This setting is best referred to as a speed limit, as it involves the limits of all timing based aspects of the emulator, not only framerate.
This allows us to differentiate it from the fps unlocker setting.
GLASM is getting good enough that we can move it out of advanced
graphics settings. This removes the setting `use_assembly_shaders`,
opting for a enum class `shader_backend`. This comes with the benefits
that it is extensible for additional shader backends besides GLSL and
GLASM, and this will work better with a QComboBox.
Qt removes the related assembly shader setting from the Advanced
Graphics section and places it as a new QComboBox in the API Settings
group. This will replace the Vulkan device selector when OpenGL is
selected.
Additionally, mark all of the custom anisotropic filtering settings as
"WILL BREAK THINGS", as that is the case with a select few games.
Mostly fixing unused *, implicit conversion, braced scalar init,
fpermissive, and some others.
Some Clang errors likely remain in video_core, and std::ranges is still
a pertinent issue in shader_recompiler
shader_recompiler: cmake: Force bracket depth to 1024 on Clang
Increases the maximum fold expression depth
thread_worker: Include condition_variable
Don't use list initializers in control flow
Co-authored-by: ReinUsesLisp <reinuseslisp@airmail.cc>
Some titles crash if the FPS limit is disabled when launching. This change ensures that titles launch with the limit in-place to avoid issues.
In order to simplify the change, the UI toggle was removed as it will always be overridden at launch to be disabled.
The setting can still be toggled during gameplay with the hotkey, and indicated by the fps label in the status bar.
Decouples the CPU debugging mode from the enumeration to its own
boolean. After this, it moves the CPU Debugging tab over to a sub tab
underneath the Debug tab in the configuration UI.
Old CPU Accuracy setting won't translate well into since we're adding
one at the beginning of the list. On first boot with the new setting,
just use the default setting.
The current CPU accuracy settings in yuzu are fairly polarized and
require more than common knowledge to know what the optimal settings for
yuzu would be. This adds a curated option called 'Auto' that applies a
few at the moment known-good unsafe optimizations to Dynarmic.
Slight improvements to readability.
Dropped suggestions for string_view (settings.h:101), pass by value
(settings.h:82), reverting double to a float (config.cpp:316), and other
smaller ones, some out of scope.
Addresses review feedback.
Co-authored-by: Ameer J <52414509+ameerj@users.noreply.github.com>
There's no point in keeping the file open after the write limit is exceeded. This allows the file to be committed to the disk shortly after it is closed and avoids redundantly checking whether or not the write limit is exceeded.
It became apparent that logging can continuously spam errors that trigger file flushing.
Since committing the files to disk is an expensive operation, this causes unnecessarily high disk usage.
As such, we will revert Flush() to the previous behavior and add a Commit() member function in the event that this behavior is needed.
Creates a new BasicSettings class in common/settings, and forces setting
a default and label for each setting that uses it in common/settings.
Moves defaults and labels from both frontends into common settings.
Creates a helper function in each frontend to facillitate reading the
settings now with the new default and label properties.
Settings::Setting is also now a subclass of Settings::BasicSetting. Also
adds documentation for both Setting and BasicSetting.
Removes common_sizes.h in favor of having `_KiB`, `_MiB`, `_GiB`, etc
user-literals within literals.h.
To keep the global namespace clean, users will have to use:
```
using namespace Common::Literals;
```
to access these literals.
This check was preventing files with the Write or Append file access modes from being created, as per the documented behavior in FileAccessMode.
This amends the check to test for the existence of a filesystem object prior to checking whether it is a regular file.
Thanks to liushuyu for pointing out that removing the check altogether would not guard against attempting to open non-regular files such as directories, symlinks, FIFO (pipes), sockets, block devices, or character devices.
The documentation has also been updated for these functions to clarify that a file refers to a regular file.
Similarly, Flush() is typically called to attempt to flush a file into the disk. In the one case where this is used, we do not care whether the flush has succeeded or not, making [[nodiscard]] unnecessary.
There are a lot of scenarios where we don't particularly care whether or not the removal operation and just simply attempt a removal.
As such, removing the [[nodiscard]] attribute is best for these functions.
