This is a correction for 0cab1731d6 (part of #14665), where it was
overlooked that the condition to enable the RAM override was flawed: if
the user had enabled the emulated memory size override in Dolphin, but
the running game did not define a simulated memory size greater than the
retail memory size, the RAM override in IOS was wrongly misconfigured,
causing issues with Wii games.
Test plan:
- Launch Dolphin.
- In the **Settings > Advanced** section, tick the **Enable Emulated
Memory Size Override** check box, and set the maximum memory size for
both **MEM1** and **MEM2** (i.e. 64 MiB and 128 MiB respectively).
- Play any Wii game (e.g. _Mario Kart Wii_ (RMCE01)).
**Without** the patch, the emulator will hit an invalid read from
`0x83FEE78C` at `0x8019FE6C`:
<img alt="Invalid read when starting Wii game after having enabled the emulated memory override in Dolphin" title="Invalid read when starting Wii game after having enabled the emulated memory override in Dolphin" src="https://github.com/user-attachments/assets/1777485e-a421-4303-8d3d-dd49f1aaa9bf" />
**With** the patch, the emulator can execute the game as usual.
For some reason, #14738 removed the headers and allowed return codes which broke using the Netplay index for users.
Also, bools converted to a string using std::to_string would become 0/1, while fmt::format by default turns them into false/true, which also broke the index.
This reverts commit e4349ae179.
The 0x prefix is included in the characters count (unlike e.g. in Rust
where there is no 0x prefix), so that commit was wrong. Sorry about
that!
Fixes regression introduced in #14842.
Dolphin would previously crash while accessing &buffer[got], as if we
were trying to access that element instead of just obtaining a pointer
to it. The proper solution is to obtain a pointer to the first element,
and only add the offset we want to get the range.
Here is the printed message:
/usr/include/c++/16/array:210: constexpr std::array<_Tp, _Nm>::value_type& std::array<_Tp, _Nm>::operator[](size_type) [with _Tp = char; long unsigned int _Nm = 128; reference = char&; size_type = long unsigned int]: Assertion '__n < this->size()' failed.
The stack trace originates from:
Source/Core/Core/HW/EXI/EXI_DeviceGecko.cpp:139
Large anonymous mappings are split into 128 MB chunks. Without MAP_MEM_VM_SHARE,
mach_make_memory_entry_64 will only return an entry spanning the first chunk.
Attempting to map through that entry will fail if it extends beyond the 128 MB
boundary, which can happen when the sizes of MEM1/MEM2 are overridden.
TypedArray doesn't implement AutoCloseable on older devices so .use{} caused a crash.
It turned out the special handling for background colours was unnecessary anyway.
SDL 3 has more misc buttons for "additional controller buttons" (e.g. left/right trigger clicks on a GameCube controller, or for the second touchpad click on the Steam Controller) which use these additional named Misc buttons. Taken from [SDL_GamepadButton](https://wiki.libsdl.org/SDL3/SDL_GamepadButton)
- Switch from ActivityResultContracts.GetContent() to OpenDocument()
which restores ACTION_OPEN_DOCUMENT behavior, granting write access
to file descriptors opened in C++ via ContentHandler.openFd().
- Replace unsafe force-unwraps (!!) in EmulationActivity with null checks.
- Improve the column measurement logic so it updates the column width when text changes.
- Use matchParentSize on the clickable target in OutlinedBox so that it is clickable when the OutlinedBox contains contents with dynamic height.
200e26c added controller handling to MovieConfigLayerLoader's
LoadFromDTM but not SaveToDTM. This caused a regression. When starting a
new input recording, MovieManager::BeginRecordingInput calls SaveToDTM,
which doesn't set the controller values in the DTM header to anything.
Then it loads the newly created movie config layer, causing LoadFromDTM
to be called, which in turn causes zeroed out controller values to be
loaded from the DTM header. This leaves the player with all controllers
set to None, making it impossible to control the game.
To fix this, let's move the saving of controllers from
MovieManager::SaveRecording to ConfigLoaders::SaveToDTM.
This game contains a special episode of the Pokémon anime as an FMV.
