Add direct softlinks for compiler in local bin

.gitignore

@@ -167,10 +167,6 @@ python/.eggs
 tests/Makefile
 tests/mxnet_unit_tests
 
-# generated wrappers for ccache
-cc
-cxx
-
 # Code coverage related
 .coverage
 *.gcov

Makefile

@@ -463,7 +463,7 @@ build/src/%.o: src/%.cc | mkldnn
 
 build/src/%_gpu.o: src/%.cu | mkldnn
 	@mkdir -p $(@D)
-	$(NVCC) $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS)" -M -MT build/src/$*_gpu.o $< >build/src/$*_gpu.d
+	$(NVCC) $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS)" --generate-dependencies -MT build/src/$*_gpu.o $< >build/src/$*_gpu.d
 	$(NVCC) -c -o $@ $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS)" $<
 
 # A nvcc bug cause it to generate "generic/xxx.h" dependencies from torch headers.
@@ -479,7 +479,7 @@ build/plugin/%.o: plugin/%.cc
 
 %_gpu.o: %.cu
 	@mkdir -p $(@D)
-	$(NVCC) $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS) -Isrc/operator" -M -MT $*_gpu.o $< >$*_gpu.d
+	$(NVCC) $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS) -Isrc/operator" --generate-dependencies -MT $*_gpu.o $< >$*_gpu.d
 	$(NVCC) -c -o $@ $(NVCCFLAGS) $(CUDA_ARCH) -Xcompiler "$(CFLAGS) -Isrc/operator" $<
 
 %.o: %.cc $(CORE_INC)
@@ -660,7 +660,7 @@ jnilint:
 
 rclean:
 	$(RM) -r R-package/src/image_recordio.h R-package/NAMESPACE R-package/man R-package/R/mxnet_generated.R \
-		R-package/inst R-package/src/*.o R-package/src/*.so mxnet_*.tar.gz
+		R-package/inst R-package/src/*.o R-package/src/*.so mxnet_*.tar.gz deps
 
 ifneq ($(EXTRA_OPERATORS),)
 clean: rclean cyclean $(EXTRA_PACKAGES_CLEAN)

ci/docker/install/centos7_adduser.sh

@@ -34,4 +34,9 @@ then
     mkdir /work/mxnet
     mkdir /work/build
     chown -R jenkins_slave /work/
+
+    # Later on, we have to override the links because underlying build systems ignore our compiler settings. Thus,
+    # we have to give the process the proper permission to these files. This is hacky, but unfortunately 
+    # there's no better way to do this without patching all our submodules.
+    chown -R jenkins_slave /usr/local/bin
 fi

ci/docker/install/ubuntu_adduser.sh

@@ -40,4 +40,9 @@ then
     mkdir /work/mxnet
     mkdir /work/build
     chown -R jenkins_slave /work/
+
+    # Later on, we have to override the links because underlying build systems ignore our compiler settings. Thus,
+    # we have to give the process the proper permission to these files. This is hacky, but unfortunately 
+    # there's no better way to do this without patching all our submodules.
+    chown -R jenkins_slave /usr/local/bin
 fi

ci/docker/runtime_functions.sh

@@ -36,35 +36,37 @@ clean_repo() {
     git submodule update --init --recursive
 }
 
-build_ccache_wrappers() {
-    set -ex
-
-    rm -f cc
-    rm -f cxx
-
-    touch cc
-    touch cxx
-
-    if [ -z ${CC+x} ]; then
-        echo "No \$CC set, defaulting to gcc";
-        export CC=gcc
-    fi
-
-    if [ -z ${CXX+x} ]; then
-       echo "No \$CXX set, defaulting to g++";
-       export CXX=g++
-    fi
-
-    # this function is nessesary for cuda enabled make based builds, since nvcc needs just an executable for -ccbin
-
-    echo -e "#!/bin/sh\n/usr/local/bin/ccache ${CC} \"\$@\"\n" >> cc
-    echo -e "#!/bin/sh\n/usr/local/bin/ccache ${CXX} \"\$@\"\n" >> cxx
-
-    chmod +x cc
-    chmod +x cxx
-
-    export CC=`pwd`/cc
-    export CXX=`pwd`/cxx
+install_ccache_wrappers() {
+    set -ex
+
+    # Recommended by CCache: https://ccache.samba.org/manual.html#_run_modes
+    # Add to the beginning of path to ensure this redirection is picked up instead
+    # of the original ones. Especially CUDA/NVCC appends itself to the beginning of the
+    # path and thus this redirect is ignored. This change fixes this problem
+    # This hacky approach with symbolic links is required because underlying build
+    # systems of our submodules ignore our CMake settings. If they use Makefile,
+    # we can't influence them at all in general and NVCC also prefers to hardcode their
+    # compiler instead of respecting the settings. Thus, we take this brutal approach
+    # and just redirect everything of this installer has been called.
+    # In future, we could do these links during image build time of the container.
+    # But in the beginning, we'll make this opt-in. In future, loads of processes like
+    # the scala make step or numpy compilation and other pip package generations
+    # could be heavily sped up by using ccache as well.
+    export PATH=/usr/local/bin:$PATH
+    ln -s ccache /usr/local/bin/gcc
+    ln -s ccache /usr/local/bin/g++
+    ln -s ccache /usr/local/bin/g++-8
+    ln -s ccache /usr/local/bin/gcc-8
+    ln -s ccache /usr/local/bin/nvcc
+    ln -s ccache /usr/local/bin/clang++-3.9
+    ln -s ccache /usr/local/bin/clang-3.9
+    ln -s ccache /usr/local/bin/clang++-5.0
+    ln -s ccache /usr/local/bin/clang-5.0
+    ln -s ccache /usr/local/bin/clang++-6.0
+    ln -s ccache /usr/local/bin/clang-6.0
+
+    # TODO: Remove
+    export CCACHE_LOGFILE=/work/mxnet/ccache-log
 }
 