If this is not waited on, the synchronization primitives are destroyed
whe main exits and the detached task ends up signalling garbage and not
properly finishing.
This introduces a new setting Enable FS Access Log which saves the filesystem access log to sdmc:/FsAccessLog.txt
If this setting is not enabled, this will indicate to FS to not call OutputAccessLogToSdCard.
Fixes softlocks during loading in Xenoblade Chronicles 2 when certain DLC is enabled.
std::fflush does not guarantee that file buffers are flushed to the disk.
Use _commit on Windows and fsync on all other OSes to ensure that the file is flushed to the disk.
MSVC's implementation of recursive_directory_iterator throws an exception on an error despite a std::error_code being passed into its constructor. This is most likely a bug in MSVC's implementation since directory_iterator does not throw an exception on an error.
We can replace the usage of recursive_directory_iterator for now until MSVC fixes their implementation of it.
This falls back to the old approach of using a virtual buffer.
Windows is untested, but this build should fix support for Windows < 10 v1803. However without fastmem support at all.
These macros all interact with the result code type, so they should
ideally be within this file as well, so all the common_funcs machinery
doesn't need to be pulled in just to use them.
* common: fs: fs_types: Create filesystem types
Contains various filesystem types used by the Common::FS library
* common: fs: fs_util: Add std::string to std::u8string conversion utility
* common: fs: path_util: Add utlity functions for paths
Contains various utility functions for getting or manipulating filesystem paths used by the Common::FS library
* common: fs: file: Rewrite the IOFile implementation
* common: fs: Reimplement Common::FS library using std::filesystem
* common: fs: fs_paths: Add fs_paths to replace common_paths
* common: fs: path_util: Add the rest of the path functions
* common: Remove the previous Common::FS implementation
* general: Remove unused fs includes
* string_util: Remove unused function and include
* nvidia_flags: Migrate to the new Common::FS library
* settings: Migrate to the new Common::FS library
* logging: backend: Migrate to the new Common::FS library
* core: Migrate to the new Common::FS library
* perf_stats: Migrate to the new Common::FS library
* reporter: Migrate to the new Common::FS library
* telemetry_session: Migrate to the new Common::FS library
* key_manager: Migrate to the new Common::FS library
* bis_factory: Migrate to the new Common::FS library
* registered_cache: Migrate to the new Common::FS library
* xts_archive: Migrate to the new Common::FS library
* service: acc: Migrate to the new Common::FS library
* applets/profile: Migrate to the new Common::FS library
* applets/web: Migrate to the new Common::FS library
* service: filesystem: Migrate to the new Common::FS library
* loader: Migrate to the new Common::FS library
* gl_shader_disk_cache: Migrate to the new Common::FS library
* nsight_aftermath_tracker: Migrate to the new Common::FS library
* vulkan_library: Migrate to the new Common::FS library
* configure_debug: Migrate to the new Common::FS library
* game_list_worker: Migrate to the new Common::FS library
* config: Migrate to the new Common::FS library
* configure_filesystem: Migrate to the new Common::FS library
* configure_per_game_addons: Migrate to the new Common::FS library
* configure_profile_manager: Migrate to the new Common::FS library
* configure_ui: Migrate to the new Common::FS library
* input_profiles: Migrate to the new Common::FS library
* yuzu_cmd: config: Migrate to the new Common::FS library
* yuzu_cmd: Migrate to the new Common::FS library
* vfs_real: Migrate to the new Common::FS library
* vfs: Migrate to the new Common::FS library
* vfs_libzip: Migrate to the new Common::FS library
* service: bcat: Migrate to the new Common::FS library
* yuzu: main: Migrate to the new Common::FS library
* vfs_real: Delete the contents of an existing file in CreateFile
Current usages of CreateFile expect to delete the contents of an existing file, retain this behavior for now.
* input_profiles: Don't iterate the input profile dir if it does not exist
Silences an error produced in the log if the directory does not exist.
* game_list_worker: Skip parsing file if the returned VfsFile is nullptr
Prevents crashes in GetLoader when the virtual file is nullptr
* common: fs: Validate paths for path length
* service: filesystem: Open the mod load directory as read only
Allows setting CPU accuracy to Accurate or Unsafe per-game, as well as
the accuracy options for Unsafe. Debug is not allowed here as a per-game
CPU accuracy.
