Commit graph

3662 commits

Author SHA1 Message Date
ReinUsesLisp
3142f1b597 fixed_pipeline_state: Add depth clamp 2020-01-06 22:02:26 -03:00
ReinUsesLisp
9c548146ca vk_rasterizer: Add placeholder 2020-01-06 22:02:26 -03:00
bunnei
5be00cba15
Merge pull request #3276 from ReinUsesLisp/pipeline-reqs
vk_update_descriptor/vk_renderpass_cache: Add pipeline cache dependencies
2020-01-06 17:03:34 -05:00
ReinUsesLisp
5aeff9aff5 vk_renderpass_cache: Initial implementation
The renderpass cache is used to avoid creating renderpasses on each
draw. The hashed structure is not currently optimized.
2020-01-06 18:28:32 -03:00
ReinUsesLisp
322d6a0311 vk_update_descriptor: Initial implementation
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.
2020-01-06 18:28:32 -03:00
ReinUsesLisp
5b01f80a12 vk_stream_buffer/vk_buffer_cache: Avoid halting and use generic cache
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.
2020-01-06 18:13:41 -03:00
ReinUsesLisp
ceb851b590 vk_memory_manager: Misc changes
* 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.
2020-01-06 18:13:41 -03:00
ReinUsesLisp
85bb6a6f08 vk_buffer_cache: Temporarily remove buffer cache
This is intended for a follow up commit to avoid circular dependencies.
2020-01-06 17:58:46 -03:00
bunnei
89fc75d769
Merge pull request #3257 from degasus/no_busy_loops
video_core: Block in WaitFence.
2020-01-06 00:09:57 -05:00
Fernando Sahmkow
56e450a3f7
Merge pull request #3264 from ReinUsesLisp/vk-descriptor-pool
vk_descriptor_pool: Initial implementation
2020-01-05 15:54:41 -04:00
bunnei
cd0a7dfdbc
Merge pull request #3258 from FernandoS27/shader-amend
Shader_IR: add the ability to amend code in the shader ir.
2020-01-04 14:05:17 -05:00
Fernando Sahmkow
3dd6b55851 Shader_IR: Address Feedback 2020-01-04 14:40:57 -04:00
Rodrigo Locatti
6e347d8d1b
Update src/video_core/renderer_vulkan/vk_descriptor_pool.cpp
Co-Authored-By: Mat M. <mathew1800@gmail.com>
2020-01-03 17:34:30 -03:00
ReinUsesLisp
0d6d8129c4 yuzu: Remove Maxwell debugger
This was carried from Citra and wasn't really used on yuzu. It also adds
some runtime overhead. This commit removes it from yuzu's codebase.
2020-01-02 23:09:44 -03:00
bunnei
ae0e481677
Merge pull request #3243 from ReinUsesLisp/topologies
maxwell_to_gl: Implement missing primitive topologies
2020-01-01 20:33:33 -05:00
ReinUsesLisp
1fe7df4517 vk_descriptor_pool: Initial implementation
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.
2020-01-01 16:44:06 -03:00
bunnei
028b2718ed
Merge pull request #3239 from ReinUsesLisp/p2r
shader/p2r: Implement P2R Pr
2019-12-31 20:37:16 -05:00
Fernando Sahmkow
b3371ed09e Shader_IR: add the ability to amend code in the shader ir.
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.
2019-12-30 15:31:48 -04:00
Fernando Sahmkow
7bd447355f
Merge pull request #3248 from ReinUsesLisp/vk-image
vk_image: Add an image object abstraction
2019-12-30 14:25:14 -04:00
Rodrigo Locatti
4cbb363d3f
vk_image: Avoid unnecesary equals 2019-12-30 13:28:23 -03:00
Fernando Sahmkow
287d5921cf
Merge pull request #3249 from ReinUsesLisp/vk-staging-buffer-pool
vk_staging_buffer_pool: Add a staging pool for temporary operations
2019-12-30 12:25:59 -04:00
Markus Wick
cb9dd01ffd video_core: Block in WaitFence.
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.
2019-12-30 13:04:53 +01:00
Rodrigo Locatti
f2c61bbe13
vk_staging_buffer_pool: Initialize last epoch to zero 2019-12-29 19:19:43 -03:00
Fernando Sahmkow
f846e3d6d0
Merge pull request #3250 from ReinUsesLisp/empty-fragment
gl_rasterizer: Allow rendering without fragment shader
2019-12-28 14:33:53 -04:00
bunnei
8a76f816a4
Merge pull request #3228 from ReinUsesLisp/ptp
shader/texture: Implement AOFFI and PTP for TLD4 and TLD4S
2019-12-26 21:43:44 -05:00
ReinUsesLisp
5b989f189f
gl_rasterizer: Allow rendering without fragment shader
Rendering without a fragment shader is usually used in depth-only
passes.
