OpenVX 源码分析-- 图的执行(TI / Sample)

TI openVX

Graph的执行

关键源代码:
vx_graph_pipeline.c

ownGraphScheduleGraph()

            /* trigger graph execution by scheduling the head nodes
             * Head nodes will trigger further nodes execution after
             * their completion
             * This will continue until leaf nodes executes
             * After a leaf node executes, it will send a completion
             * event.
             * After all completion events are received, a graph is
             * considered to have
             * executed
             */
            for(node_id=0; node_idnum_head_nodes; node_id++)
            {
                ownNodeKernelSchedule(graph->head_nodes[node_id], graph_obj_desc->pipeline_id);
            }
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

vx_target.c

    while(target->targetExitRequest == (vx_bool)vx_false_e)
   {
           status = tivxTargetDequeueObjDesc(target,
                   &obj_desc_id, TIVX_EVENT_TIMEOUT_WAIT_FOREVER);
               switch(obj_desc->type)
               {
                   case (vx_enum)TIVX_OBJ_DESC_CMD:
                       if( tivxObjDescIsValidType( obj_desc, TIVX_OBJ_DESC_CMD) != 0)
                       {
                           tivxTargetCmdDescHandler((tivx_obj_desc_cmd_t*)obj_desc);
                       }
                       break;
                   case (vx_enum)TIVX_OBJ_DESC_NODE:
                       if( tivxObjDescIsValidType( obj_desc, TIVX_OBJ_DESC_NODE) != 0)
                       {
                           tivxTargetNodeDescNodeExecute(target, (tivx_obj_desc_node_t*)obj_desc);
                       }
                       break;
                   default:
                       /* unsupported obj_desc received at target */
                       break;
               }
   }
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

Node 的执行 ownNodeUserKernelExecute

Node Exe源代码

ownNodeUserKernelExecute()

vx_reference params[TIVX_KERNEL_MAX_PARAMS];	//TIVX_KERNEL_MAX_PARAMS:64
vx_reference parent_ref[TIVX_KERNEL_MAX_PARAMS];
status = node->kernel->function(node, prm_ref, num_params);
1
2
3
4
5

解读：
Node 的执行参数是 最大64个，必须是固定格式的（vx_type_e中的一个）。