[TensorFlow] Add a bitbake recipe for TensorFlow Lite v.1.12

This patch adds a bitbake recipe for tensorflow lite v.1.12. Note that since this dev package contains header files from flatbuffers needed for using tensorflow lite v.1.12, it may create a conflict with another flatbuffers package from other layers. Signed-off-by: Wook Song <[email protected]>
nnstreamer · Apr 17, 2019 · 842aee4 · 842aee4
1 parent 149c233
commit 842aee4
Show file tree

Hide file tree

Showing 3 changed files with 142 additions and 0 deletions.
diff --git a/recipes-tensorflow/tensorflow-lite/files/apply-modification-for-tflite-1.12-to-eigen.patch b/recipes-tensorflow/tensorflow-lite/files/apply-modification-for-tflite-1.12-to-eigen.patch
@@ -0,0 +1,13 @@
+diff --git a/tensorflow/contrib/lite/tools/make/downloads/eigen/Eigen/src/Core/arch/NEON/Complex.h b/tensorflow/contrib/lite/tools/make/downloads/eigen/Eigen/src/Core/arch/NEON/Complex.h
+index 306a309beb..4527a64c91 100644
+--- a/tensorflow/contrib/lite/tools/make/downloads/eigen/Eigen/src/Core/arch/NEON/Complex.h
++++ b/tensorflow/contrib/lite/tools/make/downloads/eigen/Eigen/src/Core/arch/NEON/Complex.h
+@@ -295,7 +295,7 @@ ptranspose(PacketBlock<Packet2cf,2>& kernel) {
+   static uint64x2_t p2ul_CONJ_XOR = {0x0, 0x8000000000000000};
+ #else
+   const uint64_t  p2ul_conj_XOR_DATA[] = { 0x0, 0x8000000000000000 };
+-  static uint64x2_t p2ul_CONJ_XOR = vld1q_u64( p2ul_conj_XOR_DATA );
++  static uint64x2_t p2ul_CONJ_XOR;// = vld1q_u64( p2ul_conj_XOR_DATA ); - Removed by script
+ #endif
+
+ struct Packet1cd
diff --git a/recipes-tensorflow/tensorflow-lite/files/tensorflow-lite.pc.in b/recipes-tensorflow/tensorflow-lite/files/tensorflow-lite.pc.in
@@ -0,0 +1,6 @@
+ Name: tensorflow lite
+ Description: tensorflow lite static library
+ Version: @version@
+ Requires:
+ Libs: -L@libdir@ -ltensorflow-lite
+ Cflags: -I@includedir@
diff --git a/recipes-tensorflow/tensorflow-lite/tensorflow-lite_1.12.bb b/recipes-tensorflow/tensorflow-lite/tensorflow-lite_1.12.bb
@@ -0,0 +1,123 @@
+DESCRIPTION = "TensorFlow’s lightweight solution for mobile and embedded devices"
+AUTHOR = "Google Inc. and Yuan Tang"
+HOMEPAGE = "https://www.tensorflow.org/lite"
+LICENSE = "Apache-2.0"
+LIC_FILES_CHKSUM = "file://LICENSE;md5=01e86893010a1b87e69a213faa753ebd"
+
+MD5SUM_EIGEN = "4c884968ede816a84c70e2cd2c81de8d"
+SHA256SUM_EIGEN = "d956415d784fa4e42b6a2a45c32556d6aec9d0a3d8ef48baee2522ab762556a9"
+MD5SUM_FARMHASH = "f039a65a7f62bdb6c4b4c8a732638d80"
+SHA256SUM_FARMHASH = "6560547c63e4af82b0f202cb710ceabb3f21347a4b996db565a411da5b17aba0"
+MD5SUM_GEMMLOWP = "a90a2161ad8ee4ccf724f56e07637601"
+SHA256SUM_GEMMLOWP = "b87faa7294dfcc5d678f22a59d2c01ca94ea1e2a3b488c38a95a67889ed0a658"
+MD5SUM_GTEST = "16877098823401d1bf2ed7891d7dce36"
+SHA256SUM_GTEST = "58a6f4277ca2bc8565222b3bbd58a177609e9c488e8a72649359ba51450db7d8"
+MD5SUM_ABSLCPP = "7c9945f256d3733a8eade255afcbd072"
+SHA256SUM_ABSLCPP = "7dd09690ae7ca4551de3111d4a86b75b23ec17445f273d3c42bdcdc1c7b02e4e"
+MD5SUM_NEON2SSE = "13695d77fba3973a27d98096e8d78773"
+SHA256SUM_NEON2SSE = "f1d5e84b30cae38f4164f73f211ce075b0d09f84d27cf4dc0411d647eaf422a8"
