Commit graph

158 commits

Author SHA1 Message Date
Seungha Yang
c0a02affa4 d3d12: Add support for resource copy between d3d11 and d3d12
If driver can support cross-api resource sharing, use device-to-device
resource copy in d3d12upload/download elements.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7119>
2024-07-01 20:18:42 +00:00
Seungha Yang
c013e88295 d3d12decoder: Use sub-allocated bitstream buffer
Since a buffer resource will occupy at least 64KB,
allocating upload resource per decoding command might not be
an optimal approach. Instead, use sub-region of a upload resource
for multiple decoding command if sub-regions are not overlapped
each other.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7108>
2024-06-29 20:47:39 +09:00
Seungha Yang
69ea20b739 d3d12videosink: Present on GstVideoOverlay::expose()
... so that updated backbuffer can be swapped and presented

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7079>
2024-06-24 08:46:24 +00:00
Seungha Yang
dd4d85272e d3d12converter: Fix Y410 conversion
Adding format conversion helper and use compute shader in case that
output format does not support RTV.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7088>
2024-06-23 10:53:18 +00:00
Seungha Yang
327fb9924c d3d12basefilter: Add adapter property
Allows initial GPU adapter selection

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7066>
2024-06-20 00:03:28 +09:00
Seungha Yang
2ec7d82bd7 d3d12: Move fence setter helper method to gst-libs
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:48 +00:00
Seungha Yang
fa7c4a2e39 d3d12converter: Update API signature
Always use device's main direct queue, and control gpu waiting
behavior by using boolean value

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:48 +00:00
Seungha Yang
f6eb3a01c0 d3d12memory: Hide fence value from header
Instead of exposing fence value to wait in header, user setter/getter
methods.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:47 +00:00
Seungha Yang
49c4247cb3 d3d12device: Add helper method for getting fence handle
Add get_fence_handle() method so that caller can get command queue's
dedicated fence handle from device

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:47 +00:00
Seungha Yang
289bc1d440 d3d12: Remove notify_com and notify_mini_object helper methods
Use private macros instead of exposing multiple APIs for the same thing

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:47 +00:00
Seungha Yang
6942868bda d3d12commandqueue: Update API name and arguments
Accepts multiple fences since single command list may have
multiple dependent resources which are associated with
different GPU engines

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:47 +00:00
Seungha Yang
713ea313f8 d3d12device: Use HRESULT return code if possible
Make function signature consistent with that of command queue

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7057>
2024-06-19 13:19:47 +00:00
Seungha Yang
339c81bac8 d3d12: Promote decoder and videosink rank to primary
It's proven that d3d12 performs better than d3d11 while
consumes less resources in various cases.
Assign primary+ rank to decoder and videosink in case of Windows10/11,
so that it can be tested widely

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7045>
2024-06-18 14:40:37 +00:00
Seungha Yang
e5d2dd8e8d d3d12videosink: Add direct-swapchain property
Because DXGI flip mode swapchain will disallow GDI operation
to a HWND once swapchain is configured, videosink has been creating
child window of application's window. However, since window creation
would take a few milliseconds, it can cause performance issue such as
UI freezing. Adding a property so that videosink can attach
DXGI swapchain diretly to application's window in order to improve
performance.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7013>
2024-06-17 16:05:00 +00:00
Seungha Yang
dd1d2f5ff7 d3d12videosink: Add external-window-only property
Adding a new property in order to avoid unintended interanl window
creation.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7013>
2024-06-17 16:05:00 +00:00
Seungha Yang
37e1847464 d3d12videosink: Add support for window handle update
A large refactoring commit for adding features and improve performance

* Reuse internal converter and overlay compositor:
  Converter can be reused as long as input and display formats are not
  changed. Also overlay compositor reconstruction is required only if
  display format is changed

* Don't wait for full GPU flush on resize or close:
  D3D12 swapchain requires GPU idle in order to resize backbuffer.
  Thus CPU side waiting is required for swapchain related commands
  to be finished. However, don't need to wait for full GPU flushing.

* Support multiple sink on a single external window
  Keep installed subclass window procedure even if there's no associated
  our internal HWND. This will make window procedure hooking less racy.
  Then parent HWND's message will be transferred to our internal HWNDs
  if needed.

