thalixte wrote: Have you checked the "Statistics" tab ? Which shader has the heaviest impact ?

I guess i didn't make it clear enough despite mentioning it few times in my posts. No shaders are used and Reshade is toggled off in GUI. That's the thing.

Here's clean SOTTR:
prntscr.com/pp5uzk

Here's SOTTR with Reshade injected, but turned off. Notice the framerate etc.
prntscr.com/pp5wvf

For good measure. SOTTR with Reshade injected and turned on.
prnt.sc/pp5ymy

Similar pattern in ACO and Doom.

pyewacket wrote:

thalixte wrote: Have you checked the "Statistics" tab ? Which shader has the heaviest impact ?

I guess i didn't make it clear enough despite mentioning it few times in my posts. No shaders are used and Reshade is toggled off in GUI. That's the thing.

Here's clean SOTTR:
prntscr.com/pp5uzk

Here's SOTTR with Reshade injected, but turned off. Notice the framerate etc.
prntscr.com/pp5wvf

For good measure. SOTTR with Reshade injected and turned on.
prnt.sc/pp5ymy

Similar pattern in ACO and Doom.

However, can you show me the "statistics" tab ?

Assuming it's the same issue then the method SpecialK's ReShade build utilizes is down to a lockless design for the depth buffer while also disabling said depth buffer if no effects are utilizing depth.
steamcommunity.com/app/582010/discussion...c1735462352474503227

That's from older versions of ReShade though and a lot has changed since then and 4.0.0 to 4.3.x now.

EDIT: Just finding the info again on what the changes were down to, not entirely sure as to how the whole thing goes but multiple render threads and later functionality of D3D11 or extensions through API's and other fancy and complex features.

I would think this means ReShade holds the depth buffer to the end of frame and this affects performance until it's finished and can switch to the next instead of whatever dynamic method could be used for reducing the performance impact but locking it could also have advantages or be simpler or overall more compatible so just changing it for a performance gain and no adverse effects what so ever is probably unrealistic or it would likely have been designed as such already.


EDIT: And then there's likely also the matter of how much of a performance hit and if it's related to just how the threading in the game or game engine itself is and what said game engine is doing, only one context or multiple or just a few threads or several and commands and concurrency and well I don't know the tech too well.
(Plus also differences in D3D11.x where it's more of a optional and then D3D12 and I assume also VLK where it's designed with this in mind reducing bottlenecks and other gains from how it works in these newer API's instead of a update to D3D11 and certain limitations.)

I find it curious and a good thing to be aware of as well far as ReShade and performance goes, might not be much but then it might still be a measurable framerate drop and could also depend on a number of different factors and what the final difference in performance with and without ReShade then results in.