- "fat" version uses repacked (zip -0) archives; also separate "data"
and "themes" folders
- "slim" version doesn't use any external themes (for speed reasons)
- consolidate public #define's (just ATARI)
- cpu compiler flags are specified in the script
- allow explicit move16, SV and SV Blitter features enabled/disabled
Provide two build scripts:
1. "Fat" one targeted at 040/060 machines (possibly with SuperVidel)
This one is optimized for 68020-60 (so it's still possible to try
highres engines on 68030 machines).
2. "Slim" one targeted at 030 machines (Falcon030+DFB/CT2 or TT030)
This one is optimized for 68030 and stripped from even more features:
"fancy" (highres) themes, move16 & SuperVidel routines and most
importantly the highres engines.
This was long overdue. From my observation dirty rects don't need much
care, usually game engines already prepare them in a good shape (no
overlapping).
Also the overhead of calling the rect version of C2P isn't as big as I
was fearing, 256 64x64 rectangles are blitted roughly as quickly
as blitting one 1 MB block.
Removal of the rects traversal significantly speeds up Eco Quest for
instance.
ScummVM engines are not built for synchronous screen locking (i.e.
updateScreen() can't be really used for any sort of blocking updates
otherwise performance suffers).
Single buffering now means deliberate tearing as a speed optimization
and triple buffering provides a tearingless picture with slight overhead.
Some engines are too bound to linear surfaces in 8bpp that it is very
hard to repair them. So instead of polluting the engine with
Surface::init() & delete[] Surface::getPixels() just use this
workaround.
It is also useful as an early fix for newly found engines which may be
fixable later.
Toonstruck, Discworld and others manage their own cursor and are
interested only in *mouse* position, not in *cursor* drawing. And
indeed, atari backend doesn't go further than rectangle drawing.
- fix restoration of dirty rects from overlay
- fix vsync handling (don't try to be too smart; we can't store "vsync"
from anywhere because that wouldn't play nicely if a game has its own
"vsync" set) ... so keep the checkbox as is
- overlay restore for 640x400
- never wait for vbl in overlay
- readme.txt
Instead of shifting the destination to the left, if was moving the source to
the right, which had the effect of shifting the dirty rect. So part of the
surface that should have been copied to the screen was not, and instead it
was copying part of the surface that did not need to be copied.
Partially rolls back a change introduced in
2caa338dbddd9429c766a281231c8fb98701d968; Director relies on the notion
of updateScreen() being near-zero cost if there are no changes.
For example, starting director:henachoco05 went from taking a few
milliseconds to taking 11 seconds, and opening the top menu in
director:easternmind only works on 10% of the screens.
When scaler shaders are used, cursor size is adapted to scale on the
game screen texture and not on the back buffer so the size is scaled
twice and is imprecise.
Using a float gives more precision without any performance impact as the
OpenGL code already expects floats.
This allows the software upscaler to kick in, which looks
much better than IPU upscaling on the RG99
Signed-off-by: Christophe Branchereau <cbranchereau@gmail.com>
sdl_video.h names all of SDL_RLEACCEL, SDL_SRCCOLORKEY and SDL_SRCALPHA
as private and read-only. Documentation mentions them in
SDL_CreateRGBSurface but also says they are automatically set on
either SDL_SetColorKey or on non-zero Amask.
Previously passing them created no problem but miyoo mini SDL gets confused,
so let's never pass the private flags.
The device has a 320x480 screen and aspect of gui is wrong if we use
the full resolution
Signed-off-by: Christophe Branchereau <cbranchereau@gmail.com>
Currently we have bunch of else-if and only one of them triggered. When
e.g. aspect ratio correction and resolution changes only aspect ratio is
covered by restoration. If this fails the code retries the same mode endlessly
