Some compilers we care about (mostly g++) do not support std::make_unique yet,
but we still want to use it in our codebase to make unique_ptr code more
readable. This commit introduces an implementation derivated from the libc++
code in the Dolphin codebase so we can use it right now everywhere.
Adapted from delroth's pull request.
Set the x87 precision, even on x64. Since we are using x87 instructions
in the JIT now, we can't guarantee that x87 precision will never
influence Dolphin on x64.
The new NOP emitter breaks when called with a negative count. As it
turns out, it did happen when deoptimizing 8 bit MOVs because they are
only 4 bytes long and need no BSWAP.
Fixes issue 6990.
This uses a bit of templating to remove the duplicate code that is the CodeBlocks in each emitter headers.
No actual functionality change in this.
When creating a Fixupbranch we were swapping the BL and B targets.
I think this was found by PPSSPP a while ago, but they never send PRs to merge their changes upstream.
Between C++11 and C++14, volatile types stopped being trivially
copyable. The serializer has no reason to care about this distinction,
so tack on remove_volatile.
The underlying storage type of a bitfield can be any intrinsic integer type,
but also any enumeration.
Custom storage types are supported if the following things are defined on the storage type:
- casting 0 to the storage type
- bit shift operators (in both directions)
- bitwise & operator
- bitwise ~ operator
- std::make_unsigned specialization
Previously he function was misbehaving because of a missing check for
whether an 8-bit operand was a register operand; it would therefore
emit unnecessary REX prefixes, incorrectly assert on 32-bit targets, and
could potentially emit wrong code in rare cases (like a memory to register
operation involving AH.)
Also, some cleanup while I was in the area to make the function easier to
read.
The workaround of using fixed underlying types produces lots of warnings
in GCC because now the bit-fields are too small for the value range used
for conversion semantics.
Breakpoints have one, but memchecks don't, despite being cleared directly in the breakpoint window.
Now DolphinWX should call the interface functions and not the direct functions of the breakpoints or memchecks for clearing.
Our defines were never clear between what meant 64bit or x86_64
This makes a clear cut between bitness and architecture.
This commit also has the side effect of bringing up aarch64 compiling support.
- remove unused variables
- reduce the scope where it makes sense
- correct limits (did you know that strcat()'s last parameter does not
include the \0 that is always added?)
- set some free()'d pointers to NULL
This breaks Linux stdout logging.
This reverts commit 7ac5b1f2f8, reversing
changes made to 9bc14012fc.
Revert "Merge pull request #77 from lioncash/remove-console"
This reverts commit 9bc14012fc, reversing
changes made to b18a33377d.
Conflicts:
Source/Core/Common/LogManager.cpp
Source/Core/DolphinWX/Frame.cpp
Source/Core/DolphinWX/FrameAui.cpp
Source/Core/DolphinWX/LogConfigWindow.cpp
Source/Core/DolphinWX/LogWindow.cpp
Note I do not mean the Logging window, but the console window.
It's literally rarely, if at all used, and offers less advantages over the built-in logging window (ie. it breaks on different locales: http://i.imgur.com/Cs92tQE.png)
This commit should remove all of the console logging.
Floating-point is complicated...
Some background: Denormals are floats that are too close to zero to be
stored in a normalized way (their exponent would need more bits). Since
they are stored unnormalized, they are hard to work with, even in
hardware. That's why both PowerPC and SSE can be configured to operate
in faster but non-standard-conpliant modes in which these numbers are
simply rounded ('flushed') to zero.
Internally, we do the same as the PowerPC CPU and store all floats in
double format. This means that for loading and storing singles we need a
conversion. The PowerPC CPU does this in hardware. We previously did
this using CVTSS2SD/CVTSD2SS. Unfortunately, these instructions are
considered arithmetic and therefore flush denormals to zero if non-IEEE
mode is active. This normally wouldn't be a problem since the next
arithmetic floating-point instruction would do the same anyway but as it
turns out some games actually use floating-point instructions for
copying arbitrary data.
