The update descriptor is used to store in flat memory a large chunk of
staging data used to update descriptor sets through templates. It
provides a push interface to easily insert descriptors following the
current pipeline. The order used in the descriptor update template has
to be implicitly followed. We can catch bugs here using validation
layers.
The stream buffer before this commit once it was full (no more bytes to
write before looping) waiting for all previous operations to finish.
This was a temporary solution and had a noticeable performance penalty
in performance (from what a profiler showed).
To avoid this mark with fences usages of the stream buffer and once it
loops wait for them to be signaled. On average this will never wait.
Each fence knows where its usage finishes, resulting in a non-paged
stream buffer.
On the other side, the buffer cache is reimplemented using the generic
buffer cache. It makes use of the staging buffer pool and the new
stream buffer.
* Allocate memory in discrete exponentially increasing chunks until the
128 MiB threshold. Allocations larger thant that increase linearly by
256 MiB (depending on the required size). This allows to use small
allocations for small resources.
* Move memory maps to a RAII abstraction. To optimize for debugging
tools (like RenderDoc) users will map/unmap on usage. If this ever
becomes a noticeable overhead (from my profiling it doesn't) we can
transparently move to persistent memory maps without harming the API,
getting optimal performance for both gameplay and debugging.
* Improve messages on exceptional situations.
* Fix typos "requeriments" -> "requirements".
* Small style changes.
Create a large descriptor pool where we allocate all our descriptors
from. It has to be wide enough to support any pipeline, hence its large
numbers.
If the descritor pool is filled, we allocate more memory at that moment.
This way we can take advantage of permissive drivers like Nvidia's that
allocate more descriptors than what the spec requires.
This commit introduces a mechanism by which shader IR code can be
amended and extended. This useful for track algorithms where certain
information can derived from before the track such as indexes to array
samplers.
This function is called rarely and blocks quite often for a long time.
So don't waste power and let the CPU sleep.
This might also increase the performance as the other cores might be allowed to clock higher.
The job of this abstraction is to provide staging buffers for temporary
operations. Think of image uploads or buffer uploads to device memory.
It automatically deletes unused buffers.
This object's job is to contain an image and manage its transitions.
Since Nvidia hardware doesn't know what a transition is but Vulkan
requires them anyway, we have to state track image subresources
individually.
To avoid the overhead of tracking each subresource in images with many
subresources (think of cubemap arrays with several mipmaps), this commit
tracks when subresources have diverged. As long as this doesn't happen
we can check the state of the first subresource (that will be shared
with all subresources) and update accordingly.
Image transitions are deferred to the scheduler command buffer.
The intention behind this hasheable structure is to describe the state
of fixed function pipeline state that gets compiled to a single graphics
pipeline state object. This is all dynamic state in OpenGL but Vulkan
wants it in an immutable state, even if hardware can edit it freely.
In this commit the structure is defined in an optimized state (it uses
booleans, has paddings and many data entries that can be packed to
single integers). This is intentional as an initial implementation that
is easier to debug, implement and review. It will be optimized in later
stages, or it might change if Vulkan gets more dynamic states.
This commit adds a series of HLE methods for handling 3D textures in
general. This helps games that generate 3D textures on every frame and
may reduce loading times for certain games.
Remove false commentary. Not dividing by 4 the size of shared memory is
not a hack; it describes the number of integers, not bytes.
While we are at it sort the generated code to put preprocessor lines on
the top.
ExprCondCode visit implements the generic Visit. Use this instead of
that one.
As an intended side effect this fixes unwritten memory usages in cases
when a negation of a condition code is used.
This allows us to put VKFenceWatch inside a std::vector without storing
it in heap. On move we have to signal the fences where the new protected
resource is, adding some overhead.
VK_NV_device_diagnostic_checkpoints allows us to push data to a Vulkan
queue and then query it even after a device loss. This allows us to push
the current pipeline object and see what was the call that killed the
device.
Some games write from fragment shaders to an unexistant framebuffer
attachment or they don't write to one when it exists in the framebuffer.
