IS CAPI multi-thread predict interface is actually exactly running sequence ?
Created by: veridone
AS the example in paddle/capi/examples/model_inference/multi_thread
I use all the default parameter make an executable program
time ./multi_thread trainer_config.bin models/ 2>/dev/null #it has 2 threads
real 0m0.332s
user 0m0.420s
sys 0m0.271s
time ./multi_thread trainer_config.bin models/ 4>/dev/null #it has 4 threads
real 0m0.684s
user 0m0.802s
sys 0m0.302s
0.684s/0.332s = 2 that means if I has four threads to run the program use just use the two times time of two threads it means is actually running sequence of the four threads ?