Commit graph

6518 commits

Author SHA1 Message Date
Fernando S
72a14a6817
Merge pull request #10422 from liamwhite/gc
video_core: tune garbage collection aggressiveness
2023-05-24 03:58:49 +02:00
Fernando S
76f6388969
Merge pull request #10398 from liamwhite/bcn
video_core: add ASTC recompression
2023-05-24 03:55:45 +02:00
Liam
3b1172c10f video_core: tune garbage collection aggressiveness 2023-05-23 12:55:14 -04:00
Liam
415c78b87c textures: add BC1 and BC3 compressors and recompression setting 2023-05-23 12:54:40 -04:00
liamwhite
a5d4c3e5ad
Merge pull request #10388 from GPUCode/fence-wait
vk_master_semaphore: Move fence wait on separate thread
2023-05-23 09:42:56 -04:00
liamwhite
58be9b12f4
Merge pull request #10402 from liamwhite/uh
renderer_vulkan: barrier attachment feedback loops
2023-05-23 09:42:49 -04:00
Liam
8bba9f7dea vulkan_device: Enable VK_KHR_push_descriptor on newer ANV 2023-05-22 19:53:20 -04:00
Liam
8758932031 renderer_vulkan: barrier attachment feedback loops 2023-05-22 18:10:16 -04:00
scorpion81
f8e7b44d28
Limit the device access memory to 4 GB
Hardly limiting the device access memory to 4 GB for integrated vulkan devices here. This works for the Steam Deck in order not to go above 4 GB VRAM usage any more (above this value the likelihood to crash when the RAM exceeds 12 GB as well raises).

