Merge commit '12b619343b0e8e35cda6335f8230c4168277b8dd'

Author: reyoung

.gitignore

@@ -1,2 +1,6 @@
 *.DS_Store
 build/
+*.user
+
+.vscode
+.idea

.travis.yml

@@ -0,0 +1,38 @@
+language: cpp
+cache: ccache
+sudo: required
+dist: trusty
+env:
+  - JOB=DOCS
+  - JOB=BUILD_AND_TEST
+addons:
+  apt:
+    packages:
+      - gcc-4.8
+      - g++-4.8
+      - wget
+      - git
+      - build-essential
+      - libatlas-base-dev
+      - python
+      - python-pip
+      - python2.7-dev
+      - m4
+      - libprotobuf-dev
+      - doxygen
+      - protobuf-compiler
+      - python-protobuf
+      - python-numpy
+      - python-wheel
+      - libgoogle-glog-dev
+      - libgflags-dev
+      - libgtest-dev
+before_install:
+  - pip install wheel protobuf sphinx breathe recommonmark
+  - sudo paddle/scripts/travis/before_install.sh
+script:
+  - paddle/scripts/travis/main.sh
+notifications:
+  email:
+    on_success: change
+    on_failure: always

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 2.8)
 project(paddle CXX C)
 set(PADDLE_MAJOR_VERSION 0)
 set(PADDLE_MINOR_VERSION 8)
-set(PADDLE_PATCH_VERSION 0b)
+set(PADDLE_PATCH_VERSION 0b1)
 set(PADDLE_VERSION ${PADDLE_MAJOR_VERSION}.${PADDLE_MINOR_VERSION}.${PADDLE_PATCH_VERSION})
 
 set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/cmake")
@@ -14,8 +14,10 @@ find_package(CUDA QUIET)
 find_package(Protobuf REQUIRED)
 find_package(PythonLibs 2.7 REQUIRED)
 find_package(PythonInterp 2.7 REQUIRED)
-find_package(NumPy)
+find_package(ZLIB REQUIRED)
+find_package(NumPy REQUIRED)
 find_package(Threads REQUIRED)
+find_package(AVX QUIET)
 find_package(Glog)
 find_package(Gflags QUIET)
 find_package(GTest)
@@ -27,7 +29,7 @@ find_program(M4_EXECUTABLE m4)
 option(WITH_DSO "Compile PaddlePaddle with dynamic linked libraries" ON)
 option(WITH_GPU "Compile PaddlePaddle with gpu" ${CUDA_FOUND})
 option(WITH_DOUBLE "Compile PaddlePaddle with double precision, otherwise use single precision" OFF)
-option(WITH_AVX "Compile PaddlePaddle with avx intrinsics" ON) # TODO(yuyang18): Check AVX is supported or not as default value
+option(WITH_AVX "Compile PaddlePaddle with avx intrinsics" ${AVX_FOUND})
 option(WITH_PYTHON "Compile PaddlePaddle with python interpreter" ON)
 option(WITH_STYLE_CHECK "Style Check for PaddlePaddle" ${PYTHONINTERP_FOUND})
 option(WITH_RDMA "Compile PaddlePaddle with rdma support" OFF)
@@ -37,6 +39,7 @@ option(WITH_TIMER "Compile PaddlePaddle use timer" OFF)
 option(WITH_TESTING "Compile and run unittest for PaddlePaddle" ${GTEST_FOUND})
 option(WITH_DOC "Compile PaddlePaddle with documentation" OFF)
 option(WITH_SWIG_PY "Compile PaddlePaddle with py PaddlePaddle prediction api" ${SWIG_FOUND})
+option(ON_TRAVIS "Running test on travis-ci or not." OFF)
 if(NOT CMAKE_BUILD_TYPE)
     set(CMAKE_BUILD_TYPE "RelWithDebInfo" CACHE STRING 
         "Choose the type of build, options are: Debug Release RelWithDebInfo MinSizeRel"
@@ -99,8 +102,8 @@ if(NOT WITH_TIMER)
 endif(NOT WITH_TIMER)
 
 if(WITH_AVX)
-    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mavx")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mavx")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${AVX_FLAGS}")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${AVX_FLAGS}")
 else(WITH_AVX)
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse3")
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse3")

README.md

@@ -1,4 +1,11 @@
 # PaddlePaddle
+[![Build Status](https://travis-ci.org/baidu/Paddle.svg?branch=master)](https://travis-ci.org/baidu/Paddle)
+
+Welcome to the PaddlePaddle GitHub.
+
+The software will be released on Sept. 30 with full documentation and installation support. 
+
+A pre-release version is available now for those who are eager to take a look.
 
