Commit Graph

216 Commits

Author SHA1 Message Date
Geoffrey McRae
30c577beeb [host] all: make the downsample rule matching common 2023-11-10 06:28:05 +11:00
Geoffrey McRae
6c7f3c4197 [host] nvfbc: make the downsampleParser available outside of NvFBC 2023-11-10 06:28:05 +11:00
Tudor Brindus
ea5b6b4026 [host] DXGI: implement damage-aware RGB24 copy 2023-11-10 06:28:05 +11:00
Tudor Brindus
6329779893 [host] DXGI: rescale RGB24 texture to 3/4ths the input width
Now that data isn't packed across rows, we can decrease the amount of
texture memory we require.
2023-11-10 06:28:05 +11:00
Tudor Brindus
3106d0e3e2 [host] DXGI: stop packing data across rows
This is a precursor to allowing damage-aware RGB24 copies.
2023-11-10 06:28:05 +11:00
Tudor Brindus
d44fc36fc4 [host] DXGI: stop rescaling RGB24 texture height
For the moment, this just increases texture memory usage, but does not
affect behavior.

In a future commit, I will modify the shaders to not pack data across
rows, in order to enable damage copies.
2023-11-10 06:28:05 +11:00
Tudor Brindus
c29404eea6 [host] DXGI: fixed swapped rows and cols variables
In practice this worked out because `rows = cols`, but this will change
in future commits as I implement RGB24 damage support.
2023-11-10 06:28:05 +11:00
Geoffrey McRae
578d98fd22 [host] DXGI: initial implementation of RGB24 support
This commit breaks damage tracking and the dx12 backend and is not in
a state where it should be used by the general public.
2023-11-10 06:28:05 +11:00
Geoffrey McRae
b3879ff1d7 [host] windows: the vertex shader is common to all post-processors 2023-11-10 06:28:05 +11:00
Geoffrey McRae
1e30539fb2 [dxgi] cache the input shader resource view of the src texture 2023-10-29 21:53:48 +11:00
Geoffrey McRae
52410beea7 [host] dxgi: check for correct comRef usage 2023-10-29 20:29:14 +11:00
Geoffrey McRae
21cd380cad [host] dxgi: seperate out and implement a post processor chain 2023-10-29 20:27:17 +11:00
Geoffrey McRae
e225f66cee [host] dxgi: explicitly flush the pipeline to reduce latency 2023-10-27 22:08:33 +11:00
Geoffrey McRae
2206752b66 [host] dxgi: fix d3d12 backend resource leak
The handle is only needed so we can open the resource, once we have
it we can close the handle. We then cache the shared resource for
future reuse if possible.
2023-10-27 21:56:49 +11:00
Geoffrey McRae
0510d06c4b [host] dxgi: fix d3d12 texture sharing when in HDR 2023-10-27 21:22:00 +11:00
Geoffrey McRae
699d95818d [host] dxgi: upate d3d12 backend to use comRef 2023-10-27 21:01:44 +11:00
Geoffrey McRae
fffac35300 [host] dxgi: update d3d11 backend to make use of comRef 2023-10-27 18:44:17 +11:00
Geoffrey McRae
35b0f8edf3 [host] dxgi: allow the HDR texture to be shared with other backends 2023-10-27 18:43:26 +11:00
Geoffrey McRae
544164f637 [host] dxgi: allow the backend to shutdown before freeing comRef globals 2023-10-27 18:42:38 +11:00
Geoffrey McRae
b94166177f [host] dxgi: update to make use of comRef (part 1 of 2) 2023-10-27 17:34:34 +11:00
Geoffrey McRae
69b984aa2c [host] dxgi: add helper to manage COM object memory 2023-10-27 17:33:40 +11:00
Geoffrey McRae
47329ebd89 [host] dxgi: move utility functions into util.c 2023-10-27 11:28:37 +11:00
Geoffrey McRae
b41840b010 [host] all: tell the client if HDR PQ is in use or not 2023-10-27 01:11:57 +11:00
Geoffrey McRae
200b7b732c [host] dxgi: properly convert HDR16 to HDR10
DXGI doesn't take into account the SDRWhiteLevel that has already been
applied to the monitor when it converts to HDR10 which results in clipping.
This change set implements a HLSL shader to reverse this while at the same
time converting to HDR10.