Fix this by skipping writes or adding zeroes.
RASTERIZE_ENABLE is the opposite of GL_RASTERIZER_DISCARD. Implement it
naturally using this.
NVN games expect rasterize to be enabled by default, reflect that in our
initial GPU state.
LDG can load single bytes instead of full integers or packs of integers.
These have the advantage of loading bytes that are not aligned to 4
bytes.
To emulate these this commit gets the byte being referenced (by doing
"address & 3" and then using that to extract the byte from the loaded
integer:
result = bitfieldExtract(loaded_integer, (address % 4) * 8, 8)
I2F's byte selector is used to choose what bytes to convert to float.
e.g. if the input is 0xaabbccdd and the selector is ".B3" it will
convert 0xaa. The default (when it's not shown in nvdisasm) is ".B0", in
that example the default would convert 0xdd to float.
When a image format mismatches we were inserting zeroes to the texture
itself. This was not handling cases were the mismatch uses less
coordinates than the guest shader code. Address that by resizing the
vector.
These shaders are used to specify code that is not dynamically generated
in the Vulkan backend. Instead of packing it inside the build system,
it's manually built and copied to the C++ file to avoid adding
unnecessary build time dependencies.
quad_array should be dropped in the future since it can be emulated with
a memory pool generated from the CPU.
Add an extra argument to query device capabilities in the future. The
intention behind this is to use native quads, quad strips, line loops
and polygons if these are released for Vulkan.
The OpenGL spec defines GL_CLAMP's formula similarly to CLAMP_TO_EDGE
and CLAMP_TO_BORDER depending on the filter mode used. It doesn't
exactly behave like this, but it's the closest we can get with what
Vulkan offers without emulating it by injecting shader code.
Introduce a worker thread approach for delegating Vulkan work derived
from dxvk's approach. https://github.com/doitsujin/dxvk
Now that the scheduler is what handles all Vulkan work related to
command streaming, store state tracking in itself. This way we can know
when to reupload Vulkan dynamic state to the queue (since this one is
invalidated between command buffers unlike NVN). We can also store the
renderpass state and graphics pipeline bound to avoid redundant binds
and renderpass begins/ends.
Previously we naively checked for "Intel" in GL_VENDOR, but this
includes both Intel's proprietary driver and the mesa driver. Re-enable
compute shaders for mesa.
Add missing new-line. This caused shaders using local memory and shared
memory to inject a preprocessor GLSL line after an expression (resulting
in invalid code).
It looked like this:
shared uint smem[8];#define LOCAL_MEMORY_SIZE 16
It should look like this (addressed by this commit):
shared uint smem[8];
\#define LOCAL_MEMORY_SIZE 16
Update Sirit and its usage in vk_shader_decompiler. Highlights:
- Implement tessellation shaders
- Implement geometry shaders
- Implement some missing features
- Use native half float instructions when available.
- Setup more features and requirements.
- Improve logging for missing features.
- Collect telemetry parameters.
- Add queries for more image formats.
- Query push constants limits.
- Optionally enable some extensions.
Some texture views were being created out of bounds (with more layers or
mipmaps than what the original texture has). This is because of a
miscalculation in mipmap bounding. end_layer and end_mipmap are out of
bounds (e.g. layer 6 in a cubemap), there's no need to add one more
there.
Fixes OpenGL errors and Vulkan crashes on Splatoon 2.
Pack color attachment enumerations into a single u32. To determine the
number of buffers, the highest color attachment with a shared pointer
that doesn't point to null is used.
Amends a few interfaces to be able to handle the migration over to the
new Memory class by passing the class by reference as a function
parameter where necessary.
Notably, within the filesystem services, this eliminates two ReadBlock()
calls by using the helper functions of HLERequestContext to do that for
us.
The heuristic to detect AMD's driver was not working properly since it
also included Intel. Instead of using heuristics to detect it, compare
the GL_VENDOR string.
SSBOs and other resources are limited per pipeline on Intel and AMD.
