If you're using ninjhax2/tubehax/ironhax/oot3dhax/etc, there's a known bug causing Caelina to hang during start up. If you call consoleInit() before setting up Caelina, it will run correctly without hanging. It's very odd and due to my current interests in other projects, I don't plan to fix it at this time.Thanks a lot! i now successfully compiled, but it now hangs when I try to run it, top screen black, bottom one is red. I guess I am just extremely unlucky (or stupid) person...
Homebrew [Release] Caelina - OpenGL Driver
- Thread starter machinamentum
- Start date
- Views 27,758
- Replies 118
- Likes 33
You can already use 3D using this. It's still a little complicated because you need to draw the scene, call gfxFlush with the left framebuffer then draw the scene again at a slight angle/displacement and call gfxFlush with the right framebuffer. Don't forget to call gfxSet3D!
Also I've uploaded a new prebuilt that seems to not hang on ninjhax2 when building the nehe lessons and is updated for the latest ctrulib. There still doesn't seem to be full compatibility with ninjhax2 though.
So i got to turn on the 3d on, but i don't know how to set the angle and how big it should be.
now my code is
while DrawGlScene takes bool for left/ right side, there I should have something like glRotatef(5, 0.0f, 0.0f, 0.0f); for one of the sides, right? How big exactly the angle should be?
Sorry that I keep asking. It would be nice to have some caelina wiki for dummies. (Perhaps I'll write one after I'll understand everything)
now my code is
Code:
DrawGLScene(0);
gfxFlush(gfxGetFramebuffer(GFX_TOP, GFX_LEFT, NULL, NULL));
DrawGLScene(1);
gfxFlush(gfxGetFramebuffer(GFX_TOP, GFX_RIGHT, NULL, NULL));
gfxSet3D(1);while DrawGlScene takes bool for left/ right side, there I should have something like glRotatef(5, 0.0f, 0.0f, 0.0f); for one of the sides, right? How big exactly the angle should be?
Sorry that I keep asking. It would be nice to have some caelina wiki for dummies. (Perhaps I'll write one after I'll understand everything
)Choosing the displacement is sort of complicated and it seems everyone has a different approach based on testing. I've attached an example project that uses most of Caelina's features, including 3D using the 3DS' slider. This one uses a translation as a displacement instead of a rotation.So i got to turn on the 3d on, but i don't know how to set the angle and how big it should be.
now my code is
while DrawGlScene takes bool for left/ right side, there I should have something like glRotatef(5, 0.0f, 0.0f, 0.0f); for one of the sides, right? How big exactly the angle should be?
Sorry that I keep asking. It would be nice to have some caelina wiki for dummies. (Perhaps I'll write one after I'll understand everything
EDIT: Playing around with ninjhax2 some more, it seems like there's some global initialization somewhere that seems to cause the issues on these entry points, which is weird because I believe none of the code uses global initializers. When compiling the lib with options to remove dead code (which supposedly disables global initialization if i read correctly), it doesn't cause the usual issues such as breaking netloader.
Last edited by machinamentum,
while DrawGlScene takes bool for left/ right side, there I should have something like glRotatef(5, 0.0f, 0.0f, 0.0f); for one of the sides, right?
The goal should be to reproduce human stereo vision. Shouldn't you just keep a fixed distance between the 2 points of view and let the angle be adjusted by the 3D slider?
Maybe this helps a bit?
http://www.gali-3d.com/archive/articles/StereoOpenGL/StereoscopicOpenGLTutorial.php
Small progress update: a new feature of this project is disabling error checking. Although relatively useful, for projects that are entirely free of errors, this more or less just wastes CPU cycles and fills icache lines with code that never runs. Error checking can be disabled for source builds only, by adding "-DDISABLE_ERRORS" to CFLAGS
I'll likely add this same mechanism for display lists as well soon.
EDIT: DISABLE_LISTS will disable the overhead incurred by display list building/checking/etc...
Also, it seems that the GPU actually stores some amount of GL state across command buffers, so I've started a new branch that takes advantage of this by using less memory and managing less state on the client-side. This of course will be available as a source build only and will not be uploaded as pre-builts.
Code:
CFLAGS="-DDISABLE_ERRORS" make
make installI'll likely add this same mechanism for display lists as well soon.
