Gsdx future discussion

[refraction]

I don't see a lot of other games that have bugs due to no/improper mipmapping though so maybe it wouldn't be high priority.

Destroy all Humans and Rachet and Clank series are 2 sets of games that would benefit greatly from mipmapping support (so about 5 or 6 games)

[CK1]

Destroy all Humans and Rachet and Clank series are 2 sets of games that would benefit greatly from mipmapping support (so about 5 or 6 games)

Spiderman 2 needs it pretty bad as well. As far as my games, J&D1 needs mipmapping for foliage and transparencies to render correctly.

[gregory]

some people would want some kind of mipmapping support in the HW renderer so they can play Ace Combat 5 in higher resolution without the landscape turning to garbage blocks

The purpose of Mipmap is to reduce the resolution of textures. So instead to send a texture of 256x256, you send several layers of the texture (256x256, 128x128, 64x64, 32x32, ...)

Most of GSdx issues is related to a game optimization not mipmap per se. All 3d games use mipmap. Yet few of them suffer of issues. The issue is that instead to have a complete texture (with all layers). You have random data for the layer 256x256, 128x128, and you only begin to see real data in 64x64 and below. Currently GSdx always upload the layer 0 (or the highest resolution texture).

[Blyss Sarania]

Is there any way to detect what size texture the primitive needs and generate our own fake mipmaps? I mean assuming we can figure out the needed size, running layer 0(or whichever has the "real" data)through a bilinear downscale and outputting that size should be trivial.

[gregory]

my bad I make a couple of shortcut. For GS a texture is a pointer. Layer N could be a different object than layer 0. I don't think we need to compute our own mipmap (and we can't as a matter of fact). We just need to upload the layer 2 (for example, based on the Qmin vertex information). And only do the rendering with the layer 2. I think it will work on a couple of basic case.

Complex case is that GS could render directly into a layer of a texture. Or several of them in the same time.

[FlatOut]

Spiderman 2 needs it pretty bad as well.

Spider-Man 2 is very different, it actually looks better with mipmapping turned off in software mode.
The Spider-man 2 texture issues look more similar to Stuntman(partly similar to Driv3r) and the cars in Ridge Racer V (texturecache related?).

[Nobbs66]

Spider-Man 2 is very different, it actually looks better with mipmapping turned off in software mode.
The Spider-man 2 texture issues look more similar to Stuntman(partly similar to Driv3r) and the cars in Ridge Racer V (texturecache related?).

Ridge Racer V is a texture cache issue.

[FlatOut]

Ridge Racer V is a texture cache issue.

I believe Spider-man 2 and Stuntman are too, I'm pretty sure I've tested them both without it. Though I think Spider-Man 2 does have more issues in HW than just that.

[Blyss Sarania]

my bad I make a couple of shortcut. For GS a texture is a pointer. Layer N could be a different object than layer 0. I don't think we need to compute our own mipmap (and we can't as a matter of fact). We just need to upload the layer 2 (for example, based on the Qmin vertex information). And only do the rendering with the layer 2. I think it will work on a couple of basic case.

Complex case is that GS could render directly into a layer of a texture. Or several of them in the same time.

Okay, I see now.

That explains why this isn't super straightforward. I've wondered.

[Parotaku]

I may be wrong but I think 'Saru gechu 2/Ape escape 2' texture bug is Mipmap related too...