Gods Eater Burst (ULUS10563): Excessive bloom present in all PPSSPP builds since v0.8.1-361-gc71ae64 (updated, resolved in v0.8.1-401-g76c3f16, has since returned in v0.8.1-499-g72b13d9)

[solarmystic]

UPDATE:- The bloom issue has returned again in v0.8.1-499-g72b13d9 since @raven02's pull request #2721 was merged into master 72b13d9 by @hrydgard

UPDATE 2:- New responsible commits are dca4b1e and 78f85db by @raven02

OLD ISSUE:-

The issue is as per the title. There is a severe case of bloom in the lighting especially evident in the main mission lobby.

The responsible commit is e362bc8

First problematic build to have it is v0.8.1-361-gc71ae64.

Last build not to have it is v0.8.1-356-g989f791.

Temporary Solution/Workaround:-

Disable Buffered Rendering and the issue will disappear,

[solarmystic]

Resolved in v0.8.1-401-g76c3f16. Closed.

[triglav1024]

Recurrent on v0.8.1-503-gbf1e076.

[solarmystic]

@triglav1024

You're right, it has returned, but actually started returning again in v0.8.1-499-g72b13d9 since @raven02's #2721 pull request was merged into master 72b13d9 without any hacks and with BR ON.

Temporary solution:-

OFF BR and it will go away again

Last working revision not to have this issue with BR ON is v0.8.1-496-g7c3e171

Reopened issue since it has returned in all revisions after v0.8.1-499-g72b13d9 again, without any Framebufferhacks applied to the game.

[solarmystic]

In the process of pinpointing the direct responsible commit down, I decided to compile 2 separate builds from each of @raven02's commits, 1 from v0.8.1-497-gdca4b1e and 1 from v0.8.1-498-g78f85db, and then test them both separately to see which one was the responsible one.

To my surprise, v0.8.1-497-gdca4b1e dca4b1e still looks somewhat okay:-

v0.8.1-498-g78f85db 78f85db was the one that made it excessive:-

So both dca4b1e and 78f85db by @raven02 are responsible, but by different degrees.

No Framebuffer hacks were used in these tests. Just BR ON.

[dbz400]

I think @hrydgard have been rewritten all of these for both SW & HW here bf1e076

Basically those effect are 'side effect' of correct FBO sizing and that's why it is not happened before .

I did checked this game and it has bloom effect all around the games and similar to 1st screenshot .

[solarmystic]

@raven02

The point is that the excessive bloom re-occured again on git revision v0.8.1-499-g72b13d9 which you were responsible for (using HW and SW) and before @hrydgard rewritten them for HW in v0.8.1-503-gbf1e076

Which is why I traced it back to your precise commits in v0.8.1-497-gdca4b1e and v0.8.1-498-g78f85db.

They still happen in the latest revisions btw.

[solarmystic]

Still present in the latest Orphis build (v0.8.1-557-gb039768), with BR on:-

BR off:-

[dbz400]

I think we are come close to be correct . In real PSP , those bloom effect is emitted from those windows above .

This is PPSSPP at same angle in-game

[dbz400]

I'll goto revisit that render-to-texture sizing tonight . Seems to me that we still have bit wrong sizing .

This screenshot looks to me that renders the bloom effects more correctly (provided by @solarmystic)

[unknownbrackets]

Yeah, that's weird that that one commit would've changed things.

-[Unknown]

[dbz400]

I do wonder the above screenshot only renders the bloom effect at the top left corner only as the right console is no bloom effect at all .

[thedax]

The excessive bloom is gone in buffered rendering mode now, so this can be closed, I'd think.

[solarmystic]

Yeah. Will reopen if the issue recurs.

[unknownbrackets]

And broken again now because apparently we want to not zero when the pixels we don't draw are supposed to be not zero, which I don't understand.

-[Unknown]

[hrydgard]

Reverted commit that broke it. Need to have another look at that stencil workaround.