 build_wheel() {
@@ -128,7 +130,6 @@ build_armv6() {
     # to be linked additionally.
 
     # We do not need OpenMP, since most armv6 systems have only 1 core
-
     cmake \
         -DCMAKE_TOOLCHAIN_FILE=${CMAKE_TOOLCHAIN_FILE} \
         -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
@@ -227,6 +228,8 @@ build_android_armv8() {
     cd /work/build
     cmake\
         -DANDROID=ON \
+        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+        -DCMAKE_C_COMPILER_LAUNCHER=ccache \
         -DUSE_CUDA=OFF\
         -DUSE_SSE=OFF\
         -DUSE_LAPACK=OFF\
@@ -242,9 +245,8 @@ build_android_armv8() {
 build_centos7_cpu() {
     set -ex
     cd /work/mxnet
-    export CC="ccache gcc"
-    export CXX="ccache g++"
 
+    install_ccache_wrappers
     make \
         DEV=1 \
         USE_LAPACK=1 \
@@ -257,6 +259,7 @@ build_centos7_cpu() {
 
 build_amzn_linux_cpu() {
     cd /work/build
+    install_ccache_wrappers
     cmake \
         -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
         -DCMAKE_C_COMPILER_LAUNCHER=ccache \
@@ -277,9 +280,8 @@ build_amzn_linux_cpu() {
 build_centos7_mkldnn() {
     set -ex
     cd /work/mxnet
-    export CC="ccache gcc"
-    export CXX="ccache g++"
-
+
+    install_ccache_wrappers
     make \
         DEV=1 \
         ENABLE_TESTCOVERAGE=1 \
@@ -293,8 +295,7 @@ build_centos7_mkldnn() {
 build_centos7_gpu() {
     set -ex
     cd /work/mxnet
-    # unfortunately this build has problems in 3rdparty dependencies with ccache and make
-    # build_ccache_wrappers
+    install_ccache_wrappers
     make \
         DEV=1                                     \
         ENABLE_TESTCOVERAGE=1                     \
@@ -315,8 +316,7 @@ build_ubuntu_cpu() {
 
 build_ubuntu_cpu_openblas() {
     set -ex
-    export CC="ccache gcc"
-    export CXX="ccache g++"
+    install_ccache_wrappers
     make \
         DEV=1                         \
         ENABLE_TESTCOVERAGE=1         \
@@ -330,6 +330,7 @@ build_ubuntu_cpu_cmake_debug() {
     set -ex
     pushd .
     cd /work/build
+    install_ccache_wrappers
     cmake \
         -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
         -DCMAKE_C_COMPILER_LAUNCHER=ccache \
@@ -351,6 +352,8 @@ build_ubuntu_cpu_cmake_asan() {
 
     pushd .
     cd /work/build
+    install_ccache_wrappers
+    # TODO: Check if we can remove the g++-8 override
     cmake \
         -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
         -DCMAKE_C_COMPILER_LAUNCHER=ccache \
@@ -377,10 +380,10 @@ build_ubuntu_cpu_cmake_asan() {
 
 build_ubuntu_cpu_clang39() {
     set -ex
-     export CXX=clang++-3.9
+    export CXX=clang++-3.9
     export CC=clang-3.9
-     build_ccache_wrappers
-     make \
+    install_ccache_wrappers
+    make \
         ENABLE_TESTCOVERAGE=1         \
         USE_CPP_PACKAGE=1             \
         USE_BLAS=openblas             \
@@ -395,7 +398,7 @@ build_ubuntu_cpu_clang60() {
     export CXX=clang++-6.0
     export CC=clang-6.0
 
-    build_ccache_wrappers
+    install_ccache_wrappers
 
     make  \
         ENABLE_TESTCOVERAGE=1         \
@@ -415,6 +418,7 @@ build_ubuntu_cpu_clang_tidy() {
 
     pushd .
     cd /work/build
+    install_ccache_wrappers
     cmake \
         -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
         -DCMAKE_C_COMPILER_LAUNCHER=ccache \
@@ -438,8 +442,7 @@ build_ubuntu_cpu_clang39_mkldnn() {
     export CXX=clang++-3.9
     export CC=clang-3.9
 
