This manages two kinds of streaming buffers: one for unified memory
models and one for dedicated GPUs. The first one skips the copy from the
staging buffer to the real buffer, since it creates an unified buffer.
This implementation waits for all fences to finish their operation
before "invalidating". This is suboptimal since it should allocate
another buffer or start searching from the beginning. There is room for
improvement here.
This could also handle AMD's "pinned" memory (a heap with 256 MiB) that
seems to be designed for buffer streaming.
The scheduler abstracts command buffer and fence management with an
interface that's able to do OpenGL-like operations on Vulkan command
buffers.
It returns by value a command buffer and fence that have to be used for
subsequent operations until Flush or Finish is executed, after that the
current execution context (the pair of command buffers and fences) gets
invalidated a new one must be fetched. Thankfully validation layers will
quickly detect if this is skipped throwing an error due to modifications
to a sent command buffer.
The NVFlinger service is already passed into services that need to
guarantee its lifetime, so the BufferQueue instances will already live
as long as they're needed. Making them std::shared_ptr instances in this
case is unnecessary.
Like the previous changes made to the Display struct, this prepares the
Layer struct for changes to its interface. Given Layer will be given
more invariants in the future, we convert it into a class to better
signify that.
With the display and layer structures relocated to the vi service, we
can begin giving these a proper interface before beginning to properly
support the display types.
This converts the display struct into a class and provides it with the
necessary functions to preserve behavior within the NVFlinger class.
* Fixes Unicode Key File Directories
Adds code so that when loading a file it converts to UTF16 first, to
ensure the files can be opened. Code borrowed from FileUtil::Exists.
* Update src/core/crypto/key_manager.cpp
Co-Authored-By: Jungorend <Jungorend@users.noreply.github.com>
* Update src/core/crypto/key_manager.cpp
Co-Authored-By: Jungorend <Jungorend@users.noreply.github.com>
* Using FileUtil instead to be cleaner.
* Update src/core/crypto/key_manager.cpp
Co-Authored-By: Jungorend <Jungorend@users.noreply.github.com>
These are more closely related to the vi service as opposed to the
intermediary nvflinger.
This also places them in their relevant subfolder, as future changes to
these will likely result in subclassing to represent various displays
and services, as they're done within the service itself on hardware.
The reasoning for prefixing the display and layer source files is to
avoid potential clashing if two files with the same name are compiled
(e.g. if 'display.cpp/.h' or 'layer.cpp/.h' is added to another service
at any point), which MSVC will actually warn against. This prevents that
case from occurring.
This also presently coverts the std::array introduced within
f45c25aaba back to a std::vector to allow
the forward declaration of the Display type. Forward declaring a type
within a std::vector is allowed since the introduction of N4510
(http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4510.html) by
Zhihao Yuan.
As fetching command list headers and and the list of command headers is a fixed 1:1 relation now, they can be implemented within a single call.
This cleans up the Step() logic quite a bit.
Fetching every u32 from memory leads to a big overhead. So let's fetch all of them as a block if possible.
This reduces the Memory::* calls by the dma_pusher by a factor of 10.
A fairly trivial change. Other sections of the codebase use nested
namespaces instead of separate namespaces here. This one must have just
been overlooked.
Gets rid of the largest set of mutable global state within the core.
This also paves a way for eliminating usages of GetInstance() on the
System class as a follow-up.
Note that no behavioral changes have been made, and this simply extracts
the functionality into a class. This also has the benefit of making
dependencies on the core timing functionality explicit within the
relevant interfaces.
Previously, we were completely ignoring for screenshots whether the game uses RGB or sRGB.
This resulted in screenshot colors that looked off for some titles.
There are some potential edge cases where gl_state may fail to track the
state if a related state changes while the toggle is disabled or it
didn't change. This addresses that.
Handles a pool of resources protected by fences. Manages resource
overflow allocating more resources.
This class is intended to be used through inheritance.
