Posts: 6.069
Threads: 68
Joined: May 2010
Reputation:
167
Location: Grenoble, France
I'm quite busy with various stuff so I won't be able to have a look soon. But it feels like a core bug rather than a GSdx bug. I guess disabled alpha testing drop some draw call in zerogs and likely broke the GS rendering.
Posts: 6.069
Threads: 68
Joined: May 2010
Reputation:
167
Location: Grenoble, France
Hum, hum. I have the feeling that rendering bug is due to core rounding. I'm currently looking at your trace. I can see the draw call that ought to draw the missing element. However the configuration is quite strange.
* Alpha of the fragment is 0, so GS will always (for any configuration) output 0 for alpha
* Alpha test is enabled. All fragments that have a 0 alpha will be discarded
=> conclusion the full draw call is discarded.
My guess is that normal PS2 ought to output very small alpha values (range between 0-2). I really don't know. However maybe it would be possible to add a CRC hack to disable alpha test on this game when all pixels are rendered. It would avoid the null drawcall.
Posts: 6.069
Threads: 68
Joined: May 2010
Reputation:
167
Location: Grenoble, France
I pushed a CRC hack for this game. Which basically disable alpha testing for this game. The fix is only for the HW renderer (hope it won't have extra issues).
Posts: 1.977
Threads: 25
Joined: Sep 2009
Reputation:
31
09-05-2016, 05:03 AM
(This post was last modified: 09-05-2016, 05:07 AM by tsunami2311.)
NVM I fiqure out my own question
Posts: 6.069
Threads: 68
Joined: May 2010
Reputation:
167
Location: Grenoble, France
With dichotomy, you can find the bad build with 10 tries. Are you sure that you configuration is correct. You need CRC on minimal at least
Posts: 3.031
Threads: 4
Joined: Aug 2015
Reputation:
134
I did a quick test based on a GS dump and it regressed between dev-1389 and dev-1396.