Tool/software:
Hello, we have been trying to implement 2d bit block transfer module utilizing tiovx. In order to achieve the goal,
we have referenced edgeai-tiovx-apps package's multi_graph_test.c on the github repository. The specific path to the source code is following.
https://github.com/TexasInstruments/edgeai-tiovx-apps/blob/develop/tests/app_tiovx_linux_multi_graph_test.c
We have successfully gotten the result from the code we devised that consists of 4 graphs. (Color convert - Multi Scale - Mosaic - Color convert) to be exact.
The thing is, inside the source code,
There are lines where
/*
* Note that in below loop graph on and 2 are running sequentially
* to run them in parellel, you need to decuple them by creating
* seperate threads for enqueue/dequeue of each graph
*/
for (int i = 0; i < APP_NUM_ITERATIONS; i++) {
those are stated. So, from my understanding, in order to achieve the optimal performance,
we need to somehow decouple them and call the queue/dequeue parts below on the separate threads for each graph.
/*
* Note that in below loop graph on and 2 are running sequentially
* to run them in parellel, you need to decuple them by creating
* seperate threads for enqueue/dequeue of each graph
*/
for (int i = 0; i < APP_NUM_ITERATIONS; i++) {
//Graph1 Execution
inbuf1 = v4l2_decode_dqueue_buf(v4l2_decode_handle);
outbuf1 = tiovx_modules_acquire_buf(out_buf_pool1);
tiovx_modules_enqueue_buf(inbuf1);
tiovx_modules_enqueue_buf(outbuf1);
inbuf1 = tiovx_modules_dequeue_buf(in_buf_pool1);
outbuf1 = tiovx_modules_dequeue_buf(out_buf_pool1);
v4l2_decode_enqueue_buf(v4l2_decode_handle, inbuf1);
//Graph2 Execution
inbuf2 = tiovx_modules_acquire_buf(in_buf_pool2);
outbuf2 = tiovx_modules_acquire_buf(out_buf_pool2);
//Swap input2 mem with output1 to feed graph1 out to graph2
tiovx_modules_buf_swap_mem(outbuf1, inbuf2);
tiovx_modules_release_buf(outbuf1);
tiovx_modules_enqueue_buf(inbuf2);
tiovx_modules_enqueue_buf(outbuf2);
inbuf2 = tiovx_modules_dequeue_buf(in_buf_pool2);
outbuf2 = tiovx_modules_dequeue_buf(out_buf_pool2);
kms_display_render_buf(kms_display_handle, outbuf2);
tiovx_modules_release_buf(inbuf2);
tiovx_modules_release_buf(outbuf2);
}
Here is the question.
1. What does it mean exactly with 'by creating separate threads'?
is the API compatible / thread safe with pthread or std::thread and such?
2. openvx graph should always operate with the vx_context, Isn't it? So, if I really
make several different threads to queue/dequeue the data for each graph, how should
I manage the context for it?
3. Is there any example/tutorial/sample code that I could reference to implement the multi thread
settings with multi graphs operation with tiovx?