- This is a developer-only setting and no longer needs to be enabled by default.
- Also adds "use_auto_stub" setting to SDL frontend while we are here.
- Supersedes #1340.
It shall block until there is something to consume in the queue.
And use it for the GPU emulation instead of the spin loop.
This is only in booting the emulator, however in BOTW this is the case for about 1 second.
Advantage: Altering the handler does not need a full recompilation.
Disadvantage: noreturn is droped, so the caller is a bit slower.
We quite often run yuzu with a YOLO assertion handler. In fact, only very few
games run at all with asserts. This patch allows developers to patch the handler
without recompiling everything. The overhead of the missing "noreturn" attribute
shoul be negletable.
`network.cpp` has several error paths which either:
- report "Unhandled host socket error=n" and return `SUCCESS`, or
- switch on a few possible errors, log them, and translate them to
Errno; the same switch statement is copied and pasted in multiple
places in the code
Convert these paths to use a helper function `GetAndLogLastError`, which
is roughly the equivalent of one of the switch statements, but:
- handling more cases (both ones that were already in `Errno`, and a few
more I added), and
- using OS functions to convert the error to a string when logging, so
it'll describe the error even if it's not one of the ones in the
switch statement.
- To handle this, refactor the logic in `GetLastErrorMsg` to expose a
new function `NativeErrorToString` which takes the error number
explicitly as an argument. And improve the Windows version a bit.
Also, add a test which exercises two random error paths.
- With using unique_ptr instead of shared_ptr, we have more explicit ownership of the context.
- Fixes a memory leak due to circular reference of the shared pointer.
- Previous optimized impl. resulted in an integer overflow, so revert.
- This is our slow/fallback path that should never be really be used, so the optimization in unimportant.
Allow sharing return types with the rest of the code base. For example,
we use 'u128 = std::array<u64, 2>', meanwhile Google's code uses
'uint128 = std::pair<u64, u64>'.
While we are at it, use size_t instead of std::size_t.
We can use the standardized CLZ facilities to perform this. This also
allows us to make utilizing functions constexpr and eliminate the
inclusion of an intrinsics header.
Setting __GL_SHADER_DISK_CACHE_PATH we can force the cache directory to
be in yuzu's user directory to stop commonly distributed malware from
deleting our driver shader cache. And by setting
__GL_SHADER_DISK_CACHE_SKIP_CLEANUP we can have an unbounded shader
cache size.
This has only been implemented on Windows, mostly because previous tests
didn't seem to work on Linux.
Disable the precompiled cache on Nvidia's driver. There's no need to
hide information the driver already has in its own cache.
Reworks the tree header to operate off of templates as opposed to a
series of defines.
This allows all tree facilities to obey namespacing rules, and also
allows this code to be used within modules once compiler support is in
place.
This also gets rid to use a macro to define functions and structs for
necessary data types. With templates, these will be generated when
they're actually used, eliminating the need for the separate
declaration.
Squash attributes into the pointer's integer, making them an uintptr_t
pair containing 2 bits at the bottom and then the pointer. These bits
are currently unused thanks to alignment requirements.
Configure Dynarmic to mask out these bits on pointer reads.
While we are at it, remove some unused attributes carried over from
Citra.
Read/Write and other hot functions use a two step unpacking process that
is less readable to stop MSVC from emitting an extra AND instruction in
the hot path:
mov rdi,rcx
shr rdx,0Ch
mov r8,qword ptr [rax+8]
mov rax,qword ptr [r8+rdx*8]
mov rdx,rax
-and al,3
and rdx,0FFFFFFFFFFFFFFFCh
je Core::Memory::Memory::Impl::Read<unsigned char>
mov rax,qword ptr [vaddr]
movzx eax,byte ptr [rdx+rax]
- For `std::same_as`, add missing include of `<concepts>`.
- For `std::convertible_to`, create a replacement in `common/concepts.h`
and use that instead.
