Tales of Radiant Mythology 2 Missing Map Graphic and some other issues

[Sendayu]

Heya PPSSPP Team

i am writing this to let the team know of another bug with Tales of Radiant Mythology 2 game(The First one is the missing music bug that is being investigated here: #5323 )

This bug is where the in-game map in any area of the game never show no matter what the player do even if the player enable it in the option setting.Seeing how as more dungeon revealed and some share the same wall design without a map it;s very easy to get lost.Here;s a screen of it:

How the map should show based on the trailer of the game:

Is hard for me to trace what causing this issues as I don;t see any red word of the error though I did notice weird word when the in-game dungeon load :

There is other error that I found though I don;t know where this belong to so maybe this could help.

This word shows when the game load:

This error happen after game finish loading an in-game save:

This error happen when the music gone missing bug happen(I don;t know whether this is the one caused the music to go missing or not as this error happen when the music go missing)

Sorry for the long posts.hope this problem can be fixed.

[dbz400]

How about in framebuffer to memory rendering mode?

[Sendayu]

I tried it in the framebuffer(GPU) setting and it causes the game to have many glitches with the menu and all the bug to still happen.Then I decide to try with Framebuffer(CPU) and it causes the game to crash when loading any area along causing the menu in-game to glitch too.

Sorry for the late reply as I went to sleep after posting this issues.

[dbz400]

Can you share a savestate here ? then we can take a look the GE debugger to see if this map is being rendered.

[Sendayu]

Ok here's a save state of the game:
http://www66.zippyshare.com/v/43389868/file.html

[dbz400]

Thanks .Will test it out shortly

[Sendayu]

Just wondering when you mean by savestate is it the save data of the game or the quick save/load data?

The one I provided is the save data of the game ,if you want the quick save/load data,where do you want me to use it?

[dbz400]

Quick savestate (F2) would be much better to take exactly at the point you show above with map .

[Sendayu]

Ah ok,sorry about that as I easily got confused with that.Here;s a Quick savestate at the same map as I shown above.:
http://www25.zippyshare.com/v/60070194/file.html

[dbz400]

Thanks .I can reproduce it . Looks like the map exists but somehow not rendered

[dbz400]

I tested framebuffer to memory mode , it is not showing the map as well.

(Cannot test in softgpu , it crashes)

[Sendayu]

Hmm I wonder what could cause this problem to happen.I tried previous version from 0.95 and the problem still happen.Maybe there is something missing that cause this to happen like the music missing problem.

[unknownbrackets]

Does the log show any block transfers? It could also be a memcpy or dmac copy.

-[Unknown]

[Sendayu]

Hmm I only shows what was colored yellow and red in the logs.None of the coloured word has that name of block transfers and memcpy and dmac copy.Do you want me to provide a logs with using the log.bat just like getting the the logs when musing gone missing and the now loading error bugs?

[dbz400]

i try to log block transfer as well as dmac copy but seems not seeing it .

[Sendayu]

Tested with v0.9.6-889-gb1d6eef (12 Feb 2014)
The missing map problem and the music gone missing problem still happen though what I observed is that more sound effect that never on before this version start happen so the music is improved in a way where more sound effect played.From my logs on Tales of the world 3 the missing music problem show;s this when it happen though for some reason nothing new is been shown at all in Tow2:

[Sendayu]

Tested with the version v0.9.7.2-32-g106656f and both the missing map and music problem still happen.The sound effect in the game is getting nicer from what I observed.

[unknownbrackets]

How is this in the latest git build with "simulate block transfers" enabled?

-[Unknown]

[daniel229]

still missing with "simulate block transfers" enabled?

[unknownbrackets]

Does it work in softgpu? What if you change this, in GPU/GPUState.h:

u32 getFrameBufRawAddress() const { return (fbptr & 0xFFFFFF) | ((fbwidth & 0xFF0000) << 8); }

To:

u32 getFrameBufRawAddress() const { return fbptr & 0xFFFFFF; }

Does it work?

-[Unknown]

[daniel229]

With softgpu it crashes when entrying the cave.Change the code,it works

[unknownbrackets]

Hmm, the arrow doesn't seem to be drawn properly...

-[Unknown]

[daniel229]

Arrows are drawn,the roads are not drawn.

[unknownbrackets]

#6383 should at least make it do that. Does the GE debugger show anything interesting when drawing the roads?

-[Unknown]

[daniel229]

when it's wrong.

[unknownbrackets]

Hmm. I wonder if it's stencil/alpha related, since it looks transparent. Is that when it draws the actual path? It seems like that's already when it's drawing the completed map to the screen.

It seems to use 8888 when it draws the map to the screen, what's the blending mode?

-[Unknown]

[daniel229]

Sorry,that's the last step before the minimap showing on the screen.
This is the path drawing

blending mode

[unknownbrackets]

Aha. Stencil test increment + alpha test > 0. Also normal blending.. but the format is 3, how could a few increments make it opaque?

Wait, I just looked more closely at the PSP screenshot - is it actually supposed to be transparent? I just assumed it was incorrect.

-[Unknown]

[daniel229]

It's not transparent in psp.

[hrydgard]

A few increments could make it effectively opaque if it's for example using only alpha testing instead of normal blending to draw the final composited map to the screen.

[unknownbrackets]

I guess we could try blending an INCR as GL_FUNC_ADD (GL_CONSTANT_ALPHA, GL_ONE) and set the constantAlpha to 1.0f / depth? And use GL_FUNC_SUB for DECR? This is assuming we can replace/dualsource the alpha.

Really don't know what we can do with INVERT. Maybe GL_FUNC_SUB (GL_ONE, GL_DST_ALPHA)?

-[Unknown]

[hrydgard]

If we haven't seen INVERT used this way, it might not be necessary, but that might work. But yeah, should be able to do INCR and DECR that way as long as it doesn't conflict with some other blend mode set...

[unknownbrackets]

Fixed by #6935, yay (and map graphic fixed by #6777.)

-[Unknown]

[daniel229]

The music issue return broke by unknownbrackets/ppsspp@60378dd.

[unknownbrackets]

Is anything logged?

-[Unknown]

[daniel229]

28:20:843 cSoundMusic  I[ME]: HW\MediaEngine.cpp:87 FF: Frame data doesn't match channel configuration!
28:20:843 cSoundMusic  E[ME]: HLE\sceAtrac.cpp:442 avcodec_decode_audio4: Error decoding audio -1094995529 / bebbb1b7
28:20:847 cSoundMusic  I[ME]: HLE\sceAtrac.cpp:1413 sceAtracReleaseAtracID(0)

[unknownbrackets]

Actually, going to close this in favor of #4595. I suspect #8176 will fix it anyway.

[daniel229]

#8176 does not fix it.