well then that may be a bug in melonDS

____________________
Kuribo64

You can always try, but I highly doubt it'll work. It was tested over LAN and it doesn't work that well.

____________________
Kuribo64

We might as well dedicate a thread to these. The GPU is a pile of quirks.


Edgemarking behavior at screen edges



This is what happens when rendering a fullscreen quad with edgemarking. No idea what's the logic behind that, but it might stem from the scanline cache system.

____________________
Kuribo64

Let's see...


* Anti-aliasing is a separate rendering pass, presumably done after edge marking and fog.

* Edge marking is done before fog -- marked edges don't stand out in fog.
* Fog is applied to the final frame.

* Anti-aliasing only works for the topmost polygon pixels. Suppose you're drawing two identical polygons with different colors, A over B. A's edges are antialiased against B. B gets no antialiasing.

* In the case above, you can insert a translucent polygon (C) between A and B and it renders "properly" -- A is antialiased against C blended with B.


BUT


You can draw a translucent polygon (or even several) over an antialiased polygon, and it still gets antialiased properly -- the antialiasing is somehow applied before the blending.

No idea how this shit can work.

____________________
Kuribo64

An idea I just had is that when rendering a translucent polygon, it's blended against both top and bottom pixels, but doesn't push any pixel down.

Would explain how it comes out correct in all cases.

____________________
Kuribo64

I saw that presentation.

A part of their glitches happens because non-edgemarked polygons overlap the edges. The other part is what my screenshot demonstrates.

____________________
Kuribo64

the error is brought to you by wxWidgets retardedness.


wxWidgets is getting ditched.

____________________
Kuribo64

Honestly, I can't tell. It's not possible for now, but it'd have to be tried out in the future.

____________________
Kuribo64

try playing with the "bind socket" setting in emulation settings

____________________
Kuribo64

The best thing to do there will be optimizing the core.

Loading data from the cart is done word by word, and melonDS emulates the read delays.

I don't remember getting a speed hit when I implemented that, though, so I doubt speedhacking it will make it faster. It's just that when the game is loading data, it typically uses the CPU at 100%, which puts more strain on the emulator.

____________________
Kuribo64

News of the long-awaited melonDS-with-a-UI-that-doesn't-suck-ass


Windows

* will not be compatible with XP/Vista (sorry toaster-rocking folks out there)
* working, minus a few details (screen area needs to be refocused to ensure it receives keyboard input)


Linux

* displays screen fine (after a bugfix) but colors are wrong
* I still need to get the hang of cairo, it's a bit weird
* GTK isn't thread-safe and will shit itself if you do things wrong. caused a spurious random freeze. I think I've figured it out now and I need to work around it.
* keyboard input needs to be done still


Mac

* will not be done unless someone shows up who happens to have the knowledge


In general

* resetting and loading a new game works fine
* config dialogs still need to be done
* final framebuffer is xBGR (with alpha forced to 255). might want to change to xRGB as it appears to be more compatible (is the format cairo expects, more compatible with Direct2D...).

____________________
Kuribo64

'Profile Windows' is deprecated and will be axed in favor of a 'fast debug' build or something.


As for the rest... it's being worked on, so keep it in mind.

____________________
Kuribo64

you already posted in the wifi thread, no need to post the same thing a billion times

____________________
Kuribo64

yeah, it's planned, but not supported yet

____________________
Kuribo64

Have you tried rebuilding the project from scratch?

____________________
Kuribo64

ok, this is bad.


I can't test it.


can you try building it in debug mode and debugging that?

____________________
Kuribo64

I guess you'll have to debug via CodeBlocks, if that works.

____________________
Kuribo64

might want to add them to the bin/Debug directory, or to the root directory. if that doesn't work, dunno where they'd go.


can't check here, the debugger refuses to do anything and I can't figure out why.

____________________
Kuribo64

ok can you test the latest revision? it should be fixed

____________________
Kuribo64

Good then! I was afraid to end up unable to fix that issue. Mostly because I avoid any Windows version above 7 like the plague.

____________________
Kuribo64