Move MKLDNN debug message under MXNET_MKLDNN_DEBUG

[pengzhao-intel]

Description

Move the debug message (Allocate * bytes with malloc directly) under the environment variable to avoid the confusion for the end user.

@TaoLv

Before change:

(base) [patric@mlt-gpu206 image-classification]$ python benchmark_score.py --network=resnetv1-50 --batch-size=16
INFO:root:network: resnetv1-50
INFO:root:device: cpu(0)
/home/patric/develop/mkldnn-update/python/mxnet/module/base_module.py:67: UserWarning: Data provided by label_shapes don't match names specified by label_names ([] vs. ['softmax_label'])
warnings.warn(msg)
[17:09:25] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 147456 bytes with malloc directly
[17:09:25] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 589824 bytes with malloc directly
[17:09:25] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 2359296 bytes with malloc directly
[17:09:25] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 9437184 bytes with malloc directly
INFO:root:batch size 16, dtype float32, images/sec: 209.347207

After change:

(base) [patric@mlt-gpu206 image-classification]$ python benchmark_score.py --network=resnetv1-50 --batch-size=16
INFO:root:network: resnetv1-50
INFO:root:device: cpu(0)
/home/patric/develop/mkldnn-update/python/mxnet/module/base_module.py:67: UserWarning: Data provided by label_shapes don't match names specified by label_names ([] vs. ['softmax_label'])
warnings.warn(msg)
INFO:root:batch size 16, dtype float32, images/sec: 206.424535

With the ENV setting:

base) [patric@mlt-gpu206 image-classification]$ export MXNET_MKLDNN_DEBUG=1
(base) [patric@mlt-gpu206 image-classification]$ python benchmark_score.py --network=resnetv1-50 --batch-size=16
INFO:root:network: resnetv1-50
INFO:root:device: cpu(0)
/home/patric/develop/mkldnn-update/python/mxnet/module/base_module.py:67: UserWarning: Data provided by label_shapes don't match names specified by label_names ([] vs. ['softmax_label'])
warnings.warn(msg)
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Convolution
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test BatchNorm
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Activation
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Pooling
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Convolution
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test BatchNorm
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Activation
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:74: Allocate 147456 bytes with malloc directly
[17:11:51] src/operator/nn/mkldnn/mkldnn_base.cc:516: test Convolution
....

[pengzhao-intel]

@xcgoner who is the original requester :)

[pengzhao-intel]

FYI, @ChangMarkLiu

[TaoLv]

Do we want to expose MXNET_MKLDNN_DEBUG to front users?

[pengzhao-intel]

@TaoLv NO, for the developer rather than the end user.

[pengzhao-intel]

@zheng-da could you help take a look?