My idea for fixing this problem was to use x87 instructions since the
x87 FPU never supported flush-to-zero and thus doesn't mangle denormals.
However, there is one more problem to deal with: SNaNs are automatically
converted to QNaNs (by setting the most-significant bit of the
fraction). I opted to fix this by manually resetting the QNaN bit of all
values with all-1s exponent.
I give up. Merging the ppc_fp branch has caused issues in numerous games
and I can't find the bug. I'm leaving this merged to enable easy
recompilation for people who would like to play games that benefit from
non-IEEE mode emulation (e.g. Starfox Assault).
MemArena mmaps the emulated memory from a file in order to get the same
mapping at multiple addresses. A file which, formerly, was located at a
static filename: it was unlinked after creation, but the open did not
use O_EXCL, so if two instances started up on the same system at just
the right time, they would get the same memory. Naturally, this caused
extremely mysterious crashes, but only in Netplay, where the game is
automatically started when the client receives a broadcast from the
server, so races are actually quite likely.
And switch to shm_open, because it fits the bill better and avoids any
issues with using /tmp.
bDAZ is now called bFlushToZero to better reflect what it's actually
used for.
I decided not to support any hardware-based flush-to-zero on systems
that don't support this for both inputs _and_ outputs. It makes the code
cleaner and the intersection of CPUs that support SSE2 but not DAZ
should be very small.
- Add support for std::set and std:pair.
- Switch from std::is_pod to std::is_trivially_copyable, to allow for
types that have constructors but trivial copy constructors. Easy,
except there are three different nonstandard versions of it required
on different platforms, in addition to the standard one.
* Currently there is no DEBUGFAST configuration. Defining DEBUGFAST as a preprocessor definition in Base.props (or a global header) enables it for now, pending a better method. This was done to make managing the build harder to screw up. However it may not even be an issue anymore with the new .props usage.
* D3DX11SaveTextureToFile usage is dropped and not replaced.
* If you have $(DXSDK_DIR) in your global property sheets (Microsoft.Cpp.$(PlatformName).user), you need to remove it. The build will error out with a message if it's configured incorrectly.
* If you are on Windows 8 or above, you no longer need the June 2010 DirectX SDK installed to build dolphin. If you are in this situation, it is still required if you want your built binaries to be able to use XAudio2 and XInput on previous Windows versions.
* GLew updated to 1.10.0
* compiler switches added: /volatile:iso, /d2Zi+
* LTCG available via msbuild property: DolphinRelease
* SDL updated to 2.0.0
* All Externals (excl. OpenAL and SDL) are built from source.
* Now uses STL version of std::{mutex,condition_variable,thread}
* Now uses Build as root directory for *all* intermediate files
* Binary directory is populated as post-build msbuild action
* .gitignore is simplified
* UnitTests project is no longer compiled
Note that before pushing those changes, they were initially tested in a branch, and passed the compilation testing. Sorry that I didn't catch this before.
This implements a partial JITIL based off of the JIT64IL. It's enough to run most games, albiet at a slow speed.
Implementing instructions for this IL is really simple since it basically is just enabling based on what is already in JIT64IL, and then enabling each individual IL instruction.
And fix some stuff up. It would probably be good to unify the stack
handling some more rather than having ABI_PushRegistersAndAdjustStack do
part of it and ABI_AlignStack the rest, causing unnecessary subtract
instructions on Linux x86 (only).
As part of that, change SafeLoadToEAX to SafeLoadToReg, and have JitIL
use that, which should fix fastmem on JitIL.
This should also fix a potential stack corruption issue with x86.
Also define _M_* in a common location, and clean up code that these
changes break (including DSPJit files that assume X86 yet are compiled
on ARM for some reason...)
- For GCC, use intrinsics that will work on ARM.
- Add AtomicExchangeAcquire.
- Make Atomic{Load,LoadAcquire,Store,StoreRelease} work for any suitable type.
- Call ABI_AlignStack even on x86-64.
- Have ABI_AlignStack respect the difference in current alignment
between the root JIT function, which has a prolog, and
ProtectFunction thunks, which do not. This was causing many games
to crash on start on OS X. Since this might otherwise mean changing
the stack pointer before every call...