EDIT: DISABLE_LISTS will disable the overhead incurred by display list building/checking/etc...
Also, it seems that the GPU actually stores some amount of GL state across command buffers, so I've started a new branch that takes advantage of this by using less memory and managing less state on the client-side. This of course will be available as a source build only and will not be uploaded as pre-builts.
Last edited by machinamentum,
Small project update: a long time missing feature was using a single context to draw to both displays. This new update changes gfxFlush to use the output display's width and height so a project could draw to both displays using something along the following:
Get it here
EDIT: it seems that it currently doesn't scale the image to the framebuffer. The documentation in ctrulib is somewhat ambiguous on how anti-aliasing and scaling works. I'm looking into it.
EDIT2: this can likely just mitigated by just using glViewport to the size of the framebuffer you need to flush to
Code:
DrawGLScene();
gfxFlush(gfxGetFramebuffer(GFX_TOP, GFX_LEFT, NULL, NULL), 240, 400);
DrawGLScene2();
gfxFlush(gfxGetFramebuffer(GFX_BOTTOM, GFX_LEFT, NULL, NULL), 240, 320);Get it here
EDIT: it seems that it currently doesn't scale the image to the framebuffer. The documentation in ctrulib is somewhat ambiguous on how anti-aliasing and scaling works. I'm looking into it.
EDIT2: this can likely just mitigated by just using glViewport to the size of the framebuffer you need to flush to
Last edited by machinamentum,
Woah, this is awesome! Hopefully this will lead to more quality homebrew games. Maybe someone will use this to port Doom! That would be sweet.
Doom is already ported
- Joined
- Aug 9, 2015
- Messages
- 421
- Trophies
- 0
- Age
- 24
- Location
- Seattle
- Website
- kje123.github.io
- XP
- 524
- Country
v0.5 now available.
Change log:
This update brings more stability, and features, when using multiple contexts. A context can be used to render an image then flushed out to a buffer and submitted to a texture in the same context or another context. This essentially can work in place of render-to-texture until render targets are actually implemented. Note that GX_TRANSFER_FMT_RGBA8 is equivalent to GL_BGRA when submitting as a texture.
Attached is an example of these new features working.
EDIT: NeHe lesson06 example now available (uses stb_image, included in the repo)
Change log:
This update brings more stability, and features, when using multiple contexts. A context can be used to render an image then flushed out to a buffer and submitted to a texture in the same context or another context. This essentially can work in place of render-to-texture until render targets are actually implemented. Note that GX_TRANSFER_FMT_RGBA8 is equivalent to GL_BGRA when submitting as a texture.
Attached is an example of these new features working.
EDIT: NeHe lesson06 example now available (uses stb_image, included in the repo)
Last edited by machinamentum,
v0.6 now available. This release marks the beginning of GL ES support.
Change log:
GL ES support is still very new and currently doesn't support the fixed-function vertex attributes (glTexCoordPointer, glNormalPointer, glColorPointer, etc) which makes it very limited.
Change log:
A new example is now available demonstrating the GL ES library.
GL ES support is still very new and currently doesn't support the fixed-function vertex attributes (glTexCoordPointer, glNormalPointer, glColorPointer, etc) which makes it very limited.
If I understand correctly, you need to set the stride bits of GPUREG_ATTRIBUFFERi_CONFIG2. ctrulib's GPU_SetAttributeBuffers function is overkill in that it assumes a packed buffer, so it calculates that stride based on the number of components and the size of the attributes you specify (see gpu.c#L470). You'd probably want to use a masked write to overwrite just the stride bits after calling SetAttributeBuffers or just stop using SetAttributeBuffers all together and add the writes yourself.
Last edited by machinamentum,
Similar threads
- No one is chatting at the moment.
-
-
-
-
-
-
-
-
-
-
-
-
-
@
Psionic Roshambo:
32GBs is the baseline for 4K video editing these days and lots of recommendations for 64GBs but for games 16,GBs is honestly plenty for a long time. -
-
-
-
-
-
-
-
-
-
@
Psionic Roshambo:
Some sort of police thing going on near me lol "Get out of the car with your hands up and walk backwards towards us" over a loud speaker thing
-
-
like a 250 dollar kit back when DDR4 was Intel only
