When BindToRegister is called, the register cache marks the relevant
guest register as no longer containing an immediate. However, subfcx was
calling GetImm after BindToRegister. This led to a lot of panic alerts
after 2995aa5be4 added an assert to GetImm to check that the passed-in
register is an immediate.
Both before and after 2995aa5be4, the actual value of the immediate
wasn't overwritten by BindForRegister, only the fact that the register
is an immediate. Because of this, the emitted code happened to work
correctly.
We were previously excluding this folder from Android builds because it
didn't contain any files that were used on Android. However, we now have
an OSD font file that we do want to use on Android, and there's also a
few PNG files that will be needed by the RetroAchievements integration.
In terms of file size, this is what gets added:
OSD font: 48.1 KiB
RetroAchievements graphics: 3.5 KiB
Unused graphics: 116.8 KiB
We're still excluding Sys/Themes/, which is 1.1 MiB and entirely unused.
Like Jit64, JitArm64 now keeps track of the location of a guest register
using three booleans: Whether it is in ppcState, whether it is in a host
register, and whether it is a known immediate. The RegType enum remains
only for the purpose of keeping track of what format FPRs are stored in
in host registers.
Like the previous commit did for Jit64, JitArm64 can now handle the
combination of a value simultaneously being in a host register and being
a known immediate.
Unlike with Jit64, I've put the codegen-affecting changes in this commit
and the move away from the RegType enum in a follow-up commit. This is
in part because the design of JitArm64 made it easy to implement the
codegen-affecting changes without combining it with a big bang
refactorization, and in part because we need to keep RegType around for
keeping track of different float formats in Arm64FPRCache, complicating
the refactorization a bit.
They're now stored in ConstantPropagation instead.
I've also removed the LocationType enum. The location of each guest
register is now tracked using three booleans: Whether it is in ppcState,
whether it is in a host register, and whether it is a known immediate.
The first two of these booleans are stored in the register cache, and
the last one is stored in ConstantPropagation. This new model allows us
to handle the combination of a value simultaneously being in a host
register and being a known immediate. It also keeps track of which
registers are dirty, which was previously kept track of in X64CachedReg.
The old model maps to the new model as follows:
default host_reg immediate
Default true false false
Discarded false false false
Bound (!dirty) true false
Immediate false false true
SpeculativeImmediate true false true
[previously unrepresentable] (!dirty) true true
This commit makes the JIT set/clear the individual registers of
ConstantPropagation immediately instead of at the end of the
instruction. This is needed to prevent Jit64::ComputeRC, which reads
from a register written to earlier during the same instruction, from
reading back stale register values from ConstantPropagation in the next
commit.
To find out whether a host register needs to be unlocked, FlushRegisters
checks if the guest register is known to be a zero immediate. This works
right now, but it will stop working correctly once we gain the ability
to have a guest register be a known immediate and be in a host register
at the same time, because a register that's known to be a zero immediate
may have had a host register allocated prior to the call to
FlushRegisters. Instead, we should check whether the register is
RegType::Register after we're done calling BindForRead.
Restructuring things in this way brings two immediate benefits:
* Code is deduplicated between Jit64 and JitArm64.
* Materializing an immediate value in a register no longer results in us
forgetting what the immediate value was.
As a more long-term benefit, this lets us also run constant propagation
as part of PPCAnalyst, which could let us do cool stuff in the future
like statically determining whether a conditional branch will be taken.
But I have nothing concrete planned for that right now.
Sliders can trigger change listeners very rapidly, so let's add some
rate limiting so dragging a slider doesn't cause the whole UI to lag.
(Now the input overlay looks laggy when dragging a slider, though.)
In the dialog where you can choose what controller the input overlay
should be controlling, there's an OK button. If you change controller
but don't press OK, your selection will be saved, but the input overlay
won't refresh to show the new controller unless you perform some other
action that would cause it to refresh. This is not good UX.
This commit changes the behavior not only of this dialog but also other
dialogs spawned by EmulationActivity so that everything is properly
updated when dismissing a dialog, as if you had pressed OK.
The core no longer cares which thread is the host thread.
Cleaning up Android's HostThreadLock is left for another PR, in part
because the HostThreadLock in NativeConfig.cpp still serves a purpose,
and in part to make any issues easier to bisect.
By letting threads other than the host thread use things like
CPUThreadGuard, we can do a significant cleanup in AchievementsManager
in a later commit of this pull request.
Note: Some functions still can't be called from the CPU thread (or
threads the CPU thread might block on, like the GPU thread), but can
be called from any other thread.
When the interpreter calls MSRUpdated, we should update the membase
variable. Not because the interpreter itself needs it, but because the
JIT needs it if it's falling back to the interpreter for an instruction
that sets the MSR.
Additionally, the JIT's FallBackToInterpreter needs to read back the new
membase value afterwards.
This fixes games crashing on JitArm64 if mtmsr is set to fall back to
interpreter. I was unable to reproduce the issue on Jit64, presumably
due to a fortunate series of coincidences (instructions that set MSR are
always followed by an exception exit, and
PowerPCManager::CheckExternalExceptions was always calling
JitInterface::UpdateMembase, and Jit64::WriteExceptionExit was always
calling Jit64::EmitUpdateMembase.)
In a few places in Dolphin, we're using BottomSheetDialogFragments.
These unhelpfully tend to start out in a "collapsed" state when in
landscape mode (presumably depending on factors like screen size). The
user then has to manually expand them before they can meaningfully
interact with them.
We've been automatically setting BottomSheetDialogFragments to the
expanded state if the device Dolphin is running on doesn't support
touch, since with d-pad navigation it's impossible to expand these
sheets. But I think we should set them to expanded on devices that
support touch too. I haven't encountered a single case where you can do
anything useful with any of Dolphin's BottomSheetDialogFragments while
they're collapsed, so the user always has to expand sheets manually if
they start out collapsed. And just because a device supports touch
doesn't necessarily mean you're interacting with it through the touch
screen right now - you could be using a gamepad, for instance.
The callback mechanism AchievementManager had until now only supported
one caller registering a callback, and it didn't have any
synchronization. This isn't a problem for DolphinQt, but the PR to add
Android support for RetroAchievements exposes these problems. Let's
replace it with HookableEvent, which can handle all of this.
The instruction implementations that were shifting the size by 4 would
emit an incorrect instruction when given a size of 64. The correct
implementation is to count the number of leading or trailing zeroes in
the size parameter, which is what IntLog2 does.
No callers are affected by this, as they all use sizes other than 64.
Actually, some of these instructions are even invalid with a size of 64,
but I'm changing them anyway for consistency with the others.
We should attempt to use not only mirrored versions of the immediate as
an ORR base, but also the immediate itself. This lets us emit certain
64-bit constants using fewer instructions.
When 9d9b6d8 changed our target SDK version to Android 16, it made
Android 16 stop calling onBackPressed and stop delivering KEYCODE_BACK
events. Dolphin's code isn't ready for that yet.
Android lets us opt out of this new behavior for now, so let's do so.
But the opt-out will presumably stop working once we start targeting
Android 17, so we're going to have to update Dolphin's code within the
next one or two years to support the replacement API.
For thread safety, we shouldn't return any pointers or references that
can be used to mutate the state of the PPCSymbolDB. This should be the
final part of making PPCSymbolDB thread safe unless I've missed
something.
9395238 added locking in some PPCSymbolDB functions that access member
variables, but far from all. To ensure thread safety, this commit adds
the missing locking.
JosJuice