 PaddlePaddle (PArallel Distributed Deep LEarning) is an easy-to-use,
 efficient, flexible and scalable deep learning platform, which is originally

cmake/FindAVX.cmake

@@ -0,0 +1,65 @@
+# This file is use to check all support level of AVX on your machine
+# so that PaddlePaddle can unleash the vectorization power of muticore.
+
+INCLUDE(CheckCXXSourceRuns)
+
+SET(FIND_AVX_10)
+SET(FIND_AVX_20)
+SET(AVX_FLAGS)
+SET(AVX_FOUND)
+
+# Check AVX 2
+SET(CMAKE_REQUIRED_FLAGS)
+IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+  SET(CMAKE_REQUIRED_FLAGS "-mavx2")
+ELSEIF(MSVC AND NOT CMAKE_CL_64)  # reserve for WINDOWS
+  SET(CMAKE_REQUIRED_FLAGS "/arch:AVX2")
+ENDIF()
+
+CHECK_CXX_SOURCE_RUNS("
+#include <immintrin.h>
+int main()
+{
+    __m256i a = _mm256_set_epi32 (-1, 2, -3, 4, -1, 2, -3, 4);
+    __m256i result = _mm256_abs_epi32 (a);
+    return 0;
+}" FIND_AVX_20)
+
+# Check AVX
+SET(CMAKE_REQUIRED_FLAGS)
+IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+    SET(CMAKE_REQUIRED_FLAGS "-mavx")
+ELSEIF(MSVC AND NOT CMAKE_CL_64)
+    SET(CMAKE_REQUIRED_FLAGS "/arch:AVX")
+endif()
+
+CHECK_CXX_SOURCE_RUNS("
+#include <immintrin.h>
+int main()
+{
+    __m256 a = _mm256_set_ps (-1.0f, 2.0f, -3.0f, 4.0f, -1.0f, 2.0f, -3.0f, 4.0f);
+    __m256 b = _mm256_set_ps (1.0f, 2.0f, 3.0f, 4.0f, 1.0f, 2.0f, 3.0f, 4.0f);
+    __m256 result = _mm256_add_ps (a, b);
+    return 0;
+}" FIND_AVX_10)
+
+IF(${FIND_AVX_20})
+    IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+        SET(AVX_FLAGS "${AVX_FLAGS} -mavx2")
+    ELSEIF(MSVC)
+        SET(AVX_FLAGS "${AVX_FLAGS} /arch:AVX2")
+    ENDIF()
+ENDIF()
+
+IF(${FIND_AVX_10})
+    IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+        SET(AVX_FLAGS "${AVX_FLAGS} -mavx")
+    ELSEIF(MSVC)
+        SET(AVX_FLAGS "${AVX_FLAGS} /arch:AVX")
+    ENDIF()
+ENDIF()
+
+IF("${FIND_AVX_10}" OR "${FIND_AVX_20}")
+    SET(AVX_FOUND TRUE)
+    MESSAGE(STATUS "Find CPU supports ${AVX_FLAGS}.")
+ENDIF()

cmake/cblas.cmake

@@ -65,12 +65,14 @@ set(OPENBLAS_ROOT $ENV{OPENBLAS_ROOT} CACHE PATH "Folder contains Openblas")
 set(OPENBLAS_INCLUDE_SEARCH_PATHS
         ${OPENBLAS_ROOT}/include
         /usr/include
-        /usr/include/openblas)
+        /usr/include/openblas
+        /usr/local/opt/openblas/include)
 set(OPENBLAS_LIB_SEARCH_PATHS
         ${OPENBLAS_ROOT}/lib
         /usr/lib
         /usr/lib/blas/openblas
-        /usr/lib/openblas)
+        /usr/lib/openblas
+        /usr/local/opt/openblas/lib)
 
 find_path(OPENBLAS_INC_DIR NAMES cblas.h
   PATHS ${OPENBLAS_INCLUDE_SEARCH_PATHS})

