FaSTTUBe/GST-Tensordecoder-ov_ep
37
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nvdsdewarp: Avoid synchronization if possible

Browse Source 
If input/output memory objects have the same cuda stream,
don't need to synchronize stream

Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/8558>
This commit is contained in:
Seungha Yang 2025-02-26 02:27:38 +09:00
parent 6d2530fb37
commit 608e57e542
1 changed files with 32 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
subprojects/gst-plugins-bad/ext/nvdswrapper/gstnvdsdewarp.cpp
View File

@ -1650,11 +1650,33 @@ gst_nv_ds_dewarp_transform (GstBaseTransform * trans, GstBuffer * inbuf,
}
CUstream cuda_stream = 0;
auto stream = gst_cuda_memory_get_stream (in_cmem);
if (stream)
cuda_stream = gst_cuda_stream_get_handle (stream);
else
cuda_stream = gst_cuda_stream_get_handle (priv->stream);
auto in_stream = gst_cuda_memory_get_stream (in_cmem);
auto out_cmem = GST_CUDA_MEMORY_CAST (out_mem);
auto out_stream = gst_cuda_memory_get_stream (out_cmem);
GstCudaStream *selected_stream = nullptr;
/* If downstream does not aware of CUDA stream (i.e., using default stream) */
if (!out_stream) {
if (in_stream) {
GST_TRACE_OBJECT (self, "Use upstram CUDA stream");
selected_stream = in_stream;
} else if (priv->stream) {
GST_TRACE_OBJECT (self, "Use our CUDA stream");
selected_stream = priv->stream;
}
} else {
selected_stream = out_stream;
if (in_stream) {
if (in_stream == out_stream) {
GST_TRACE_OBJECT (self, "Same stream");
} else {
GST_TRACE_OBJECT (self, "Different CUDA stream");
gst_cuda_memory_sync (in_cmem);
}
}
}
cuda_stream = gst_cuda_stream_get_handle (selected_stream);
auto data = (guint8 *) GST_VIDEO_FRAME_PLANE_DATA (&out_frame, 0);
auto stride = GST_VIDEO_FRAME_PLANE_STRIDE (&out_frame, 0);
@ -1676,7 +1698,11 @@ gst_nv_ds_dewarp_transform (GstBaseTransform * trans, GstBuffer * inbuf,
auto ret = nvwarpWarpBuffer (priv->handle, (cudaStream_t) cuda_stream,
(cudaTextureObject_t) texture, data + offset, stride);
CuStreamSynchronize (cuda_stream);
if (selected_stream != out_stream) {
GST_MEMORY_FLAG_UNSET (out_cmem, GST_CUDA_MEMORY_TRANSFER_NEED_SYNC);
GST_TRACE_OBJECT (self, "Waiting for convert sync");
CuStreamSynchronize (cuda_stream);
}
gst_cuda_context_pop (nullptr);
GstFlowReturn flow_ret = GST_FLOW_OK;