Revert "[MXNET-43] Fix Jetson compilation" (apache#13665)

* Revert "remove omp which can cause ssd accuracy variance (apache#13622)"

This reverts commit 655f1c6.

* Revert "Fix Jetson compilation (apache#13532)"

This reverts commit 48e25c4.

ci/docker/Dockerfile.build.jetson

@@ -77,7 +77,6 @@ RUN JETPACK_DOWNLOAD_PREFIX=https://developer.download.nvidia.com/devzone/devcen
     dpkg -i --force-architecture  $ARM_NVINFER_INSTALLER_PACKAGE && \
     dpkg -i --force-architecture  $ARM_NVINFER_DEV_INSTALLER_PACKAGE && \
     apt update -y || true && apt install -y cuda-libraries-dev-9-0 libcudnn7-dev libnvinfer-dev
-RUN ln -s /usr/include/aarch64-linux-gnu/cudnn_v7.h /usr/include/aarch64-linux-gnu/cudnn.h
 ENV PATH $PATH:/usr/local/cuda/bin
 ENV NVCCFLAGS "-m64"
 ENV CUDA_ARCH "-gencode arch=compute_53,code=sm_53 -gencode arch=compute_62,code=sm_62"

ci/jenkins/Jenkinsfile_edge

@@ -34,7 +34,7 @@ utils.assign_node_labels(utility: 'utility', linux_cpu: 'mxnetlinux-cpu', linux_
 utils.main_wrapper(
 core_logic: {
   utils.parallel_stage('Build', [
-    custom_steps.compile_armv8_jetson_gpu(),
+//    custom_steps.compile_armv8_jetson_gpu(),
     custom_steps.compile_armv7_cpu(),
     custom_steps.compile_armv6_cpu(),
     custom_steps.compile_armv8_cpu(),

make/crosscompile.jetson.mk

@@ -57,10 +57,10 @@ DEBUG = 0
 USE_SIGNAL_HANDLER = 1
 
 # the additional link flags you want to add
-ADD_LDFLAGS = -L${CROSS_ROOT}/lib -L/usr/lib/aarch64-linux-gnu/
+ADD_LDFLAGS = -L${CROSS_ROOT}/lib
 
 # the additional compile flags you want to add
-ADD_CFLAGS = -I${CROSS_ROOT}/include -I/usr/include/aarch64-linux-gnu/
+ADD_CFLAGS = -I${CROSS_ROOT}/include
 
 #---------------------------------------------
 # matrix computation libraries for CPU/GPU