mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2025-01-29 18:48:44 +00:00
Code Issues 1 Projects Releases Wiki Activity

nvencoder: Add support for shared CUDA stream

Browse source 
Sets CUDA stream on CUDA buffer pool so that CUDA stream object
can be shared

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/3629>
This commit is contained in:
Seungha Yang 2022-12-20 00:54:42 +09:00 committed by GStreamer Marge Bot
parent d409c35367
commit 8d14194fdd
1 changed files with 53 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

66
subprojects/gst-plugins-bad/sys/nvcodec/gstnvencoder.cpp
View file

@ -246,7 +246,6 @@ gst_nv_encoder_reset (GstNvEncoder * self)
priv->session = NULL;
}
gst_clear_cuda_stream (&priv->stream);
g_queue_clear (&priv->free_tasks);
g_queue_clear (&priv->output_tasks);
@ -438,6 +437,8 @@ gst_nv_encoder_open (GstVideoEncoder * encoder)
GST_ERROR_OBJECT (self, "failed to create CUDA context");
return FALSE;
}
if (!priv->stream && gst_nvenc_have_set_io_cuda_streams ())
priv->stream = gst_cuda_stream_new (priv->context);
break;
default:
g_assert_not_reached ();
@ -453,6 +454,7 @@ gst_nv_encoder_close (GstVideoEncoder * encoder)
GstNvEncoder *self = GST_NV_ENCODER (encoder);
GstNvEncoderPrivate *priv = self->priv;
gst_clear_cuda_stream (&priv->stream);
gst_clear_object (&priv->context);
#ifdef GST_CUDA_HAS_D3D
gst_clear_d3d11_fence (&priv->fence);
@ -473,6 +475,7 @@ gst_nv_encoder_stop (GstVideoEncoder * encoder)
gst_nv_encoder_drain (self, FALSE);
if (priv->subclass_device_mode == GST_NV_ENCODER_DEVICE_AUTO_SELECT) {
gst_clear_cuda_stream (&priv->stream);
gst_clear_object (&priv->context);
#ifdef GST_CUDA_HAS_D3D
gst_clear_object (&priv->device);
@ -559,6 +562,7 @@ gst_nv_encoder_propose_allocation (GstVideoEncoder * encoder, GstQuery * query)
GstStructure *config;
GstCapsFeatures *features;
guint min_buffers;
gboolean use_cuda_pool = FALSE;
gst_query_parse_allocation (query, &caps, NULL);
if (!caps) {
@ -600,6 +604,7 @@ gst_nv_encoder_propose_allocation (GstVideoEncoder * encoder, GstQuery * query)
GST_CAPS_FEATURE_MEMORY_CUDA_MEMORY)) {
GST_DEBUG_OBJECT (self, "upstream support CUDA memory");
pool = gst_cuda_buffer_pool_new (priv->context);
use_cuda_pool = TRUE;
}
break;
default:
@ -616,6 +621,10 @@ gst_nv_encoder_propose_allocation (GstVideoEncoder * encoder, GstQuery * query)
size = GST_VIDEO_INFO_SIZE (&info);
gst_buffer_pool_config_set_params (config, caps, size, min_buffers, 0);
if (use_cuda_pool && priv->stream) {
/* Set our stream on buffer pool config so that CUstream can be shared */
gst_buffer_pool_config_set_cuda_stream (config, priv->stream);
}
if (!gst_buffer_pool_set_config (pool, config)) {
GST_WARNING_OBJECT (self, "Failed to set pool config");
@ -1004,6 +1013,15 @@ gst_nv_encoder_thread_func (GstNvEncoder * self)
}
GST_NV_ENCODER_LOCK (self);
/* Any pending GPU command associated with this memory must be finished
* at this point */
if (task->buffer && priv->stream) {
GstMemory *mem;
mem = gst_buffer_peek_memory (task->buffer, 0);
if (gst_is_cuda_memory (mem))
GST_MEMORY_FLAG_UNSET (mem, GST_CUDA_MEMORY_TRANSFER_NEED_SYNC);
}