* Adding support for window handle update
  Application can change target HWND even when videosink is playing or
  paused state. So, users can call gst_video_overlay_set_window_handle()
  against d3d12videosink anytime. The videosink will be able to update
  internal state and setup resource upon requested.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7013>
2024-06-17 16:05:00 +00:00
Seungha Yang
c00c36e33b d3d12overlaycompositor: Remove unused parameter
Don't need to check fence value of overlay buffer since
window uses global direct command queue

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7013>
2024-06-17 16:05:00 +00:00
Seungha Yang
f91be03550 d3d12videosink: Calculate display resolution only per caps change
Don't need to calculate it per window property update

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7013>
2024-06-17 16:05:00 +00:00
Seungha Yang
a2df44da7d d3d12: Workaround for Intel iGPU decoder crash
Observed Intel GPU driver crash when multiple decoders are
configured in a process. It might be because of frequent
command queue alloc/free or too many in-flight decoding commands.
In order to make command queue persistent and limit the number of
in-flight command lists, holds global decoding command queue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7019>
2024-06-17 15:15:07 +00:00
Seungha Yang
20852ba028 d3d12videosink: Disconnect window signal handler on dispose as intended
Fixing typo

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/7012>
2024-06-10 23:33:26 +09:00
Seungha Yang
20db3369d3 d3d12videosink: Add error-on-closed property
Adding a property to control error reporting behavior when output
window is closed in playing or paused state. This can be useful
for apps where an app wants to close window even if it's playing
a stream, and the closed window is expected.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6939>
2024-06-09 16:48:41 +00:00
Seungha Yang
96cf3d7063 d3d12encoder: Do not print error log for not-supported feature
gst_d3d12_result() will print message with ERROR level if failed.
Use FAILED/SUCCEEDED macros instead, since not-supported feature
is not a critical error

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6955>
2024-05-29 10:45:13 +00:00
Seungha Yang
84aecab150 d3d12videosink: Add overlay signal to support d3d12/d3d11/d2d overlay
Conceptually identical to the present signal of d3d11videosink.
This signal will be emitted with current render target
(i.e., swapchain backbuffer) and command queue. Signal handler
can record GPU commands for an overlay image or to blend
an image to the render target.

In addition to d3d12 resources, videosink will send
d3d11 and d2d resources depending on "overlay-mode"
property, so that signal handler can render by using
preferred/required DirectX API.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6838>
2024-05-24 15:55:17 +00:00
Seungha Yang
4d779d7de8 d3d12videosink: Use device's main direct queue
The idea of using separate command queue per videosink was that
swapchain is bound to a command queue and we need to flush the
command queue when window size is changed. But the separate
queue does not seem to improve performance a lot.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6838>
2024-05-24 15:55:17 +00:00
Seungha Yang
9d23c26027 d3d12decoder: Fix SDK debug layer warning
Address below message reported by SDK debug layer.

ID3D12Device::CheckFeatureSupport: Unsupported Decode Profile Specified.
Use ID3D12VideoDevice::CheckFeatureSupport with D3D12_FEATURE_VIDEO_DECODE_PROFILES
to retrieve a list of supported profiles

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6839>
2024-05-13 19:42:52 +00:00
Sebastian Dröge
0ef396359c gst: Move GstQueueArray as GstVecDeque to core
And change lengths and indices from guint to gsize for a more correct type.

Also deprecate GstQueueArray and implement it in terms of GstVecDeque.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6779>
2024-05-06 18:25:42 +00:00
Seungha Yang
cb20a371c2 d3d12decoder: Fix d3d12 resource copy
It was copying to self resource

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6753>
2024-04-28 15:31:19 +00:00
Seungha Yang
46131f0cb0 d3d12ipcclient: Fix deadlock when copying texture
Fixing deadlock in below case
* GC lock is taken by background thread, and the background thread calls
  gst_d3d12_ipc_client_release_imported_data() which takes ipc lock
* ipc lock is already taken in ipc thread and trying to pushing GC data
  via gst_d3d12_command_queue_set_notify()
* gst_d3d12_command_queue_set_notify() is trying to take GC lock
  but it's already taken by background thread

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
19932cf178 d3d12ipcsink: Handle external fence
Waits external fence before sending frame to peer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
b7844ef307 d3d12decoder: Remove CPU-side waiting
Sets decoder command queue's fence to memory instead of waiting
from decoder's output thread. CPU-side waiting will happen
only if download is required.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
a05961ab7b d3d12screencapturesrc: Fix output to non-d3d12 element
Configures upload/download flags to memory after write

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
e29655e9ca d3d12screencapturesrc: Release and flush d3d11 objects before d3d12
Fixing device-removed error when closing pipeline

