GSdx

[Krisan]

I mentioned awhile back some issues with the minimap in P3:FES, and from what I was told it was kind of impossible to fix, due to it requiring perfect pixel alignment. I figured oh well, and I moved on. Recently though I was told by someone to try a really old version of GSDX, which to my surprise actually fixed various graphical problems with the game, albeit introducing a minor new one in the process.

Comparison Shots:


The odd shadow bug in the menu is fixed, and the minimap is rendered perfectly, though the character portraits suffer from a small artifact whenever they blink. (It's not there normally, and since blinking happens so fast it's not that big a problem.)

It's also worth noting that earlier versions (890 and earlier or so) suffer from all kinds of graphical glitches elsewhere, like flickering and garbled floor textures, but 950 seems okay so far that I've played. Haven't noticed anything yet aside from the blinking artifact.

I just figured I'd post this for future reference in case anyone else cares\is interested. Though I'd also be curious to know (I'm a curious creature by habit) why this terribly old plugin manages to render the minimap correctly when the issue itself sounded terribly complex from how I first heard it. Were some tricky hacks in place or something? And if so, is that why the portraits artifact? Or is that a separate issue..? Ah well, no big deal if my curiosity is asking for too much, I'm just glad to have found this workaround for the problem.

[rama]

Run around a bit more. Is the map still perfect when more tiles are uncovered?

[Krisan]

The minimap? Yeah, the entire minimap is correct in that version.

Example:


It may be worth adding that Checked Texture filtering will cause linebreaks in the minimap tiles on r950, but the alignment itself remains intact. Square\Grey ticked texture filtering doesn't cause this to happen though. Also this is using the DX10 Hardware renderer. The DX9 version and all software versions (including the DX10 software renderer) have the broken minimap exactly as it is seen in the current version of GSDX.

The "artifact" in character portraits when they blink also seems to be more of an alignment issue than an actual artifact, and any renderer or plugin that gets the minimap wrong seems to have the blinking alignment correct in portraits. This leads me to suspect whatever fixes the minimap is what's breaking the blinking animation. (Although of the two, I'd rather have broken blinking considering it's fairly non-consequential, and the minimap by comparison is much more practical..)

As a sidenote, the shadow bug on the portraits in the menus? That seems to have started at r997 or so. Prior to that revision, the portraits in the menus were fine. Interestingly enough the minimap is still correct in r997 too, it doesn't appear that the minimap has problems on the DX10 renderer until r1341.. Everything up to r1340 has the minimap looking correct in the DX10 renderer.

It'd be interesting if a P3 specific hack could be added in a newer GSDX for the minimap, based on whatever makes it work in older revisions.. but I realize that's probably asking for too much.

In case you wanted to test this yourself, I've attached a memory card containing a P3:FES save (US Version) - it's during the party's first visit to the main dungeon of the game.

[rama]

I have a save to test thise, thanks. Also thanks for testing so far
The shadow draw order is unrelated to the aligment problem.

[KrossX]

So a "bug" in the Dx10 HW render had nice mipmap without weird shadows and stuff? Cool.
Just gotta bring the bug back... and to every render. =D

[Krisan]

I have a save to test thise, thanks. Also thanks for testing so far
The shadow draw order is unrelated to the aligment problem.

Oh yeah I figured it was unrelated. Just mentioning it all the same.
And no problem. As I said earlier I'm a curious creature by habit.. If my curiosity helps in some way though, awesome. ^^

So a "bug" in the Dx10 HW render had nice mipmap without weird shadows and stuff? Cool.
Just gotta bring the bug back... and to every render. =D

Haha, it's weird isn't it? ^^

[dvstudio3]

From what I understand, there are certain functions that can't be produced such as the infamous D3D Blending due to the limitations of the Direct X API.

Since the DX11.1 release is fast approaching and opens more hardware accessibility do you think that these functions (D3D Blending) can be produced?

Also, in the mean time are there any work arounds that can enable D3D blending besides using the software renderer? Pseudonym suggested in June this year that such a work around exists but hasn't been implemented however he didn't specify what the work around was and where it's to be implemented in the source.

[KrossX]

I don't see any new blending modes: Direct3D 11.1 Features. Lots of functions that must NOT be used when deploying Windows Store apps though. (hmm...)

[Squall Leonhart]

not sure if the blending couldn't be faked with D3D10.1 shaders anyway :<

[VIRGIN KLM]

One question. May have been asked already but I will rephrase it. Since my PC is an Ivy Bridge and it supports both SSE4 and AVX I can notice a gain in speed in AVX in software mode. But in hardware mode, should it be the same or I should swap to SSE4? Or better, AVX may be superior to software rendering to SSE4 but is it equal or inferior to hardware mode in any aspect? Also I think the AVX implementations on AMD and Intel processors are a bit different right?