Commit graph

211 commits

Author SHA1 Message Date
Clownacy
53c0e8470f Add an optimisation to the OpenGL renderers
Since we track the currently-bound "source texture", we don't need
to poll the OpenGL driver for the same info.
2020-02-03 15:53:59 +00:00
Clownacy
dfacd62662 Fix text-flickering in the OpenGL renderers
This would occur in CSE2E's options menu.

It was caused by cute_spritebatch destroying a texture atlas that
was being used by the current unflushed vertex buffer. To solve
this, we now track what textures are being used by current buffer,
and flush the buffer when the texture are about to be
modified/deleted.

As you can guess, this issue doesn't affect the SDLTexture backend,
since its batching system is half-decent.
2020-02-03 15:44:09 +00:00
Clownacy
b894543f3b Shut up a C++11 error from the enhanced branch
Nice one, C++. Adding yet another C-incompatibility?
2020-02-03 14:43:02 +00:00
Clownacy
4190493fd8 More documentation 2020-02-02 00:34:03 +00:00
Clownacy
514d9f8739 Document the OpenGL renderer a bit more
This makes it easier to find things
2020-02-02 00:21:28 +00:00
Clownacy
d84f64a5a9 Tweak cute_spritebatch settings
1024x1024 seems like overkill
2020-02-02 00:10:01 +00:00
Clownacy
eb6ab89b85 Fixes and improvements
Made texture-atlasing occur immediately, notably.
2020-02-02 00:01:45 +00:00
Clownacy
1b2d4fdb4d Fix the other rendering backends
Also finished ripping-out the sub-pixel support
2020-02-01 22:57:07 +00:00
Clownacy
eeed719c16 Get the SDLTexture renderer working again 2020-02-01 22:47:13 +00:00
Clownacy
e3cd7ce353 Add glyph-batching to OpenGL renderer
Unfortunately, sub-pixel font rendering had to go. It wasn't portable
anyway (not every display is an LCD with R->G->B ordering).
2020-02-01 22:37:59 +00:00
Clownacy
eaf2de05b6 Add glyph-batching to the SDLTexture backend
Hope it will be this easy for the OpenGL backend
2020-02-01 17:49:26 +00:00
Clownacy
639039ce3a Added Backend_PrepareToDrawGlyphs
This is to reduce OpenGL context changes, and help pave the way for
glyph-batching
2020-02-01 16:40:23 +00:00
Clownacy
e3db7749ba Add verbose error-reporting to backends
Should fix #98
2020-02-01 12:22:44 +00:00
Clownacy
ed1ff2f275 Tweak MIN/MAX macros in software renderer 2020-01-29 23:45:22 +00:00
Clownacy
788697dc5d Fix a possible hang in the audio mixer
It was likely a race condition between the game thread writing to
'organya_timer', and the audio thread reading it.

...I really need to rethink the API for this Organya-synchronisation
thing.
2020-01-29 23:21:27 +00:00
Clownacy
e4394aea1f Whoops, forgot to remove this #include 2020-01-29 22:16:21 +00:00
Clownacy
d0defa4a7b This cast shouldn't be necessary 2020-01-29 22:15:53 +00:00
Clownacy
638e96be11 Move and rename attrHot
Like the MIN and MAX macros, I'd rather keep this in the source file
than a common header.
2020-01-29 22:14:35 +00:00
Gabriel Ravier
5e6658847c Re-add missing "(double)"
Signed-off-by: Gabriel Ravier <gabravier@gmail.com>
2020-01-29 23:04:17 +01:00
Gabriel Ravier
88a8362f23 Reverted fmodf stuff and tried to make it so C++ compilers would go for the float version in appropriate circumstances (fmodf directly is unavailable due to not being in C89)
Signed-off-by: Gabriel Ravier <gabravier@gmail.com>
2020-01-29 23:01:31 +01:00
Gabriel Ravier
1442299924 Made MixSounds be marked as hot
Signed-off-by: Gabriel Ravier <gabrielravier@gabrielAncientIBMv2>
2020-01-29 23:01:31 +01:00
Gabriel Ravier
c15ecbf728 Used fmodf instead of fmod for a 0.5% performance increase in MixSounds
Signed-off-by: Gabriel Ravier <gabrielravier@gabrielAncientIBMv2>
2020-01-29 23:01:31 +01:00
Clownacy
e33bd9c8f9 Fix visual artefacting on the Pi
I'm not sure why there was linear filtering when I was rendering at
1:1 pixel ratio, but it did happen. This fixes it by forcing
nearest-neighbour. The artefacting was caused by the linear filtering
blending with pixels outside the specified texture coordinates,
creating lines around everything.