+MD5SUM_FLATBUFFER = "3811552512049fac3af419130904bc55"
+SHA256SUM_FLATBUFFER = "b2bb0311ca40b12ebe36671bdda350b10c7728caf0cfe2d432ea3b6e409016f3"
+MD5SUM_FFT = "4255dd8a74949d123216b1ab91520469"
+SHA256SUM_FFT = "52bb637c70b971958ec79c9c8752b1df5ff0218a4db4510e60826e0cb79b5296"
+
+SRC_URI = "git://github.com/tensorflow/tensorflow.git;branch=r1.12 \
+    https://bitbucket.org/eigen/eigen/get/fd6845384b86.tar.gz;md5sum=${MD5SUM_EIGEN};sha256sum=${SHA256SUM_EIGEN} \
+    https://mirror.bazel.build/github.com/google/gemmlowp/archive/38ebac7b059e84692f53e5938f97a9943c120d98.zip;md5sum=${MD5SUM_GEMMLOWP};sha256sum=${SHA256SUM_GEMMLOWP} \
+    https://github.com/google/googletest/archive/release-1.8.0.tar.gz;md5sum=${MD5SUM_GTEST};sha256sum=${SHA256SUM_GTEST} \
+    https://github.com/abseil/abseil-cpp/archive/48cd2c3f351ff188bc85684b84a91b6e6d17d896.tar.gz;md5sum=${MD5SUM_ABSLCPP};sha256sum=${SHA256SUM_ABSLCPP} \
+    https://github.com/intel/ARM_NEON_2_x86_SSE/archive/master.zip;md5sum=${MD5SUM_NEON2SSE};sha256sum=${SHA256SUM_NEON2SSE} \
+    https://mirror.bazel.build/github.com/google/farmhash/archive/816a4ae622e964763ca0862d9dbd19324a1eaf45.tar.gz;md5sum=${MD5SUM_FARMHASH};sha256sum=${SHA256SUM_FARMHASH} \
+    https://github.com/google/flatbuffers/archive/1f5eae5d6a135ff6811724f6c57f911d1f46bb15.tar.gz;md5sum=${MD5SUM_FLATBUFFER};sha256sum=${SHA256SUM_FLATBUFFER} \
+    https://mirror.bazel.build/www.kurims.kyoto-u.ac.jp/~ooura/fft.tgz;md5sum=${MD5SUM_FFT};sha256sum=${SHA256SUM_FFT} \
+    file://apply-modification-for-tflite-1.12-to-eigen.patch \
+    file://tensorflow-lite.pc.in"
+
+SRCREV = "${AUTOREV}"
+
+S = "${WORKDIR}/git"
+
+DEPENDS = "zlib"
+TARGET_CFLAGS_remove = "-O2"
+TARGET_CPPLAGS_remove = "-O2"
+TARGET_CXXLAGS_remove = "-O2"
+CCFLAGS_append = " -O3 -DNDEBUG -fPIC -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK \
+    -I${STAGING_INCDIR}"
+CXXFLAGS_append = " -O3 -DNDEBUG -fPIC -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK \
+    -I${STAGING_INCDIR}"
+LDFLAGS_remove = "-Wl,-O1"
+CXXFLAGS_append = " -O3 -DNDEBUG -fPIC -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK \
+    -I${STAGING_INCDIR}"
+LIBS = "-lstdc++ -lpthread -lm -lz -ldl"
+BUILD_DEPS_DOWNLOAD_DIR_PREFIX = "${S}/tensorflow/contrib/lite/tools/make/downloads/"
+
+do_cp_downloaded_build_deps() {
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}eigen
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}gemmlowp
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}googletest
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}absl
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}farmhash
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}neon_2_sse
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}flatbuffers
+    mkdir -p ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}fft2d
+
+    cp -rf ${WORKDIR}/eigen-eigen-fd6845384b86/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}eigen
+    cp -rf ${WORKDIR}/gemmlowp-38ebac7b059e84692f53e5938f97a9943c120d98/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}gemmlowp