This would also be found in `<concepts>`, but unlike `std::same_as`,
`std::convertible_to` is not yet implemented in libc++, LLVM's STL
implementation - not even in master. (In fact, `std::same_as` is the
*only* concept currently implemented. For some reason.)
Fixes regression by 91c13721de, causing
yuzu to not build on Linux with any version of Boost except a cached
1.73 Conan version from before about a day ago.
Moves the Boost requirement out of the `REQUIRED_LIBS` psuedo-2D-array
for Conan to instead be manually configured, using Conan as a fallback
solution if the system does not meet our requirements.
Requires any update from the linux-fresh container in order to build.
**DO NOT MERGE** until someone with the MSVC toolchain can verify this
works there, too.
Fix CreateFullPath to have its intended previous behavior (whatever
that was), and deprecate it in favor of the new CreateDirs function.
Unlike CreateDir, CreateDirs is marked as [[nodiscard]] to avoid new
code ignoring its result value.
Converts creation and deletion functions over to std::filesystem,
simplifying our file-handling code.
Notably with this, CopyDir will now function on Windows.
Add a std::bit_cast-like function archiving the same runtime results as
the standard function, without compile time support.
This allows us to use bit_cast while we wait for compiler support, it
can be trivially replaced in the future.
VirtualBuffer makes use of VirtualAlloc (on Windows) and mmap() (on
other platforms). Neither of these ensure that non-trivial objects are
properly constructed in the allocated memory.
To prevent potential undefined behavior occurring due to that, we can
add a static assert to loudly complain about cases where that is done.
Makes page tables and virtual buffers able to be moved, but not copied,
making the interface more flexible.
Previously, with the destructor specified, but no move assignment or
constructor specified, they wouldn't be implicitly generated.
Hides all of the implementation details for users of the class. This has
the benefit of reducing includes and also making the fiber classes
movable again.
Allows our CI to catch more potential bugs. This also removes the
[[nodiscard]] attribute of IOFile's Open member function. There are
cases where a file may want to be opened, but have the status of it
checked at a later time.
This commit aims to implement the NVDEC (Nvidia Decoder) functionality, with video frame decoding being handled by the FFmpeg library.
The process begins with Ioctl commands being sent to the NVDEC and VIC (Video Image Composer) emulated devices. These allocate the necessary GPU buffers for the frame data, along with providing information on the incoming video data. A Submit command then signals the GPU to process and decode the frame data.
To decode the frame, the respective codec's header must be manually composed from the information provided by NVDEC, then sent with the raw frame data to the ffmpeg library.
Currently, H264 and VP9 are supported, with VP9 having some minor artifacting issues related mainly to the reference frame composition in its uncompressed header.
Async GPU is not properly implemented at the moment.
Co-Authored-By: David <25727384+ogniK5377@users.noreply.github.com>
Makes our error coverage a little more consistent across the board by
applying it to Linux side of things as well. This also makes it more
consistent with the warning settings in other libraries in the project.
This also updates httplib to 0.7.9, as there are several warning
cleanups made that allow us to enable several warnings as errors.
From -fsanitize=address, this code wasn't calling the proper destructor.
Adding virtual destructors for each inherited class and the base class
fixes this bug.
While we are at it, mark the functions as final.
I made a request on the Xbyak issue tracker to allow some constructors
to be constexpr in order to avoid static constructors from needing to
execute for some of our register constants.
This request was implemented, so this updates Xbyak so that we can make
use of it.
The extended logging option is automatically disabled on boot but can be enabled afterwards, allowing the log file to go up to 1 GB during that session.
This commit also fixes a few errors that are present in the general debug menu.
This is the only place it's actively used. It's also more appropriate
for web-related structures to be within the web service target.
Especially given this one doesn't rely on anything in the common
library.
Migrates a remaining common file over to the Common namespace, making it
consistent with the rest of common files.
This also allows for high-traffic FS related code to alias the
filesystem function namespace as
namespace FS = Common::FS;
for more concise typing.
These are intentionally discarded internally, since the rest of the
public API allows querying success. We want all non-internal uses of
these functions to be explicitly checked, so we can signify that we
intentionally want to discard the return values here.
