This fixes a race condition that happens whenever the ROM cache is flushed but
the RAM one is not, causing any SWI calls (implemented as direct branches) to
jump to random instructions.
The fix could be to flush both caches at the same time (~expensive on
low mem platforms), use indirect jumps (a bit expensive) or emit the SWI
handler below the watermark to ensure it is never flushed. This is cheap
and effective, requires minimal changes.
This mis-emits CMN instead of TEQ and TST in the reg-shift operand mode.
This is never used by gpsp directly but translating real tst opcodes,
hence it only affects games using such instruction.
This fixes video players that previously crashed, many games that had
graphical glitches in ARM mode (but not on other CPUs) usually in menus
or other dialogues. Also fixes games that either crashed or went blank
or similar issues. The extent of fixing is hard to determine but could
affect many games in different levels.
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.
This fixes unallowed BIOS accesses (outside from BIOS) which fixes some
games like Silent Scope but also Zelda (fixes rolling as reported by
neonloop!)
This removes ram_block_ptrs and encodes the pointer directly in the
block tag. Saves ~256KB at no performance cost.
Drawback is that it limits the ram cache size to 512KB (we were using
768KB before). Should not be a problem since most games use less than
32KB of cache anyway.
Fixed ARM routines accordingly.
This might make a handful games slightly slower (but on the upper side
they work now instead of crashing or restarting).
Also while at it, fix some minor stuff in arm stubs for speed.
Cleans up a ton of whitespace in cpu.c (like 100KB!) and improves
readability of some massive decode statements.
Added an optimization for PC-relative loads (pool load) in ROM (since
it's read only and cannot possibily change) that directly emits an
immediate load. This is way faster, specially in MIPS/x86, ARM can be
even faster if we rewrite the immediate load macros to also use a pool.
Using an invalid SP makes Vita crash (for an unkown reason) and makes
things like C signal handlers crash (luckily Retroarch doesn't use
them). It is also a violation of the ABI and not a great idea.
Recycled some little used registers to free SP. Perf should be roughly
the same.
Seems that using the __atribute__ magic for sections is not the best way
of doing this, since it injects some default atributtes that collide
with the user defined ones. Using assembly is far easier in this case.
Reworked definitions a bit to make it easier to import from assembly.
Also wrapped stuff around macros for easy and less verbose
implementation of the symbol prefix issue.
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.
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 removes libco and all the usages of it (+pthreads).
Rewired all dynarecs and interpreter to return after every frame so that
libretro can process events. This required to make dynarec re-entrant.
Dynarecs were updated to check for new frame on every update (IRQ, cycle
exhaustion, I/O write, etc). The performance impact of doing so should
be minimal (and definitely outweight the libco gains). While at it,
fixed small issues to get a bit more perf: arm dynarec was not idling
correctly, mips was using stack when not needed, etc.
Tested on PSP (mips), OGA (armv7), Linux (x86 and interpreter). Not
tested on Android though.
I think this does not make a difference at all in the code, since PC is
treated in a special way anyway (reloaded with an immediate when read
and treated as an indirect branch when written). However for the sake of
completeness I'm undoing what I did. (The comma fix stays :P)
Turns out there were a couple of very interesting and hard to track
bugs. A missing comma made the reg list too short, leaving the 31th
element at the mercy of the linker ordering algorithm, which seems to
work in some cases depending on the compiler version.
Also the cache flush code seemed not to work on my machine (OGA),
not sure why it wored in the past :/
David Guillen Fandos
twinaphex
Francisco José García García