Fun fact: the framebuffer technique CSE2 uses is demanding on the Pi
(1278x720 runs at 60 FPS when the framebuffer is forced to 852x480,
even though all the internal rendering is still 1278x720). I guess
rendering those extra 920160 pixels really takes its toll.
2020-01-25 14:26:15 +00:00
Clownacy
99a8b2bd18 Improve OpenGL performance on the Raspberry Pi
Apparently 2 VBOs wasn't enough. This bumped the framerate from 13FPS
to 20FPS in a stress-test (CSE2E at 1704x960 on a Raspberry Pi 3B
in X11 with the KMS OpenGL driver).
2020-01-25 14:22:50 +00:00
Clownacy
87b296c513 Converted SDLSurface renderer 2020-01-22 22:33:34 +00:00
Clownacy
f6f3c26a3b Converted SDLTexture renderer 2020-01-22 22:31:19 +00:00
Clownacy
48b7a878ba Convert software renderer 2020-01-22 22:24:23 +00:00
Clownacy
6dac8254e2 Begin merge of Backend_Init/Backend_CreateWindow
OpenGL3+OpenGLES2 backend done so far
2020-01-22 22:19:55 +00:00
Clownacy
3baba6a727 Make FlushVertexBuffer bail when buffer is empty 2020-01-22 21:56:11 +00:00
Clownacy
88c30442c5 Force SDL2's render-batching
SDL2 will implicitly disable batching if we force a specific
rendering driver (apparently the user setting SDL_RENDER_DRIVER
counts too) for backwards-compatibility with older programs that use
their own rendering side-by-side with SDL2's. We don't do that,
though, so this commit forces SDL2 to use batching if it's available
(2.0.10 and onwards).
2020-01-21 23:39:49 +00:00
Clownacy
2fec50ff4e Use constants for attribute locations 2020-01-21 20:16:38 +00:00
Clownacy
8d92bf2004 Double-buffer the OpenGL VBO
This should reduce stalling when the OpenGL driver is still
processing the buffer when we're about to upload to it.

Hopefully, this is what was making the OpenGL ES 2.0 renderer so much
slower than the SDLTexture renderer on the Raspberry Pi 3B (SDL uses
*8* buffers). Unfortunately, I don't have access to it right now, so
I can't test this.
2020-01-21 16:43:21 +00:00
Clownacy
f8a40318a2 OpenGL fixes and optimisations
Now the VBO isn't resized constantly, and CreateSurface preserves
the currently-bound texture.
2020-01-21 15:52:33 +00:00
Clownacy
73de1adc45 Fix a memory leak 2020-01-21 13:24:30 +00:00
Clownacy
cd7ef93f5e We don't need to initialise the VBO here 2020-01-21 13:23:33 +00:00
Clownacy
79886f5ae4 Add debug prints for listing SDL2 drivers
These list what video/render/audio backends are available, and which
are currently being used.

SDL2 allows you to choose a specific video/audio driver with the
SDL_VIDEODRIVER and SDL_AUDIODRIVER environment variables,
respectively, but there's no such option for the render driver.
2020-01-20 15:27:05 +00:00
Clownacy
1cf5df2d72 Remove a debug print 2020-01-20 14:49:18 +00:00
Clownacy
767262ed8c Handle a malloc failure better
No memory leak
2020-01-20 14:48:12 +00:00
Clownacy
0423614dca More-efficient OpenGL ES 2.0 font textures 2020-01-20 13:52:50 +00:00
Clownacy
ede541db0a Clean up the OpenGL shaders a bit 2020-01-20 13:51:32 +00:00
Clownacy
4cbc56e272 Add OpenGL ES 2.0 renderer 2020-01-20 13:49:30 +00:00
Clownacy
0352af333b Fix another typo 2020-01-17 11:08:36 +00:00
Clownacy
a40c3d7b1b Perform cleanup when OpenGL's Backend_Init fails 2020-01-17 10:54:49 +00:00
Clownacy
faa96094b4 Use SDL's GetProcAddress function for OpenGL
Also added extra sanity checks
2020-01-17 10:25:18 +00:00
Clownacy
d76b074aa0 Switched from GLEW to GLAD
One less set of license obligations to satisfy (GLAD-generated files
are public-domain).
2020-01-15 07:11:43 +00:00
Clownacy
449a09b09e Fix an occational invalid memory read
Stupid floating-point rounding errors. Had to undo a fancy
optimisation to avoid it.
2020-01-09 09:10:24 +00:00
Clownacy
4faf94e951 Add missing #include 2020-01-08 19:12:12 +00:00
Clownacy
a0eb646a1f Add ability to lock certain parts of a surface
This avoids conditional jumps based on uninitalised memory, and
should be faster, and should be lighter on memory.
2020-01-07 03:52:37 +00:00
Gabriel Ravier
2911bfda5c Removed memory leak in Backend_LoadGlyph
Signed-off-by: Gabriel Ravier <gabravier@gmail.com>
2020-01-06 18:26:25 +01:00
Gabriel Ravier
752b4cee3f Correct bug in which Backend_LockSurface would not initialize surface->pixels, leaving a bug in which Backend_UnlockSurface used uninitialized values in certain scenarios (such as in ScaleAndUploadSurface)
Signed-off-by: Gabriel Ravier <gabravier@gmail.com>
2020-01-06 18:26:25 +01:00