We can simplify this function down into a single line with the use of
fmt. A benefit with the fmt approach is that the fmt variant of
localtime is thread-safe as well, making GetOsTimeZoneOffset()
thread-safe as well.
Now that clang-format makes [[nodiscard]] attributes format sensibly, we
can apply them to several functions within the common library to allow
the compiler to complain about any misuses of the functions.
This makes it more inline with its currently unavailable standardized
analogue std::derived_from.
While we're at it, we can also make the template match the requirements
of the standardized variant as well.
Previously the constructor for all of these would run at program
startup, consuming time before the application can enter main().
This is also particularly dangerous, given the logging system wouldn't
have been initialized properly yet, yet the program would use the logs
to signify an error.
To rectify this, we can replace the literals with constexpr functions
that perform the conversion at compile-time, completely eliminating the
runtime cost of initializing these arrays.
- In `SetCurrentThreadName`, when on Linux, truncate to 15 bytes, as (at
least on glibc) `pthread_set_name_np` will otherwise return `ERANGE` and
do nothing.
- Also, add logging in case `pthread_set_name_np` returns an error
anyway. This is Linux-specific, as the Apple and BSD versions of
`pthread_set_name_np return `void`.
- Change the name for CPU threads in multi-core mode from
"yuzu:CoreCPUThread_N" (19 bytes) to "yuzu:CPUCore_N" (14 bytes) so it
fits into the Linux limit. Some other thread names are also cut off,
but I didn't bother addressing them as you can guess them from the
truncated versions. For a CPU thread, truncation means you can't see
which core it is!
On DragonFly and NetBSD build fails with
src/common/virtual_buffer.cpp
src/common/virtual_buffer.cpp:16:10: fatal error: sys/sysinfo.h: No such file or directory
#include <sys/sysinfo.h>
^~~~~~~~~~~~~~~
* ipc: Allow all trivially copyable objects to be passed directly into WriteBuffer
With the support of C++20, we can use concepts to deduce if a type is an STL container or not.
* More agressive concept for stl containers
* Add -fconcepts
* Move to common namespace
* Add Common::IsBaseOf
In file included from src/core/hle/kernel/memory/page_table.cpp:5:
src/./common/alignment.h:67:68: error: no member named 'align_val_t' in namespace 'std'
return static_cast<T*>(::operator new (n * sizeof(T), std::align_val_t{Align}));
~~~~~^
src/./common/alignment.h:71:51: error: no member named 'align_val_t' in namespace 'std'
::operator delete (p, n * sizeof(T), std::align_val_t{Align});
~~~~~^
In cases where the size is not a known constant when inlining, AlignUp<std::size_t> currently generates two 64-bit div instructions.
This generates one div and a cmov which is significantly cheaper.
This is a new attempt at #4206 that shouldn't break windows builds.
If someone else could test on windows, it would be much appreciated.
Previously, the build bot passed but the actual builds failed.
On gcc/ld, and clang/lld, fmt::v6 symbols are excluded, so linking
fails. This fixes the issue.
Note: This was included in the FindBoost changes I shared with
BlinkHawk, however only they were merged. I'm not sure if it was missed,
or if there was an issue with this part of the change.
This commit: Implements CPU Interrupts, Replaces Cycle Timing for Host
Timing, Reworks the Kernel's Scheduler, Introduce Idle State and
Suspended State, Recreates the bootmanager, Initializes Multicore
system.
Emit code compatible with NV_gpu_program5.
This should emit code compatible with Fermi, but it wasn't tested on
that architecture. Pascal has some issues not present on Turing GPUs.
* Remove git submodules that will be loaded through conan
* Move custom Find modules to their own folder
* Use conan for downloading missing external dependencies
* CI: Change the yuzu source folder user to the user that the containers run on
* Attempt to remove dirty mingw build hack
* Install conan on the msvc build
* Only set release build type when using not using multi config generator
* Re-add qt bundled to workaround an issue with conan qt not downloading prebuilt binaries
* Add workaround for submodules that use legacy CMAKE variables
* Re-add USE_BUNDLED_QT on the msvc build bot
It's undefined behavior to pass a null pointer to std::fread and
std::fwrite, even if the length passed in is zero, so we must perform
the precondition checking ourselves.
