Number of threads used during auto-tunning

Questions
#1

Can we config the number of threads of remote devices during auto-tunning, or it is just the value determined by MaxConcurrency in tvm/src/runtime/threading_backend.cc?
@merrymercy

2 Likes
autotvm.RPCRunner and TVM_NUM_THREADS
Limit CPU cores for Auto tuned Model
#2

There is a config function in

from python side, we can set this by

tracker = tvm.rpc.connect_tracker(args.host, args.port)                 
remote = tracker.request(args.rpc_key)
config_func = remote.get_function('runtime.config_threadpool')

config_func(1, 1) # use 1 big core
config_func(1, 2) # use 2 big cores
config_func(-1, 1) # use 1 small core
config_func(-1, 2) # use 2 small cores

For auto-tuning, there is no interface for this. Maybe you can add these code after this line.

3 Likes
Android big.LITTLE core control
Can TVM split work into different layers, and assign layers into different cores?
Use all cores in a big.LITTLE architecture
#3

I know this thread is old, but I have found another solution. Maybe it will help others. I have a setup with android device which runs rpc server(binary executable), and host machine. To specify number of threads in android runtime, I used env var: TVM_NUM_THREADS, example:

TVM_NUM_THREADS=1 LD_LIBRARY_PATH=. ./tvm_rpc server --host=10.70.225.253

Then tvmc run will print following output:

Configuration
-------------
Number of threads: 1
Executor: Graph

Execution time summary: