RPi3 ARMv8

rpi3-arm64.patch

diff --git a/tensorflow/core/platform/default/build_config.bzl b/tensorflow/core/platform/default/build_config.bzl
index 48ef8dfa8..be831d5c0 100644
--- a/tensorflow/core/platform/default/build_config.bzl
+++ b/tensorflow/core/platform/default/build_config.bzl
@@ -8,7 +8,7 @@ load("//tensorflow:tensorflow.bzl", "if_not_mobile")
 WITH_GCP_SUPPORT = False
 WITH_HDFS_SUPPORT = False
 WITH_XLA_SUPPORT = False
-WITH_JEMALLOC = True
+WITH_JEMALLOC = False
 
 # Appends a suffix to a list of deps.
 def tf_deps(deps, suffix):
diff --git a/tensorflow/tools/graph_transforms/quantize_nodes.cc b/tensorflow/tools/graph_transforms/quantize_nodes.cc
index f460f31d3..7e7b4abe0 100644
--- a/tensorflow/tools/graph_transforms/quantize_nodes.cc
+++ b/tensorflow/tools/graph_transforms/quantize_nodes.cc
@@ -684,6 +684,11 @@ Status QuantizeNodes(const GraphDef& input_graph_def,
         const NodeDef& float_node = match.node;
         const QuantizedOpInfo& op_info = op_map[float_node.op()];
 
+	bool are_under_while_if = false;
+        if ( (float_node.name().find("/while/") != string::npos) || (float_node.name().find("/if/") != string::npos) ) {
+            are_under_while_if = true;
+        }
+
         DataTypeVector input_types;
         DataTypeVector output_types;
         TF_RETURN_IF_ERROR(
@@ -723,10 +728,15 @@ Status QuantizeNodes(const GraphDef& input_graph_def,
           string unique_input_name =
               namespace_prefix + "/" + UniqueNodeNameFromInput(input_name);
 
+	  fprintf(stderr, "float_node.op()=%s input_name=%s are_under_while_if=%d\n", float_node.op().c_str(), input_name.c_str(), are_under_while_if);
+
           // Add some common constants we need for reshaping inputs.
           NodeDef reshape_dims;
           reshape_dims.set_op("Const");
           reshape_dims.set_name(unique_input_name + "/reshape_dims");
+          if (are_under_while_if) {
+              AddNodeInput("^" + input_name, &reshape_dims);
+          }
           SetNodeAttr("dtype", DT_INT32, &reshape_dims);
           Tensor reshape_dims_tensor(DT_INT32, {1});
           reshape_dims_tensor.flat<int32>()(0) = -1;
@@ -736,6 +746,9 @@ Status QuantizeNodes(const GraphDef& input_graph_def,
           NodeDef reduction_dims;
           reduction_dims.set_op("Const");
           reduction_dims.set_name(unique_input_name + "/reduction_dims");
+          if (are_under_while_if) {
+              AddNodeInput("^" + input_name, &reduction_dims);
+          }
           SetNodeAttr("dtype", DT_INT32, &reduction_dims);
           Tensor reduction_dims_tensor(DT_INT32, {1});
           reduction_dims_tensor.flat<int32>()(0) = 0;
diff --git a/tools/arm_compiler/BUILD b/tools/arm_compiler/BUILD
index 92699753f..58fa200b4 100644
--- a/tools/arm_compiler/BUILD
+++ b/tools/arm_compiler/BUILD
@@ -13,6 +13,7 @@ cc_toolchain_suite(
     "linaro-armeabi|gcc": ":cc-compiler-linaro-armeabi",
     "linaro64-armeabi|gcc": ":cc-compiler-linaro64-armeabi",
     "rpi-armeabi|gcc": ":cc-compiler-rpi-armeabi",
+    "rpi3-aarch64|gcc": ":cc-compiler-rpi3-aarch64",
   },
 )
 