But there will be perhaps a detection mechanism necessary for detecting the real memory limit for integrated vulkan devices. Those likely might have small limits anyway, but what about integrated GPUs on machines with > 16 GB RAM, aka larger amounts ?
2023-05-22 16:48:55 +02:00
Danila Malyutin
7701a00a02 Add support for deinterlaced videos playback
This is a follow up to #10254 to improve the playback of cut scenes in Layton's Mystery Journey.
It uses ffmpeg's yadif filter for deinterlacing.
2023-05-22 01:43:44 +04:00
GPUCode
4e491ab59b vk_master_semaphore: Move fence wait on separate thread 2023-05-20 19:23:53 +03:00
Liam
de7c92d7c4 renderer_vulkan: remove wrong constexpr 2023-05-18 18:01:01 -04:00
lat9nq
c1a8a508bc vulkan_device: Disable VK_KHR_push_descriptor on ANV
Mesa commit ff91c5ca42bc80aa411cb3fd8f550aa6fdd16bdc breaks
VK_KHR_push_descriptor usage on ANV drivers 22.3.0, so disable it
and allow games to boot.
2023-05-17 22:19:57 -04:00
bunnei
9d4a2de72b
Merge pull request #10262 from liamwhite/depth-clamp
vulkan_common: disable depth clamp dynamic state for older radv
2023-05-17 12:19:03 -07:00
liamwhite
47c5c37bed
Merge pull request #10217 from Kelebek1/clear_value
Use the rendertarget format of the correct RT rather than the first valid
2023-05-16 10:06:30 -04:00
liamwhite
bbb6b58aa4
Merge pull request #10181 from lat9nq/intel-compute-toggle
configure_graphics: Add option to enable compute pipelines for Intel proprietary
2023-05-15 12:05:24 -04:00
liamwhite
6060685609
Merge pull request #10249 from FernandoS27/sorry-i-am-late
Buffer Cache: Clear sync code.
2023-05-15 12:03:25 -04:00
liamwhite
5b85925469
Merge pull request #10254 from danilaml/fix-h264-decode
Fix missing pic_order_present_flag in h264 header
2023-05-15 12:03:14 -04:00
Fernando Sahmkow
c4bfbc6d25 Buffer Cache: Clear sync code. 2023-05-15 01:50:21 +02:00
liamwhite
e9069dfe76
Merge pull request #10288 from liamwhite/vram-limits
vulkan_device: reserve extra memory to prevent swaps
2023-05-14 17:02:15 -04:00
Liam
2be751100b vulkan_device: reserve extra memory to prevent swaps 2023-05-14 16:49:59 -04:00
Liam
122435e080 vulkan_common: fix incompatible property flags 2023-05-14 01:13:11 -04:00
Liam
856838f7ce vulkan_common: disable depth clamp dynamic state for older radv 2023-05-13 00:37:17 -04:00
Danila Malyutin
9367769fe7 Fix missing pic_order_present_flag in h264 header
Fixes #9635
2023-05-12 22:30:59 +04:00
Kelebek1
cd0ded7771 Correctly track RT indexes for image aspect lookup during clears 2023-05-12 01:40:21 +01:00
liamwhite
182221b9ff
Merge pull request #10132 from Kelebek1/fermi_blit2
Allow Fermi blit accelerate to work without images in cache
2023-05-11 10:45:59 -04:00
liamwhite
2643ea80df
Merge pull request #10216 from Kelebek1/buffer_cache_region_checks
Swap order of checking/setting region modifications in the buffer_cache
2023-05-11 10:45:47 -04:00
Kelebek1
bf08bc3c0f Allow Fermi blit accelerate to add src/dst to the cache if they don't exist already. Use ScratchBuffers in the software blit path. 2023-05-11 06:42:38 +01:00
Liam
67fd1df762 renderer_vulkan: separate guest and host compute descriptor queues 2023-05-10 13:46:48 -04:00
Kelebek1
a386003b64 Use the rendertarget format of the correct RT rather than the first valid 2023-05-09 22:13:15 +01:00
Kelebek1
b3691fc33c Swap order of checking/setting region modifications in the buffer_cache 2023-05-09 20:21:08 +01:00
Fernando Sahmkow
8a214e5530 Texture Cache: Fix ASTC textures 2023-05-09 02:42:10 +02:00
Fernando Sahmkow
8014dd8259 Texture cache: Only force flush the dma downloads 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
2df19ef0fd Buffer Cache: disable reactive flushing in it. 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
016c6feb49 Texture cache: reverse inmediate flush changes 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
36c302fa32 Buffer cache: always use async buffer downloads and fix regression. 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
6f90dff293 Address feedback, add CR notice, etc 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
ab0c0a469c Query cache: stop updating pages as it's not affected by cpu writes 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
92da86290c Settings: add option to enable / disable reactive flushing 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
0f4f18265f Texture cache: sync the first flush. 2023-05-07 23:46:12 +02:00
Fernando Sahmkow
c6cac2ffaa GPU: Add Reactive flushing 2023-05-07 23:46:12 +02:00
liamwhite
0c5bdc7241
Merge pull request #10081 from Kelebek1/copy_overlap_tick
Sort overlap_ids by modification tick before copy
2023-05-07 14:09:10 -04:00
liamwhite
cb49c3b64c
Merge pull request #10172 from Kelebek1/debug_validation_names
Log object names with debug renderer, add a GPU address to ImageViews
2023-05-07 14:09:03 -04:00
lat9nq
6ed6e6e18e vk_pipeline_cache: Use setting to disable intel compute 2023-05-07 01:06:22 -04:00
bunnei
3547248ec2
Merge pull request #10125 from lat9nq/vsync-select
configuration: Expose separate swap present modes
2023-05-06 21:55:39 -07:00
Kelebek1
ca6bf06ef7 Log object names with debug renderer, add a GPU address to ImageViews 2023-05-06 04:48:32 +01:00
liamwhite
60d54d911e
Merge pull request #10145 from Kelebek1/code_size
Fix shader code resize to use word size rather than byte size
2023-05-04 14:44:02 -04:00
Fernando S
2506594c50
Merge pull request #10153 from FernandoS27/a-quickie-fixie
Memory manager: Fix possible softlock
2023-05-04 03:56:53 +02:00
bunnei
055ee84024
Merge pull request #10142 from FernandoS27/missing-astc
GPU: implement missing ASTC
2023-05-03 16:49:27 -07:00
Fernando Sahmkow
1c13c74295 Memory manager: Fix possible softlock 2023-05-04 00:15:21 +02:00
bunnei
a661c547d8
Merge pull request #10088 from FernandoS27/100-gelato-flavor-test-builds-later
Y.F.C Implement Asynchronous Fence manager and Rework Query async downloads
2023-05-03 15:10:22 -07:00
Fernando Sahmkow
87a9be8dec GPU: implement missing ASTC 2023-05-03 11:33:28 -04:00
liamwhite
ffeb5cdd8d
Merge pull request #10151 from GPUCode/no-softlocks-please
Fix softlocks when disabling async present
2023-05-03 10:54:24 -04:00
Morph
8dd3baa562
Merge pull request #10144 from liamwhite/dont-turbo
vulkan: disable turbo when debugging tool is attached
2023-05-03 10:53:03 -04:00
Morph
daf7936095
Merge pull request #10143 from liamwhite/fruit-company-moment
video_core: fix build on Apple Clang
2023-05-03 10:52:56 -04:00
GPUCode
f3fcc15ad5 vk_present_manager: Fix softlocks when disabling async present 2023-05-03 07:50:10 +03:00
lat9nq
952b271092 vk_swapchain: Use certain modes for unlocked
Uses mailbox, then immediate for unlocked framerate depending on
support for either. Also adds support for FIFO_RELAXED.

