Naruto Ultimate Ninja Impact Hachibi White Colour

[ghost]

Real PSP

[daniel229]

Debug build is different from release build and It's not be affect by this.comment out either one won't help,have to comment out both,Change the log types don't make any different.

debug build coment out or not is the same.

release build not comment out

after comment out,same as before

[daniel229]

The game does not hit these logs. #pragma optimize( "",off ) does not help release build.

[unknownbrackets]

The codegen changes with commenting it out from about the viewport line down. But I don't see any substantive differences in the math. Hrm. I wonder if it's some inf/nan issue somehow... seems unlikely.

Can you log the values if left, right, bottom, top, depthRangeMin, and depthRangeMax?

-[Unknown]

[daniel229]

Add the log NOTICE_LOG(G3D, "left = %f, right = %f, bottom = %f, top = %f, depthRangeMin = %f, depthRangeMax = %f", left, right, bottom, top, depthRangeMin, depthRangeMax); above #ifndef MOBILE_DEVICE ,it works agains.

34:40:688 drawThread   N[G3D]: GLES\StateMapping.cpp:870 left = 0.000000, right = 1920.000000, bottom = 0.000000, top = 1088.000000, depthRangeMin = 0
.875013, depthRangeMax = 0.062501

[NeoDarkness]

if i turn off hardware transform and software skinning its looking good but get more lagg hardware transform and software skinning on

[unknownbrackets]

How does this look in the software renderer? It sounds like this is a lighting issue, and want to confirm if the software renderer (not the same as hardware transform off) has the issue - since it has different lighting code.

If it doesn't, it probably means an inf/nan type of issue, I think. If it does, it probably means the hardware transform and software renderer code share a bug, since they are fairly similar now.

-[Unknown]

[unknownbrackets]

Has this improved or changed at all in the latest version?

If not, a GE dump would help a lot: https://github.com/hrydgard/ppsspp/wiki/How-to-create-a-frame-dump

-[Unknown]

[ghost]

ppsspp-v1.7.1-143-g79d16f7b9-android

[unknownbrackets]

Obtained a dump: ULUS10582_#4155_naruto_white_enemy.zip

Drawing the eight-tails:

• Draws with lighting disabled: 0x0891a4e0, 0x0891ff00, 0x0892abe0. Look like octopus parts.
• Draws without any texture or color, lighting still off. This adds some outlines.
• Draws from 0x08940480, a two-tone black texture. This applies some shading and definition, no lighting. Uses slope/bias mip mapping with an additional three-tone mip level.
• Draws from 0x089528b0, still lighting disabled. This looks like a white/black gradient.

The later two draws use tex env mapping which seems to be not working right here.

Compare with this image from @daniel229 above:

In softgpu, we get the purple, but it's more flat and ugly than the PSP screenshot. This appears to be related to slope mipmapping. Changing the bias from slope + 0.0625 to const 1.0 gives us the smoother shading expected:

Compared to the more harsh (different backends since I wrote this as testing the different backends, but they're the same):

Notice the white is still missing. That's where we get to bug 2. In softgpu, that texenv gradient makes it draw nothing, which is wrong. In Vulkan/etc. it fills it in with white - more wrong.

Texture settings (same for both dark and light shading passes):

• gen: tex env map, proj: pos / src s: 3, t: 3.
• no mips, linear/linear, clamp s/t.
• verts are normals/positions only.
• lighting off, but all light flags are enabled.

Various issues here, but the tex env stuff at least has a straight forward fix, afaict. That makes it much better. Maybe should open a separate issue for the harsh shading.

-[Unknown]