1
0
mirror of https://github.com/doitsujin/dxvk.git synced 2025-02-23 19:54:16 +01:00

6592 Commits

Author SHA1 Message Date
WinterSnowfall
4a89b75bb7 [d3d8] Fix x64 crash on shader validation 2025-02-17 15:11:05 +01:00
WinterSnowfall
274c590ad6 [d3d9] Set priority only for D3DPOOL_MANAGED/DEFAULT resources 2025-02-17 15:11:05 +01:00
WinterSnowfall
84b2ac3f97 [d3d8] Set priority only for D3DPOOL_MANAGED resources 2025-02-17 15:11:05 +01:00
WinterSnowfall
8efa3ed84a [d3d8] A few minor formatting adjustments 2025-02-17 15:11:05 +01:00
Robin Kertels
3269f92138 [d3d9] Fix mismatching texture type mask updates 2025-02-17 14:32:34 +01:00
Jeff
0aebedae16 [d3d8] Add shadow perspective divide hack for Splinter Cell 2025-02-17 14:10:54 +01:00
Philip Rebohle
f2802dd2ff [dxvk] Fix some potential image layout bugs 2025-02-17 01:26:03 +01:00
Philip Rebohle
9389a0ca96 [dxvk] Don't transition image layout in clearImageViewCs unless necessary 2025-02-17 01:26:03 +01:00
Philip Rebohle
6d9e0baa27 [dxvk] Don't hold queue lock when invoking periodic memory tasks 2025-02-14 14:34:33 +01:00
Philip Rebohle
9573c389de [dxvk] Don't create GPL pipeline if we need to patch the vertex shader
Works around a game bug in Kingdom Come Deliverance and silences a validation
error.
2025-02-13 18:46:49 +01:00
Philip Rebohle
3453732bba [dxvk] Fix clear optimization edge case with multi-planar images 2025-02-12 21:42:27 +01:00
Philip Rebohle
0a84dbb787 [dxvk] Implement resolves with a render target clear if possible
Same idea as for copies, if the last use of the source image was a clear
then we can simply clear the destination image instead of performing a
much more expensive resolve operation.
2025-02-12 21:42:27 +01:00
Philip Rebohle
f5a5ec7c88 [dxvk] Factor out method to find pending deferred clears 2025-02-12 21:42:27 +01:00
Philip Rebohle
1cabbee2bb [dxvk] Rework render pass resolves
In an attempt to also support batching layered resolves.
2025-02-12 21:42:27 +01:00
Philip Rebohle
563129c863 [dxvk] Fix potential resolve attachment invalidation bug 2025-02-12 14:44:54 +01:00
Philip Rebohle
c5a3aa73a0 [util] Enable dummy composition swap chain for an upcoming game 2025-02-12 01:36:10 +01:00
Philip Rebohle
80bfd2ed97 [util] Add functionality to support hashed app profiles 2025-02-12 01:34:40 +01:00
Philip Rebohle
813ae020b6 [d3d11] Flush more around non-multisampled render passes 2025-02-11 16:14:53 +01:00
Philip Rebohle
7e3c6f819a [util] Add new flush type for render pass boundaries 2025-02-11 16:14:53 +01:00
Philip Rebohle
94b5cfc3d8 [dxvk] Don't use secondary command buffers for certain render passes
If we can't use any store/resolve op optimizations on a render pass, there
is no reason to pay for the overhead of secondary command buffers.
2025-02-11 16:14:53 +01:00
Philip Rebohle
beaf6dd2a6 [dxvk] Add context flag for active secondary command buffers
Makes some checks more explicit.
2025-02-11 16:14:53 +01:00
Philip Rebohle
cf9ccf711b [dxvk] Add perf hint to avoid secondary command buffers 2025-02-11 16:14:53 +01:00
Philip Rebohle
94254f556e [dxvk] Fix stencil resolve mode 2025-02-11 15:26:08 +01:00
Philip Rebohle
70492b2954 [dxvk] Avoid locking built-in latency tracker when calling into presenter
Same reason as for the Reflex one.
2025-02-11 15:12:13 +01:00
Philip Rebohle
b1174a1bdf [dxvk] Avoid locking Reflex latency tracker when calling into presenter
There are complex deadlock conditions during swap chain destruction.
2025-02-11 15:12:13 +01:00
Philip Rebohle
efeb15edbd [dxvk] Fix lack of forward progress guarantee in presenter
Turns out that relying on two threads to unblock another was a bad idea.
Fixes a potential deadlock w.r.t. swapchain recreation with Reflex.
2025-02-11 15:12:13 +01:00
Philip Rebohle
ac17d5f9b1 [dxvk] Enable tiler mode for more drivers 2025-02-10 18:36:01 +01:00
Philip Rebohle
11ec65d516 [include] Update Vulkan headers 2025-02-10 17:44:36 +01:00
Neal Gompa
4d3b35ae53 meson: Fix version construct by dropping the "v" prefix
The "v" prefix was causing the version to be incorrectly set for
generated files (such as pc files for pkg-config).

