Commit Graph

54 Commits

Author SHA1 Message Date
Geoffrey McRae
0ca760fad6 [c-host] revert locking series, this needs more thought
Revert "[c-host] make pointer post function thread safe"

This reverts commit 3feed7ba07.

Revert "[c-hots] fix incorrect unlock timing"

This reverts commit 57f1f2d1fe.

Revert "[c-host] increase the queue length and remove debug output"

This reverts commit b0f9f15a60.

Revert "[c-host] dxgi: use low level mouse input by default"

This reverts commit dc4d820666.

Revert "[c-host] nvfbc: no need for a cursor position event with LGMP"

This reverts commit e30b54ddb2.
2020-01-29 23:23:31 +11:00
Geoffrey McRae
57f1f2d1fe [c-hots] fix incorrect unlock timing 2020-01-29 22:12:59 +11:00
Geoffrey McRae
dc4d820666 [c-host] dxgi: use low level mouse input by default
This is known to prevent cursor updates on the secure desktop
(UAC) but DXGI DD does not provide us with the real mouse
coordinates when applications have 'captured' the cursor.
2020-01-29 21:58:39 +11:00
Geoffrey McRae
e30b54ddb2 [c-host] nvfbc: no need for a cursor position event with LGMP 2020-01-29 21:58:00 +11:00
Geoffrey McRae
939bb07603 [all] cleanup use of atomic locking and switch to C11 stdatomic 2020-01-29 19:06:09 +11:00
Geoffrey McRae
e30e5da75a [c-host] nvfbc: correct frame change check logic 2020-01-27 16:16:43 +11:00
Geoffrey McRae
fc6681306e [c-host] nvfbc: do not send frames that have not changed 2020-01-27 16:01:31 +11:00
Geoffrey McRae
bced5f95ff [all] make cursor visible a flag and send it seperate to position 2020-01-27 02:07:32 +11:00
Geoffrey McRae
fea0a98b9e [c-host] dxgi: invisible cursors do not have position information 2020-01-27 01:47:40 +11:00
Geoffrey McRae
b7e4426002 [c-host] inform the client if we have positional cursor information 2020-01-26 17:25:14 +11:00
Geoffrey McRae
17e05c6fd5 [all] expose the FrameBuffer struct for correct sizeof calculations 2020-01-13 19:30:49 +11:00
Geoffrey McRae
7de030bb69 [c-host] nvfbc: free event on deinit 2020-01-12 18:09:11 +11:00
Geoffrey McRae
b5d91ccc21 [c-host] nvfbc: fixed invalid nvfbc init 2020-01-11 22:28:52 +11:00
Geoffrey McRae
0eafa7de5d [c-host] update NvFBC to use new capture interface 2020-01-11 21:51:59 +11:00
Geoffrey McRae
e7239c53fd [c-host] cleanup dxgi cursor code a bit 2020-01-09 21:20:01 +11:00
Geoffrey McRae
7a98a886b6 [c-host] use the correct buffer for the cursor shape 2020-01-09 20:27:55 +11:00
Geoffrey McRae
b0fb7177bb [c-host] improved intial connection sync 2020-01-09 19:49:47 +11:00
Geoffrey McRae
0b8f1a18b2 [LGMP] start of c-host conversion to use LGMP 2020-01-09 15:42:32 +11:00
Geoffrey McRae
5db4c32035 [c-host] dont use DX12 feature levels on Windows8
Fixes #218
2020-01-06 20:53:15 +11:00
Geoffrey McRae
0dc0e6490c [c-host] dxgi: check for failure of getDesc1 2020-01-03 17:29:07 +11:00
Geoffrey McRae
0c6ff6822d [common/c-host] move agnostic code into common library 2020-01-02 22:21:42 +11:00
Geoffrey McRae
0bd19cfd38 [c-host] dxgi: fix segfault with maxTextures=1 on client reconnect 2019-12-17 20:56:14 +11:00
Geoffrey McRae
8ada29e25f [c-host] nvfbc: fix build attempt 2 :) 2019-12-17 16:42:48 +11:00
Geoffrey McRae
3b5c1bd09c [c-host] nvfbc: fix failure to build due to new event interface 2019-12-17 16:41:02 +11:00
Geoffrey McRae
c82a5e0523 [c-host] dxgi: futher event improvements 2019-12-17 16:36:43 +11:00
Geoffrey McRae
9c5f9906fa [c-host] add spinlock support to events and alter dxgi to use them 2019-12-17 14:59:58 +11:00
Geoffrey McRae
db2f5b85a9 [c-host] dxgi: added new useAcquireLock option for quirked GPUs 2019-12-17 13:45:08 +11:00
Geoffrey McRae
711fbc549a [c-host] dxgi: interlock so we can map outside of the capture thread 2019-12-16 15:18:26 +11:00
Geoffrey McRae
f85c017184 [c-host] DXGI profiled and tuned again :) 2019-12-15 16:21:21 +11:00
Geoffrey McRae
bca54ab1f6 [client/host] added new asyncronous memory copy
This changes the method of the memory copy from the host application to
the guest. Instead of performing a full copy from the capture device
into shared memory, and then flagging the new frame, we instead set a
write pointer, flag the client that there is a new frame and then copy
in chunks of 1024 bytes until the entire frame is copied. The client
upon seeing the new frame flag begins to poll at high frequency the
write pointer and upon each update copies as much as it can into the
texture.