- Have one prolog/epilog function rather than two (one of which
definitely did not do what it was thought to do), and make it
actually work like a normal one, so that the stack frame shows up
properly in the debugger. There should be no performance impact.
Changes a lot of parsing code which previously was not aware of the notion of
key/value, and operated only with raw lines. Now key/value is the default and
lines are handled as raw only if they do not contain =, or they start with $ or
+ (for Gecko/AR compatibility).
It isn't easily accessible with sigaction or Mach exceptions (well,
requires an additional system call in the latter), and isn't necessary.
(and get rid of the enum, because it's only used once, and the comments
are more expressive than enum names)
MSVC insisted on using a copy assignment where a move was intended and
ought to be used. This would have been caught, because the class in
question inherits from NonCopyable, which declares a move assignment
operator, which is supposed to delete the implicitly declared copy
assignment operator, but of course MSVC didn't do that either, causing a
class that should have been safe to be unsafe.
(Intertwined enough that's it's easier to do in one patch.)
(1) /dev/es did not support state save, which could cause crashes and
incorrect behavior after loading.
(2) NANDContentLoader tried to read all of a title's contents into
memory when it was first opened. Two issues:
- If any contents were missing, it bailed out. However, with DLC,
only some of the contents may be downloaded, as determined by the
permission bits in the ticket. Instead, return an appropriate error
when a content is accessed that doesn't exist on the filesystem
(don't bother checking the permission bits though).
- Everything was loaded into memory - even if it consisted of 3 GB of
songs, which caused Dolphin to lag out for quite a while (and would
fail on 32-bit). Instead, open content on demand.
This is required to be able to move objects that inherit from it.
(Note that this patch also #ifs out the class for the externals that
include it yet are compiled in pre-C++11 mode. It shouldn't matter,
since those externals don't use it.)
It's not enough to check for the CPUID bit to know if AVX is supported since
AVX requires OS support (new set of registers == more registers to be saved
when context switching). If the OS does not support, the cpuid bit will still
be set but using YMM registers will cause an illegal exception fault.
The MS INI parser and most other INI parsing libraries APIs only support
comments at the beginning of lines. Right now, some Game INI files use sections
like:
[OnFrame]#Add memory patches here
But these section headers are parsed separately, so this should not break
them.
This reverts commit 7aae9ccbc0.
Reasons:
- no test results have been provided to prove the usefulness of the patch
- broken coding style
- the author hasn't replied to any criticism
Also, some tab/space mismatches removed from VideoOGL, and some places I missed in VideoDX[number] projects.
Now, the Core is literally the only project with tab/space mismatches (on a large scale).
Probably better to keep that same-register-MOV assert useful. Also,
explicit type extension documents whats happening.
Internally, this boils down to the original MOV, but without the assert.
This commit mainly elaborates on some messages a little more. Also fixes some typos that slipped through the last commit.
A large change in text can be seen in EXI_DeviceMemoryCard.cpp. I added more info as to why a write to a memory card may fail. (This actually was a reason I was unable to write to a memcard recently).
Elaborations can be seen in WGL.cpp
I did change some comments in some files that I was correcting logging messages in, however this is only if I spot a typo or if an abbreviation is lower-cased. Even in that case, the amount of changes done to comments is very minimal.
# By Ryan Houdek (185) and others
# Via degasus (12) and others
* master: (625 commits)
Revert "Don't open/close file for every file operation." as it was crashing PokePark in Windows builds.
Array overrun fixed in VertexShaderCache for the DX11 plugin.
Fixed DSPTool build.
Windows build fix
Go back to assuming every HID device is a wiimote on Windows. Fixed issue 6117. Unfixed issue 6031.
VideoSoftware: Improve fog range adjustment by using less magic and more comments.
revert RasterFont for VideoSoftware
ogl: fix virtual xfb
Windows build fix from web interface...
Adjusted the audio loop criteria, using >= on the Wii and == on GC. This fixes the audio static that occurred in Wii games after hours of play.