This is still not perfect but far better then doing nothing.
2023-10-26 22:57:41 +11:00
Geoffrey McRae
c5923b9b4d [host] dxgi: obtain HDR color space and monitor information 2023-10-22 17:38:42 +11:00
Geoffrey McRae
bde2eef175 [host] dxgi: fix format for HDR10 when it needs correcting 2023-10-22 04:01:51 +11:00
Geoffrey McRae
ccd0a0bcf9 [host] dxgi: use HDR10 instead of HDR16 to save bandwidth 2023-10-22 03:47:32 +11:00
Geoffrey McRae
742e41c2c3 [host] dxgi: fix HDR damage aware copy 2023-10-22 03:15:12 +11:00
Geoffrey McRae
12d051d8c0 [host] nvfbc: do not try to init on non NVIDIA hardware 2023-10-22 01:46:57 +11:00
Geoffrey McRae
a6a6b8779a [host] nvfbc: fix use of initialized variable 2023-10-22 01:21:20 +11:00
Geoffrey McRae
128a8938c6 [host] nvfbc: remove useless update of member 2023-10-20 22:12:42 +11:00
Geoffrey McRae
2e515657dd [all] update/add license headers 2023-10-20 15:36:34 +11:00
Geoffrey McRae
99333a03c1 [host] nvfbc: fix failure to store last hdr state 2023-04-26 15:58:22 +10:00
Geoffrey McRae
45318aa653 [host] nvfbc: bump the format version when HDR status changes 2023-04-26 15:55:26 +10:00
Geoffrey McRae
f84165ac66 [host] app: report to the client if the frame is actually HDR 2023-04-26 15:48:01 +10:00
Geoffrey McRae
e4a8424fad [host] nvfbc: fix ARGB10 support
There is no performance penalty to using ARGB10 as it's still a 32-bit copy per
pixel and the nvidia driver internally handles any conversions required to make
this work for both SDR and HDR.
2023-04-19 19:52:30 +10:00
Geoffrey McRae
188f25c6bf [host] nvfbc: increase retry timeout to 1 second 2022-12-09 08:00:03 +11:00
Geoffrey McRae
8619f787b9 [host] nvfbc: retry on failure to init
@quantum has observed nvfbc under rare circumstances fail to initialize,
this adds a retry to the init with a short delay to hopefully recover
from this situation.
2022-12-08 21:24:11 +11:00
Quantum
a8521b821e [all] cmake: standardize indentation to 2 spaces 2022-05-29 11:16:16 +10:00
Geoffrey McRae
a41ab81a90 [host] nvfbc: try all NVIDIA adapters in the system
This will allow LG to start if the VM still has a virtual device as
the primary output.
2022-05-18 16:27:17 +10:00
Geoffrey McRae
0f8c0b5fb3 [host] dxgi: fix incorrect skip logic 2022-05-18 15:59:31 +10:00
Geoffrey McRae
57e20007db [host] dxgi: don't try to use devices without d3d support
This change should allow LG to work even if a virtual device is still
attached to the VM even though it might be capturing the wrong display.
2022-05-18 15:38:42 +10:00
Geoffrey McRae
87077dfe6e [host] nvfbc: add downscale support 2022-05-04 13:38:49 +10:00
Geoffrey McRae
60834a5719 [host] xcb/pipewire/nfbc: use min and correct inverted logic 2022-05-04 12:23:20 +10:00
Geoffrey McRae
d5e2689d64 [host] nvfbc: fix inverted truncated frame logic 2022-05-04 11:58:04 +10:00
Geoffrey McRae
e4a4e2331a [host] nvfbc: fix build due to frame info structure changes 2022-05-04 10:37:52 +10:00
Geoffrey McRae
8682ec207e [host] dxgi: check for invalid usage of d3d12 & downsampling 2022-05-04 10:23:42 +10:00
Geoffrey McRae
c8a5293645 [host] dxgi: refactor dxgi_copyFrame 2022-05-04 09:27:52 +10:00
Geoffrey McRae
15334c89d6 [host] dxgi: match the last entry in the downsample rule list 2022-05-03 17:58:41 +10:00
Geoffrey McRae
2eec459b47 [host] dxgi: finish downsample support for d3d11 backend 2022-05-03 12:15:24 +10:00
Geoffrey McRae
947325e00d [host] dxgi: fix failure to copy to texture with sub resources
`CopyResource` will silently fail when trying to copy from a texture
with no subresources to one with. Instead we must use `ResolveSubresource`
2022-05-02 15:53:20 +10:00
Geoffrey McRae
3134ec84de [host] dxgi: add support for downsampling the capture before sending
This is an experimental & incomplete feature for those using
supersampling. Anything > 1200p will be downsampled by 50% before
copying out of the GPU to save on memory bandwidth.