@@ -146,3 +147,18 @@ cc_toolchain(
   supports_param_files = 1,
   visibility = ["//visibility:public"],
 )
+
+cc_toolchain(
+  name = "cc-compiler-rpi3-aarch64",
+  all_files = ":gcc_linux_all_files",
+  compiler_files = ":gcc_linux_compiler_files",
+  cpu = "gcc-aarch64",
+  dwp_files = ":empty",
+  dynamic_runtime_libs = [":empty"],
+  linker_files = ":gcc_linux_linker_files",
+  objcopy_files = "//tools/arm_compiler/gcc_arm_rpi:objcopy",
+  static_runtime_libs = [":empty"],
+  strip_files = "//tools/arm_compiler/gcc_arm_rpi:strip",
+  supports_param_files = 1,
+  visibility = ["//visibility:public"],
+)
diff --git a/tools/arm_compiler/CROSSTOOL b/tools/arm_compiler/CROSSTOOL
index e8f855b8a..c09182065 100644
--- a/tools/arm_compiler/CROSSTOOL
+++ b/tools/arm_compiler/CROSSTOOL
@@ -17,6 +17,11 @@ default_toolchain {
   toolchain_identifier: "gcc_rpi_linux_armhf"
 }
 
+default_toolchain {
+  cpu: "rpi3-aarch64"
+  toolchain_identifier: "gcc_rpi3_linux_aarch64"
+}
+
 toolchain {
   abi_version: "armeabi"
   abi_libc_version: "glibc_2.13"
@@ -352,8 +357,6 @@ toolchain {
     linker_flag: "-Wl,--gc-sections"
   }
 }
-
-
 toolchain {
   abi_version: "armeabi"
   abi_libc_version: "glibc_2.19"
@@ -412,6 +415,8 @@ toolchain {
   compiler_flag: "external/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot/usr/include/arm-linux-gnueabihf"
   compiler_flag: "-isystem"
   compiler_flag: "external/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/lib/gcc/arm-linux-gnueabihf/4.9.3/include"
+  compiler_flag: "-isystem"
+  compiler_flag: "DEEPSPEECH_ROOT/multistrap-raspbian-jessie/usr/include/"
 
   cxx_flag: "-std=c++11"
   cxx_flag: "-isystem"
@@ -433,6 +438,7 @@ toolchain {
   cxx_builtin_include_directory: "%package(@GccArmRpi//arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/lib/gcc/arm-linux-gnueabihf/4.9.3/include)%"
   cxx_builtin_include_directory: "%package(@GccArmRpi//arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/lib/gcc/arm-linux-gnueabihf/4.9.3/include-fixed)%"
   cxx_builtin_include_directory: "%package(@GccArmRpi//arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/include)%/c++/4.9.3"
+  cxx_builtin_include_directory: "DEEPSPEECH_ROOT/multistrap-raspbian-jessie/usr/include/"
 
   # Anticipated future default.
   # This makes GCC and Clang do what we want when called through symlinks.
@@ -473,6 +479,7 @@ toolchain {
   linker_flag: "-Lexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot/lib"
   linker_flag: "-Lexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot/usr/lib"
   linker_flag: "-Bexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/bin"
+  linker_flag: "-LDEEPSPEECH_ROOT/multistrap-raspbian-jessie/usr/lib"
   linker_flag: "-pie"
   linker_flag: "-lstdc++"
   # linker_flag: "-lm"
@@ -497,7 +504,7 @@ toolchain {
     # Maybe we should enable https://gcc.gnu.org/wiki/DebugFission for opt or
     # even generally? However, that can't happen here, as it requires special
     # handling in Bazel.
-    compiler_flag: "-g0"
+    compiler_flag: "-g"
 
