Philip Rebohle
412d79c8c1
[d3d11] Use new query implementation
2019-03-14 21:16:41 +01:00
Philip Rebohle
a8144370c8
[dxvk] Create new query pool and forward it to the context
2019-03-14 21:16:41 +01:00
Philip Rebohle
6b9653d261
[dxvk] Create GPU event pool and forward it to the context
2019-03-14 21:16:41 +01:00
Philip Rebohle
2a6d4fa2ba
[dxvk] Implement DxvkDevice::getShaderPipelineStages
2019-03-02 19:56:00 +01:00
Philip Rebohle
10123ce9ab
[dxvk] Pass description of the client API to the DXVK device
2019-02-15 17:38:52 +01:00
Philip Rebohle
8b5db80fbd
[dxvk] Reimplement DxvkBuffer
...
Avoids the DxvkPhysicalBuffer indirection and keeps all buffer
objects alive until the DxvkBuffer itself gets destroyed.
2019-01-09 18:04:46 +01:00
Philip Rebohle
092cad0e76
[dxvk] Remove DxvkSemaphore
2018-11-28 12:44:18 +01:00
Philip Rebohle
9b923bb386
[dxvk] Remove DxvkSwapChain
2018-11-28 12:44:17 +01:00
Philip Rebohle
81306cdc11
[dxvk] Support new Vulkan presenter in DxvkDevice
2018-11-28 12:44:17 +01:00
Philip Rebohle
746529b71b
[dxvk] Pass raw semaphore handles to submitCommandList
2018-11-28 12:44:17 +01:00
Philip Rebohle
4057937d2d
[dxvk] Refactor descriptor set allocation
...
With this new approach, descriptor pools are decoupled from the
command list they are used with. Instead, the DXVK context takes
ownership of a single descriptor pool until it runs out of memory.
This reduces the amount of memory wasted for under-utilized pools
and should this reduce an application's memory footprint.
2018-11-27 11:42:13 +01:00
Philip Rebohle
fc8f49899a
[dxvk] Create instance of depth-stencil packing objects
2018-11-08 18:10:56 +01:00
Philip Rebohle
509e6f1abf
[dxvk] Use new meta copy for depth <> color image copies
2018-09-27 11:44:09 +02:00
Philip Rebohle
4a72cae1e6
[dxvk] Enable state cache
2018-09-23 14:41:43 +02:00
Philip Rebohle
0f52ec2796
Revert "[dxvk] Move command submission to separate thread"
...
This reverts commit f61ff5d345fdbd1627b2fd833f57ff303015c747.
Apparently, this causes extremely poor frame times on some
configurations. Reason is currently unknown.
2018-09-20 19:58:00 +02:00
Philip Rebohle
f61ff5d345
[dxvk] Move command submission to separate thread
...
- Reduces load on the CS thread and the main rendering thread.
- This takes into account that VR apps access the queue as well.
2018-09-20 14:43:06 +02:00
Philip Rebohle
c7ddaea5a6
[dxvk] Add method to retrieve current frame number
2018-09-10 18:05:33 +02:00
Philip Rebohle
657093c14e
[dxvk] Refactor query management
...
Moved all query-related state tracking and management into a
separate class. This allows for new query types to be added
in the future, and makes less dodgy assumptions about the
current state when beginning or ending a query.
2018-08-31 15:38:09 +02:00
Philip Rebohle
b2c4855490
[dxvk] Use global user config for backend options
2018-08-07 16:59:49 +02:00
Philip Rebohle
0ba00b3f59
[dxvk] Add extended device feature structure
...
This allows the client API to query and enable extended
features in the future, should it become necessary, much
like the extende feature queries.
2018-07-31 16:58:25 +02:00
Philip Rebohle
656cb2e4aa
[dxvk] Refactor Vulkan extension management
...
Adding an extension will now require an additional step, but
this evolved version of the extension list code is more sane
and the structures can be more easily copied around.
2018-07-23 20:07:21 +02:00
Philip Rebohle
5d1f00be34
[dxvk] Implement shader-based meta-resolve operation
...
This should work around a variety of potential driver issues
when games resolve typeless images with a different format.
2018-06-23 13:17:01 +02:00
Philip Rebohle
aa3e718fcd
[dxvk] Add DxvkOptions struct for quick lookup of features/limits
2018-06-22 00:29:44 +02:00
Philip Rebohle
cfe99368fb
[dxvk] Make number of queued submissions available to DXVK
2018-06-04 23:24:42 +02:00
Philip Rebohle
4b37590e14
[dxvk] Use new mip map generator
2018-05-25 17:45:41 +02:00
Philip Rebohle
37456d583e
[dxvk] Move DxvkPipelineCache to DxvkPipeManager
...
Since the pipeline cache isn't used for anything else but compiling
pipelines, keeping this stuff together is much more useful.
