Commit graph

648 commits

Author SHA1 Message Date
ameerj
e0977af861 astc_decoder: Refactor for style and more efficient memory use 2021-03-25 16:53:51 -04:00
Rodrigo Locatti
bca641cf8a astc_decoder: Reimplement Layers
Reimplements the approach to decoding layers in the compute shader. Fixes multilayer astc decoding when using Vulkan.
2021-03-13 12:16:03 -05:00
ameerj
cac341dbc7 renderer_vulkan: Accelerate ASTC decoding
Co-Authored-By: Rodrigo Locatti <reinuseslisp@airmail.cc>
2021-03-13 12:16:03 -05:00
Rodrigo Locatti
1f0cb72f40 Merge pull request #5891 from ameerj/bgra-ogl
renderer_opengl: Use compute shaders to swizzle BGR textures on copy
2021-03-09 02:47:51 -03:00
ameerj
fb79ff4cab texture_cache: Blacklist BGRA8 copies and views on OpenGL
In order to force the BGRA8 conversion on Nvidia using OpenGL, we need to forbid texture copies and views with other formats.

This commit also adds a boolean relating to this, as this needs to be done only for the OpenGL api, Vulkan must remain unchanged.
2021-03-04 14:14:49 -05:00
ReinUsesLisp
da2876ad7b vk_command_pool: Reduce the command pool size from 4096 to 4
This allows drivers to reuse memory more easily and preallocate less.
The optimal number has been measured booting Pokémon Sword.
2021-02-23 19:08:24 -03:00
bunnei
adc9097952 Merge pull request #5936 from Kelebek1/Offsets
Offsets for TexelFetch and TextureGather in Vulkan
2021-02-21 21:23:45 -07:00
bunnei
c9770f92d8 Merge pull request #5924 from ReinUsesLisp/inline-bindings
vk_update_descriptor: Inline and improve code for binding buffers
2021-02-19 12:27:10 -08:00
ReinUsesLisp
76e2d40963 vk_rasterizer: Fix loading shader addresses twice
This was recently introduced on a wrongly rebased commit.
2021-02-15 21:34:13 -03:00
bunnei
0b63701ebf Merge pull request #5923 from ReinUsesLisp/vk-dirty-pipeline
fixed_pipeline_cache: Use dirty flags to lazily update key
2021-02-15 13:17:27 -08:00
Kelebek1
16a5c56b7c Review 1 2021-02-15 05:26:28 +00:00
Kelebek1
4e04e95a8e Implement texture offset support for TexelFetch and TextureGather and add offsets for Tlds
Formatting
2021-02-15 00:36:37 +00:00
ReinUsesLisp
ec1854363e vk_resource_pool: Load GPU tick once and compare with it
Other minor style improvements. Rename free_iterator to hint_iterator,
to describe better what it does.
2021-02-13 17:53:58 -03:00
ReinUsesLisp
7fa30ea272 vk_update_descriptor: Inline and improve code for binding buffers
Allow compilers with our settings inline hot code.
2021-02-13 17:46:24 -03:00
ReinUsesLisp
261380d2b6 fixed_pipeline_cache: Use dirty flags to lazily update key
Use dirty flags to avoid building pipeline key from scratch on each draw
call. This saves a bit of unnecesary work on each draw call.
2021-02-13 17:44:47 -03:00
ReinUsesLisp
898de871a9 vk_master_semaphore: Mark gpu_tick atomic operations with relaxed order 2021-02-13 05:57:28 -03:00
ReinUsesLisp
6f5d45aecc vk_staging_buffer_pool: Inline tick tests
Load the current tick to a local variable, moving it out of an atomic
and allowing us to compare the value without going through a pointer
each time. This should make the loop more optimizable.
2021-02-13 05:14:11 -03:00
ReinUsesLisp
2f40ef90c5 gl_stream_buffer/vk_staging_buffer_pool: Fix size check
Fix a tragic off-by-one condition that causes Vulkan's stream buffer to
think it's always full, using fallback memory. The OpenGL was also
affected by this bug to a lesser extent.
2021-02-13 05:11:48 -03:00
ReinUsesLisp
3f190b946c video_core: Fix clang build issues 2021-02-13 02:26:47 -03:00
ReinUsesLisp
50d8c1eb35 vk_staging_buffer_pool: Fix softlock when stream buffer overflows
There was still a code path that could wait on a timeline semaphore tick
that would never be signalled.