This function now assumes vsync_mode was originially configured to a value
that the driver supports.

vk_swapchain: ChooseSwapPresentMode determines updates

Simplifies swapchain a bit and allows us to change the present mode
during guest runtime.

vk_swapchain: Fix MSVC error

vk_swapchain: Enforce available present modes

Some frontends don't check the value of vsync_mode before comitting it.
Just as well, since a driver update or misconfiguration could problems
in the swap chain.

vk_swapchain: Silence warnings

Silences GCC warnings implicit-fallthrough and shadow, which apparently
are not enabled on clang.
2023-05-02 21:52:43 -04:00
lat9nq
c6c11c1553 vulkan_surface: Pass only window info for surface creation
We don't need the whole EmuWindow when creating a surface,
and it creates onerous requirements outside of typical usage for
creating a surface elsewhere.
2023-05-02 21:51:30 -04:00
lat9nq
6f0929df82 configuration: Expose separate swap present modes
Previously, yuzu would try and guess which vsync mode to use given
different scenarios, but apparently we didn't always get it right. This
exposes the separate modes in a drop-down the user can select.

If a mode isn't available in Vulkan, it defaults to FIFO.
2023-05-02 21:51:29 -04:00
Kelebek1
f902cc2a2b Fix code resize to use word size rather than byte size 2023-05-02 23:52:21 +01:00
Liam
4df49631de vulkan: disable turbo when debugging tool is attached 2023-05-02 18:14:57 -04:00
Liam
e1c74cea10 video_core: fix build on Apple Clang 2023-05-02 18:05:30 -04:00
GPUCode
f403d27941 vk_present_manager: Add toggle for async presentation 2023-05-01 23:13:24 +03:00
GPUCode
1d7abac84b vk_blit_screen: Recreate FSR when frame is recreated
* Depends on the layout dimentions and thus should be recreated as well
2023-05-01 23:13:24 +03:00
GPUCode
50791cb974 renderer_vulkan: Fix crashing when updating descriptors
* During pipeline configure the function would acquire some payload space from the descriptor update queue,
  write the descriptor data on the GPU thread and give the scheduler a pointer to the beginning of said space to update it later.
  TickFrame resets the payload cursor, used to track acquires, back to the beginning of the buffer.
  This wasn't a problem before since WaitWorker was called at the end of the frame but now it is.
  If a frame writes to a cursor before the scheduler catches up, it will crash

* To fix this the payload buffer has been increased to account for the in flight frames that are allowed to exist now.
  TickFrame will switch between the payload spaces instead of resetting
2023-05-01 23:13:24 +03:00
GPUCode
2ad9acf795 renderer_vulkan: Async presentation 2023-05-01 23:13:24 +03:00
Morph
47938541c2
Merge pull request #10084 from FernandoS27/yuzu-goes-broom-broom
Y.F.C Buffer Cache Revamp
2023-05-01 11:08:02 -04:00
Fernando Sahmkow
d6f565e5da BufferCache: Fixes and address feedback 2023-05-01 11:43:26 +02:00
bunnei
689f9a75a7
Merge pull request #10110 from Morph1984/intel-disable-compute
vk_pipeline_cache: Skip compute pipelines on Intel proprietary drivers
2023-04-29 23:02:45 -07:00
Fernando Sahmkow
4bc5469f52 Texture Cache: Release stagging buffers on tick frame 2023-04-29 15:31:38 +02:00
Fernando Sahmkow
f8d31d1ae1 Buffer Cache: Release stagging buffers on tick frame 2023-04-29 00:46:31 +02:00
Fernando Sahmkow
80480fe3f7 Clang: format and ficx compile errors. 2023-04-29 00:46:31 +02:00
Fernando Sahmkow
ed4553806a Implement Async downloads in normal and fix a few issues. 2023-04-29 00:46:31 +02:00
Fernando Sahmkow
f2d3212de9 Buffer Cache rework: Setup async downloads. 2023-04-29 00:46:31 +02:00
Fernando Sahmkow
a16c261131 Buffer Cache: Fully rework the buffer cache. 2023-04-29 00:46:31 +02:00
Fernando Sahmkow
58d1c7c77a Address Feedback & Clang Format 2023-04-29 00:18:21 +02:00
Fernando Sahmkow
56c9730a16 Maxwell3D: only update parameters on High 2023-04-29 00:18:21 +02:00
Fernando Sahmkow
e3a2ca96bd Accelerate DMA: Use texture cache async downloads to perform the copies
to host.