Forced the exception check only for ARAM DMA transfers. Removed the Eternal Darkness boot hack and replaced it with an exception check.
VideoSoftware: Implement fog range adjustment, fixing issue 6147.
implement 4xSSAA for OGL
move ogl-only settings into backend
Fix description of disable fog, and move it to enhancements tab.
Reverted rd76ca5783743 as it was made obsolete by r1d550f4496e4.
Removed the tracking of the FIFO Writes as it was made obsolete by r1d550f4496e4.
Forced the external exception check to occur sooner by changing the downcount.
Mark the Direct3D9 backend deprecated.
Prefer D3D11 and OpenGL over D3D9 by default.
...
Conflicts:
CMakeLists.txt
Source/Core/Common/Common.vcxproj.filters
Source/Core/Common/Src/CommonPaths.h
Source/Core/Core/Core.vcxproj.filters
Source/Core/Core/Src/Core.cpp
Source/Core/Core/Src/IPC_HLE/WII_IPC_HLE_Device_FileIO.cpp
Source/VSProps/Dolphin.Win32.props
Source/VSProps/Dolphin.x64.props
Some cleanup throughout related code. (try to make logic in ISOFile understandable by a human)
Encode strings in UTF-8 rather than somehow trying to determine the encoding in the GUI code.
Non-windows OSes temporarily broken.
The real file was never kept open for longer than a single operation so there was no point in dealing with it in DoState.
Saving the real path in the savestate was also probably a bad idea. Savestates should be a bit more portable now.
* OpenAL:
Changed SoundTouch to use float samples, allowing SSE to be used. Made the DPL2 decoder disabled by default. Re-added the audio hack used by the Accurate VBeam emulation option.
Added a latency setting to the audio settings. Removed the Sample Rate setting. It is now hardcoded to 48000hz (accurate audio timing).
Skipped timestretching if the emulator is running below 10% speed to prevent buffer overflows.
Removed the synchronisation between the CPU thread and the audio thread. Added code to detect and resume from buffer underruns. Disabled the ability to change the DPL2 option after the game has started. Fixed a memory leak that occurred in the DPL2 decoder. Fixed the OSX build.
Build fix
Added a Dolby Pro Logic II (DPL2) decoder in the OpenAL backend. DPL2 audio is decoded to 5.1. Code adapted from ffdshow. Added an option in the DSP settings to disable the DPL2 decoder in case Dolphin incorrectly detects a 5.1 audio system. Updated the OpenAL files to OpenAL Soft 1.15.1 in the Windows build.
Removed the system timing hack which was activated when the Accurate VBeam option was enabled.
Fixed the include directories in Audio Common for the Windows 32bit build.
Fixed the include directories in Audio Common for the Windows build.
Messed up the static include line
Fix include paths and compiling in Linux. Externals soundtouch is 1.7.1, while Ubuntu 12.10 is 1.6.x. Externals soundtouch is compiled with integer samples, while ubuntu is compiled with float samples. Float samples is probably the more common route. If you're going to use soundtouch, you should probably use SAMPLETYPE instead of explicitly choosing short. This probably breaks the windows build since its includes aren't setup.
OSX: typedef signed char BOOL
OSX build fix
Build fix
Added audio time stretching by using the SoundTouch library.
Implemented correct audio timing.
OpenAL for Windows initial commit
Also changed the size parameter from DoVoid to size_t. As far as I know, that should never be negative.
Also changed the the count parameter of DoArray to size_t, this also should never be negative.
Got rid of some typecasts.
Depth calculations are always done in the pixel shader now.
Due to the unpredictability of our zcomploc hacks this commit probably changes the behavior of some games which use zcomploc.
- In icon retrieving I removed the "format check" as it shouldn't really matter to have mixed icon formats. Also removed the "Time splitters hack" as there's no reason for it since we are only checking the last 3 bits and I'm pretty sure having bits 1 and 2 set is the same as having them unset.