2019-12-26 16:38:49 -03:00
ReinUsesLisp
3813af2f3c
vk_staging_buffer_pool: Add a staging pool for temporary operations
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.
2019-12-25 18:12:17 -03:00
ReinUsesLisp
c83bf7cd1e
vk_image: Add an image object abstraction
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.
2019-12-25 18:00:16 -03:00
Fernando Sahmkow
5619d24377
Merge pull request #3244 from ReinUsesLisp/vk-fps
fixed_pipeline_state: Define structure and loaders
2019-12-25 14:31:29 -04:00
bunnei
4af569ee47
Merge pull request #3236 from ReinUsesLisp/rasterize-enable
gl_rasterizer: Implement RASTERIZE_ENABLE
2019-12-24 22:54:10 -05:00
ReinUsesLisp
b9e3f5eb36
fixed_pipeline_state: Define symetric operator!= and mark as noexcept
Marks as noexcept Hash, operator== and operator!= for consistency.
2019-12-24 18:24:08 -03:00
ReinUsesLisp
4a3026b16b
fixed_pipeline_state: Define structure and loaders
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.
2019-12-22 22:59:11 -03:00
ReinUsesLisp
5770418fb3
maxwell_3d: Add depth bounds registers 2019-12-22 22:55:06 -03:00
ReinUsesLisp
91d35559e5
maxwell_to_gl: Implement missing primitive topologies
Many of these topologies are exclusively available in OpenGL.
2019-12-22 22:33:01 -03:00
bunnei
e976d0e924
Merge pull request #3241 from ReinUsesLisp/gl-shader-cache
gl_shader_cache: Style changes
2019-12-22 16:23:46 -05:00
bunnei
1e76655f83
Merge pull request #3238 from ReinUsesLisp/vk-resource-manager
vk_resource_manager: Catch device losses and other changes
2019-12-22 15:57:16 -05:00
bunnei
0f3ac9cfeb
Merge pull request #3203 from FernandoS27/tex-cache-fixes
Texture Cache: Add HLE methods for building 3D textures
2019-12-22 14:25:13 -05:00
Fernando Sahmkow
3dc585d011
Merge pull request #3237 from ReinUsesLisp/vk-shader-decompiler
vk_shader_decompiler: Misc changes
2019-12-22 12:36:56 -04:00
Fernando Sahmkow
218ee18417 Texture Cache: Improve documentation 2019-12-22 12:29:23 -04:00
Fernando Sahmkow
a3916588b6 Texture Cache: Address Feedback 2019-12-22 12:24:34 -04:00
Fernando Sahmkow
51c9e98677 Texture Cache: Add HLE methods for building 3D textures within the GPU in certain scenarios.
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.
2019-12-22 12:24:34 -04:00
Fernando Sahmkow
aea978e037
Merge pull request #3230 from ReinUsesLisp/vk-emu-shaders
renderer_vulkan/shader: Add helper GLSL shaders
2019-12-22 11:23:09 -04:00
Fernando Sahmkow
27efcc15e9
Merge pull request #3240 from ReinUsesLisp/decomp-cond-code
vk_shader_decompiler: Use Visit instead of reimplementing it
2019-12-22 11:20:55 -04:00
bunnei
16dcfacbfc
Merge pull request #3235 from ReinUsesLisp/ldg-u8
shader/memory: Implement LDG.U8 and unaligned U8 loads
2019-12-21 22:50:28 -05:00
ReinUsesLisp
1e16023d60
gl_shader_cache: Update commentary for shared memory
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.
2019-12-20 22:51:21 -03:00
ReinUsesLisp
486c6a5316
gl_shader_cache: Remove unused entry in GetPrimitiveDescription 2019-12-20 22:49:30 -03:00
ReinUsesLisp
af93909c9c
vk_shader_decompiler: Use Visit instead of reimplementing it
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.
2019-12-20 21:36:25 -03:00
ReinUsesLisp
38d3a48873
shader/p2r: Implement P2R Pr
P2R dumps predicate or condition codes state to a register. This is
useful for unit testing.