This should improve latency but also slightly increase CPU usage on the
client due to the high frequency polling.
2019-10-09 13:53:02 +11:00
Geoffrey McRae
6d24dd52d6 [c-host] not all versions of mingw support wcstombs_s
While the _s functions are for security as they avoid exceeding the
supplied buffer, in our case they are not really required as we are
allocating a buffer large enough to store the entire result.

Fixes #171
2019-06-12 15:31:18 +10:00
Geoffrey McRae
780bb248f7 [c-host] dxgi: fix invalid cursor type define 2019-05-28 15:17:11 +10:00
Geoffrey McRae
7d26027752 [c-host] resend the last on client reconnect if a timeout occurs 2019-05-28 14:24:48 +10:00
Geoffrey McRae
3d426ccef8 [all] fix missing cursor when client reconnects 2019-05-28 14:06:15 +10:00
Geoffrey McRae
5e201a32ca [c-host] dxgi: allow out of order frame mapping 2019-05-27 15:26:58 +10:00
Geoffrey McRae
21b02efb4d [c-host] dxgi: don't stall the GPU pipeline to map textures to ram
ID3D11DeviceContext_Map by default will force a CPU sync if the prior call to
CopyResource has not completed, this change defers the mapping and sets the
D3D11_MAP_FLAG_DO_NOT_WAIT when attempting to map the texture allowing the
capture to continue without incurring an expensive CPU/GPU sync.

A new tuneable has also been added

  * dxgi:maxTextures
2019-05-26 23:36:17 +10:00
Geoffrey McRae
1b031582a4 [c-host] nvfbc: expose option to disable cursor decoupling 2019-05-23 15:12:28 +10:00
Geoffrey McRae
afe072adf1 [c-host] nvfbc: print out the SDK version and enable NvFBC 2019-05-23 14:49:38 +10:00
Geoffrey McRae
773dd7773b [c-host] try to be more compatible with mingw headers 2019-05-22 19:46:18 +10:00
Geoffrey McRae
108c7d3aaa [c-host] fixed project for cross compliation for Win on Linux
Thanks @fatalis for your guidance on this
2019-05-22 14:59:19 +10:00
Geoffrey McRae
cc6dd58778 [c-host] windows: fix dxgi option struct syntax 2019-05-17 09:27:04 +10:00
Geoffrey McRae
5677117c0d [c-host] nvfbc: remove debug line from prior commit 2019-05-09 22:53:02 +10:00
Geoffrey McRae
558ae5dc45 [c-host] dxgi: initialize option to NULL for consistancy 2019-05-09 22:51:12 +10:00
Geoffrey McRae
83f63f4c42 [c-host] dxgi: add the ability to specify the adapter and/or output.
Fixes #132
2019-05-09 22:48:39 +10:00
Geoffrey McRae
338bc2e0dc [c-host] nvfbc: disable ARGB10 until NVIDIA fix the API (if ever) 2019-04-11 19:30:42 +10:00
Geoffrey McRae
8cedad8241 [c-host] fix NvFBC build after moving headers 2019-04-11 17:15:17 +10:00
Geoffrey McRae
28b12c85f4 [build] make "common" a static library (part 1/2) 2019-04-11 11:03:30 +10:00
Geoffrey McRae
bee221c18d [c-host] add ability to provide NvFBC privData 2019-04-10 22:23:56 +10:00
Geoffrey McRae
a03075416c [c-host] cmake: fix bad method of setting CFLAGS 2019-04-10 21:20:15 +10:00
Geoffrey McRae
e4d8cf2d76 [c-host] remove extra NvFBC noise 2019-04-10 21:14:11 +10:00