     # Disable assertions
     compiler_flag: "-DNDEBUG"
@@ -530,3 +537,155 @@ toolchain {
   }
   linking_mode_flags { mode: DYNAMIC }
 }
+
+toolchain {
+  abi_version: "aarch64"
+  abi_libc_version: "glibc_2.19"
+  builtin_sysroot: ""
+  compiler: "gcc"
+  host_system_name: "aarch64"
+  needsPic: true
+  supports_gold_linker: false
+  supports_incremental_linker: false
+  supports_fission: false
+  supports_interface_shared_objects: false
+  supports_normalizing_ar: false
+  supports_start_end_lib: false
+  supports_thin_archives: false
+  target_libc: "glibc_2.19"
+  target_cpu: "rpi3-aarch64"
+  target_system_name: "arm64-v8a"
+  toolchain_identifier: "gcc_rpi3_linux_aarch64"
+
+  tool_path { name: "ar" path: "/usr/bin/aarch64-linux-gnu-ar" }
+  tool_path { name: "compat-ld" path: "/usr/bin/aarch64-linux-gnu-ld" }
+  tool_path { name: "cpp" path: "/usr/bin/aarch64-linux-gnu-cpp" }
+  tool_path { name: "dwp" path: "/usr/bin/aarch64-linux-gnu-dwp" }
+  tool_path { name: "gcc" path: "/usr/bin/aarch64-linux-gnu-gcc" }
+  tool_path { name: "gcov" path: "/usr/bin/aarch64-linux-gnu-gcov" }
+  # C(++) compiles invoke the compiler (as that is the one knowing where
+  # to find libraries), but we provide LD so other rules can invoke the linker.
+  tool_path { name: "ld" path: "/usr/bin/aarch64-linux-gnu-ld" }
+  tool_path { name: "nm" path: "/usr/bin/aarch64-linux-gnu-nm" }
+  tool_path { name: "objcopy" path: "/usr/bin/aarch64-linux-gnu-objcopy" }
+  objcopy_embed_flag: "-I"
+  objcopy_embed_flag: "binary"
+  tool_path { name: "objdump" path: "/usr/bin/aarch64-linux-gnu-objdump" }
+  tool_path { name: "strip" path: "/usr/bin/aarch64-linux-gnu-strip" }
+
+  compiler_flag: "-march=armv8-a+crc"
+  #compiler_flag: "--sysroot=external/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot"
+  compiler_flag: "--sysroot=DEEPSPEECH_ROOT/multistrap-debian_arm64-sid/"
+  #compiler_flag: "-Wl,--sysroot=external/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot"
+  #compiler_flag: "-mfloat-abi=hard"
+  # Local change to disable IS_MOBILE_PLATFORM
+  compiler_flag: "-D__ARM_RPI__"
+  # compiler_flag: "-nostdinc"
+  # compiler_flag: "-isystem"
+  cxx_flag: "-std=c++11"
+  cxx_builtin_include_directory: "DEEPSPEECH_ROOT/multistrap-debian_arm64-sid/usr/include/"
+
+  # Anticipated future default.
+  # This makes GCC and Clang do what we want when called through symlinks.
+  unfiltered_cxx_flag: "-no-canonical-prefixes"
+
+  # Make C++ compilation deterministic. Use linkstamping instead of these
+  # compiler symbols.
+  unfiltered_cxx_flag: "-Wno-builtin-macro-redefined"
+  unfiltered_cxx_flag: "-D__DATE__=\"redacted\""
+  unfiltered_cxx_flag: "-D__TIMESTAMP__=\"redacted\""
+  unfiltered_cxx_flag: "-D__TIME__=\"redacted\""
+
+  # Security hardening on by default.
+  compiler_flag: "-fstack-protector"
+  compiler_flag: "-fPIE"
+  # All warnings are enabled. Maybe enable -Werror as well?
+  compiler_flag: "-Wall"
+  # Enable a few more warnings that aren't part of -Wall.
+  compiler_flag: "-Wunused-but-set-parameter"
+  # But disable some that are problematic.
+  compiler_flag: "-Wno-free-nonheap-object" # has false positives
+  # Keep stack frames for debugging, even in opt mode.
+  compiler_flag: "-fno-omit-frame-pointer"
+  # Enable coloring even if there's no attached terminal. Bazel removes the
+  # escape sequences if --nocolor is specified.
