[BUG] Display: Ripple/shifting effect

[Dart Zaidyer]

For me, the editor used to display perfectly. But once tweaks were made to make it work better for other people, I am now experiencing an even ripple/shift effect every few tile columns as seen in this screenshot:


It appears to distort and then go back to normal with every other pixel passed on the scroll bars. Horizontal scrolling creates the stretching, vertical scrolling seems to augment it.
Tiles on layers also noticably shift upward or downward by a single pixel when this happens.

After further testing, this only occurs when maps get to a certain level of detail complexity. Simply applying "Random noise" followed by "Auto fringe" is enough to make it happen.
_________________

Remember: Friends don't let moons eat monkeys.

[Diaftau]

Huh. The only difference between the old maps and the new ones is that the old ones use the legacy tilesets which are bmps, and new ones (should) be using the non-legacy tilesets which are pngs. Although that shouldn't make a difference.

If you create a new map and use the legacy tilesets, does the problem occur?

When Ant was experiencing the shimmering grid, it was caused by the device not being reset when the window was resized. Could you also try resizing the window to see if it makes any difference?
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Diaftau]

*Double post*. Grr edits.

Ok, so scratch my comment about tilesets. Clearly it's (probably) doing it all the time, it's just not noticeable unless the map reaches a certain level of noise.

Try the resize thing though.

Oh, and what graphics card are you running?
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Dart Zaidyer]

"Shimmering grid" doesn't seem to have reoccurred yet, though. Could have just been tiles on layers shifting around by one pixel.
Resizing the window does nothing, and trying to make it smaller actually serves to crunch the effect together and make it worse.

I'm running on an NVidia GeForce FX 5200 Ultra.
_________________

Remember: Friends don't let moons eat monkeys.

[Diaftau]

I expect the problem is being caused by vertex positions not aligning directly to pixel positions post transformation - possibly due to a rounding error somewhere.

Another change that I'd made (to fix Pixie's bug) is that when the D3D device is created, it now has the option of dropping to software vertex processing instead of just using hardware. If, for some reason, that's happening to you, that might account for the rounding error - SVP likely works at a lower precision to increase performance.

Did it definitely not ever occur before 1.0.0.31?
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Dart Zaidyer]

This definitely never occurred before 1.0.0.31, I just tested an older version.
_________________

Remember: Friends don't let moons eat monkeys.

[Diaftau]

In that case the most likely culprit is falling back to SVP. I'll check that the way that I'm looking for hardware vertex processing support is correct (I did kinda just guess), as well as add a mechanism to let you find out whether HVP or SVP was used within the editor itself.

I'll be releasing a new version soon with full undo history support, so I'll shoehorn those things into it as well. You can let me know the results then.
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Diaftau]

I've released the next version.

Try it and see if you're still getting the shimmery problem. If you are, then click View > Renderer Information, and let me know what the device create flags are. (It'll either be HardwareVertexProcessing, MixedVertexProcessing or SoftwareVertexProcessing).
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Dart Zaidyer]

Bug still occurs.

Renderer information shows it's using HardwareVertexProcessing.
_________________

Remember: Friends don't let moons eat monkeys.

[Diaftau]

Oh. Snap.

Presumably it's not stopping you from actually using it though, is it? Only, I'm a little stumped as to the cause. Nothing has changed with regard to how textures are loaded, and how tiles are drawn through those releases, so it's a little odd.

I'm going to keep it in mind though, hopefully the solution will present itself over the next couple of days.
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Dart Zaidyer]

Coming back to this bug now, it seems to me that the problem has something to do with the graphics card drivers or DirectX, since it also occurs on other completely different programs. Switching to a different renderer such as OpenGL seems to work around the problem. I suggest that if possible, you try to work it in for the next revision.
_________________

Remember: Friends don't let moons eat monkeys.

[Diaftau]

Write a completely seperate renderer that runs on OpenGL? I don't believe that's necessary. In fact, it's overkill . The problem will be due to a device setting that OpenGL defaults to that differs to D3D's default. I suspect it's a texture filter or clamping setting, but I haven't found out which yet.

It's possible that it's a driver controlled setting - I know that on most drivers you can tweak the profile of OpenGL and D3D settings independently. In this instance, tweaking the driver profile would resolve the issue, but as yet I wouldn't know what value to tweak .
_________________
My Portfolio Site | My Development Log | Schools on the Net - Website Design for Schools

[Runefox]

Latest graphics drivers, anyone?
_________________