2019-12-20 18:02:41 -03:00
ReinUsesLisp
cf27b59493
shader/r2p: Refactor P2R to support P2R 2019-12-20 17:55:42 -03:00
bunnei
7be65c6a68
Merge pull request #3234 from ReinUsesLisp/i2f-u8-selector
shader/conversion: Implement byte selector in I2F
2019-12-19 22:36:26 -05:00
bunnei
6d55b14cc0
Merge pull request #3233 from ReinUsesLisp/mismatch-sizes
shader/texture: Properly shrink unused entries in size mismatches
2019-12-19 20:40:27 -05:00
ReinUsesLisp
e41da22c8d
vk_resource_manager: Add entry to VKFence to test its usage 2019-12-19 16:31:34 -03:00
ReinUsesLisp
ec983a2451
vk_reosurce_manager: Add assert for releasing fences
Notify the programmer when a request to release a fence is invalid
because the fence is already free.
2019-12-19 16:31:34 -03:00
ReinUsesLisp
6ddffa010a
vk_resource_manager: Implement VKFenceWatch move constructor
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.
2019-12-19 16:31:34 -03:00
ReinUsesLisp
54747d60bc
vk_device: Add entry to catch device losses
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.
2019-12-19 16:31:33 -03:00
ReinUsesLisp
2a63b3bdb9
vk_shader_decompiler: Fix full decompilation
When full decompilation was enabled, labels were not being inserted and
instructions were misused. Fix these bugs.
2019-12-19 16:24:45 -03:00
ReinUsesLisp
de918ebeb0
vk_shader_decompiler: Skip NDC correction when it is native
Avoid changing gl_Position when the NDC used by the game is [0, 1]
(Vulkan's native).
2019-12-19 16:24:45 -03:00
ReinUsesLisp
485c21eac3
vk_shader_decompiler: Normalize output fragment attachments
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.
2019-12-19 16:24:45 -03:00
bunnei
1eb4a95d2b
Merge pull request #3232 from ReinUsesLisp/gl-decompiler-images
gl_shader_decompiler: Add missing DeclareImages
2019-12-19 11:32:47 -05:00
bunnei
253aa52351
Merge pull request #3231 from ReinUsesLisp/tld4s-encoding
shader_bytecode: Fix TLD4S encoding
2019-12-19 11:32:25 -05:00
ReinUsesLisp
f4a25f854c
vk_device: Add query for RGBA8Uint 2019-12-19 02:08:29 -03:00
ReinUsesLisp
abb33d4aec
vk_shader_decompiler: Update sirit and implement Texture AOFFI 2019-12-19 01:42:13 -03:00
bunnei
d53cf05513
Merge pull request #3221 from ReinUsesLisp/vk-scheduler
vk_scheduler: Delegate commands to a worker thread and state track
2019-12-18 22:04:08 -05:00
ReinUsesLisp
da0aa4da6b
gl_rasterizer: Implement RASTERIZE_ENABLE
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.
2019-12-18 19:28:23 -03:00
ReinUsesLisp
ae8d4b6c0c
shader/memory: Implement LDG.U8 and unaligned U8 loads
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)
2019-12-18 01:21:46 -03:00
ReinUsesLisp
a7d6bd1ef1
shader/conversion: Implement byte selector in I2F
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.
2019-12-18 00:41:22 -03:00
ReinUsesLisp
15a753b9a5
shader/texture: Properly shrink unused entries in size mismatches
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.
2019-12-17 23:38:10 -03:00
ReinUsesLisp
e438079b50
gl_shader_decompiler: Add missing DeclareImages 2019-12-17 23:34:15 -03:00
ReinUsesLisp
8b26b4228b
shader_bytecode: Fix TLD4S encoding 2019-12-17 23:32:10 -03:00
ReinUsesLisp
b52297767e
renderer_vulkan/shader: Add helper GLSL shaders
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.