+  compiler_flag: "-fdiagnostics-color=always"
+
+  # compiler_flag: "-D__GCC_HAVE_SYNC_COMPARE_AND_SWAP_1"
+  # compiler_flag: "-D__GCC_HAVE_SYNC_COMPARE_AND_SWAP_2"
+  # compiler_flag: "-D__GCC_HAVE_SYNC_COMPARE_AND_SWAP_4"
+  # compiler_flag: "-D__GCC_HAVE_SYNC_COMPARE_AND_SWAP_8"
+
+  cxx_builtin_include_directory: "/usr/aarch64-linux-gnu/include/"
+  cxx_builtin_include_directory: "/usr/lib/gcc-cross/aarch64-linux-gnu/6/include/"
+  cxx_builtin_include_directory: "/usr/lib/gcc-cross/aarch64-linux-gnu/6/include-fixed/"
+
+  # linker_flag: "-target"
+  # linker_flag: "arm-linux-gnueabihf"
+  linker_flag: "--sysroot=DEEPSPEECH_ROOT/multistrap-debian_arm64-sid"
+  linker_flag: "-pass-exit-codes"
+  #linker_flag: "-Lexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/lib"
+  #linker_flag: "-Lexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot/lib"
+  #linker_flag: "-Lexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/sysroot/usr/lib"
+  #linker_flag: "-LDEEPSPEECH_ROOT/multistrap-debian_arm64-sid/lib"
+  #linker_flag: "-LDEEPSPEECH_ROOT/multistrap-debian_arm64-sid/usr/lib"
+  #linker_flag: "-Bexternal/GccArmRpi/arm-bcm2708/arm-rpi-4.9.3-linux-gnueabihf/arm-linux-gnueabihf/bin"
+  linker_flag: "-pie"
+  linker_flag: "-lstdc++"
+  # linker_flag: "-lm"
+  # linker_flag: "-lpthread"
+  linker_flag: "-Wl,--dynamic-linker=/lib/ld-linux-armhf.so.3"
+  linker_flag: "-Wl,-no-as-needed"
+  linker_flag: "-Wl,-z,relro,-z,now"
+  linker_flag: "-no-canonical-prefixes"
+  # Stamp the binary with a unique identifier.
+  linker_flag: "-Wl,--build-id=md5"
+  linker_flag: "-Wl,--hash-style=gnu"
+
+  compilation_mode_flags {
+    mode: DBG
+    # Enable debug symbols.
+    compiler_flag: "-g"
+  }
+  compilation_mode_flags {
+    mode: OPT
+
+    # No debug symbols.
+    # Maybe we should enable https://gcc.gnu.org/wiki/DebugFission for opt or
+    # even generally? However, that can't happen here, as it requires special
+    # handling in Bazel.
+    compiler_flag: "-g"
+
+    # Disable assertions
+    compiler_flag: "-DNDEBUG"
+
+    # Removal of unused code and data at link time (can this increase binary size in some cases?).
+    compiler_flag: "-ffunction-sections"
+    compiler_flag: "-fdata-sections"
+    linker_flag: "-Wl,--gc-sections"
+
+    # Conservative choice for -O
+    # -O3 can increase binary size and even slow down the resulting binaries.
+    # Profile first and / or use FDO if you need better performance than this.
+    compiler_flag: "-O2"
+
+    # Conservative choice; -D_FORTIFY_SOURCE=2 may be unsafe in some cases.
+    # We need to undef it before redefining it as some distributions now have
+    # it enabled by default.
+    # Also depends on -O.
+    compiler_flag: "-U_FORTIFY_SOURCE"
+    compiler_flag: "-D_FORTIFY_SOURCE=1"
+
+    # Basic optims for RPi3, from Gentoo/ArchLinux Wiki
+    compiler_flag: "-mtune=cortex-a53"
+    #compiler_flag: "-mfpu=crypto-neon-fp-armv8"
+    #compiler_flag: "-mfloat-abi=hard"
+    # Inference time on small frozen LDC93S1 model goes from 30s to 20s with
+    # that flag, but generates instructions that valgrind chokes on.
+    #compiler_flag: "-mfpu=neon-fp-armv8"
+    #compiler_flag: "-funsafe-math-optimizations"
+    compiler_flag: "-ftree-vectorize"
+    compiler_flag: "-pipe"
+  }
+  linking_mode_flags { mode: DYNAMIC }
+}

thanks for help !