Fences take ownership of objects, protecting them from GPU-side or
driver-side concurrent access. They must be commited from the resource
manager. Their usage flow is: commit the fence from the resource
manager, protect resources with it and use them, send the fence to an
execution queue and Wait for it if needed and then call Release. Used
resources will automatically be signaled when they are free to be
reused.
Makes it consistent with the regular standard containers in terms of
size representation. This also gets rid of dependence on our own
type aliases, removing the need for an include.
The necessity of this parameter is dubious at best, and in 2019 probably
offers completely negligible savings as opposed to just leaving this
enabled. This removes it and simplifies the overall interface.
VKDevice contains all the data required to manage and initialize a
physical device. Its intention is to be passed across Vulkan objects to
query device-specific data (for example the logical device and the
dispatch loader).
We already store a reference to the system instance that the renderer is
created with, so we don't need to refer to the system instance via
Core::System::GetInstance()
This file is intended to be included instead of vulkan/vulkan.hpp. It
includes declarations of unique handlers using a dynamic dispatcher
instead of a static one (which would require linking to a Vulkan
library).
Places all of the timing-related functionality under the existing Core
namespace to keep things consistent, rather than having the timing
utilities sitting in its own completely separate namespace.
When I originally added the compute assert I used the wrong
documentation. This addresses that.
The dispatch register was tested with homebrew against hardware and is
triggered by some games (e.g. Super Mario Odyssey). What exactly is
missing to get a valid program bound by this engine requires more
investigation.
This was originally included because texture operations returned a vec4.
These operations now return a single float and the F4 prefix doesn't
mean anything.
Previous code relied on GLSL parameter order (something that's always
ill-formed on an IR design). This approach passes spatial coordiantes
through operation nodes and array and depth compare values in the the
texture metadata. It still contains an "extra" vector containing generic
nodes for bias and component index (for example) which is still a bit
ill-formed but it should be better than the previous approach.
i965 (and probably all mesa drivers) require GL_PROGRAM_SEPARABLE when using
glProgramBinary. This is probably required by the standard but it's ignored by
permisive proprietary drivers.
This commit it automatically generated by command in zsh:
sed -i -- 's/BitField<\(.*\)_le>/BitField<\1>/g' **/*(D.)
BitField is now aware to endianness and default to little endian. It expects a value representation type without storage specification for its template parameter.
This is compromise for swap type being used in union. A union has deleted default constructor if it has at least one variant member with non-trivial default constructor, and no variant member of T has a default member initializer. In the use case of Bitfield, all variant members will be the swap type on endianness mismatch, which would all have non-trivial default constructor if default value is specified, and non of them can have member initializer
Converts many of the Find* functions to return a std::optional<T> as
opposed to returning the raw return values directly. This allows
removing a few assertions and handles error cases like the service
itself does.
Some games search conditionally use global memory instructions. This
allows the heuristic to search inside conditional nodes for the source
constant buffer.
Some games call LDG at the top of a basic block, making the tracking
heuristic to fail. This commit lets the heuristic the decoded nodes as a
whole instead of per basic blocks.
This may lead to some false positives but allows it the heuristic to
track cases it previously couldn't.
A holdover from citra, the Horizon kernel on the switch has no
prominent kernel object that functions as a timer. At least not
to the degree of sophistication that this class provided.
As such, this can be removed entirely. This class also wasn't used at
all in any meaningful way within the core, so this was just code sitting
around doing nothing. This also allows removing a few things from the
main KernelCore class that allows it to use slightly less resources
overall (though very minor and not anything really noticeable).
No inheritors of the WaitObject class actually make use of their own
implementations of these functions, so they can be made non-virtual.
It's also kind of sketchy to allow overriding how the threads get added
to the list anyways, given the kernel itself on the actual hardware
doesn't seem to customize based off this.
This was previously causing:
warning C4828: The file contains a character starting at offset 0xa33
that is illegal in the current source character set (codepage 65001).
warnings on Windows when compiling yuzu.
This functions almost identically to DecodeInterleavedWithPerfOld,
however this function also has the ability to reset the decoder context.