When playing it, if only a tiny part of the image has changed between
frames, Dolphin's texture hashing can fail to notice that anything
changed, or it can notice that the chroma changed but not the luma, or
vice versa.
512 samples reduces the problem but doesn't get rid of it entirely, but
with 2048 I'm unable to see the problem.
This makes Android players show up as "Android" instead of "Lin".
(This is only visible in DolphinQt, not in the Android GUI.)
I'm also unabbreviating all the existing operating system names:
* Win -> Windows
* Mac -> macOS
* Lin -> Linux
Independently spotted by both @Dentomologist and me while reviewing
PR #14805
The previous limit was correct for valid VPBs, but an invalid VPB
controlled by a malicious game could contain a non-fractional value
in current_pos_frac, which would allow writing an extra 15 samples
(30 bytes) into the stack. With AFC encoding, this is rounded up to 16
samples, but with much less control over which bytes are written.
Maybe we should be doing some validation, or bounds checking, but I'm
pretty sure this issue was copied from the original ucode, and we kinda
want to stay compatible.
The simpler fix is to just increase the size of raw_input_samples.
I've checked other code paths, and 0x514 samples seems to be the limit.
glMemoryBarrier comes from ARB_shader_image_load_store, so on a desktop context
below GL 4.2 the pointer is null. bSupportsBBox is set from
bSupportsFragmentStoresAndAtomics, which such a driver can still have, so saving
a state reads the bounding box and calls a null pointer.
UsePersistentStagingBuffers() in OGLTexture.cpp already guards the same call with
bSupportsImageLoadStore.
This started as a fixing a misalignment issue, but I got carried away.
Was originally 41 bytes (overflowing it's expected 40 bytes by 1 bit).
Now it's 27 bytes plus 8 bits of padding (ready for future expansion).
The savings come from:
- Removing UV usage from components, it can be reconstructed from
texcoord_elem_count (saved 8 bits)
- Removed the 8 unused bits from texMtxInfo_n_projection (saved 8 bits)
- Removing unused bit from from start of components (saved 1 bit)
- Removed extra bit from inputform, texgentype and sourcerow
- packed texMtxInfo_n_projection and texcoord_elem_count back into
the per texgen info space freed up above. (saved 24 bits)
- Overlayed postMtx index and Emboss mode shifts into a union based on
texgentype (saved 56 bits)
- Move to a single-bit union tag, freeing up an extra bit for regular
texgens.
- Move PotMtx normalize into the freed up space (saved 8 bits)
Total savings: 105 bits of data
EDIT: Even worse, MSVC wasn't respecting pack(1) for bitfields at all,
so on windows this struct was actually quite a bit larger.
Something like 80 bytes, if not more.
Fixed this by dropping the size of any enums used in these
bit structs to u8. Our Common::BitField allows for the underlying
type to be explicitly specified to something larger.
msvc's bitfield packing appears to be completely braindead, can't
mix sizes at all.
msg_in and msg_out were ending up with only 4-byte alignment, which
is undefined behaviour, since they contain 64-bit fields. Probably
doesn't cause any issues, but it showed up in ubsan.
A malicious GCZ file could probably force this to be negative.
Shouldn't cause any issues other than file read failures, but need to fix
because it is causing errors on MSVC.
SectorReader::ReadChunk does do some validation on it, but it only
checks against the original disc size (reported by the GCZ file).
It has no idea how many blocks the header claimed the disc had.
A maliciously crafted GCZ file could trigger read overflows off the end
of the m_block_pointers/m_hashes arrays.
A malicious dol could theoretically use integer wraparound to bypass
bounds checking and cause DolReader to read past the end of m_bytes.
Could result in crashes, wasting large amounts of memory, or even the
disclosure of heap memory contents.
We already had code to close open host files when reading or writing a
savestate, but due to d35fe1b we also need to close open guest files
when reading a savestate, otherwise DoStateRead fails to delete them.
I was considering an alternative solution where instead of copying and
clearing m_handles, we just set `handle.opened = false;` for each handle
before reading a savestate (but not before writing a savestate).