While we are at it, make use of more STL algorithms.
2021-02-13 02:18:38 -03:00
ReinUsesLisp
0cc70777ca vk_buffer_cache: Add support for null index buffers
Games can bind a null index buffer (size=0) where all indices are
evaluated as zero. VK_EXT_robustness2 doesn't support this and all
drivers segfault when a null index buffer is passed to
vkCmdBindIndexBuffer.

Workaround this by creating a 4 byte buffer and filling it with zeroes.
If it's read out of bounds, robustness takes care of returning zeroes as
indices.
2021-02-13 02:18:38 -03:00
ReinUsesLisp
e83c4bd65d vk_staging_buffer_pool: Get a staging buffer instead of waiting
Avoids waiting idle while the GPU finishes to do work, and fixes an
issue where we'd wait forever if a single command buffer (logic tick)
all the data.
2021-02-13 02:18:05 -03:00
ReinUsesLisp
48a391810c buffer_cache: Skip cache on small uploads on Vulkan
Ports from OpenGL the optimization to skip small 3D uniform buffer
uploads. This will take advantage of the previously introduced stream
buffer.

Fixes instances where the staging buffer offset was being ignored.
2021-02-13 02:17:24 -03:00
ReinUsesLisp
c210923ccb vk_staging_buffer_pool: Add stream buffer for small uploads
This uses a ring buffer similar to OpenGL's stream buffer for small
uploads. This stops us from allocating several small buffers, reducing
memory fragmentation and cache locality.

It uses dedicated allocations when possible.
2021-02-13 02:17:24 -03:00
ReinUsesLisp
2dfce2fca6 video_core: Reimplement the buffer cache
Reimplement the buffer cache using cached bindings and page level
granularity for modification tracking. This also drops the usage of
shared pointers and virtual functions from the cache.

- Bindings are cached, allowing to skip work when the game changes few
  bits between draws.
- OpenGL Assembly shaders no longer copy when a region has been modified
  from the GPU to emulate constant buffers, instead GL_EXT_memory_object
  is used to alias sub-buffers within the same allocation.
- OpenGL Assembly shaders stream constant buffer data using
  glProgramBufferParametersIuivNV, from NV_parameter_buffer_object. In
  theory this should save one hash table resolve inside the driver
  compared to glBufferSubData.
- A new OpenGL stream buffer is implemented based on fences for drivers
  that are not Nvidia's proprietary, due to their low performance on
  partial glBufferSubData calls synchronized with 3D rendering (that
  some games use a lot).
- Most optimizations are shared between APIs now, allowing Vulkan to
  cache more bindings than before, skipping unnecesarry work.

This commit adds the necessary infrastructure to use Vulkan object from
OpenGL. Overall, it improves performance and fixes some bugs present on
the old cache. There are still some edge cases hit by some games that
harm performance on some vendors, this are planned to be fixed in later
commits.
2021-02-13 02:17:22 -03:00
ReinUsesLisp
70b3c29534 gpu: Report renderer errors with exceptions
Instead of using a two step initialization to report errors, initialize
the GPU renderer and rasterizer on the constructor and report errors
through std::runtime_error.
2021-02-13 02:16:19 -03:00
ReinUsesLisp
4f807afdc4 vk_scheduler: Fix unaligned placement new expressions
We were accidentaly creating an object in an unaligned memory address.
Fix this by manually aligning the offset.
2021-01-27 22:28:22 -03:00
ReinUsesLisp
8778d4a743 vk_graphics_pipeline: Fix narrowing conversion on MSVC 2021-01-24 21:41:29 -03:00
LC
04da2ac057 Merge pull request #5807 from ReinUsesLisp/vc-warnings
video_core: Silence the remaining gcc warnings and enforce them
2021-01-24 17:36:43 -05:00
Rodrigo Locatti
bc244a3231 Merge pull request #5363 from ReinUsesLisp/vk-image-usage
vk_texture_cache: Support image store on sRGB images with VkImageViewUsageCreateInfo
2021-01-24 18:44:51 -03:00
ReinUsesLisp
1d3c309eac vk_texture_cache: Support image store on sRGB images with VkImageViewUsageCreateInfo
Vulkan 1.0 didn't support creating sRGB image views on an ABGR8 VkImage
with storage usage bits. VK_KHR_maintenance2 addressed this allowing to
reduce the usage bits on a VkImageView.

To allow image store on non-sRGB image views when the VkImage is created
with sRGB, always create VkImages without sRGB and add the sRGB format
on the view.
2021-01-24 18:16:43 -03:00
ReinUsesLisp
7afc85ba3a video_core: Silence -Wmissing-field-initializers warnings 2021-01-24 04:32:19 -03:00
ReinUsesLisp
9d882e0000 maxwell_to_vk: Silence -Wextra warnings about using different enum types 2021-01-24 04:03:36 -03:00
ReinUsesLisp
23a5aed188 vk_pipeline_cache: Properly bypass VertexA shaders
The VertexA stage is not yet implemented, but Vulkan is adding its
descriptors, causing a discrepancy in the pushed descriptors and the
template. This generally ends up in a driver side crash.

