Frontier Gate Boost+ graphic problem

[sum2012]

orignal reporter:qq 测试PPSSPP edit2:save status https://drive.google.com/file/d/0B3OaSdeV0L8kVGpuNGxLTDhpNkE/edit?usp=sharing

tag @daniel229 as he has this game

edit:soft gpu no problem

[daniel229]

For the demo,using this save.(rename jpg to rar)

Choose this mission.

Walk through to the area and move up the camera.

[daniel229]

By the way,stil random hangs with demo and retail by pressing start twice or completing mission.

[daniel229]

I have found a consistent hangup after mission clear with demo. Using above save,choose this man as a partner,he only have a mission on the list.

Running to the area as the following pictures showing.

talk to the man to clear the mission.

[LunaMoo]

I can certainly confirm the graphic glitch after seeing where to look, but didn't got any crash in full game yet through. Dunno if I'm not just extremely lucky since I do open map at times(pressing start twice) and did quite a few quests already. I do have all experimental/cpu settings off through and do savestate like once per location, so maybe it hangs only with some setting or resets something during savestate.(see alot of ReplaceTables in the log every time I savestate)

Edit: Just passed a mission without a single savestate and nothing, but then like a moment after I used savestate after the mission in the player house it did hang while I was changing tabs in quest list. Funny thing reloading a savestate made like a second earlier prevented this crash, so basically it's the same issue as earlier with jit off? (~when the camera was going weird, but in theory game could be played with jit off when doing savestates and reloading them often)

[unknownbrackets]

From what I can tell, this is an issue of depth clamping. It works in Vulkan and in softgpu, and I see a few vertex coords that are negative (although none look to be so far out of range that the existing accurate depth path shouldn't be catching them, not sure...)

-[Unknown]

[unknownbrackets]

A GE dump of this would help to confirm it's related to depth clamp.

-[Unknown]

[sum2012]

Please note that I use old save status https://drive.google.com/file/d/14dGrInaVRe1XeybNQvq-whnrZboQigo5/view?usp=sharing

[Saramagrean]

Test with PPSSPP v1.9.3-960-gb23f94993.

Vulkan and D3D11 is look fine.

Graphic glitch still present in OpenGL and D3D9 only.

GE Dump. GE_DUMP.zip

[ghost]

ppsspp latest git build.

HW rendering

SW rendering

[ghost]

Both OpenGL and Vulkan have this issue on my Redmi Note 9..

[Panderner]

I tested My POCO M3 for game Vulkan is fine but OpenGL:

OpenGL:

Vulkan:

[unknownbrackets]

Hm, it's not obvious from the dump (I think I looked at it before and came to the same conclusion too) whether this is depth clamp related or not. Might need better GE debugging facilities on mobile...

-[Unknown]

[Saramagrean]

https://github.com/hrydgard/ppsspp/issues/6531#issuecomment-633974020

Try latest git build v1.11.3-1124-g77558bf18 seem like this issue in gone, OpenGL only but D3D9 still exit.

[unknownbrackets]

This reproduces on desktop when setting simulateGLES = true (code change only), but not using desktop GL. It also happens in D3D9, so basically everywhere without depth clamp...

Initial draws are supposed to result in this:

But instead result in this:

Depth clamp is not enabled, according to renderdoc, and neither is clip distance. If I enable EXT_depth_clamp, everything works fine. Unfortunately, most mobile devices do not support depth clamp on GLES.

Specifically here it's the depth TEST failing due to depth clamp. It uses src >= dst, but it assumes depth will be clamped to zero prior to that test (existing values are all zero.)

-[Unknown]