Heuristically reserve resources per stage having in mind the reported
OpenGL limits.
The current shared memory size seems to be smaller than what the game
actually uses. This makes Nvidia's driver consistently blow up; in the
case of FE3H it made it explode on Qt's SwapBuffers while SDL2 worked
just fine. For now keep this hack since it's still progress over the
previous hardcoded shared memory size.
Drop the usage of ARB_compute_variable_group_size and specialize compute
shaders instead. This permits compute to run on AMD and Intel
proprietary drivers.
Some games like "Fire Emblem: Three Houses" bind 2D textures to offsets
used by instructions of 1D textures. To handle the discrepancy this
commit uses the the texture type from the binding and modifies the
emitted code IR to build a valid backend expression.
E.g.: Bound texture is 2D and instruction is 1D, the emitted IR samples
a 2D texture in the coordinate ivec2(X, 0).
This commit aims to redo the full setup of invalid textures and
guarantee correct behavior across backends in the case of finding one by
using black dummy textures that match the target of the expected
texture.
While DEPBAR is stubbed it doesn't change anything from our end. Shading
languages handle what this instruction does implicitly. We are not
getting anything out fo this log except noise.
Nvidia has sane default output values for varyings, but the other
vendors don't apply these. To properly emulate this we would have to
analyze the shader header. For the time being, apply the same default
Nvidia applies so we get the same behaviour on non-Nvidia drivers.
format_lookup_table: Drop bitfields
format_lookup_table: Use std::array for definition table
format_lookup_table: Include <limits> instead of <numeric>
Use a large flat array to look up texture formats. This allows us to
properly implement formats with different component types. It should
also be faster.
Abstracted ComponentType was not being used in a meaningful way.
This commit drops its usage.
There is one place where it was being used to test compatibility between
two cached surfaces, but this one is implied in the pixel format.
Removing the component type test doesn't change the behaviour.
`boost::make_iterator_range` is available when `boost/range/iterator_range.hpp` is included.
Also include `boost/icl/interval_map.hpp` and `boost/icl/interval_set.hpp`.
Emulates negative y viewports with ARB_clip_control. This allows us to
more easily emulated pipelines with tessellation and/or geometry shader
stages. It also avoids corrupting games with transform feedbacks and
negative viewports (gl_Position.y was being modified).
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Update src/video_core/shader/control_flow.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
Nvidia's OpenGL driver maps gl(Named)BufferSubData with some requirements
to a fast. This path has an extra memcpy but updates the buffer without
orphaning or waiting for previous calls. It can be seen as a better
model for "push constants" that can upload a whole UBO instead of 256
bytes.
This path has some requirements established here:
http://on-demand.gputechconf.com/gtc/2014/presentations/S4379-opengl-44-scene-rendering-techniques.pdf#page=24
Instead of using the stream buffer, this commits moves constant buffers
uploads to calls of glNamedBufferSubData and from my testing it brings a
performance improvement. This is disabled when the vendor is not Nvidia
since it brings performance regressions.
Originally on the last commit I thought TLD4 acted the same as TLD4S and
didn't have a mask. It actually does have a component mask. This commit
corrects that.
This commit fixes an issue where not all 4 results of tld4 were being
written, the color component was defaulted to red, among other things.
It also implements the bindless variant.
Bindless textures were using u64 to pack the buffer and offset from
where they come from. Drop this in favor of separated entries in the
struct.
Remove the usage of std::set in favor of std::list (it's not std::vector
to avoid reference invalidations) for samplers and images.
Greatly shrinks the amount of generated code for GetDecodeTable().
Collapses an assembly output of 9000+ lines down to ~3621 with Clang,
and 6513 down to ~2616 with GCC, given it's now allowed to construct all
the entries as a sequence of constant data.
Given the overall size of the maps are very small, we can use arrays of
pairs here instead of always heap allocating a new map every time the
functions are called. Given the small size of the maps, the difference
in container lookups are negligible, especially given the entries are
already sorted.