cmake/cudnn.cmake

@@ -15,7 +15,7 @@ list(APPEND CUDNN_CHECK_LIBRARY_DIRS
     $ENV{CUDNN_ROOT}/lib64
     $ENV{CUDNN_ROOT}/lib
     /usr/lib)
-find_library(CUDNN_LIBRARY NAMES libcudnn.so # libcudnn_static.a
+find_library(CUDNN_LIBRARY NAMES libcudnn.so libcudnn.dylib # libcudnn_static.a
     PATHS ${CUDNN_CHECK_LIBRARY_DIRS} ${CUDNN_INCLUDE_DIR} ${__libpath_hist}
           NO_DEFAULT_PATH
     DOC "Path to cuDNN library.")

cmake/flags.cmake

@@ -8,7 +8,7 @@ include(CheckCXXSymbolExists)
 # is_c: is C flag or C++ flag, bool type.
 # src_list: The list name which the flag name will be append to.
 # flag_name: the flag name for compiler, such as '-Werror' '-Wall' etc
-# rest arguments: not used. 
+# rest arguments: not used.
 function(safe_set_flag is_c src_list flag_name)
     string(REPLACE "-" "_" safe_name ${flag_name})
     string(REPLACE "=" "_" safe_name ${safe_name})
@@ -44,7 +44,7 @@ CHECK_CXX_SYMBOL_EXISTS(UINT64_MAX "stdint.h" UINT64_MAX_EXISTS)
 if(NOT UINT64_MAX_EXISTS)
   set(CMAKE_REQUIRED_DEFINITIONS -D__STDC_LIMIT_MACROS)
   CHECK_CXX_SYMBOL_EXISTS(UINT64_MAX "stdint.h" UINT64_MAX_EXISTS_HERE)
-  if(UINT64_MAX_EXISTS_HERE) 
+  if(UINT64_MAX_EXISTS_HERE)
     set(CMAKE_REQUIRED_DEFINITIONS)
     add_definitions(-D__STDC_LIMIT_MACROS)
   else()
@@ -74,13 +74,37 @@ endforeach()
 # Release/Debug flags set by cmake. Such as -O3 -g -DNDEBUG etc.
 # So, don't set these flags here.
 
+function(specify_cuda_arch cuda_version cuda_arch)
+    if(${cuda_version} VERSION_GREATER "8.0")
+        foreach(capability 61 62)
+          if(${cuda_arch} STREQUAL ${capability})
+            list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
+          endif()
+        endforeach()
+    elseif(${cuda_version} VERSION_GREATER "7.0" and ${cuda_arch} STREQUAL "53")
+        list(APPEND __arch_flags " -gencode arch=compute_${cuda_arch},code=sm_${cuda_arch}")
+    endif()
+endfunction()
+
+# Common gpu architectures: Kepler, Maxwell
 foreach(capability 30 35 50)
-    list(APPEND __arch_flags "-gencode arch=compute_${capability},code=sm_${capability}")
+    list(APPEND __arch_flags " -gencode arch=compute_${capability},code=sm_${capability}")
 endforeach()
 
 if (CUDA_VERSION VERSION_GREATER "7.0")
-    list(APPEND __arch_flags "-gencode arch=compute_52,code=sm_52")
+      list(APPEND __arch_flags " -gencode arch=compute_52,code=sm_52")
 endif()
 
-set(CUDA_NVCC_FLAGS ${__arch_flags} ${CUDA_NVCC_FLAGS})
+# Modern gpu architectures: Pascal
+if (CUDA_VERSION VERSION_GREATER "8.0")
+      list(APPEND __arch_flags " -gencode arch=compute_60,code=sm_60")
+endif()
 
+# Custom gpu architecture
+set(CUDA_ARCH)
+
+if(CUDA_ARCH)
+  specify_cuda_arch(${CUDA_VERSION} ${CUDA_ARCH})
+endif()
+
+set(CUDA_NVCC_FLAGS ${__arch_flags} ${CUDA_NVCC_FLAGS})

cmake/util.cmake

@@ -1,16 +1,55 @@
 # Some common routine for paddle compile.
 