Dropping the prefix fixes the issue.
2025-02-10 16:38:03 +01:00
Philip Rebohle
3aeedb9c98 [d3d11] Forward DiscardResource calls to backend for images 2025-02-10 16:13:31 +01:00
Philip Rebohle
9b5499caf2 [dxvk] Ignore deferred discards that we can't fold into load ops 2025-02-10 16:13:31 +01:00
Philip Rebohle
fcabffc1e5 [dxvk] Add store op optimization for image discards in tiler mode 2025-02-10 16:13:31 +01:00
Philip Rebohle
821386aeff [dxvk] Add config option to toggle tiler optimizations 2025-02-10 16:13:31 +01:00
Philip Rebohle
fa2d791d79 [dxvk] Use resolve attachments if possible 2025-02-10 16:13:31 +01:00
Philip Rebohle
80fb7e2294 [d3d9,d3d11] Adjust flush heuristic on tiling GPUs
Probably needs work, just disable most of the heuristic for the time being.
2025-02-10 16:13:31 +01:00
Philip Rebohle
00299b264b [dxvk] Use secondary command buffers for rendering on tiling GPUs 2025-02-10 16:13:31 +01:00
Philip Rebohle
4a92b92f58 [dxvk] Add perf hint for tiling GPUs 2025-02-10 16:13:31 +01:00
Philip Rebohle
6c1bc35264 [dxvk] Add functionality to use secondary command buffers 2025-02-10 16:13:31 +01:00
Philip Rebohle
7d35afdc37 [dxvk] Add functionality to allocate secondary command buffer 2025-02-10 16:13:31 +01:00
Philip Rebohle
f615594f22 [dxgi] Create dummy window for composition swap chains
Obviously not a valid implementation, but allows composition swap chains
to exist with some basic functionality.
2025-02-10 14:23:12 +01:00
Philip Rebohle
76f9bec7cc [dxgi] Add option to expose composition swapchains 2025-02-10 14:14:38 +01:00
WinterSnowfall
0a19e5b6db [d3d8] Store the value of D3DRS_ZVISIBLE 2025-02-10 11:25:40 +01:00
WinterSnowfall
946419cda2 [d3d8] Fix invalid casting in GetRenderState 2025-02-10 11:25:40 +01:00
WinterSnowfall
13554f18bd [d3d9] Update software cursor position using SetCursorPosition 2025-02-08 22:28:48 +01:00
Blisto91
54a26dde3d [util] Disable dcSingleUseMode for Cardfight!! Vanguard Dear Days 2 2025-02-07 19:53:45 +01:00
Philip Rebohle
357484bd4b [d3d11] Don't call Flush from swap chain code
Use internal function instead.
2025-02-06 23:26:05 +01:00
Philip Rebohle
9664e0b850 [dxvk] Fix potential race conditions w.r.t. swapchain destruction
The signaling thread can queue up a wait while the swapchain is being
destroyed, which blows up. This also fixes a related race condition
where we would queue up a wait using the wrong present mode.

This is somewhat fragile in the sense that the queue worker *must* call
signalFrame for each and every present, or this will fall apart.
2025-02-06 20:25:32 +01:00
Philip Rebohle
e5a81f8c7e [dxvk] Don't probe buffer compatibility on image-only memory types
Silences a validation error.
2025-02-06 15:29:24 +01:00
Philip Rebohle
215cebc019 [dxvk] Remove superfluous function declaration 2025-02-06 14:29:19 +01:00
Philip Rebohle
a413eb0843 [dxvk] Only end render pass in invalidateImage if currently bound
Not really needed otherwise.
2025-02-05 22:49:25 +01:00