wip: CUDA with nvjpeg2k

1 files changed, 72 insertions, 0 deletions

diff --git a/src/lib/cuda_j2k_encoder_thread.h b/src/lib/cuda_j2k_encoder_thread.h
new file mode 100644
index 000000000..245d34328
--- /dev/null
+++ b/src/lib/cuda_j2k_encoder_thread.h
@@ -0,0 +1,72 @@
+/*
+    Copyright (C) 2025 Carl Hetherington <cth@carlh.net>
+
+    This file is part of DCP-o-matic.
+
+    DCP-o-matic is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    DCP-o-matic is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License
+    along with DCP-o-matic.  If not, see <http://www.gnu.org/licenses/>.
+
+*/
+
+
+#ifndef DCPOMATIC_CUDA_J2K_ENCODER_THREAD
+#define DCPOMATIC_CUDA_J2K_ENCODER_THREAD
+
+
+#include "j2k_sync_encoder_thread.h"
+#include <dcp/array_data.h>
+#include <dcp/types.h>
+#include <nvjpeg2k.h>
+
+
+class DCPVideo;
+
+
+/** @class CUDAJ2KEncoderThread
+ *  @brief J2K encoder thread using CUDA via nvidia's nvjpeg2k library.
+ *
+ *  Each CUDAJ2KEncoderThread sends frames for encoding to a different CUDA
+ *  stream.  The thread sends the image and blocks until the result is ready.
+ *
+ *  This is different to the Grok encoder (that also uses CUDA but via an
+ *  additional paid-for tool called Grok).
+ */
+class CUDAJ2KEncoderThread : public J2KSyncEncoderThread
+{
+public:
+	CUDAJ2KEncoderThread(J2KEncoder& encoder);
+	~CUDAJ2KEncoderThread();
+
+	std::shared_ptr<dcp::ArrayData> encode(DCPVideo const& frame) override;
+
+private:
+	void allocate_pixel_data_d(dcp::Size size);
+	void free_pixel_data_d();
+
+	nvjpeg2kEncoder_t _encoder_handle;
+	nvjpeg2kEncodeState_t _encode_state;
+	nvjpeg2kEncodeParams_t _encode_params;
+	cudaStream_t _stream;
+
+	uint8_t* _pixel_data_d[3] = { nullptr, nullptr, nullptr };
+	size_t _pitch_in_bytes_d[3];
+	dcp::Size _pixel_data_d_size;
+
+	std::vector<int16_t> _xyz_x;
+	std::vector<int16_t> _xyz_y;
+	std::vector<int16_t> _xyz_z;
+};
+
+
+#endif
+