2be0d6842e
[dxvk] Add option to limit memory chunk size
2023-03-01 11:53:18 +01:00
2abc102d2c
[meta] Release 2.1
2023-01-24 14:54:09 +01:00
255ab1a63c
[hud] Display more useful driver info in HUD
...
Vulkan versions are irrelevant anyway, not sure why we
ever displayed that in the first place.
2023-01-23 00:10:39 +01:00
41ee092b97
[dxvk] Get rid of global timeline semaphore
...
Timeline semaphores are broken on 32-bit Proton, so just use the existing
command list binary semaphore to synchronize sparse binding operation.
This will introduce some additional graphics queue submissions around
sparse binding, but since that is basically unused it's not a concern.
2023-01-21 13:15:09 +01:00
d020f4451a
[dxvk] Invalidate all v16 cache files
...
We were randomly writing bogus data in an irrecoverable way,
so in order to not cause any problems for users, invalidate
all caches created with this version.
2023-01-21 01:24:02 +01:00
3d24560af7
[dxvk] Fix reading out-of-bounds memory in state cache
...
Fixes #3196 , not to mention that this was god-awful code anyway.
2023-01-21 01:14:45 +01:00
0af5ececa6
[dxvk] Also reset descriptor pool when only one pool is in use
...
This way we can often avoid allocating additional descriptor pools.
2023-01-20 15:23:25 +01:00
eacb8da9f7
[dxvk] Tweak descriptor pool numbers
...
Increases the descriptor pool size on 64-bit, but also makes
significantly better use of descriptor pool memory in situations
with very high draw counts and many UBO descriptor set updates.
2023-01-20 14:34:46 +01:00
c6668ffbaa
[dxvk] Drop shader module workaround for compute pipelines.
...
This was a bug in early Nvidia development drivers, but has been fixed since.
2023-01-18 15:56:00 +01:00
741cc493c6
[dxvk] Increase maximum number of command buffers in flight
2023-01-17 15:01:06 +01:00
f7aa310fdd
[dxvk] Disable pipeline lifetime tracking for RADV
...
Seems like this is not needed on this driver, so let's just use
the fast path by default. Makes the current implementation work
with 32-bit games as well since caching does not work yet.
2023-01-17 14:02:38 +01:00
421d7d9077
Revert "[vulkan] Use fence for acquiring a swap chain image"
...
Breaks things and we don't really need this right now anyway.
2023-01-16 23:16:38 +01:00
599357721a
[hud] Support HDR color spaces
...
Blending is broken if we need to do encoding in the shader, but we
cannot do much about that without changing the rendering process,
so this will have to do for now.
2023-01-16 18:04:18 +00:00
1482715fd1
[vulkan] Use fence for acquiring a swap chain image
...
In practice, drivers will block here anyway.
2023-01-16 16:20:56 +01:00
af05265cb6
[dxvk] Set thread priority for background workers
...
We accidentally lost this somehow.
2023-01-15 15:54:03 +01:00
d35bf455d9
[dxvk] Rename bindResourceBuffer to bindUniformBuffer
2023-01-15 15:36:05 +01:00
a8f9fdb21d
[dxvk] Rearrange descriptor sets
...
This allows us not to unnecessarily dirty the FS UBO set when
changing tetxure bindings, leading to a cleaner separation.
2023-01-15 15:36:05 +01:00
4a30933359
[dxvk] Add UBO set property to descriptor info
...
Used to explicitly propagate storage buffer bindings
to the respective UBO set.
2023-01-15 15:36:05 +01:00
82f500250f
[hud] Bind data buffers as views
2023-01-15 15:36:05 +01:00
c347bd4d8b
[dxvk] Allow creating buffer views with undefined format
2023-01-15 15:36:05 +01:00
1c2df54bdf
[dxvk] Improve logging in case of memory errors
2023-01-14 03:39:55 +01:00
2922b780c1
[dxvk] Properly handle bufferImageGranularity for images
...
Fixes validation errors about dedicated allocations being
bigger than the image's memory requirement on Nvidia.
2023-01-14 03:39:55 +01:00
3a84838ac4
[hud] Display approximate progress when compiling shaders
2023-01-13 16:43:27 +01:00
c978e62ec8
[dxvk] Implement better priority system for background shader compiles
...
Reduces the number of workers that perform background optimization,
which may reduce the performance impact when encountering a large
number of new pipelines at once.
2023-01-13 16:43:27 +01:00
070a340115
[dxvk] Remove path to look up optimized pipeline in cache
...
For some reason this sometimes takes several milliseconds, which
leads to noticeable stutter. Linking a pipeline is more consistent.
