get dead lock when using`c_predict_api` and multi-threads(>2)

[PapaMadeleine2022]

Hello, I have a problem:
I use c_predict_api to infer batch of images using multi-threads(>2), but after some minutes, I get dead lock in MXPredGetOutput. This problem is also mentioned in issue, but not solved. Anyone can give some advises? Thank you very much~

Linux distro and version:
LSB Version: :core-4.1-amd64:core-4.1-noarch
Distributor ID: CentOS
Description: CentOS Linux release 7.4.1708 (Core)

Other envirs:
mxnet: v1.0.0/v1.4.x
GPU type: Tesla v100/p40
nvidia driver version: NVIDIA-SMI 410.48/384.81
CUDA version: 9.0/8.0
CUDNN version: 7.5.0/5.0

[PapaMadeleine2022]

need help...

[vrakesh]

@IvyGongoogle MXNet in its current form is not thread safe, this is something we are looking to resolve going forward

@mxnet-label-bot add [C API, Thread Safety]

[PapaMadeleine2022]

@vrakesh thanks for your reply. Please notice when this thread safe is solved. Thank you very much~

[ddavydenko]

@mxnet-label-bot delete [C API] add [Backend].
Removing C API label as this issue is not specific to API being used but rather is a lack of thread safety in the engine.