However, this wouldn't solve the problem of DoStateWriteOrMeasure's
calls to OpenFile failing due to all handles being open. I'm not aware
of any games that have that many handles open, though.
We are updating from 3.4.8
Changes relevant to us include:
3.4.10:
- Added controller sensor support for GameInput v3
- Fixed Ipega controllers being ignored in keyboard mode
- Added support for GameCube rumble when the adapter is in PC mode and has the latest firmware
- Fixed rumble on the new Steam Controller
- Added support for the GameSir Super Nova in Xbox 360 mode
- Added support for the PDP Afterglow Wave Wireless Controller for Switch
3.4.12:
- Added hotplug detection support when using libusb for HIDAPI controllers
- Fixed flipped Xbox 360 controller axes on macOS
It's not in the changelogs, but 3.4.10 is the revision which allowed for my Steam Controller's gyro to properly work in Dolphin
Since our bump to Qt, dxcompiler.dll and dxil.dll have been added alongside other Qt dlls as part of QtGui. Those DLLs are not needed since we use QtWidgets and not QML.
This introduces a CMakePresets file for both Unix-likes and Windows that replace the old CMakeSettings.
It adds presets for **Debug** and **Release** profiles for both **x64** and **ARM64** architectures, as well as **Generic builds**.
Presets can be used using[ Visual Studio's built-in CMakePresets support](https://learn.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-170), or [Visual Studio Code's CMake Tools](https://marketplace.visualstudio.com/items?itemName=ms-vscode.cmake-tools) extension.
They can also be used from the command line, like so:
- x64/Unix-like/Ninja:
- Configure: `cmake --preset ninja-release-x64`
- Build: `cmake --build --preset ninja-build-release-x64`
- Configure + Build: `cmake --workflow --preset ninja-release-x64`
- ARM64/Windows/Visual Studio:
- Configure: `cmake --preset visualstudio-release-arm64`
- Build: `cmake --build --preset visualstudio-build-release-arm64`
- Configure + Build: `cmake --workflow --preset visualstudio-release-arm64`
The Ninja generator is available to both Windows and Unix-likes, while the Visual Studio Generator is only available on Windows.
**Cross-compiling**
On **Windows**, the Visual Studio generator automatically takes care of everything, you just need to select an ARM64 preset.
On **Unix-likes**, to cross-compile you need to install a cross-compiler and (optionally) a sysroot of the target system.
Here is an example to compile from x64 to ARM64 with a sysroot:
- `cmake --preset ninja-release-arm64 -DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++ -DCMAKE_SYSROOT=/opt/sysroots/aarch64-linux-gnu`
- `cmake --build --preset ninja-build-release-arm64`
You will need a sysroot to link against Qt, since we do not vendor it in on platforms other than Windows.
**User presets**
A `CMakeUserPresets.json` file may be created locally at the root of the project to further customize your presets.
For example, here are the user presets I used to test this PR on Arch Linux with a generic Arch Linux ARM sysroot:
```json
{
"version": 10,
"configurePresets": [
{
"name": "gcc-debug-arm64",
"inherits": "ninja-debug-arm64",
"cacheVariables": {
"CMAKE_C_COMPILER": "aarch64-linux-gnu-gcc",
"CMAKE_CXX_COMPILER": "aarch64-linux-gnu-g++",
"CMAKE_EXE_LINKER_FLAGS": "-L/opt/sysroots/ArchLinuxARM/lib",
"CMAKE_SYSROOT": "/opt/sysroots/ArchLinuxARM"
}
},
{
"name": "clang-debug-arm64",
"inherits": "ninja-debug-arm64",
"cacheVariables": {
"CMAKE_C_COMPILER": "clang",
"CMAKE_CXX_COMPILER": "clang++",
"CMAKE_C_FLAGS": "-target aarch64-linux-gnu",
"CMAKE_CXX_FLAGS": "-target aarch64-linux-gnu",
"CMAKE_SYSROOT": "/opt/sysroots/ArchLinuxARM"
}
},
{
"name": "clang-debug-x64",
"inherits": "ninja-debug-x64",
"cacheVariables": {
"CMAKE_C_COMPILER": "clang",
"CMAKE_CXX_COMPILER": "clang++"
}
}
],
"buildPresets": [
{
"name": "gcc-build-debug-arm64",
"configurePreset": "gcc-debug-arm64"
},
{
"name": "clang-build-debug-arm64",
"configurePreset": "clang-debug-arm64"
},
{
"name": "clang-build-debug-x64",
"configurePreset": "clang-debug-x64"
}
],
"workflowPresets": [
{
