nvidia capture SDK v7.1

[Antoine Martin]

​https://developer.nvidia.com/capture-sdk

What’s new in NVIDIA Capture SDK 7.1

Linux

• NvFBC clients can now specify the rate at which NvFBC should attempt to sample frames.
• Added push model support to NvFBC blocking capture, clients can receive frames as soon as they are presented by applications.
• NvIFROpenGL headers and samples are removed from the SDK. Please refer to README for more details.

Windows

• NvFBCHwEncode headers and samples are removed from the SDK. Please refer to release notes for more details.
• NvIFR headers and samples are removed from the SDK. Please refer to release notes for more details.

[Antoine Martin]

Of particular interest:

[in] Enable push model for frame capture

When set to NVFBC_TRUE, the display server will generate frames whenever it receives a damage event from applications.

Setting this to NVFBC_TRUE will ignore ::dwSamplingRateMs.

Using push model with the NVFBC_*_GRAB_FLAGS_NOWAIT_IF_NEW_FRAME_READY capture flag should guarantee the shortest amount of time between an application rendering a frame and an NvFBC client capturing it, provided that the NvFBC client is able to process the frames quickly enough.

Note that applications running at high frame rates will increase CPU and GPU loads.

NVFBC_BOOL bPushModel;

[Antoine Martin]

See also #2786

[Antoine Martin]

r27633 adds push mode, enabled by default, which can be disabled with XPRA_NVFBC_PUSH=0.

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/2387