-    build_ccache_wrappers
-
+    install_ccache_wrappers
     make \
         ENABLE_TESTCOVERAGE=1         \
         USE_CPP_PACKAGE=1             \
@@ -455,8 +458,7 @@ build_ubuntu_cpu_clang60_mkldnn() {
     export CXX=clang++-6.0
     export CC=clang-6.0
 
-    build_ccache_wrappers
-
+    install_ccache_wrappers
     make \
         ENABLE_TESTCOVERAGE=1         \
         USE_CPP_PACKAGE=1             \
@@ -469,8 +471,7 @@ build_ubuntu_cpu_clang60_mkldnn() {
 build_ubuntu_cpu_mkldnn() {
     set -ex
 
-    build_ccache_wrappers
-
+    install_ccache_wrappers
     make  \
         DEV=1                         \
         ENABLE_TESTCOVERAGE=1         \
@@ -488,7 +489,7 @@ build_ubuntu_gpu_tensorrt() {
 
     set -ex
 
-    build_ccache_wrappers
+    install_ccache_wrappers
 
     # Build ONNX
     pushd .
@@ -498,6 +499,8 @@ build_ubuntu_gpu_tensorrt() {
     mkdir -p build
     cd build
     cmake \
+        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+        -DCMAKE_C_COMPILER_LAUNCHER=ccache \
         -DCMAKE_CXX_FLAGS=-I/usr/include/python${PYVER}\
         -DBUILD_SHARED_LIBS=ON ..\
         -G Ninja
@@ -512,7 +515,10 @@ build_ubuntu_gpu_tensorrt() {
     cd 3rdparty/onnx-tensorrt/
     mkdir -p build
     cd build
-    cmake ..
+    cmake \
+        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+        -DCMAKE_C_COMPILER_LAUNCHER=ccache \
+        ..
     make -j$(nproc)
     export LIBRARY_PATH=`pwd`:$LIBRARY_PATH
     popd
@@ -543,8 +549,7 @@ build_ubuntu_gpu_tensorrt() {
 build_ubuntu_gpu_mkldnn() {
     set -ex
 
-    build_ccache_wrappers
-
+    install_ccache_wrappers
     make  \
         DEV=1                                     \
         ENABLE_TESTCOVERAGE=1                     \
@@ -561,8 +566,7 @@ build_ubuntu_gpu_mkldnn() {
 build_ubuntu_gpu_mkldnn_nocudnn() {
     set -ex
 
-    build_ccache_wrappers
-
+    install_ccache_wrappers
     make  \
         DEV=1                                     \
         ENABLE_TESTCOVERAGE=1                     \
@@ -577,8 +581,7 @@ build_ubuntu_gpu_mkldnn_nocudnn() {
 
 build_ubuntu_gpu_cuda91_cudnn7() {
     set -ex
-    # unfortunately this build has problems in 3rdparty dependencies with ccache and make
-    # build_ccache_wrappers
+    install_ccache_wrappers
     make \
         DEV=1                                     \
         ENABLE_TESTCOVERAGE=1                     \
@@ -595,6 +598,7 @@ build_ubuntu_gpu_cuda91_cudnn7() {
 build_ubuntu_amalgamation() {
     set -ex
     # Amalgamation can not be run with -j nproc
+    install_ccache_wrappers
     make -C amalgamation/ clean
     make -C amalgamation/     \
         USE_BLAS=openblas     \
@@ -604,6 +608,7 @@ build_ubuntu_amalgamation() {
 build_ubuntu_amalgamation_min() {
     set -ex
     # Amalgamation can not be run with -j nproc
+    install_ccache_wrappers
     make -C amalgamation/ clean
     make -C amalgamation/     \
         USE_BLAS=openblas     \
@@ -614,9 +619,10 @@ build_ubuntu_amalgamation_min() {
 build_ubuntu_gpu_cmake_mkldnn() {
     set -ex
     cd /work/build
+    install_ccache_wrappers
     cmake \
-        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache    \
-        -DCMAKE_C_COMPILER_LAUNCHER=ccache      \
+        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+        -DCMAKE_C_COMPILER_LAUNCHER=ccache \
         -DENABLE_TESTCOVERAGE=ON                \
         -DUSE_CUDA=1                            \
         -DUSE_CUDNN=1                           \
@@ -637,9 +643,10 @@ build_ubuntu_gpu_cmake_mkldnn() {
 build_ubuntu_gpu_cmake() {
     set -ex
     cd /work/build
+    install_ccache_wrappers
     cmake \
-        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache    \
-        -DCMAKE_C_COMPILER_LAUNCHER=ccache      \
+        -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
+        -DCMAKE_C_COMPILER_LAUNCHER=ccache \
         -DENABLE_TESTCOVERAGE=ON                \
         -DUSE_CUDA=1                            \
         -DUSE_CUDNN=1                           \