Sega Rally Revo : missing sky textures

[dbz400]

Just for record .

-Buffered mode

With some tweak to framebuffer to memory , it can render correctly (not our latest build)

[dbz400]

It is pretty funny here .

If i set the following in TextureCache::AttachFramebuffer() which is read from memory

  if (exactMatch) {
          if (g_Config.iRenderingMode == FB_BUFFERED_MODE)
             return;

The road texture appears okay in buffered mode .Wheneve it go into attach framebuffer , texture will disappear
somehow ....

If i do the same for non-buffered mode , for example

  if (exactMatch) {
          if (g_Config.iRenderingMode == FB_NON_BUFFERED_MODE)
             return;

The road texture appears okay in non-buffered mode as well.

[dbz400]

Look like it is using overlapping framebuffers with different formats simialr to Silent Hill .
#1541

13:34:061 user_main I[SCEGE]: GLES\Framebuffer.cpp:605 Creating FBO for 04000000 : 480 x 272 x 1
13:34:077 user_main I[SCEGE]: GLES\Framebuffer.cpp:605 Creating FBO for 000cc000 : 480 x 272 x 0
13:34:105 user_main I[SCEGE]: GLES\Framebuffer.cpp:605 Creating FBO for 000d4000 : 480 x 272 x 3
13:34:172 user_main I[SCEGE]: GLES\Framebuffer.cpp:605 Creating FBO for 04044000 : 480 x 272 x 1

[unknownbrackets]

Hmm, cc and d4 are relatively far apart. Are you sure they're actually overlapping?

It seems like this is an exact match, right? So the problem may not be related to overlapping.

It's possible this is an issue of postprocessing, e.g. loading into memory and then drawing on top of it, which is not supported at all. Which of those framebuffers specifically is the road (e.g. if you check the address, which one makes it black if you attach it)?

It could also be intra-framebuffer transfer or etc.

-[Unknown]

[dbz400]

I'm also guessing they may be somehow overlapped but mostly likely i'm wrong as they are so far apart.

Yes , it is exact match.It go into the same format one .

INFO_LOG(HLE, "Render to texture with same formats %d at %08x", entry->format, address);

29:29:811 user_main I[HLE]: GLES\TextureCache.cpp:197 Render to texture with same formats 2 at 040cc000
29:29:834 user_main I[HLE]: GLES\TextureCache.cpp:197 Render to texture with same formats 3 at 040d4000

[dbz400]

The address with 0x040cc000 is the framebuffer render for the road itself .

[dbz400]

I think we don't support intra-framebuffer transfer as well , right?

[unknownbrackets]

Not at the moment. Are there any block transfers?

-[Unknown]

[dbz400]

No , there is no block transfer happened throughout the game .

[dbz400]

GE Debugger .The missing road.

[unknownbrackets]

So that means it's drawing it somewhere, e.g. to another FBO. More functionality in the debugger will probably help figure this out.

-[Unknown]

[dbz400]

@unknownbrackets , that would be great .It would be great if a one off button can be provided so we don't have to keep pressing 'Break' button until all the textures drawn :)

[unknownbrackets]

But what would the button do? It can't know when you want to stop.

Planning to be able to set breakpoints at list addresses but you still have to tell it when to stop.

-[Unknown]

[dbz400]

I take the road above as an example .

I need to press around 50 times so it can draw all the textures completely .When i press 51 times , it will be white screen and redraw from 1st draw again .

[dbz400]

So the question is , is it possible we can know when 1st texture got redrawn ?

[unknownbrackets]

Oh, yes, I also want a "break on texture" feature, maybe it could do texture changes as well. There are many options, although I don't want to make it too complicated.

-[Unknown]

[dbz400]

Thanks .It is a minor thing .I should be okay to press many times to get the complete texture :)

[dbz400]

This game shows few FBO sharing depthbuffer then i tried something here and the road texture shows up .

        } else if (MaskedEqual(z_address, vfbs_[i]->z_address) && fb_address != vfbs_[i]->fb_address) {
            WARN_LOG_REPORT(SCEGE, "FBO sharing existing depthbuffer (unsupported), %08x/%08x and %08x/%08x", fb_address, z_address, vfbs_[i]->fb_address, vfbs_[i]->z_address);
            if (useBufferedRendering_) {
                vfbs_[i]->fbo = fbo_create(vfbs_[i]->renderWidth, vfbs_[i]->renderHeight, 1, true, vfbs_[i]->colorDepth);
                if (vfbs_[i]->fbo) {
                    fbo_bind_as_render_target(vfbs_[i]->fbo);
                } else 
                    ERROR_LOG(SCEGE, "Error creating FBO! %i x %i",  vfbs_[i]->renderWidth, vfbs_[i]->renderHeight);
            } else {
                    fbo_unbind();
                    // Let's ignore rendering to targets that have not (yet) been displayed.
                    gstate_c.skipDrawReason |= SKIPDRAW_NON_DISPLAYED_FB;
            }
        }

[unknownbrackets]

I'm not sure I understand what you're doing there. You replaced the other fbo becaues it was sharing the depth buffer? Obviously the above will leak, right?

-[Unknown]

[dbz400]

Yes , right but only in sharing depth buffer case. I think it just accidentally worked to show up the road textures in this game.

[jbakach]

Is there any plan for a fix of this game?

[dbz400]

It is interesting .The road shows up okay now but not the sky (It is becasue of depth clear)

[unknownbrackets]

Has this improved with the recent depth changes? How does it look in the software renderer?

-[Unknown]

[dbz400]

It is the same and btw , it got lots of invalid ADSR modes

43:39:812 user_main W[SCESAS]: HLE\sceSas.cpp:398 sceSasSetADSRMode(08d6c940, 12, 15, 00000005, 00000005, 00000005, 00000005): invalid modes
43:40:179 user_main W[SCESAS]: HLE\sceSas.cpp:398 sceSasSetADSRMode(08d6c940, 27, 15, 00000005, 00000005, 00000005, 00000005): invalid modes
43:41:013 user_main W[SCESAS]: HLE\sceSas.cpp:398 sceSasSetADSRMode(08d6c940, 12, 15, 00000005, 00000005, 00000005, 00000005): invalid modes
43:41:413 user_main W[SCESAS]: HLE\sceSas.cpp:398 sceSasSetADSRMode(08d6c940, 23, 15, 00000005, 00000005, 00000005, 00000005): invalid modes
43:42:748 user_main W[SCESAS]: HLE\sceSas.cpp:398 sceSasSetADSRMode(08d6c940, 12, 15, 00000005, 00000005, 00000005, 00000005): invalid modes

[dbz400]

softgpu seems to be working fine (but the road seems to be wrong ?)

[dbz400]

Meanwhile , there should be two videos playback but missing here

[unknownbrackets]

How is this these days with the latest git and "simulate block transfers"? At least the videos maybe?

-[Unknown]

[ppmeis]

Tested with latest build. Now both videos play fine at startup. Sky still missing and there are some artifacts on the road:

Also, car's mud is with wrong colors:

correct view:

[thedax]

According to the forum, all of these issues have been fixed.

http://forums.ppsspp.org/showthread.php?tid=2595&pid=93757#pid93757

Closing for now. Please respond if you have a different experience.

[ppmeis]

This should be revised. The dirty cars effect is missing. I mean when you drive for a while, mud appears in the cars' body (you can see that on my comment above)

[unknownbrackets]

Well, that's not really related to the sky right? I think it makes sense to create a separate issue if only that affect is missing now, maybe with GE debugger information.

-[Unknown]