Turns out there's a couple of inaccuracies that do affect a couple of
games. Most of them are buggy games but emulating these accesses
correctly helps jumping over some bugs.
This rewrites the way that CPU alerts work, making them a bitmap (since
multiple alerts can happen simultaneously, like SMC and IRQ). This
doesn't really fix many games but improves accuracy overall and improves
performance on some I/O writes (the ones without side effects).
The IRQ raising is now decoupled and explicitely called via a new
function (check_and_raise_interrupts) to avoid issues such as invalid
CPSR values (doesn't seem to bother most games!). There's more side
effects missing, so this just lays the ground for more fixes.
Whenever an interrupt is pending and interrupts are disabled (via
IME/IE), an IE/IME write that re-enables IRQs will fail to raise an IRQ.
This makes some games hang. Most games seem to use CPSR.IRQ to enable
and disable interruts, so they are not affected. However some others use
IME/IE (or all of them), causing these deadlocks and some race
conditions.
This fixes a bunch of games that did not crash but would "hang" in some
interesting ways.
This fixes ROM swapping for x86/64, arm32 and arm64. On top of that it
improves speed by removing unnecessary slow paths on small ROMs for
arm32 and mips. If the ROM can fit in RAM, it will emit more efficient
code that assumes the ROM is fully loaded.
For low-memory Linux platforms it would be better to use some mmap'ed
ROM, that way the OS would transparently handle page swapping, which is
perhaps faster. Will investigate and follow up on this in a separate
commit.
Converted the hack into another... hopefully better hack.
Check for IW/EWRAM addr bit and handle accesses to EWRAM by correcting
the base addr and the mirroring range.
Assumes that iwram and ewram buffers are at a distance multiple of 64KB
for speed (so that lw/sw offsets can be shared).
This fixes a variety of games that should have been present in
game_config.txt such as Ninja Cop, Star Wars JPB, Medal of Honor,
Spongebob, etc. Some of them were just missing regional versions of the
cart. It also fixes newer games and homebrew such as GoodBoy Galaxy.
This is just responsible for a few cycles every frame (could be around 1
and 2% depending on the game usage) but makes emulation a bit more
accurate and potentially faster.
This uses BSON as savestate format, to allow external tools to parse it
(so that we can add proper test of the states). The BSON is not 100%
correct according to spec (no ordered keys) but can be parsed by most
libraries.
This fixes also a bug in the savestate palette color recalculation that
was wrongly overwritting the original palette (which could cause some
problems on some games).
Also fixes some potential issues by serializing some more stuff and
cleans up unused stuff.
Testing shows that states look good and there's only minor differences
in audio ticks, related to buffer sizes (since buffer flushes are
de-synced from video frames due to different frequency).
This is not needed at all, since the variables are not updated between
reload and end-of-frame (where we take our savestates). Added a reload
call during gba_load_state() to initialize it from the I/O regs.
Removed the last bits of text relocations by moving all relevant RAMs to
the stub reachable area. This should be as fast or even faster than
previous code.
Seems that this could make VRAM overflow big time and overwrite IWRAM
most of the time (since it lives after that buffer). This causes a
variety of errors, some of them hidden if IWRAM not used.
Most games do not use VRAM mirror so it's not a big deal, but some do
have a off-by-one errors that trigger this.
This improves the existing on-demand ROM paging and also breaks down ROM
buffers into 1MB blocks for platforms with memory fragmentation issues.
Fixed some potential RTC reg issue in said platforms too.
Fixed page pinning on interpreter (would crash due to LRU evictions).
This patch adds big-endian compatibility in gpsp (in general but only
for the interpreter). There's no performance hit for little-endian
platforms (should be a no-op) and only add a small overhead in memory
accesses for big-endian platforms.
Most memory accesses are wrapped with a byteswap instruction and I/O reg
accesses are also rewired for proper access (using macros). Video
rendering has been fixed to also do byteswaps but there's a couple of
games and rendering modes that still seem broken (but they amount to
less than 20 games in my tests with 1K ROMs).
This also adds build rules and CI for NGC/WII/WIIU (untested)
This reduces code size more than 20% (which is 200-300KB!).
DMA handling accounts for less than 0.5% the average emulation runtime
which doesn't justify the crazy optimization level that the code has. In
fact it's more than likely that the new code runs faster due to less
I-cache trashing.
This saves a few cycles in MIPS and simplifies a bit the core.
Removed the write map, only affects interpreter performance very
minimally. Rewired ARM and x86 handlers to support direct access to
I/EWRAM (and VRAM on ARM) to compensate. Overall performance is slightly
better but code is cleaner and allows for further improvements in the
dynarecs.
Added a more thorough cache cleanup for reset/mode-change too.
Fixed the mmap initialization that ends up leaking memory.
Minor x86 asm fixes for Android.
This is not really necessary since it can share area with ROM.
Performance impact should be very minimal (haven't noticed it myself)
and could be compensated (even by a positive offset) if we bump the ROM
cache area size.
Tested with several dynarecs.
This allows us to emit the handlers directly in a more efficient manner.
At the same time it allows for an easy fix to emit PIC code, which is
necessary for libretro. This also enables more platform specific
optimizations and variations, perhaps even run-time multiplatform
support.
David Guillen Fandos