-
 # target_circle_link_libraries
 # Link libraries to target which has circle dependencies.
 #
 # First Argument: target name want to be linked with libraries
 # Rest Arguments: libraries which link together.
 function(target_circle_link_libraries TARGET_NAME)
-    target_link_libraries(${TARGET_NAME}
-        -Wl,--start-group
-        ${ARGN}
-        -Wl,--end-group)
+    if(APPLE)
+        set(LIBS)
+        set(inArchive OFF)
+        set(libsInArgn)
+
+        foreach(arg ${ARGN})
+            if(${arg} STREQUAL "ARCHIVE_START")
+                set(inArchive ON)
+            elseif(${arg} STREQUAL "ARCHIVE_END")
+                set(inArchive OFF)
+            else()
+                if(inArchive)
+                    list(APPEND LIBS "-Wl,-force_load")
+                endif()
+                list(APPEND LIBS ${arg})
+                list(APPEND libsInArgn ${arg})
+            endif()
+        endforeach()
+        if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
+            list(APPEND LIBS "-undefined dynamic_lookup")
+        endif()
+        list(REVERSE libsInArgn)
+        target_link_libraries(${TARGET_NAME}
+            ${LIBS}
+            ${libsInArgn})
+
+    else()  # LINUX
+        set(LIBS)
+
+        foreach(arg ${ARGN})
+            if(${arg} STREQUAL "ARCHIVE_START")
+                list(APPEND LIBS "-Wl,--whole-archive")
+            elseif(${arg} STREQUAL "ARCHIVE_END")
+                list(APPEND LIBS "-Wl,--no-whole-archive")
+            else()
+                list(APPEND LIBS ${arg})
+            endif()
+        endforeach()
+
+        target_link_libraries(${TARGET_NAME}
+                "-Wl,--start-group"
+                ${LIBS}
+                "-Wl,--end-group")
+    endif()
 endfunction()
 
 # compile_cu_as_cpp
@@ -41,20 +80,20 @@ function(link_paddle_exe TARGET_NAME)
     if(PADDLE_WITH_INTERNAL)
         set(INTERAL_LIBS paddle_internal_gserver paddle_internal_parameter)
         target_circle_link_libraries(${TARGET_NAME}
-            -Wl,--whole-archive
+            ARCHIVE_START
             paddle_internal_gserver
             paddle_internal_owlqn
-            -Wl,--no-whole-archive
+            ARCHIVE_END
             paddle_internal_parameter)
     else()
         set(INTERAL_LIBS "")
     endif()
 
     target_circle_link_libraries(${TARGET_NAME}
-        -Wl,--whole-archive
+        ARCHIVE_START
         paddle_gserver
         ${METRIC_LIBS}
-        -Wl,--no-whole-archive
+        ARCHIVE_END
         paddle_pserver
         paddle_trainer_lib
         paddle_network
@@ -67,9 +106,9 @@ function(link_paddle_exe TARGET_NAME)
         ${PROTOBUF_LIBRARY}
         ${CMAKE_THREAD_LIBS_INIT}
         ${CBLAS_LIBS}
-        ${CMAKE_DL_LIBS}
+        ${ZLIB_LIBRARIES}
         ${INTERAL_LIBS}
-        -lz)
+        ${CMAKE_DL_LIBS})
 
     if(WITH_PYTHON)
         target_link_libraries(${TARGET_NAME}

demo/image_classification/prediction.py

@@ -20,9 +20,8 @@
 
 import paddle.utils.image_util as image_util
 
-from py_paddle import swig_paddle, util
-from py_paddle import DataProviderWrapperConverter
-from paddle.trainer.PyDataProviderWrapper import DenseSlot
+from py_paddle import swig_paddle, DataProviderConverter
+from paddle.trainer.PyDataProvider2 import dense_vector
 from paddle.trainer.config_parser import parse_config
 
 logging.basicConfig(format='[%(levelname)s %(asctime)s %(filename)s:%(lineno)s] %(message)s')
@@ -75,8 +74,8 @@ def __init__(self,
         self.network.loadParameters(self.model_dir)
 
         data_size = 3 * self.crop_dims[0] * self.crop_dims[1]
-        slots = [DenseSlot(data_size)]
-        self.converter = util.DataProviderWrapperConverter(False, slots)
+        slots = [dense_vector(data_size)]
+        self.converter = DataProviderConverter(slots)
 
     def get_data(self, img_path):
         """