- Icon retrieving uses AnimSpeed as stop signal (every icon must have an speed set, the first speed that is 0 means there are no more icons)
- Also, in icon retrieving I added support for "blank frames"(Luigi's Mansion and Pikmin that I know of). With this the base for icon animation is complete.
- Fixed PSOIII savegame patch which was wrong before.
Signed-off-by: LPFaint99 <lpfaint99@gmail.com>
Reason:
- It's wrong, zcomploc can't be emulated perfectly in HW backends without severely impacting performance.
- It provides virtually no advantages over the previous hack while introducing lots of code.
- There is a better alternative: If people insist on having some sort of valid zcomploc emulation, I suggest rendering each primitive separately while using a _clean_ dual-pass approach to emulate zcomploc.
This reverts commit 0efd4e5c29.
This reverts commit b4ec836aca.
This reverts commit bb4c9e2205.
This reverts commit 146b02615c.
the intent is to replace the haphazard scheduling and finger-crossing associated with saving/loading with the correct and minimal necessary wait for each thread to reach a known safe location before commencing the savestate operation, and for any already-paused components to not need to be resumed to do so.
* misc-speedups:
fixed and reenabled and slightly optimized the JIT version of fcmpo/fcmpu.
slightly more precise speed percent display (this is really minor)
a small thread synchronization speedup for dual core mode. it's most noticeable in games where the CPU is running behind compared to the GPU.
Conflicts:
Source/Core/Core/Src/PowerPC/Jit64/Jit.cpp
The Fifo.cpp changes from rdaefb3b550e2 was not merged as there was no performance benefit.
please test for regressions, speed and for other issues fixed, as a example, the black color in water splash in super mario galaxy are fixed with this rev.
please as soon as yo find a bug let me know.
a little code cleaning to avoid duplicated execution of AlphaPreTest and a little correction to some comments from the previous commits.
this change must behave exactly like last revision, if something is broken please let me know
to marcosvitali.
Added an external exception check when the CPU writes to the FIFO. This allows
the CPU time to service FIFO overflows. Fixes random hangs caused by FIFO
overflows and desyncs like in "The Last Story" and "Battalion Wars 2". Thanks
to marcosvitali for the research.
Added some code to unlink invalidated blocks so that the recompiled block can be
linked (speed-up).
This release still fixed the hangs produced by fifo overflow without sacrifice
performance. For example you can test Tutorial moves at the beginning of The last history now
is fluid 30/60.
Fixed possibles random hangs in DC mode.
Fixed hangs in DC mode in (Simpsons, Monkey Island, Pokemon XD, etc)
Implemented accurate management of Pixel Engine Interrupts. Now the GPU loop
is stopped when a PE Interrupt needs to be managed and resumed when Pixel Engine
finish.
Fixed Metroid Prime 3 and 2 desync. And other games with desync because of
FIFO Reset. That happens because FIFO_RW_DISTANCE_HI must be written first, for checking
fifo.CPReadWriteDistance == 0, so some fifo resets was not managed in the right
way.
Fixed Super Monkey Ball in some cases when the game write the
WriteReadDistance need to be safe like the SafeCPRead.
Improved the CheckException for the GatherPipe writes in JIT, now only the
External Exceptions are processed.
Fixed definitely Pokemon XD in dual core mode. This game is doing something
not allowed. It attach to CPU the same fifo attached to the GPU in multibuffer
mode. I added a check to prevent overwrite the GPU FIFO with the CPU FIFO. If
the game do that on breakpoint the solution can fail.
Fixed ReadWriteDistance calc when CPRead > CPWrite.
Added Token and Finish cause to GP Jit checking.
Additional cleanup in CommandProcessor.
Fixes issue 5209
Fixes issue 5055
Fixes issue 4889
Fixes issue 4061
Fixes issue 4010
Fixes issue 3902
For example you can test Tutorial moves at the beginning of The last history now is fluid 30/60.
Shuffle2: I've delete the hacky line, I think is not necessary anymore. Additional some clean in CommandProcessor.
Please test The Last Story and others games affected in the previous commits and give me a feedback.
Pierre Bourdon