This is documented as a potentially desirable thing in the libopus
manual in some circumstances as it says for the OPUS_RESET_STATE ctl:
"This should be called when switching streams in order to prevent the
back to back decoding from giving different result from one at a time
decoding."
Constant buffer values on the shader IR were using different offsets if
the access direct or indirect. cbuf34 has a non-multiplied offset while
cbuf36 does. On shader decoding this commit multiplies it by four on
cbuf34 queries.
* Implemented the puller semaphore operations.
* Nit: Fix 2 style issues
* Nit: Add Break to default case.
* Fix style.
* Update for comments. Added ReferenceCount method
* Forgot to remove GpuSmaphoreAddress union.
* Fix the clang-format issues.
* More clang formatting.
* two more white spaces for the Clang formatting.
* Move puller members into the regs union
* Updated to use Memory::WriteBlock instead of Memory::Write*
* Fix clang style issues
* White space clang error
* Removing unused funcitons and other pr comment
* Removing unused funcitons and other pr comment
* More union magic for setting regs value.
* union magic refcnt as well
* Remove local var
* Set up the regs and regs_assert_positions up properly
* Fix clang error
Cubemaps are considered layered and to create a texture view the texture
mustn't be a layered texture, resulting in cubemaps being bound as
cubemap arrays. To fix this issue this commit introduces an extra
surface parameter called "is_array" and uses this to query for texture
view creation.
Now that texture views for cubemaps are actually being created, this
also fixes the number of layers created for the texture view (since they
have to be 6 to create a texture view of cubemaps).
Some games (like Xenoblade Chronicles 2) clear both depth and stencil
buffers while there's a depth-only texture attached (e.g. D16 Unorm).
This commit reads the zeta format of the bound surface on
ConfigureFramebuffers and returns if depth and/or stencil attachments
were set. This is ignored on DrawArrays but on Clear it's used to just
clear those attachments, bypassing an OpenGL error.
In addition to the default, external, EDID, and internal displays,
there's also a null display provided as well, which as the name
suggests, does nothing but discard all commands given to it. This is
provided for completeness.
Opening a display isn't really a thing to warn about. It's an expected
thing, so this can be a debug log. This also alters the string to
indicate the display name better.
Opening "Default" display reads a little nicer compared to Opening
display Default.
This quite literally functions as a basic setter. No other error
checking or anything (since there's nothing to really check against).
With this, it completes the pm:bm interface in terms of functionality.
This appears to be a vestigial API function that's only kept around for
compatibility's sake, given the function only returns a success error
code and exits.
Since that's the case, we can remove the stubbed notification from the
log, since doing nothing is technically the correct behavior in this
case.
std::moveing a local variable in a return statement has the potential to
prevent copy elision from occurring, so this can just be converted into
a regular return.
Looking into the implementation of the C++ standard facilities that seem
to be within all modules, it appears that they use 7 as a break reason
to indicate an uncaught C++ exception.
This was primarily found via the third last function called within
Horizon's equivalent of libcxxabi's demangling_terminate_handler(),
which passes the value 0x80000007 to svcBreak.
According to documentation, if the argument of std::exp is zero, one is returned.
However we want the return value to be also zero in this case so no audio is played.
Commercial games assume that this value is 1 but they never set it. On
the other hand nouveau manually sets this register. On
ConfigureFramebuffers we were asserting for what we are actually
implementing (according to envytools).
With shader caches on the horizon, one requirement is to provide visible
feedback for the progress. The shader cache reportedly takes several
minutes to load for large caches that were invalidated, and as such we
should provide a loading screen with progress.
Adds a loading screen widget that will be shown until the first frame of
the game is swapped. This was chosen in case shader caches are not being
used, several games still take more than a few seconds to launch and
could benefit from a loading screen.
This is a function that definitely doesn't always have a non-modifying
behavior across all implementations, so this should be made non-const.
This gets rid of the need to mark data members as mutable to work around
the fact mutating data members needs to occur.