Bypass the VertexA stage for now.
2021-01-23 03:59:59 -03:00
bunnei
02dfcb4b8f Merge pull request #5361 from ReinUsesLisp/vk-shader-comment
vk_shader_decompiler: Show comments as OpUndef with a type
2021-01-20 21:33:42 -08:00
ReinUsesLisp
3d5910da6d vk_shader_decompiler: Show comments as OpUndef with a type
Silence the new validation layer error about SPIR-V not allowing OpUndef
on a OpTypeVoid, even when the SPIR-V spec doesn't say anything against
it.

They will be inserted as an undefined int to avoid SPIRV-Cross and
validation errors, but only when a debugging tool is attached.
2021-01-15 21:12:57 -03:00
ReinUsesLisp
36597c0b38 vk_texture_cache: Use Download memory types for texture flushes
Use the Download memory type where it matters.
2021-01-15 16:19:40 -03:00
ReinUsesLisp
60bb77340f vulkan_memory_allocator: Add "download" memory usage hint
Allow users of the allocator to hint memory usage for downloads. This
removes the non-descriptive boolean passed for "host visible" or not
host visible memory commits, and uses an enum to hint device local,
upload and download usages.
2021-01-15 16:19:39 -03:00
ReinUsesLisp
fa012cc7e6 vulkan_common: Move allocator to the common directory
Allow using the abstraction from the OpenGL backend.
2021-01-15 16:19:39 -03:00
ReinUsesLisp
bb1cc99e64 renderer_vulkan: Rename Vulkan memory manager to memory allocator
"Memory manager" collides with the guest GPU memory manager, and a
memory allocator sounds closer to what the abstraction aims to be.
2021-01-15 16:19:39 -03:00
ReinUsesLisp
550a9cbcb5 vk_memory_manager: Improve memory manager and its API
Fix a bug where the memory allocator could leave gaps between commits.
To fix this the allocation algorithm was reworked, although it's still
short in number of lines of code.

Rework the allocation API to self-contained movable objects instead of
naively using an unique_ptr to do the job for us. Remove the VK prefix.
2021-01-15 16:19:36 -03:00
ReinUsesLisp
bf1d317892 vk_buffer_cache: Remove unused function 2021-01-15 02:58:55 -03:00
bunnei
cbf81948f5 Merge pull request #5311 from ReinUsesLisp/fence-wait
vk_fence_manager: Use timeline semaphores instead of spin waits
2021-01-12 21:00:05 -08:00
bunnei
fa0fc64dee Merge pull request #5231 from ReinUsesLisp/dyn-bindings
renderer_vulkan/fixed_pipeline_state: Move enabled bindings to static state
2021-01-08 12:24:46 -08:00
ReinUsesLisp
ecb015995e vk_fence_manager: Use timeline semaphores instead of spin waits
With timeline semaphores we can avoid creating objects. Instead of
creating an event, grab the current tick from the scheduler and flush
the current command buffer. When the fence has to be queried/waited, we
can do so against the master semaphore instead of spinning on an event.

If Vulkan supported NVN like events or fences, we could signal from the
command buffer and wait for that without splitting things in two
separate command buffers.
2021-01-08 02:47:28 -03:00
Morph
07f5c03885 Merge pull request #5288 from ReinUsesLisp/workaround-garbage
gl_texture_cache: Avoid format views on Intel and AMD
2021-01-06 15:39:51 +08:00
bunnei
0d3eb81f66 Merge pull request #5289 from ReinUsesLisp/vulkan-device
vulkan_common: Move device abstraction to the common directory and allow surfaceless devices
2021-01-05 17:44:56 -08:00
LC
7ba63cf225 Merge pull request #5292 from ReinUsesLisp/empty-set
vk_rasterizer: Skip binding empty descriptor sets on compute
2021-01-04 21:32:57 -05:00
ReinUsesLisp
2fb40db5fc vk_rasterizer: Skip binding empty descriptor sets on compute
Fixes unit tests where compute shaders had no descriptors in the set,
making Vulkan drivers crash when binding an empty set.
2021-01-04 17:56:39 -03:00
ReinUsesLisp
eb04c63df5 renderer_vulkan/nsight_aftermath_tracker: Move to vulkan_common 2021-01-04 02:22:22 -03:00