gst_nv_encoder_task_reset (self, task);
priv->last_flow = ret;
g_cond_broadcast (&priv->cond);
@ -1154,6 +1172,9 @@ gst_nv_encoder_create_pool (GstNvEncoder * self, GstVideoCodecState * state)
config = gst_buffer_pool_get_config (pool);
gst_buffer_pool_config_set_params (config, state->caps,
GST_VIDEO_INFO_SIZE (&state->info), 0, 0);
if (priv->selected_device_mode == GST_NV_ENCODER_DEVICE_CUDA && priv->stream)
gst_buffer_pool_config_set_cuda_stream (config, priv->stream);
if (!gst_buffer_pool_set_config (pool, config)) {
GST_ERROR_OBJECT (self, "Failed to set pool config");
gst_object_unref (pool);
@ -1216,8 +1237,25 @@ gst_nv_encoder_init_session (GstNvEncoder * self, GstBuffer * in_buf)
priv->cuda_device_id = data.cuda_device_id;
priv->dxgi_adapter_luid = data.adapter_luid;
gst_clear_object (&priv->context);
if (data.device_mode == GST_NV_ENCODER_DEVICE_CUDA)
if (data.device_mode == GST_NV_ENCODER_DEVICE_CUDA) {
GstMemory *mem = gst_buffer_peek_memory (in_buf, 0);
priv->context = (GstCudaContext *) data.device;
gst_clear_cuda_stream (&priv->stream);
if (gst_nvenc_have_set_io_cuda_streams ()) {
if (gst_is_cuda_memory (mem)) {
/* Use upstream CUDA stream */
priv->stream =
gst_cuda_memory_get_stream (GST_CUDA_MEMORY_CAST (mem));
if (priv->stream)
gst_cuda_stream_ref (priv->stream);
}
if (!priv->stream)
priv->stream = gst_cuda_stream_new (priv->context);
}
}
#ifdef GST_CUDA_HAS_D3D
gst_clear_object (&priv->device);
if (data.device_mode == GST_NV_ENCODER_DEVICE_D3D11)
@ -1269,17 +1307,13 @@ gst_nv_encoder_init_session (GstNvEncoder * self, GstBuffer * in_buf)
}
if (priv->selected_device_mode == GST_NV_ENCODER_DEVICE_CUDA &&
gst_nvenc_have_set_io_cuda_streams ()) {
priv->stream = gst_cuda_stream_new (priv->context);
if (priv->stream) {
CUstream stream = gst_cuda_stream_get_handle (priv->stream);
status = NvEncSetIOCudaStreams (priv->session,
(NV_ENC_CUSTREAM_PTR) & stream, (NV_ENC_CUSTREAM_PTR) & stream);
if (status != NV_ENC_SUCCESS) {
GST_WARNING_OBJECT (self, "NvEncSetIOCudaStreams failed, status: %"
GST_NVENC_STATUS_FORMAT, GST_NVENC_STATUS_ARGS (status));
}
gst_nvenc_have_set_io_cuda_streams () && priv->stream) {
CUstream stream = gst_cuda_stream_get_handle (priv->stream);
status = NvEncSetIOCudaStreams (priv->session,
(NV_ENC_CUSTREAM_PTR) & stream, (NV_ENC_CUSTREAM_PTR) & stream);
if (status != NV_ENC_SUCCESS) {
GST_WARNING_OBJECT (self, "NvEncSetIOCudaStreams failed, status: %"
GST_NVENC_STATUS_FORMAT, GST_NVENC_STATUS_ARGS (status));
}
}
@ -1511,6 +1545,7 @@ gst_nv_encoder_prepare_task_input_cuda (GstNvEncoder * self,
GstMemory *mem;
GstCudaMemory *cmem;
NVENCSTATUS status;
GstCudaStream *stream;
mem = gst_buffer_peek_memory (buffer, 0);
if (!gst_is_cuda_memory (mem)) {
@ -1534,6 +1569,11 @@ gst_nv_encoder_prepare_task_input_cuda (GstNvEncoder * self,
}
cmem = (GstCudaMemory *) gst_buffer_peek_memory (task->buffer, 0);
stream = gst_cuda_memory_get_stream (cmem);
if (stream != priv->stream) {
/* different stream, needs sync */
gst_cuda_memory_sync (cmem);
}
task->register_resource.version = gst_nvenc_get_register_resource_version ();
task->register_resource.resourceType =