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
2c203e0d40 d3d12encoder: Handle external fence explicitly
Waits for external fence if any

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
2a14793ee1 d3d12converter: Add support for GPU-side external fence waiting
Ideally, GPU waiting should be scheduled just before executing command list.
But handling the case outside of converter is a bit complicated.
Under an assumption that constructed command list will be executed
immediately, schedules GPU-side waiting inside of conversion method
to simplify the flow.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
478e49dd73 d3d12: Update copy_texture_region() method
Pass external fence value if any and allow passing fence
data so that dependent resources can be released
once copy is done

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6749>
2024-04-28 12:49:07 +00:00
Seungha Yang
4ac46ce82b d3d12screencapturesrc: Performance improvement
Process captured frame using d3d11 instead of d3d12, and use shared
fence when copying processed d3d11 texture to d3d12 resource.
In this way, capture CPU thread does not need to wait for fence signal.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6733>
2024-04-25 22:51:01 +00:00
Seungha Yang
cecb0f2148 d3d12decoder: Lock DPB while building command
Since DPB resource can be modified in output thread, protect
it when building command list.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6709>
2024-04-23 10:08:19 +00:00
Seungha Yang
27c02a0b80 d3d12decoder: Hold reference pictures in fence data
Keep reference pictures alive during executing decoding commands

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6709>
2024-04-23 10:08:19 +00:00
Seungha Yang
0f5f170a40 d3d12vp9dec: Disallow resolution change to larger size on non-keyframe
Intel GPU seems to be crashing if the case happens.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6709>
2024-04-23 10:08:18 +00:00
Seungha Yang
700c00eda3 d3d12decoder: Fix potential use after free
A DPB buffer held by codec picture object may not be writable
at the moment, then gst_buffer_make_writable() will unref passed buffer.

Specifically, the use after free or double free can happen if:
* Crop meta of buffer copy is required because of non-zero
  top-left crop position
* zero-copy is possible with crop meta
* A picture was duplicated, interlaced h264 stream for example

Interlaced h264 stream with non-zero top-left crop position
is not very common but it's possible configuration in theory.

Thus gst_buffer_make_writable() should be called with
GstVideoCodecFrame.output_buffer directly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6706>
2024-04-22 13:28:06 +00:00
Seungha Yang
5179cbccfa d3d12testsrc: Use shared 11on12 device
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6697>
2024-04-20 04:16:48 +09:00
Seungha Yang
cd97275a19 d3d12device: Hold d3d11on12 device to be shared
d3d11on12 device seems to be occupying a bit of GPU memory
Hold the instance in GstD3D12Device so that it can be shared

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6697>
2024-04-20 04:16:41 +09:00
Seungha Yang
8e4fe98361 d3d12videosink: Handle mouse double click and modifier
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6693>
2024-04-19 16:11:30 +00:00
Seungha Yang
dabce9635d d3d12videosink: Disconnect window's signal on dispose
Same as the commit of 7b69d1758f
but for d3d12videosink.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6692>
2024-04-19 21:25:06 +09:00
Seungha Yang
6659bd6362 d3d12videosink: Handle external HWND's mouse/keyboard events
OS will not propagate the event to child HWND if it's handled by
the parent. Thus, navigation event should be handled by parent HWND's
event handler.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6692>
2024-04-19 21:24:52 +09:00
Seungha Yang
f6ec4e6256 d3d12overlaycompositor: Add support for d3d12 memory
Don't allocate d3d12 texture if overlay is d3d12 memory already

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6620>
2024-04-14 09:55:41 +00:00
Seungha Yang
13d892d336 d3d12: Move gst_d3d12_buffer_copy_into method to library
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6620>
2024-04-14 09:55:41 +00:00
Seungha Yang
e597dcdf0d d3d12encoder: Fix buffer pool leak
Add missing buffer pool release

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6628>
2024-04-13 14:55:09 +00:00
Seungha Yang
ad6670ba81 d3d12decoder: Always output sharable texture
Because shared heap's additional costs is not significant,
use D3D12_HEAP_FLAG_SHARED for resource can be shared over process
boundary. And enables render target for d3d11 interop in the process.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6513>
2024-04-03 02:05:25 +00:00
Seungha Yang
d00c26cdc8 d3d12screencapturesrc: Use gst_d3d12_memory_get_d3d11_texture()
... and use fence to wait for GPU sync

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/6499>
2024-04-02 03:43:59 +09:00