"name": "gcc-debug-arm64",
"steps": [
{ "type": "configure", "name": "gcc-debug-arm64" },
{ "type": "build", "name": "gcc-build-debug-arm64" }
]
},
{
"name": "clang-debug-arm64",
"steps": [
{ "type": "configure", "name": "clang-debug-arm64" },
{ "type": "build", "name": "clang-build-debug-arm64" }
]
},
{
"name": "clang-debug-x64",
"steps": [
{ "type": "configure", "name": "clang-debug-x64" },
{ "type": "build", "name": "clang-build-debug-x64" }
]
}
]
}
```
They are then used like so:
Configure + Build with GCC: `cmake --workflow --preset gcc-debug-arm64`
Configure + Build with Clang: `cmake --workflow --preset clang-debug-arm64`
Configure + Build with Clang (x64): `cmake --workflow --preset clang-debug-x64`
*Addendum: It should also now be possible to cross-compile from Windows to Unix-likes, and Unix-like to other Unix-like (e.g. Linux -> FreeBSD), however this is untested.*
Log each rejected ELF header, range, and symbol reference. This
provides actionable diagnostics for malformed files without changing the
validation behavior.
ElfReader trusted table offsets and counts from standalone ELF files.
Malformed input could make it read and write past the loaded file buffer.
Validate the ELF header, table ranges, segment data, section data, and
string-table references before accessing them. Invalid files use the
existing executable boot failure path.
NetPlay save synchronization decoded remote LZO blocks with the unsafe
decoder and no output capacity. A malicious host could overflow a client
buffer with a block larger than its declared size.
Use the bounds-checking decoder, validate the declared output length, and
grow buffer results only after each checked block has been decoded.
Using dual core during netplay seems sufficiently detrimental that its worth showing this warning message and the option to turn dual core off when starting netplay.
Just copy the gamecube logic and automatically assign players to the first empty slot.
More consistent with GameCube and helps Android hosts where there is no UI for managing controller mappings currently.
Custom composable that switched between compact and medium size app bar depending on a combination of orientation and screen size. (Small app bar when horizontal on small devices, medium app bar otherwise. Medium app bar supports collapse on scroll.
Update Android dependencies and build tools to their latest available versions.
Rename the GameCube save magic constant to avoid a collision with the NDK 30 ARM64 signal context header.
We were registering the hotplug callback when starting the thread, but
deregistering it when shutting down GCAdapter as a whole. Because the
thread can start and stop potentially many times before GCAdapter shuts
down, this led to callbacks being registered when we already have
callbacks. On Android, this was making Dolphin's Kotlin code for
registering the callback throw an IllegalStateException.
To fix this, deregister the callback when stopping the thread.
Dolphin now uses the simulated memory size defined in `bi2.bin` to
adjust the memory size in the emulated console automatically.
If **Enable Emulated Memory Size Override** has been enabled by the
user, the fixed memory size specified for **MEM1** and **MEM2** are
still used as they were before.
Most retail games do not define the simulated memory size (Wii or
Triforce games), or define it to a value that matches the default 24 MiB
(GameCube games), so, in the general case, there is no behavior change.
One game that sets the simulated memory to a non-default value is the
debug build of _Mario Kart: Double Dash!!_, where the value is set to
48 MiB.
This enhancement is focused mainly to the modding community. Prior to
these changes, modded games with extended memory requirements would
fail to launch in Dolphin [with no indication of what the problem is] if
the user failed to set the emulated memory override to the correct
value. Now, modding tools can specify the simulated memory size in the
`bi2.bin` file to produce extended games that _just work_ in Dolphin,
without cumbersome instructions that can be overlooked by the user.