A common case where this can occur is when passing in the data of an
empty std::vector and size, as an empty vector will typically have a
null internal buffer.
While we're at it, we can move the implementation out of line and add
debug checks against passing in nullptr to std::fread and std::fwrite.
On Windows, network shares use paths like \\server\share\file which were
being broken by FileUtil::SanitizePath() removing double slashes.
Changed the code in SanitizePath to permit a double-backslash if it
occurs at the start of a filepath (on Windows only).
* IOFile: Make the move constructor and move assignment operator noexcept
Certain parts of the standard library try to determine whether or not a
transfer operation should either be a copy or a move. The prevalent notion
of move constructors/assignment operators is that they should not throw,
they simply move an already existing resource somewhere else.
This is typically done with 'std::move_if_noexcept'. Like the name says,
if a type's move constructor is noexcept, then the functions retrieves an
r-value reference (for move semantics), or an l-value (for copy semantics)
if it is not noexcept.
As IOFile deletes the copy constructor and copy assignment operators,
using IOFile with certain parts of the standard library can fail in
unexcepted ways (especially when used with various container
implementations). This prevents that.
* fix various instances of -1 being assigned to unsigned types
* do not assign in conditional statements
* File/IOFile: Check _tfopen_s properly
* common/file_util.cpp: address review comments
Co-authored-by: Lioncash <mathew1800@gmail.com>
Co-authored-by: Shawn Hoffman <godisgovernment@gmail.com>
Co-authored-by: Sepalani <sepalani@hotmail.fr>
This PR aims to reduce the memory usage in the CPU page table by moving
GPU specific parameters into a child class. This saves 1Gb of Memory for
most games.
An implementation of the cemuhook motion/touch protocol, this adds the
ability for users to connect several different devices to citra to send
direct motion and touch data to citra.
Co-Authored-By: jroweboy <jroweboy@gmail.com>
We relies on UNREACHABLE's noreturn attribute to eliminate parent's "no return value" warning. However, this was wrapped in a `if(!false)` block, which compilers may not unfold to recognize the noreturn nature.
Makes the header more general for other potential algorithms in the
future. While we're at it, include a missing <functional> include to
satisfy the use of std::less.
This was related to the source allocator being passed into the
constructor potentially having a different type than allocator being
constructed.
We simply need to provide a constructor to handle this case.
This resolves issues related to the allocator causing debug builds on
MSVC to fail.
Implement VOTE using Nvidia's intrinsics. Documentation about these can
be found here
https://developer.nvidia.com/reading-between-threads-shader-intrinsics
Instead of using portable ARB instructions I opted to use Nvidia
intrinsics because these are the closest we have to how Tegra X1
hardware renders.
To stub VOTE on non-Nvidia drivers (including nouveau) this commit
simulates a GPU with a warp size of one, returning what is meaningful
for the instruction being emulated:
* anyThreadNV(value) -> value
* allThreadsNV(value) -> value
* allThreadsEqualNV(value) -> true
ballotARB, also known as "uint64_t(activeThreadsNV())", emits
VOTE.ANY Rd, PT, PT;
on nouveau's compiler. This doesn't match exactly to Nvidia's code
VOTE.ALL Rd, PT, PT;
Which is emulated with activeThreadsNV() by this commit. In theory this
shouldn't really matter since .ANY, .ALL and .EQ affect the predicates
(set to PT on those cases) and not the registers.
This commit ensures that all backing memory allocated for the Guest CPU
is aligned to 256 bytes. This due to how gpu memory works and the heavy
constraints it has in the alignment of physical memory.
Instead of storing all block width, height and depths in their shifted
form:
block_width = 1U << block_shift;
Store them like they are provided by the emulated hardware (their
block_shift form). This way we can avoid doing the costly
Common::AlignUp operation to align texture sizes and drop CPU integer
divisions with bitwise logic (defined in Common::AlignBits).
Avoids potentially performing multiple reallocations (depending on the
size of the input data) by reserving the necessary amount of memory
ahead of time.
This is trivially doable, so there's no harm in it.
These can be generified together by using a concept type to designate
them. This also has the benefit of not making copies of potentially very
large arrays.