+    cp -rf ${WORKDIR}/googletest-release-1.8.0/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}googletest
+    cp -rf ${WORKDIR}/abseil-cpp-48cd2c3f351ff188bc85684b84a91b6e6d17d896/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}absl
+    cp -rf ${WORKDIR}/farmhash-816a4ae622e964763ca0862d9dbd19324a1eaf45/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}farmhash
+    cp -rf ${WORKDIR}/ARM_NEON_2_x86_SSE-master/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}neon_2_sse
+    cp -rf ${WORKDIR}/flatbuffers-1f5eae5d6a135ff6811724f6c57f911d1f46bb15/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}flatbuffers
+    cp -rf ${WORKDIR}/fft/* ${BUILD_DEPS_DOWNLOAD_DIR_PREFIX}fft2d
+}
+addtask do_cp_downloaded_build_deps after do_unpack before do_patch
+
+EXTRA_OEMAKE = "\
+    'CC=${CC}' \
+    'CXX=${CXX}' \
+    'CPP=${CPP}' \
+    'CFLAGS=${CFLAGS}' \
+    'CPPFLAGS=${CFLAGS}' \
+    'CXXFLAGS=${CXXFLAGS}' \
+    'AR=${AR}' \
+    'LD=${LD}' \
+    'LDFLAGS=${LDFLAGS}' \
+    'LIBS=${LIBS}' \
+    'TARGET=${TARGET_OS}' \
+    'TARGET_ARCH=${TUNE_ARCH}'"
+
+do_configure() {
+    oe_runmake -f tensorflow/contrib/lite/tools/make/Makefile clean
+}
+
+do_compile() {
+    oe_runmake -f tensorflow/contrib/lite/tools/make/Makefile
+}
+
+do_install() {
+    install -d ${D}${libdir}
+    install -m 0644 ${S}/tensorflow/contrib/lite/tools/make/gen/${TARGET_OS}_${TUNE_ARCH}/lib/libtensorflow-lite.a ${D}${libdir}/
+    install -d ${D}${includedir}/tensorflow/contrib/lite
+    install -m 0644 ${S}/tensorflow/contrib/lite/*.h ${D}${includedir}/tensorflow/contrib/lite/
+    install -d ${D}${includedir}/tensorflow/contrib/lite/c
+    install -m 0644 ${S}/tensorflow/contrib/lite/c/*.h ${D}${includedir}/tensorflow/contrib/lite/c/
+    install -d ${D}${includedir}/tensorflow/contrib/lite/core/api/
+    install -m 0644 ${S}/tensorflow/contrib/lite/core/api/*.h ${D}${includedir}/tensorflow/contrib/lite/core/api/
+    install -d ${D}${includedir}/tensorflow/contrib/lite/kernels
+    install -m 0644 ${S}/tensorflow/contrib/lite/kernels/*.h ${D}${includedir}/tensorflow/contrib/lite/kernels/
+    install -d  ${D}${includedir}/tensorflow/contrib/lite/profiling/
+    install -m 0644 ${S}/tensorflow/contrib/lite/profiling/*.h ${D}${includedir}/tensorflow/contrib/lite/profiling/
+    install -d ${D}${includedir}/tensorflow/contrib/lite/schema/
+    install -m 0644 ${S}/tensorflow/contrib/lite/schema/*.h ${D}${includedir}/tensorflow/contrib/lite/schema/
+    install -d ${D}${includedir}/tensorflow/contrib/lite/tools/
+    install -m 0644 ${S}/tensorflow/contrib/lite/tools/*.h ${D}${includedir}/tensorflow/contrib/lite/tools/
+    install -d ${D}${libdir}/pkgconfig
+    install -m 0644 ${WORKDIR}/tensorflow-lite.pc.in ${D}${libdir}/pkgconfig/tensorflow-lite.pc
+    sed -i 's:@version@:${PV}:g
+        s:@libdir@:${libdir}:g
+        s:@includedir@:${includedir}:g' ${D}${libdir}/pkgconfig/tensorflow-lite.pc
+    # flatbuffers
+    install -d  ${D}${includedir}/flatbuffers
+    install -m 0644 ${S}/tensorflow/contrib/lite/tools/make/downloads/flatbuffers/include/flatbuffers/*.h ${D}${includedir}/flatbuffers/
+}
+