`cudacodec`: `VideoWriter` fails to compile with Nvidia Video Codec SDK 12.2 about opencv_contrib

[vbickov]

Just FYI, maybe useful.

According to the latest ncEncodeAPI.h the parameter pixelBitDepthMinus8 has been replaced with two new ones:

`NV_ENC_BIT_DEPTH outputBitDepth; /**< [in]: Specifies pixel bit depth of encoded video. Should be set to NV_ENC_BIT_DEPTH_8 for 8 bit, NV_ENC_BIT_DEPTH_10 for 10 bit.
SW will do the bitdepth conversion internally from inputBitDepth -> outputBitDepth if bit depths differ
Support for 8 bit input to 10 bit encode conversion only*/

NV_ENC_BIT_DEPTH inputBitDepth; /**< [in]: Specifies pixel bit depth of video input. Should be set to NV_ENC_BIT_DEPTH_8 for 8 bit input, NV_ENC_BIT_DEPTH_10 for 10 bit input.*/`

I've modified NvEncoder.cpp:

Line 119:

pIntializeParams->encodeConfig->encodeCodecConfig.hevcConfig.inputBitDepth = pIntializeParams->encodeConfig->encodeCodecConfig.hevcConfig.outputBitDepth = (m_eBufferFormat == NV_ENC_BUFFER_FORMAT_YUV420_10BIT || m_eBufferFormat == NV_ENC_BUFFER_FORMAT_YUV444_10BIT) ? NV_ENC_BIT_DEPTH_10 : NV_ENC_BIT_DEPTH_8;

Line 177 to:

if (yuv10BitFormat && pEncoderParams->encodeConfig->encodeCodecConfig.hevcConfig.outputBitDepth != NV_ENC_BIT_DEPTH_10)

This seems fixed the issue. However, I'm not sure about the inputBitDepth parameter. I'm not even sure what meaning this parameter has for the encoder.