Allows for things such as:
auto rect = Common::Rectangle{0, 0, 0, 0};
as opposed to being required to explicitly write out the underlying
type, such as:
auto rect = Common::Rectangle<int>{0, 0, 0, 0};
The only requirement for the deduction is that all constructor arguments
be the same type.
nullptr was being returned in the error case, which, at a glance may
seem perfectly OK... until you realize that std::string has the
invariant that it may not be constructed from a null pointer. This
means that if this error case was ever hit, then the application would
most likely crash from a thrown exception in std::string's constructor.
Instead, we can change the function to return an optional value,
indicating if a failure occurred.
Makes the parameter ordering consistent, and also makes the filename
parameter a std::string. A std::string would be constructed anyways with
the previous code, as IOFile's only constructor with a filepath is one
taking a std::string.
We can also make WriteStringToFile's string parameter utilize a
std::string_view for the string, making use of our previous changes to
IOFile.
We don't need to force the usage of a std::string here, and can instead
use a std::string_view, which allows writing out other forms of strings
(e.g. C-style strings) without any unnecessary heap allocations.
Uses arithmetic that can be identified more trivially by compilers for
optimizations. e.g. Rather than shifting the halves of the value and
then swapping and combining them, we can swap them in place.
e.g. for the original swap32 code on x86-64, clang 8.0 would generate:
mov ecx, edi
rol cx, 8
shl ecx, 16
shr edi, 16
rol di, 8
movzx eax, di
or eax, ecx
ret
while GCC 8.3 would generate the ideal:
mov eax, edi
bswap eax
ret
now both generate the same optimal output.
MSVC used to generate the following with the old code:
mov eax, ecx
rol cx, 8
shr eax, 16
rol ax, 8
movzx ecx, cx
movzx eax, ax
shl ecx, 16
or eax, ecx
ret 0
Now MSVC also generates a similar, but equally optimal result as clang/GCC:
bswap ecx
mov eax, ecx
ret 0
====
In the swap64 case, for the original code, clang 8.0 would generate:
mov eax, edi
bswap eax
shl rax, 32
shr rdi, 32
bswap edi
or rax, rdi
ret
(almost there, but still missing the mark)
while, again, GCC 8.3 would generate the more ideal:
mov rax, rdi
bswap rax
ret
now clang also generates the optimal sequence for this fallback as well.
This is a case where MSVC unfortunately falls short, despite the new
code, this one still generates a doozy of an output.
mov r8, rcx
mov r9, rcx
mov rax, 71776119061217280
mov rdx, r8
and r9, rax
and edx, 65280
mov rax, rcx
shr rax, 16
or r9, rax
mov rax, rcx
shr r9, 16
mov rcx, 280375465082880
and rax, rcx
mov rcx, 1095216660480
or r9, rax
mov rax, r8
and rax, rcx
shr r9, 16
or r9, rax
mov rcx, r8
mov rax, r8
shr r9, 8
shl rax, 16
and ecx, 16711680
or rdx, rax
mov eax, -16777216
and rax, r8
shl rdx, 16
or rdx, rcx
shl rdx, 16
or rax, rdx
shl rax, 8
or rax, r9
ret 0
which is pretty unfortunate.
Allows the compiler to inform when the result of a swap function is
being ignored (which is 100% a bug in all usage scenarios). We also mark
them noexcept to allow other functions using them to be able to be
marked as noexcept and play nicely with things that potentially inspect
"nothrowability".
Including every OS' own built-in byte swapping functions is kind of
undesirable, since it adds yet another build path to ensure compilation
succeeds on.
Given we only support clang, GCC, and MSVC for the time being, we can
utilize their built-in functions directly instead of going through the
OS's API functions.
This shrinks the overall code down to just
if (msvc)
use msvc's functions
else if (clang or gcc)
use clang/gcc's builtins
else
use the slow path
The template type here is actually a forwarding reference, not an rvalue
reference in this case, so it's more appropriate to use std::forward to
preserve the value category of the type being moved.
Makes the return type consistently uniform (like the intrinsics we're
wrapping). This also conveniently silences a truncation warning within
the kernel multi_level_queue.
