add MKLDNNAddtoLayer

[tensor-tang]

add mkldnn_addto layer without bias

[tensor-tang]

After merged latest code, still failed of lod_tensor_test

[02:19:18] : [Step 1/1] The following tests FAILED:
[02:19:18] : [Step 1/1] 98 - lod_tensor_test (Failed)
[02:19:18]W: [Step 1/1] Errors while running CTest
[02:19:18]W: [Step 1/1] Process exited with code 8

lod_tensor_test .............................***Failed 1.15 sec
[02:14:02] : [Step 1/1] Running main() from gtest_main.cc
[02:14:02] : [Step 1/1] [==========] Running 6 tests from 2 test cases.
[02:14:02] : [Step 1/1] [----------] Global test environment set-up.
[02:14:02] : [Step 1/1] [----------] 5 tests from LoDTensorTester
[02:14:02] : [Step 1/1] [ RUN ] LoDTensorTester.NumLevels
[02:14:02] : [Step 1/1] unknown file: Failure
[02:14:02] : [Step 1/1] C++ exception with description "std::bad_alloc" thrown in SetUp().

[tensor-tang]

After rerun, it pass.

[luotao1]

下一个PR是完成有bias的addtoLayer？

[tensor-tang]

嗯，不是下一个就是再下一个了。可能先把Resnet调通下，因为Resnet里面可能还用不到bias的addto。

出于功能，也会实现带有bias的，我已经加到我们list的TODO里面了。

[luotao1]

dnnConfig.biasSize = withBias？ pm.ic * pm.ih * pm.iw ： 0；
下次改Bias的时候一起改了吧。

[tensor-tang]

哦好的，thx

[luotao1]

这里为什么要互换下位置呢

[tensor-tang]

哦，理论上是可以不要调换的，但是还是想把顺序统一下，并且我默认第一个参数是参考。