2018-05-09 14:26:45 +02:00
Philip Rebohle
5e02c1bb2f
[dxvk] Move DxvkPipelineManager instance back to DxvkDevice
2018-05-01 16:56:33 +02:00
Philip Rebohle
9aa73d1148
[dxvk] Set non-zero framebuffer size when there are no attachments
...
This is required for UAV-only rendering and will be required for
stream output in the future as well.
2018-04-29 11:19:18 +02:00
Philip Rebohle
a6a22cd00a
[dxvk] Export DxvkDeviceQueue from DxvkDevice
...
Access to the Vulkan queues provided by the DXVK device
is required for external Vulkan libraries to work.
2018-04-19 15:42:48 +02:00
Philip Rebohle
676b0cb476
[dxvk] Added DxvkMetaClearObjects stub
2018-04-11 17:05:12 +02:00
Philip Rebohle
4797645bd2
[dxvk] Removed unused DxvkOptions
2018-04-11 13:06:43 +02:00
Philip Rebohle
8aa8610221
[dxvk] Implement basic stat counters for draw calls, queue ops
2018-04-03 11:56:12 +02:00
Philip Rebohle
6e27f12e22
[dxvk] Move DxvkPipelineManager object to DxvkContext
...
Since we create only one DxvkContext per D3D11Device, rather than
per D3D11DeviceContext as originally planned, there is no need to
keep the pipeline manager as a global thread-safe object. This may
slightly reduce CPU overhead.
2018-03-29 12:32:20 +02:00
Philip Rebohle
0bdae4f930
[dxvk] Move fence object into DxvkCommandList
...
Reduces command submission overhead by reusing fence objects
instead of creating new ones for each submission. Improves
error reporting in case the submission cannot be complete.
2018-03-22 18:57:33 +01:00
Philip Rebohle
fb36b56891
[dxvk] Removed stat counters
2018-03-10 17:37:10 +01:00
Philip Rebohle
dd237d866d
[dxvk] Make query pools more usable as query allocators
2018-02-15 13:25:18 +01:00
Philip Rebohle
54cfdbdc51
[dxvk] Implemented more query and query pool methods
2018-02-13 13:43:27 +01:00
Philip Rebohle
c66492a90f
[dxvk] Initialize dummy resources to zero
...
Read access to unbound images and buffers usually yields
a result vector that contains zeroes in all components.
2018-02-01 14:26:38 +01:00
Philip Rebohle
d4a0581f8f
[dxvk] Bind dummy resource for unbound vertex and index buffers
...
Allows GTA V to run, although heavy rendering artifacts remain.
2018-02-01 13:29:57 +01:00
Philip Rebohle
d74be35e8c
[dxvk] Refactored physical buffer
...
Class is now much more generic, which should make things
easier for the upcomping deferred context implementation
2018-01-29 00:01:00 +01:00
Philip Rebohle
d1f76b96af
[dxvk] Added dummy resources for descriptors
...
This should finally shut up validation layers and also allows
the branches in texture sample operations to be removed.
2018-01-27 19:25:41 +01:00
Philip Rebohle
a0acbeec72
[dxvk] Optimized buffer renaming
...
Helps applications which frequently update buffers
through either Map()/Unmap() or UpdateSubresource.
2018-01-19 00:20:05 +01:00
Philip Rebohle
a87ae8aba4
[dxvk] Added DxvkPhysicalBuffer to back virtual buffers
...
This is the first step to optimizing buffer updates for applications
that frequently invalidate buffers. The goal is to reduce the number
of buffer allocations per frame and reduce the cost of invalidation.
2018-01-18 15:52:57 +01:00
Philip Rebohle
27573e9b25
[dxvk] Added app and device specific options
2018-01-16 13:58:57 +01:00
Philip Rebohle
bc5dfc1cad
[dxvk] Refactored device extension handling
...
Support for extensions can now be queried from the device
object in an efficient way. This will allow the backend to
use optional extensions for the purpose of optimization.
2018-01-16 13:24:36 +01:00
Philip Rebohle
c1f16d36bd
[dxvk] Implemented local pipeline cache
2018-01-13 22:18:32 +01:00
Philip Rebohle
5dd9fea011
[dxvk] Implemented input layout validation
...
Checks whether all input slots consumed by the vertex shader
are provided by the input layout, and disables rendering in
case the state validation fails. This should hopefully fix
GPU lockups in Nier:Automata.
2018-01-12 14:25:26 +01:00
Philip Rebohle
30eb43a284
[dxvk] Minor memory management tweaks
2018-01-11 19:06:58 +01:00
Philip Rebohle
ee796b966d
[dxvk] Reduced staging buffer size
...
Allocating as much memory as we did was wasteful. The new limit may
still be too generous for most applications, but this will require
more testing.
2018-01-10 10:22:33 +01:00
Philip Rebohle
1fe5b74762
Optimized command submission
...
Command submission now does not synchronize with the device every single
time. Instead, the command list and the fence that was created for it are
added to a queue. A separate thread will then wait for the execution to
complete and return the command list to the device.
2017-12-16 18:10:55 +01:00