This is the 32-bit unsigned integer that is seen at offset `0x04` in the
`bi2.bin` file.
The intent is to study what games define values that departure from the
norm. Common values are 0 MiB (Wii games) and 24 MiB (GameCube games),
with some debug and demo builds featuring 48 MiB (e.g. the debug build
of _Mario Kart: Double Dash!!_).
The data structure of the `bi2.bin` file in which the simulated memory
size is set:
> **13.2 Disk header Information**
>
> this is loaded to the Address in `0x800000f4` when a disc is initialized by the IPL
>
> | offset | end | size | Description |
> | :------: | --- | :--: | --------------------- |
> | `0x0000` | | 4 | Debug-monitor Size |
> | `0x0004` | | 4 | Simulated Memory Size |
> | `0x0008` | | 4 | Argument offset |
> | `0x000c` | | 4 | Debug flag |
> | `0x0010` | | 4 | Track Location |
> | `0x0014` | | 4 | Track size |
> | `0x0018` | | 4 | Countrycode |
> | `0x001c` | | 4 | ? |
See https://hitmen.c02.at/files/yagcd/yagcd/chap13.html#sec13.2.
Most GameCube games set the value to `24 MiB`, which matches the
physical memory size. Most Wii games set the value to `0 MiB` (unset).
The debug build of _Mario Kart: Double Dash!!_ is one game that sets the
value to `48 MiB`.
Erase a removed pack from ResourcePack::<anonymous namespace>::packs
instead of only deleting the pack file from the disk.
This causes the repopulated table to actually display that the pack has
been removed (which previously only happened after restarting Dolphin or
hitting `Refresh`), and also removes the pack from the `Packs.ini` file.
Avoid accessing destroyed `QTableWidgetItem` by retrieving the desired
path from the item before it's destroyed.
In `ResourcePackManager::Remove` pointers to the selected items in `m_table_widget`
were saved in the local variable `items` before calling `Uninstall`.
`Uninstall` called `RepopulateTable` which called
`m_table_widget->clear()`, destroying the table's descendants. Upon
returning to `Remove` `items` then pointed to some of those destroyed descendants, and
passing `items[0]` to `GetResourcePackIndex` resulted in a call to
`item->row()` which was a use-after-free.
Fixes https://bugs.dolphin-emu.org/issues/14095.
The ifdef made it so this was never called on Android, but we do want it
to be called on Android.
There are ifdefs inside the function implementation but not around the
function declaration or definition, so there's no need for ifdefs when
calling it.
Previously we would vibrate for 100 ms on every vibration, which was
especially annoying if a game was doing a bunch of vibrations that were
supposed to be much shorter than that. Now we tell Android to vibrate
for 10 s, then cancel the vibration as soon as the game turns the
vibration off.
Some of the speed hacks were not approved for RetroAchievements, meaning that users had to choose between not getting hardcore achievements, or having poor performance.
Also, renamed the instances of `$Limit internal FPS to VPS (improves performance)` to `$Limit internal frame rate (speed hack)` for consistency. They are enabled by default, so unless the user disabled them, it won't break the configuration.
WiimoteScannerAndroid::FindAttachedWiimotes() added every device
connected through the DolphinBar to the Wii Remote results, with a
TODO noting that balance boards were not differentiated. As a result,
a Wii Balance Board connected through the DolphinBar was always
assigned to a Wii Remote slot (Player 1), and games that request a
Balance Board never detected it.
Use Wiimote::IsBalanceBoard() to route balance boards into the
balance_boards results, matching the Linux and Windows scanner
backends. Tested on an AYN Thor (Android 13) with a Mayflash
DolphinBar in Mode 4 running Wii Fit Plus: the board is now detected
in the Balance Board slot.
