2 Copyright (C) 2022 Carl Hetherington <cth@carlh.net>
4 This file is part of DCP-o-matic.
6 DCP-o-matic is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2 of the License, or
9 (at your option) any later version.
11 DCP-o-matic is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with DCP-o-matic. If not, see <http://www.gnu.org/licenses/>.
23 #include "cuda_j2k_frame_encoder.h"
24 #include "dcpomatic_log.h"
25 #include "dcp_video.h"
26 #include "exceptions.h"
27 #include "player_video.h"
28 #include <dcp/openjpeg_image.h>
35 using boost::optional;
38 CUDAJ2KFrameEncoder::CUDAJ2KFrameEncoder()
40 nvjpeg2kEncoderCreateSimple(&_encoder_handle);
41 nvjpeg2kEncodeStateCreate(_encoder_handle, &_encoder_state);
42 nvjpeg2kEncodeParamsCreate(&_encoder_params);
44 cudaStreamCreateWithFlags(&_stream, cudaStreamNonBlocking);
48 CUDAJ2KFrameEncoder::~CUDAJ2KFrameEncoder()
50 cudaStreamDestroy(_stream);
54 CUDAJ2KFrameEncoder::Input::Input(DCPVideo const& vf, cudaStream_t stream)
58 _xyz = convert_to_xyz(vf.frame(), boost::bind(&Log::dcp_log, dcpomatic_log.get(), _1, _2));
60 for (int i = 0; i < 3; ++i) {
61 _pixel_data_h[i] = reinterpret_cast<uint8_t*>(_xyz->data(i));
64 auto const pitch = _xyz->size().width * 2;
66 for (int i = 0; i < 3; ++i) {
67 _pitch_in_bytes[i] = pitch;
68 auto status = cudaMallocPitch(
69 reinterpret_cast<void**>(&_pixel_data_d[i]),
75 if (status != cudaSuccess) {
76 throw CUDAError("cudaMallocPitch", status);
79 status = cudaMemcpy2DAsync(
86 cudaMemcpyHostToDevice,
90 if (status != cudaSuccess) {
91 throw CUDAError("cudaMemcpy2D", status);
95 _device_image.num_components = 3;
96 _device_image.pixel_data = reinterpret_cast<void**>(_pixel_data_d);
97 _device_image.pixel_type = NVJPEG2K_UINT16;
98 _device_image.pitch_in_bytes = reinterpret_cast<size_t*>(_pitch_in_bytes);
102 CUDAJ2KFrameEncoder::Input::Input(Input&& other)
103 : _index(other._index)
106 for (int i = 0; i < 3; ++i) {
107 _pixel_data_d[i] = other._pixel_data_d[i];
108 other._pixel_data_d[i] = nullptr;
109 _pitch_in_bytes[i] = other._pitch_in_bytes[i];
112 _device_image.num_components = other._device_image.num_components;
113 _device_image.pixel_data = reinterpret_cast<void**>(_pixel_data_d);
114 _device_image.pixel_type = NVJPEG2K_UINT16;
115 _device_image.pitch_in_bytes = reinterpret_cast<size_t*>(_pitch_in_bytes);
119 CUDAJ2KFrameEncoder::Input::~Input()
121 cudaFree(_pixel_data_d[0]);
122 cudaFree(_pixel_data_d[1]);
123 cudaFree(_pixel_data_d[2]);
127 optional<dcp::ArrayData>
128 CUDAJ2KFrameEncoder::encode(DCPVideo const& vf)
130 auto input = Input(vf, _stream);
132 auto const size = vf.frame()->out_size();
133 DCPOMATIC_ASSERT(!_size || size == *_size);
136 DCPOMATIC_ASSERT(!_resolution || vf.resolution() == *_resolution);
137 _resolution = vf.resolution();
139 nvjpeg2kImageComponentInfo_t info[3];
140 for (int i = 0; i < 3; ++i) {
141 info[i].component_width = _size->width;
142 info[i].component_height = _size->height;
143 info[i].precision = 12;
147 nvjpeg2kEncodeConfig_t config;
148 memset(&config, 0, sizeof(config));
149 config.stream_type = NVJPEG2K_STREAM_J2K;
150 config.color_space = NVJPEG2K_COLORSPACE_SRGB;
151 config.image_width = _size->width;
152 config.image_height = _size->height;
153 config.num_components = 3;
154 config.image_comp_info = reinterpret_cast<nvjpeg2kImageComponentInfo_t*>(&info);
155 config.code_block_w = 32;
156 config.code_block_h = 32;
157 config.irreversible = 0;
159 config.prog_order = NVJPEG2K_CPRL;
160 config.num_resolutions = *_resolution == Resolution::FOUR_K ? 7 : 6;
162 auto status = nvjpeg2kEncodeParamsSetEncodeConfig(_encoder_params, &config);
163 if (status != NVJPEG2K_STATUS_SUCCESS) {
164 throw CUDAError("nvjpeg2kEncodeParamsSetEncodeConfig", status);
168 status = nvjpeg2kEncodeParamsSetQuality(_encoder_params, 30);
169 if (status != NVJPEG2K_STATUS_SUCCESS) {
170 throw CUDAError("nvjpeg2kEncodeParamsSetQuality", status);
173 status = nvjpeg2kEncode(_encoder_handle, _encoder_state, _encoder_params, input.device_image(), _stream);
174 if (status != NVJPEG2K_STATUS_SUCCESS) {
175 throw CUDAError("nvjpeg2kEncode", status);
178 size_t compressed_size;
179 status = nvjpeg2kEncodeRetrieveBitstream(_encoder_handle, _encoder_state, nullptr, &compressed_size, _stream);
181 dcp::ArrayData output(compressed_size);
182 status = nvjpeg2kEncodeRetrieveBitstream(_encoder_handle, _encoder_state, output.data(), &compressed_size, _stream);
183 if (status != NVJPEG2K_STATUS_SUCCESS) {
184 throw CUDAError("nvjpeg2kEncodeRetrieveBitstream", status);
192 CUDAJ2KFrameEncoder::log_thread_start ()
194 LOG_TIMING("start-encoder-thread thread=%1", thread_id());
199 CUDAJ2KFrameEncoder::flush()