WIP
2023-04-29 00:18:21 +02:00
Fernando Sahmkow
3fbee093b2 TextureCache: refactor DMA downloads to allow multiple buffers. 2023-04-29 00:18:21 +02:00
Morph
cb092af3f0 vk_pipeline_cache: Skip compute pipelines on Intel proprietary drivers
Intel's SPIR-V shader compiler is broken. For now, skip compiling any compute pipelines until they fix this issue.
This is not a perfect workaround, as there are a small subset of non-compute pipelines that still cause it to crash, but this should cover the majority of crashes.
It is unfortunate that even with a test case reported 6 months ago the issue has not been fixed in favor of fixing "the most popular games and apps".
Intel, you can do better than this.
2023-04-28 17:59:36 -04:00
Fernando Sahmkow
2f15876524 QueryCache: Fix write invalidation. 2023-04-28 23:53:46 +02:00
Fernando Sahmkow
9a7c172f76 MemoryManager: Fix race conditions. 2023-04-28 23:53:02 +02:00
Fernando Sahmkow
e4dc73f61e Clang format and ddress feedback 2023-04-24 12:38:47 +02:00
Fernando S
9bf19b04f6
Merge pull request #10051 from liamwhite/surface-capabilities
vulkan: pick alpha composite flags based on available values
2023-04-24 12:37:13 +02:00
Fernando S
2311fa7c84
Merge pull request #10069 from liamwhite/log
maxwell_3d: fix out of bounds array access in size estimation
2023-04-24 12:36:24 +02:00
Fernando Sahmkow
e29ced29fa QueryCache: rework async downloads. 2023-04-23 22:04:14 +02:00
Fernando Sahmkow
7e76c1642c Accuracy Normal: reduce accuracy further for perf improvements in Project Lime 2023-04-23 22:03:44 +02:00
Fernando Sahmkow
fca72beb2d Fence Manager: implement async fence management in a sepparate thread. 2023-04-23 04:48:50 +02:00
Liam
eb7c2314f6 maxwell_3d: fix out of bounds array access in size estimation 2023-04-22 10:35:26 -04:00
Kelebek1
3346de365a Sort overlap_ids by modification tick before copy 2023-04-22 14:02:10 +01:00
Kelebek1
4e14b64bfc Account for a pre-added offset when using Corner sample mode for 2D blits 2023-04-21 19:08:21 +01:00
Liam
e3fb9b5e00 vulkan: use plain fences when timeline semaphores are not available 2023-04-14 22:53:37 -04:00
bunnei
e0895a8581
Merge pull request #10030 from Wollnashorn/botw-amd-fix
shader_recompiler: Fix ImageGather rounding on AMD/Intel
2023-04-14 16:56:34 -07:00
Liam
e37e1d24f9 vulkan: pick alpha composite flags based on available values 2023-04-13 16:38:20 -04:00
Wollnashorn
c0e5ecc399 video_core: Enable ImageGather rounding fix on AMD open source drivers 2023-04-12 17:11:02 +02:00
liamwhite
54b4c84ab6
Merge pull request #10008 from vonchenplus/texture_cache
video_core: update imageinfo implement
2023-04-11 11:59:18 -04:00
Wollnashorn
fe91066f46 video_core: Enable ImageGather with subpixel offset on Intel 2023-04-08 16:12:44 +02:00
Wollnashorn
780240e697 shader_recompiler: Add subpixel offset for correct rounding at ImageGather
On AMD a subpixel offset of 1/512 of the texel size is applied to the texture coordinates at a ImageGather call to ensure the rounding at the texel centers is done the same way as in Maxwell or other Nvidia architectures.
See https://www.reedbeta.com/blog/texture-gathers-and-coordinate-precision/ for more details why this might be necessary.

This should fix shadow artifacts at object edges in Zelda: Breath of the Wild (#9957, #6956).
2023-04-08 16:12:30 +02:00
liamwhite
ff386e794b
Merge pull request #10004 from Kelebek1/cubemap
[texture_cache] Only upload GPU-modified overlaps
2023-04-03 13:05:52 -04:00
Jan Beich
083d913eab externals: update Vulkan-Headers to v1.3.246 2023-04-01 05:38:54 +00:00