1. Added a new game option for linear interpolation when scaling
overlay-mode video in ScummVM builds with USE_RGB_COLOR;
2. Implemented SCI2.1-variant of the VMD player renderer (fixes
Trac#9857), which bypasses the engine's normal rendering
pipeline;
3. Improved accuracy of the SCI3-variant of the VMD player by
writing HunkPalettes into the VMD's CelObjMem instead of
submitting palettes directly to GfxPalette32.
More than one call to OSystem::updateScreen per frame on systems
with vsync ruins performance because the call is blocked until
the next vsync interval.
This also fixes bad rendering performance with the OpenGL backend.
The approach to video benchmarking used by SCI engine does not
translate very well to modern video devices -- it will either be
so slow that the games think the system is not capable of showing
normal visual effects, or so fast that the benchmarks overflow
their counters. So, game scripts that perform video benchmarking
are now patched to unconditionally return the highest speed value.
A pleasant but subtle side-effect of this change is that the extra
time sitting at a blank screen before the start of a game (while
benchmarks ran) is now gone.
Fixes Trac#9741.
This flag was used in SSCI to read from VRAM instead of from the
back buffer when a mouse interrupt was received in the middle of
a back buffer update. Since ScummVM controls when mouse events
are received via polling, it is not possible to receive a mouse
event in the middle of back buffer updates, so this code is
unnecessary for the engine to work properly.
This also fixes Valgrind warnings about use of uninitialized
memory at the start of the game, caused by not filling the cursor
memory buffers because `_frameNowVisible` was false until the first
frame was rendered.
PQ4CD and several other games contain a hack in two renderer
methods to avoid rendering invalid screen items with zero or
negative-dimension target rects. This prevents PQ4CD from
crashing during the fifth phase of target practice.
This is most noticeable at the beginning of the game during
benchmarking, where the benchmarking loop used to cause the mouse
to get stuck for the duration of the benchmark.
SSCI transitions code sends a large number of small show rects
to the graphics manager, one at a time, for each division of a
transition. Each time a rect is submitted, a call to showBits
is made. This design was used when transitions for SCI32 were
first implemented in ScummVM, and it worked OK because the
hardware surface was updated by EventManager::getSciEvent,
not showBits, so the large number of calls to showBits from the
transitions code did not adversely affect engine performance.
Later in SCI32 engine development, hardware surface updates
were changed to occur in showBits so that the hardware surface
would be updated at frame-out time, instead of at input-in time.
This change meant that now the large number of calls to showBits
from transitions became very expensive, and the engine would
stall constantly refreshing the entire hardware surface.
To fix this problem, the transitions code now (1) maximises the
size of rects coming from transitions, when possible, and (2) only
calls showBits when all the rects from one frame of a transition
have been calculated and added to the show rects list.
Additionally, there were some arithmetic errors in the
implementation of pixel dissolve that have been corrected in this
changeset.
Fixes Trac#9614.
This may come back in the future to deduplicate some gfx code,
but SCI32 had two different inlined ways of doing coordinate
conversions with different rounding methods, so CoordAdjuster32
didn't get used when the graphics system was rewritten.
At the moment, SCI32 code uses the mulru/mulinc methods from
helper.h for scaling up/down coordinates.
In the original engine this would have simply resulted in no draw,
but ScummVM is more strict about it. It is not 100% clear whether
these are normal and should be allowed or not, so for the moment
we'll emit a warning whenever a zero-dimension show rect is seen.
For now they only seem to be generated by plane transitions, and
these zero-dimension screen items cannot simply be omitted because
the 2.1early transitions code requires a fixed number of screen
items per step.
This commit implements all of the known plane transitions from
SCI2 through SCI2.1mid games. Because kSetShowStyle is always
called indirectly via the Styler game script, it is difficult to
find all the places where transitions are used. As such,
transitions that appeared to never be used have been added as
stubs which will trigger a game crash with a message to report
what was being done, so any missed transition types can be
identified quickly and then implemented.
1. Use the same throttling speed as normal frameouts. The
old throttling speed seemed a bit too slow.
2. Exit the event loop immediately if the engine is supposed to
quit, instead of forcing the user to wait until the transition
has finished before quitting.
Avoid forcing the screen to refresh after a palette change if the
screen is also about to be drawn to, as the palette change + draw
is intended to be an atomic operation.
Controls that manage their own event loops and call frameOut
directly generally need to sleep in order to avoid 100% CPU,
just like the main VM event loop.
This pass exposed two bugs, which have been fixed:
1. Checks of `_updated` and `_moved` were reversed in some areas,
which lead to rendering bugs. In SQ6 the rendering bugs were
subtle or non-existant, but in e.g. PQ:SWAT the Sierra logo and
title screen animations were totally missing.
2. The renderer formerly kept reading from ScreenItemLists when
new items were added in decrementScreenItemArrayCounts, but
this was determined to be unnecessary.
Remap would crash SCI2.1early games with 19 remap slots, and
did not actually work in most cases in SCI2.1mid+ games.
The SCI16 implementation was moved to its own separate file but
was otherwise touched as little as possible, so may still have
similar problems to the SCI32 code.
1. Split SCI16 and SCI32 code into separate files
2. Use -32 prefixes for SCI32 code and no prefix for SCI16 code,
where possible, to match other existing code
3. Avoid accidental corruption of values from the VM that may be
valid when signed or larger than 8 bits
4. Added documentation
5. Add missing remap CelObj calls
6. Inline where possible in performance-critical code paths
7. Fix bad `matchColor` function, and move it from GfxPalette to
GfxRemap32 since it is only used by GfxRemap32
8. Fix bad capitalisation in getCycleMap
9. Remove unnecessary initialisation of SingleRemaps
10. Update architecture to more closely mirror how SSCI worked
11. Clarify the purpose of each type of remap type (and
associated variable names)
12. Split large `apply` function into smaller units
13. Fix buffer overrun when loading a SCI2.1early game with remap
14. Remove use of `#define` constants
15. Warn instead of crashing with an error on invalid input (to
match SSCI more closely)
16. Change the collision avoidance mechanism between the RemapType
enum and remap kernel functions
17. Add save/load function