Unfinished! Has issues with damage tracking and currently can not
be configured. Only dx11 has been tested at this point, everything
else will likely have problems/crash.
2022-05-01 19:45:44 +10:00
vmfortress
36f97f08ad [host] dxgi: add nanosecond-scale sleep capability to d3d12
The nsleep() call lets d3d12 sleep for a more precise amount of
time while maintaining the current millisecond-scale sleep
interface in the configuration file.
2022-03-06 17:21:14 +11:00
Tudor Brindus
eeefc15e46 [host] build with -Wstrict-prototypes 2022-02-25 20:38:44 +11:00
Quantum
75ec3c0478 [host] nvfbc: shorten nvfbc_getName() result
To avoid client showing "Using    : NVFBC (NVidia Frame Buffer Capt".
This happens because the string is truncated to 31 characters to fit
in the char capture[32]; member of KVMFRRecord_VMInfo.
2022-01-26 23:06:47 +11:00
Geoffrey McRae
ccdf7b7c0e [host] app: report the full capture name to the client 2022-01-26 19:32:55 +11:00
Geoffrey McRae
99536eaf9d dxgi: check for memory leaks in d3d12 2022-01-19 09:58:14 +11:00
Geoffrey McRae
f8b4874799 dxgi: cosmetics 2022-01-19 09:58:14 +11:00
Geoffrey McRae
cff64ee7d3 dxgi: cache shared handles instead of re-creating them
This will cache up to 10 handles, in practice I have never seen DXGI
return anything but the same resource each time but we allow for more
anyway should MS change something in the future.

Should the cache get over filled it is disabled entirely and we revert
to the original behaviour.
2022-01-19 09:58:07 +11:00
Quantum
afd5e2d057 [host] dxgi: remove duplicate AcquireLock print 2022-01-17 15:12:54 +11:00
Quantum
508c491967 [host] dxgi: allow the debug layer to be turned on via config 2022-01-17 15:12:54 +11:00
Quantum
b117bbafe5 [host] dxgi: add ability to capture OutputDebugMessage
This is how Direct3D 12's debug layer outputs messages. Normally, these
are read by the debugger, but we want them logged instead.
2022-01-17 15:12:54 +11:00
Quantum
4c271f8744 [host] dxgi: move d3d12.h to vendor/directx 2022-01-17 15:12:54 +11:00
Chris Spencer
3d0a8f6987 [host] dxgi: fix frame damage method parameters
The buffer input sizes to the `IDXGIOutputDuplication` methods are measured
in bytes. This dramatically increases the number of dirty/move rects that
can be handled.
2022-01-13 09:34:14 +11:00
Chris Spencer
f145225dbc [host] dxgi: fix d3d11 assertion failure
`DEBUG_ASSERT(!this)` in `d3d11_create` is firing on the second
instantiation because we are not clearing `this` in `d3d11_free`.
2022-01-13 08:17:59 +11:00
vmfortress
7da2becfbd [host] dxgi: Replace standard asserts with DEBUG_ASSERT 2022-01-11 11:11:42 +11:00
Quantum
042a7d0925 [host] dxgi: add configurable sleep before D3D12 copy 2022-01-10 14:45:51 +11:00
Quantum
c69b19e68f [host] dxgi: add option to disable damage-aware copies 2022-01-10 14:45:51 +11:00
Quantum
cf7d501bc4 [host] dxgi: allow copy backend selection 2022-01-10 14:45:51 +11:00
Quantum
68e5b812a9 [host] dxgi: add preRelease callback
This is meant to avoid freeing the texture before the copy has finished.
2022-01-10 14:45:51 +11:00
Quantum
5a93f1e00c [host] dxgi: implement Direct3D 12 texture copy backend 2022-01-10 14:45:51 +11:00
Quantum
891f00a011 [host] dxgi: add d3d12.h from latest MinGW
This header was added in late 2020 and hasn't made its way into the GitHub
Actions image yet.
2022-01-10 14:45:51 +11:00
Quantum
137171a8a2 [host] dxgi: refactor to support additional copy backends 2022-01-10 14:45:51 +11:00
Quantum
a391e271c3 [host] dxgi: damage all textures when skipping frame 2022-01-09 16:36:26 +11:00
Johnathon Weaver
0f998582b9 [host] nvfbc: Fix dwmapi linking error
Fixed linking for DwmFlush and also rearranged as per how DXGI is.
2022-01-07 01:46:35 +11:00
Geoffrey McRae
2f8b139131 [host] windows: set DwmFlush default to off
This new feature while helps on some systems, others using freesync or
higher refresh rates where the capture can't keep up will limit to
fractions of the refresh rate. Better to disable and allow users to
opt-in.
2022-01-06 19:20:08 +11:00
Geoffrey McRae
b058cbe9fe [host] nvfbc: add DwmFlush here too as it makes a large difference 2022-01-06 19:01:29 +11:00
Geoffrey McRae
92f27cc0f0 [host] dxgi: use DwmFlush to sync to presentation interval
This change reduces the host GPU and CPU load by a large margin
improving guest system performance along with removing latency spikes
when moving the mouse. This is default enabled but can be disabled with
the new option `dxgi:dwmFlush=no` as it limits the capture rate to the
refresh rate of the guests output which may not be desireable.
2022-01-06 18:39:08 +11:00
Geoffrey McRae
952ebea2c5 [all] refresh copyright dates 2022-01-05 19:42:46 +11:00
Geoffrey McRae
ebf20dd108 [host] nvfbc: fix failure to startup 2022-01-05 19:31:47 +11:00
Geoffrey McRae
ba9f2b85b6 [host/client] kvmfr: update to include extra user data about the VM
This change allows the host to provide information to the client about
how the VM is configured, information such as the UUID, CPU
configuration and capture method both for informational display in the
client as well as debugging in the client's logs.