This lets rumble work even if the user has turned off the Android
setting Sound and vibration > Vibration and haptics > Interactive
haptics > Touch feedback. Rumble can still be disabled through Dolphin's
controller bindings and through in-game settings (GameCube) or SYSCONF
(Wii).
fb4ff3e put all the code for the show internal resolution setting right
next to the code for the show FPS setting, presumably because an earlier
version of that commit had them next to each other in the GUI. It makes
more conceptual sense to put the code next to the code for the show
statistics settings, matching what the GUI looks like now.
fb4ff3e added a statistics option to show the internal resolution, but
it just showed the total size of the EFB (which is always 640x528) times
the IR scale, so it didn't convey any useful information.
This commit instead makes the option use the size of the last XFB copy
(not multiplied by the IR scale), which changes based on the game's
rendering resolution.
Check if Dolphin already has a handle for stdout or stderr before trying
to attach to the console of the parent process.
Previously, running Dolphin via Git Bash with options like `--help` or
`--version` wouldn't produce any output. This was the result of the fix
(1ce75ce217) for
https://bugs.dolphin-emu.org/issues/11042, before which output did work
for `Git Bash` but didn't for `Command Prompt` or `PowerShell`.
We don't need to see Compose previews when making a release build, but
androidx.compose.ui.tooling.preview.Preview is referenced in the source
code nonetheless, so the build system needs to know what it is.
Example:
```
nand. r1, r2, r3
```
Assuming that none of these registers will be used again:
- Before #14278, all of them would be flushed on both sides of the branch.
- After #14278, r2 and r3 would have been flushed earlier, but r1 won't.
- With this, r1 is flushed in ComputeRC.
Now SerialInterfaceManager::UpdateDevices reads the configured SI
devices directly from Config instead.
The main reason why I wanted to do this is so that Android can change SI
devices while emulation is running. (Android didn't have the code for
calling ChangeDevice.) But when implementing the change, I noticed that
NetPlay and Movie were using ChangeDevice as a way of overriding the SI
devices configured by the user. Replacing this ended up making the change
larger than I first anticipated.
For Wii Remotes, there was no equivalent to ChangeDevice, so NetPlay and
Movie were using Config::SetCurrent to override the Wii Remote source
configured by the user. If we can use the config system to override Wii
Remote sources, why not do the same for SI devices? This commit makes
NetPlay and Movie set SI devices and Wii Remote sources in the NetPlay
and Movie config layers, as that is the conceptually appropriate place
to do it.
As far as I can tell, the old Movie code for overriding SI devices and
Wii Remote sources didn't actually work. This new code does. I didn't
investigate exactly why it didn't work, but maybe it's because loading
a movie happens before emulation actually starts.
This fixes https://bugs.dolphin-emu.org/issues/14076. The issue report
more or less already says it all, but to provide a shorter summary:
We were fetching a list of vibrator IDs, but instead of passing the
vibrator ID to the vibrator manager, we passed the index of the ID in
the list. This happened to work fine on many devices, including all
devices that use DolphinVibratorManagerCompat, due to the only Vibrator
having both an index and ID of 0. But on some devices, it failed due to
the ID of the Vibrator being 1.
This fix makes us correctly pass the ID to the vibrator manager. We
still use indices in controller INI files, both for compatibility with
the controller mappings shipped with Dolphin (which use index 0) and for
backwards compatibility with older controller INI files.
These changes originate from testing the behaviour of the PI registers
in Linux on my Wii.
I started by testing the masks, writing 0xFFFF_FFFF to the registers on
the Wii, and then got told about their value on a GameCube, and then
figured out most of the registers through a discussion with Extrems.
Found out when attempting to make a speed hack to help Dolphin detect an idle loop. The inserted `nop` was preventing it from considering it an idle loop.
A few actions were missing from shortcuts in the Tools menu. This allows using e.g. `ALT`+(`T`>`E`>`S`) to open the Skylanders Manager, or `ALT`+(`C`>`C`>`1`) to toggle the first Wii Remote's connection.
Extrems said that it's unnecessary because GBIHF doesn't use the GPU.
Video output doesn't work in GBIHF in Dolphin regardless of this
setting, so I can't actually test it.
NetplayScreen was not drawing under the bottom nav bar like other screens. In landscape mode half of the screen scrolls and half is fixed, to make this look natural with edge to edge, provide a fade on the bottom of the scrolling side so things look natural when the screens loads and is resting at 0 scroll offset.