2019-12-16 17:59:08 -03:00
bunnei
65b1b05e05
Merge pull request #3182 from ReinUsesLisp/renderer-opengl
renderer_opengl: Miscellaneous clean ups
2019-12-16 13:01:04 -05:00
ReinUsesLisp
e09c1fbc1f
shader/texture: Implement TLD4.PTP 2019-12-16 04:09:24 -03:00
ReinUsesLisp
844e4a297b
shader/texture: Enable arrayed TLD4 2019-12-16 02:37:21 -03:00
ReinUsesLisp
a87c85eba2
gl_shader_decompiler: Rename "sepparate" to "separate" 2019-12-16 02:12:51 -03:00
ReinUsesLisp
3d2c44848b
shader/texture: Implement AOFFI for TLD4S 2019-12-16 02:06:42 -03:00
ReinUsesLisp
3d9fff82c0
shader/texture: Remove unnecesary parenthesis 2019-12-16 01:52:33 -03:00
Rodrigo Locatti
eac075692b
Merge pull request #3219 from FernandoS27/fix-bindless
Corrections and fixes to TLD4S & bindless samplers failing
2019-12-16 01:26:11 -03:00
bunnei
3d51153611
Merge pull request #3222 from ReinUsesLisp/maxwell-to-vk
maxwell_to_vk: Use VK_EXT_index_type_uint8 and misc changes
2019-12-14 22:30:12 -05:00
bunnei
035ec7d9de
Merge pull request #3213 from ReinUsesLisp/intel-mesa
gl_device: Enable compute shaders for Intel Mesa drivers
2019-12-14 16:04:31 -05:00
bunnei
2b650543c6
Merge pull request #3212 from ReinUsesLisp/fix-smem-lmem
gl_shader_cache: Add missing new-line on emitted GLSL
2019-12-13 21:35:29 -05:00
ReinUsesLisp
e3ea583893
maxwell_to_vk: Improve image format table and add more formats
A1B5G5R5 uses A1R5G5B5. This is flipped with image view swizzles;
flushing is still not properly implemented on Vulkan for this particular
format.
2019-12-13 03:12:29 -03:00
ReinUsesLisp
f27b21077d
maxwell_to_vk: Implement more vertex formats 2019-12-13 03:12:28 -03:00
ReinUsesLisp
8db8631d81
maxwell_to_vk: Implement more primitive topologies
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.
2019-12-13 03:12:28 -03:00
ReinUsesLisp
15513f0801
maxwell_to_vk: Approach GL_CLAMP closer to the GL spec
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.
2019-12-13 03:12:28 -03:00
ReinUsesLisp
f845df8651
maxwell_to_vk: Use VK_EXT_index_type_uint8 when available 2019-12-13 02:37:23 -03:00
ReinUsesLisp
2df9a2dcaf
vk_scheduler: Delegate commands to a worker thread and state track
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.
2019-12-13 02:24:48 -03:00
bunnei
8fc49a83b6
Merge pull request #3217 from jhol/fix-boost-include
Added missing include
2019-12-11 22:21:24 -05:00
Fernando Sahmkow
c0ee0aa1a8 Shader_IR: Correct TLD4S Depth Compare. 2019-12-11 19:53:17 -04:00
Fernando Sahmkow
af89723fa3 Shader_Ir: Correct TLD4S encoding and implement f16 flag. 2019-12-11 19:53:17 -04:00
Fernando Sahmkow
84a158c977 Gl_Shader_compiler: Correct Depth Compare for Texture Gather operations. 2019-12-11 19:53:16 -04:00
Fernando Sahmkow
271a3264f3 Shader_Ir: default failed tracks on bindless samplers to null values. 2019-12-11 19:53:16 -04:00
Fernando Sahmkow
1d2ba3cc97 Gl_Rasterizer: Skip Tesselation Control and Eval stages as they are un implemented.
This commit ensures the OGL backend does not execute tesselation shader 
stages as they are currently unimplemented.
2019-12-11 15:41:26 -04:00
bunnei
1a66cde175
Merge pull request #3210 from ReinUsesLisp/memory-barrier
shader: Implement MEMBAR.GL
2019-12-11 14:24:39 -05:00
Joel Holdsworth
e9faa1617c Added missing include 2019-12-11 18:11:49 +00:00
ReinUsesLisp
f564eaebed
gl_device: Enable compute shaders for Intel Mesa drivers
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.
2019-12-11 00:00:30 -03:00
ReinUsesLisp
48e16c4c49
gl_shader_cache: Add missing new-line on emitted GLSL
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
2019-12-10 23:52:51 -03:00
Fernando Sahmkow
7ffb672f61 Maxwell3D: Implement Depth Mode.
This commit finishes adding depth mode that was reverted before due to
other unresolved issues.
2019-12-10 19:51:46 -04:00
ReinUsesLisp
425a254fa2
shader: Implement MEMBAR.GL
Implement using memoryBarrier in GLSL and OpMemoryBarrier on SPIR-V.
2019-12-10 16:45:03 -03:00
ReinUsesLisp
233ed96a5c
vk_shader_decompiler: Fix build issues on old gcc versions 2019-12-10 01:55:38 -03:00
ReinUsesLisp
d30cf51d7d
vk_shader_decompiler: Reduce YNegate's severity 2019-12-09 23:52:28 -03:00