Commit graph

13 commits

Author SHA1 Message Date
Seungha Yang
39b9f79e11 d3d11: Implement memory pool
Major changes:
* GstD3D11Allocator: This allocator is now device-independent object
  which can allocate GstD3D11Memory object for any GstD3D11Device.
  User can get this object via gst_allocator_find(GST_D3D11_MEMORY_NAME)
* GstD3D11PoolAllocator: A new allocator implementation for texture pool.
  From now on GstD3D11BufferPool will make use of this memory pool allocator
  to avoid frequent texture reallocation. That usually happens because
  of buffer copy (gst_buffer_make_writable for example)

In addition to that, GstD3D11BufferPool will provide GstBuffer with
GstVideoMeta, because CPU access to a GstD3D11Memory without GstVideoMeta
is almost impossible since GPU drivers needs padding for stride alignment.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2097>
2021-03-24 15:44:42 +00:00
Seungha Yang
4e7e390cab d3d11decoder: Temporarily remove zero-copy related code
We will re-implement it based on memory pool

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2097>
2021-03-24 15:44:42 +00:00
Seungha Yang
e006366206 d3d11decoder: Enable high precision clock if needed
We've been doing retry with 1ms sleep if DecoderBeginFrame()
returned E_PENDING which means application should call
DecoderBeginFrame() again because GPU is busy.
The 1ms sleep() during retry would result in usually about 15ms delay
in reality because of bad clock precision on Windows.
To improve throughput performance, this commit will enable
high precision clock only for NVIDIA platform since
DecoderBeginFrame() call on the other GPU vendors seems to
succeed without retry.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2099>
2021-03-23 12:41:43 +00:00
Seungha Yang
824b0ce0f4 Revert "d3d11: Enable native multi-thread protection layer and make use of it"
This reverts commit 872b7f503c.

Native multi-thread protection layer seems to be consuming more CPU
resource than application side protection approach in some cases

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2095>
2021-03-20 17:53:58 +09:00
Seungha Yang
872b7f503c d3d11: Enable native multi-thread protection layer and make use of it
... instead of our own GRecMutex locking. In this way, any other
Direct3D11 client (MediaFoundation for example) can safely call
any Direct3D11 API even when we are sharing our Direct3D11 device
with others.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2092>
2021-03-18 16:37:37 +00:00
Seungha Yang
3f08fcb4ff d3d11decoder: WARNING if ID3D11VideoDevice is unavailable, not ERROR
gst_d3d11_decoder_new() method is also used for device capability
checking during plugin init. Although we are checking hardware
flag prior to that, it doesn't guarantee ID3D11VideoDevice interface.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2088>
2021-03-16 12:39:23 +00:00
Seungha Yang
309a940614 d3d11decoder: Do negotiation again per forward/reverse playback mode change
For reverse playback, we are always copying decoded
frame to downstream buffer. So the pool size can be
and need to be large enough.

In case that forward playback, however, we need to restrict
the max pool size for performance reason. Otherwise decoder
will keep copying decoded texture to downstream buffer pool
if decoding is faster than downstream throughput
performance and also there are queue element between them.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2083>
2021-03-16 05:54:41 +00:00
Seungha Yang
72345875b1 d3d11decoder: Refactor device open step and negotiation
* Remove redundant method arguments
* Don't allocate staging texture if downstream supports d3d11 memory

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
2021-03-14 17:34:20 +09:00
Seungha Yang
347d9ceb4e d3d11decoder: Move profile GUID handling into decoder object
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
2021-03-14 16:31:31 +09:00
Seungha Yang
1c1cfc4ba7 d3d11decoder: Get rid of private struct
Completely hide member variables

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
2021-03-14 16:30:19 +09:00
Seungha Yang
78e4251750 Revert "d3d11vp9dec: Add support for internal frame resizing"
This reverts commit 58a4c33a0e

We should use ID3D11VideoProcessor instead of shader
to avoid copy. We need to revisit this topic later

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
2021-03-14 16:30:15 +09:00
Seungha Yang
615f52f2f7 d3d11device: Hold ID3D11VideoDevice and ID3D11VideoContext object
... instead of QueryInterface-ing per elements. Note that
ID3D11VideoDevice and ID3D11VideoContext objects might not be available
if device doesn't support video interface.
So GstD3D11Device object will create those objects only when requested.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2079>
2021-03-14 14:44:55 +09:00
Seungha Yang
5b3e316039 d3d11: Port to C++
Direct3D11 objects are COM, and most COM C APIs are verbose
(C++ is a little better). So, by using C++ APIs, we can make code
shorter and more readable.
Moreover, "ComPtr" helper class (which is C++ only) can be
utilized, that is very helpful for avoiding error-prone COM refcounting
issue/leak.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/2077>
2021-03-14 13:05:22 +09:00
Renamed from sys/d3d11/gstd3d11decoder.c (Browse further)