2023-01-13 11:40:01 +01:00
a22d70e184
[dxvk] Remove unused compileComputePipeline function
2023-01-12 17:35:05 +01:00
e5157a5360
[dxvk] Add pre-rasterization pipeline libraries to the state cache
...
This allows compiling tessellation or geometry shader pipelines
early while still using the pipeline library path.
Also removes compute shaders. Since API-provided compute shaders
are always compiled early, supporting them is no longer needed.
2023-01-11 11:48:16 +01:00
28ae85b7ab
[dxvk] Enable creating full pre-rasterization pipeline libraries
2023-01-11 11:48:16 +01:00
17529101d5
[dxvk] Allow creating pipeline libraries with more than one shader
2023-01-11 11:48:16 +01:00
f9ff96d727
[dxvk] Implement checks for pre-raster pipeline library compatibility
2023-01-11 11:48:16 +01:00
5e42230b95
[dxvk] Add patch vertex count to shader info struct
2023-01-11 11:48:16 +01:00
f76a7c285c
[dxvk] Rework DxvkShaderPipelineLibrary to work with multiple shaders
2023-01-11 11:48:16 +01:00
8052347a22
[dxvk] Keep more empty system memory chunks alive
2023-01-10 03:42:37 +01:00
06fb93daf0
[dxvk] Only create state cache file on demand
...
If no pipelines are ever written to it, we should not create
the cache file in the first place.
2023-01-09 18:31:18 +01:00
6a4fe06ac6
[dxvk] Fix incorrect flags for null fs pipeline layout
2023-01-09 18:31:18 +01:00
c11a63f5e5
[wsi] Add GLFW backend
2023-01-08 10:36:25 +01:00
3491895960
[dxvk] Remove SDL dummy window
...
SDL >= 2.0.9 permits passing a nullptr window to
SDL_Vulkan_GetInstanceExtensions, so there's no
point in going though the work of creating a
window just to call this function.
2023-01-07 22:43:32 +00:00
196fefec4c
[dxvk] Enable dynamic multisample state if supported by the device
...
Eliminates stutter in situations where sample rate shading is used with MSAA.
2023-01-06 23:48:31 +01:00
f269cde749
[dxvk] Enable device features for dynamic multisample state
2023-01-06 23:48:31 +01:00
12901b52f1
[dxvk] Split barriers with very large VkDependencyInfo structures
...
Works around an AMD driver bug. Fixes #3138 .
2022-12-20 03:10:44 +01:00
Joshua Ashton
731bf84edd
[dxvk] Enable VK_EXT_hdr_metadata
2022-12-19 22:15:18 +01:00
Joshua Ashton
4335eccae9
[dxvk] Enable VK_EXT_swapchain_colorspace
2022-12-19 22:15:18 +01:00
18b0ef697c
[dxvk] Rework SetEventOnCompletion
...
* Create the waiter thread on demand
* Don't wake up the waiter thread every 10ms
when no events are queued
* Wait on caller thread when hEvent = null
2022-12-16 17:37:19 +01:00
0b9f785bc5
[dxvk] Remove broken memory budget logic for UMA systems
...
All heaps on an UMA systems are by definition device-local, so this
just leads to unnecessary issues.
2022-11-01 12:48:03 +01:00
fea86ef116
[dxvk] Use dynamic depth clip enable for linked pipelines if supported
...
This way we won't have to compile any vertex shader pipelines twice.
2022-10-24 16:39:24 +02:00
8e7ea899d2
[dxvk] Enable VK_EXT_extended_dynamic_state_3 if available
2022-10-24 16:39:24 +02:00
859de7e828
[dxvk] Always enable depth clip feature if supported
2022-10-24 16:39:24 +02:00
53a0c3726c
[dxvk] Don't crash immediately on pipeline library compile error
...
Things will blow up down the line but there's no good reason to crash
immediately. Also ignore returned pipeline on error so we'll always
return null.
2022-10-18 00:07:41 +02:00
f1f8d45fcd
[dxvk] Rework allocation logic for large resources
...
This may reduce internal fragmentation with very large resources.
We previously changed behaviour to not do this in order to reduce memory
pressure in the average case, however by trying to suballocate from existing
chunks and falling back to a dedicated allocation on failure, rather than
allocating a new chunk, we can mostly avoid that situation.
2022-09-24 17:06:00 +02:00
15588004b4
[dxvk] Bump memory chunk size to 256 MiB.
2022-09-24 17:06:00 +02:00
Philip Rebohle
Beyley Thomas
Hunter Kvalevog
Joshua Ashton
Robin Kertels