The format of the records allows this to be extended later with new
record types without needing to bump the KVMFR version.
2022-01-05 19:18:43 +11:00
WYF
b21d842f0e [host] nvfbc: add an option to specify adapter 2021-12-26 11:14:17 +11:00
Tudor Brindus
9872d2e407 [host] dxgi: fix typo in debug log message 2021-12-26 09:49:03 +11:00
Quantum
12461196c3 [host] nvfbc: fix comments in updateDamageRects 2021-12-24 15:30:04 +11:00
Geoffrey McRae
b515fa80d5 [host] nvfbc: be more intellegent when unionizing disjointed sets 2021-10-27 00:00:38 +11:00
Quantum
1f24ab0742 [host] nvfbc: avoid waking up pointer thread for no reason
If the wait times out, we used to simply restart the loop, which causes
it to check this->stop and exit if set to true. However, nvfbc_stop
already calls lgSignalEvent, which would wake up the pointer thread to
perform the check, so there is no need to set a timeout on the wait.
2021-09-29 17:47:02 +10:00
Quantum
4b3aaa7e0c [host] cmake: report capture backends enabled 2021-08-20 21:03:50 +10:00
Quantum
4ecf749f7e [host] remove all casts around malloc 2021-08-16 16:26:58 +10:00
Quantum
cdda89cef7 [host] use correct argument order for calloc 2021-08-16 16:25:59 +10:00
Tudor Brindus
982b4e6625 [host] use variable-relative sizeof where possible 2021-08-16 16:22:55 +10:00
Quantum
4f7ce91e7f [host] capture: switch all asserts to DEBUG_ASSERT 2021-08-14 12:19:07 +10:00
Quantum
10ee6cd031 [host] nvfbc: read nvfbc:diffRes option with the correct type 2021-08-14 09:21:34 +10:00
Quantum
e5d252290d [common] array: add ALIGN_PAD macro for common logic
ALIGN_PAD(x, a) returns x rounded up to the nearest multiple of a.
2021-08-14 08:05:29 +10:00
Quantum
acd5ce51db [host] dxgi: use FAILED macro instead of comparing against S_OK 2021-08-13 20:21:50 +10:00
Quantum
d3ea9662bf [host] nvfbc: remove rectangles that are entirely contained in others
This makes nvfbc report less useless damage and makes the client run faster.
2021-08-13 20:21:27 +10:00
Quantum
566c89e9d8 [host] dxgi: correctly count moved rectangles 2021-08-13 20:21:08 +10:00
Quantum
b3173bdddc [host] dxgi: include correct DXGI headers
The declarations in dxgi_extra.h are not missing, they are in dxgi1_2.h and
dxgi1_5.h, which exist in MinGW-w64 since 2017.
2021-08-12 12:35:45 +10:00
Quantum
61a4b0744d [host] dxgi: use standard MinGW libd3d11.a
MinGW has been shipping this file since 2014, and that version contains the
only function from the dll that we call: D3D11CreateDevice.
2021-08-12 11:58:34 +10:00
Quantum
9bded74543 [host] dxgi: use CopySubresourceRegion when possible
This commit adds damage tracking to the DXGI textures, and only copies the
damaged areas to the textures with ID3D11DeviceContext::CopySubresourceRegion.

The sleep logic in waitFrame makes it difficult for this to reduce the
latency, but removing it shows significant improvements (6-7 ms to ~3 ms)
when a tiny portion of the screen is damaged, while showing no difference on
full screen damage.
2021-08-11 19:01:52 +10:00
Quantum
cab95c5eed [common] rects: refactor rect buffer copy code to common module
This also fixes a bug of accidentally multiplying the stride by 4 when
the stride is already in bytes and not pixels.
2021-08-08 08:30:11 +10:00