To avoid using dual pane layout on very small devices, ensure that the current width available for display is at least medium (600dp).
It also has to be landscape, the dual pane mode looks weird on a tablet in portrait mode since there's not that much content.
We just about get away with using a StateFlow in NetplaySession since the host sends AbortGameDigest when closing their own dialog. Without that it would be harder for the UI to distinguish between subsequent dialogs. If that wasn't the case then NetplaySession might need to expose the individual progress and result updates and have the view model assemble it into the overall GameDigestProgress.
Create a new NetplaySession each time we try to join a netplay game. Hold onto it in NetplayManager so its available to the different activities that need to access it. Close the session when backing out of the netplay UI. Some guardrails in case things go out of sync: creating a session closes the old one if it is still around for some reason, finalizer in NetplaySession to release native resources if not closed explicitly for some reason. Profiling done to ensure all kotlin and native objects are successfully cleared / garbage collected.
An orientation change can trigger this code path immediately when a game starts. e.g. dolphin is portrait when launching the game but settings force gameplay to landscape. We want to avoid this and continue to the netplay launch code below.
If the user backgrounds dolphin during netplay and then resumes after the process has died it will actually resume from the save state in single player mode, not sure if thats good or bad but fine for now.
Netplay doesnt handle rotation very well, seems to go more and more out of sync the more rotations.
Boot session data is already handled when the game is booted so this is just fallback in case the game launch fails in some weird way.
Add missing @Keep annotations to functions called from C++
If the last write to a register comes before the last read of it,
we can write the register to ppcState after the last write instead of
after the last read. This will hopefully help spread out m_ppc_state
writes across a code block, improving pipelining. Also, if there's a
conditional branch that's after the last write but before the last read,
instead of needing to emit one m_ppc_state write on each side of the
branch, we now only need to emit one m_ppc_state write.
A note about the changes made to stmw and mfcr: These instructions don't
write to any GPRs or CRs respectively – they only read from them.
With this commit, there are no longer any cases where registers get
written back to m_ppc_state after an instruction that just reads from
them, so we can get rid of all STP logic from these two instructions.
lmw still needs its STP logic, since that one does write to registers.
This jumps us up two releases, both of which only introduce bug fixes. Changes potentially relevant to us are:
[3.4.6](https://github.com/libsdl-org/SDL/tree/release-3.4.6):
- Fixed Xbox controllers being unresponsive on preview release of Windows 11
- Fixed the menu button state on controllers on Apple platforms
[3.4.8](https://github.com/libsdl-org/SDL/tree/release-3.4.8) (hotfix release):
- Fixed a controller related crash on Windows, introduced in 3.4.6
The 0x8000 alignment in DirectoryBlob is needed for Wii disc group encryption,
but for GC/Triforce it inflates file offsets unnecessarily. Use 0x20 alignment
for Triforce (matching original disc layout) while keeping 0x8000 for GC due
to DTK audio streaming requirements.
On Triforce games with many files, the inflated offsets can land in the
AMMediaboard DIMM memory range (0x1F000000-0x1F800000). Reads from that
region return SRAM data instead of disc data, causing the game to hang.
Skip the DIMM range when assigning per-file data offsets if any portion of
the file would overlap [0x1F000000, 0x1F800000).
I am not certain what this was used for, but the last non-formatting related change made to it was over 16 years ago.
The project targets Windows Vista, uses a Visual Studio 2010 project.
Looking at the entry point in `DSP_InterC.cpp`, the program opens a hardcoded path to a DSP rom in `C:\_\dsp_rom.bin` without any documentation on it whatsoever, and decodes something between two mystery addresses before exiting.
This PR provides 60 FPS code additions to Sonic Unleashed.
NTSC-U:
* Added 60 FPS code, coming from the Dolphin wiki.
* Followed my code naming convention for Sega games.
This PR provides 60 FPS code additions to Sonic and the Secret Rings & Sonic and the Black Knight.
All regions:
* Formatting fixes.
NTSC-U/PAL:
* Added 60 FPS code, coming from the Dolphin wiki.
* Followed my code naming convention for Sega games.
This PR provides 60 FPS code additions to NiGHTS: Journey of Dreams.
NTSC-U/PAL/NTSC-J:
* Added 60 FPS code, coming from the Dolphin wiki.
* Followed my code naming convention for Sega games.
* The wiki says that the code is for NTSC-U but it's actually region free. It's been confirmed via testing.
* The wiki also says that a certain part in the final level softlocks, but that's actually not true and can be bypassed. An explanation has been included with the code.
In the past, the register cache would contain either a host register or
an immediate (or neither) for a given guest register. Jit64::cmpXX would
roll with whichever one it got. But now that PR 12134 is merged, the
register cache can contain a host register at the same time as an
immediate. This gives us an optimization opportunity in Jit64::cmpXX.
Some paths are more efficient when we use an immediate and some paths
are more efficient when we don't, so to generate the best code, we
should carefully choose if we want an immediate or not rather than
leaving it to the register cache.
Tested with a binary that has VMA != (LMA | 0x80000000), unlike most
libogc binaries. This is indeed a valid setup for ELFs, and one that is
generally expected to work properly.
Test code: https://github.com/Wii-Linux/NPLL/tree/reloc
In either case, you still need to set the PC in the debugger to get it
to boot properly, though that is a different issue that I will be fixing
in a future patch.
Without this patch, the code to be runtime-relocated ends up already
loaded at the desired final address, and then when the code tries to
relocate from what it thinks is the source address (where the code
should be, where the LMA is) to the destination address (the VMA, where
Dolphin mistakenly already put the code), it ends up overwriting the
code that it is about to execute with garbage, and then promptly
crashing.
After this patch is applied, the behavior now matches the GameCube with
Swiss (assuming a [recent patch](https://github.com/emukidid/swiss-gc/commit/bb4a57186c1481e291b7b0ea6d8e722bab816042)
is applied), and the Wii with the Homebrew Channel.
# like add_library(new ALIAS old) but avoids add_library cannot create ALIAS target "new" because target "old" is imported but not globally visible. on older cmake
# This can be replaced with a direct alias call once our minimum is cmake 3.18
function(dolphin_alias_librarynewold)
@@ -138,3 +141,31 @@ function(dolphin_find_optional_system_library_pkgconfig library search alias bun
message(FATAL_ERROR"Requires ${COMPILER} ${${COMPILER}_min_version} or higher")
endif()
# libstdc++ is almost always used on Linux, even when using clang as the compiler.
# libc++ is used on the likes of Android, Apple devices, FreeBSD, and a few (very) rare Linux distros like Chimera Linux.
# Windows uses its own standard library named STL, which we check as part of the toolset above. (outside of MinGW which can use either libstdc++ or libc++)
message(WARNING"APPROVED_VENDORED_DEPENDENCIES is deprecated. Please migrate to setting USE_SYSTEM_LIBS to ON and setting USE_SYSTEM_<dependency> to either AUTO or OFF to allow bundled libs.")
endif()
option(USE_UPNP"Enables UPnP port mapping support"ON)
option(ENABLE_NOGUI"Enable NoGUI frontend"ON)
option(ENABLE_NOGUI"Enable NoGUI frontend"OFF)
option(ENABLE_QT"Enable Qt (Default)"ON)
option(ENABLE_LTO"Enables Link Time Optimization"OFF)
option(ENABLE_GENERIC"Enables generic build that should run on any little-endian host"OFF)
@@ -97,6 +118,7 @@ option(USE_DISCORD_PRESENCE "Enables Discord Rich Presence, show the current gam
option(USE_MGBA"Enables GBA controllers emulation using libmgba"ON)
option(ENABLE_AUTOUPDATE"Enables support for automatic updates"ON)
option(USE_RETRO_ACHIEVEMENTS"Enables integration with retroachievements.org"ON)
# Make sure we get DSP LLE recompiler and not DSP LLE interpreter, for performance.
EnableJIT=True
[Video_Settings]
# Fixes old frames being shown when a game only updates a small part of the screen. Examples include gradually revealed text and the HP bars in Pokémon.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
