Modify compiling flags to support neon and add the definition of PADDLE_USE_LAPACK.

cmake/cblas.cmake

@@ -14,31 +14,6 @@
 #
 
 set(CBLAS_FOUND OFF)
-set(LAPACK_FOUND OFF)
-
-macro(find_lapack)
-  ## Find clapack
-  set(CLAPACK_ROOT $ENV{CLAPACK_ROOT} CACHE PATH "Folder contain clapack")
-  find_path(CLAPACK_INC_DIR NAMES clapack.h PATHS
-    ${CLAPACK_ROOT}/include)
-  find_path(CLAPACK_F2C_INC_DIR NAMES f2c.h PATHS
-    ${CLAPACK_ROOT}/include)
-  find_library(CLAPACK_F2C_LIB NAMES f2c PATHS
-    ${CLAPACK_ROOT}/lib)
-  find_library(CLAPACK_CBLASWR_LIB NAMES cblaswr PATHS
-    ${CLAPACK_ROOT}/lib)
-  find_library(CLAPACK_CLAPACK_LIB NAMES clapack PATHS
-    ${CLAPACK_ROOT}/lib)
-  if(CLAPACK_INC_DIR AND CLAPACK_F2C_INC_DIR AND CLAPACK_F2C_LIB AND
-     CLAPACK_CBLASWR_LIB AND CLAPACK_CLAPACK_LIB AND NOT LAPACK_FOUND)
-    set(LAPACK_PROVIDER CLAPACK)
-    set(LAPACK_INC_DIR ${CLAPACK_INC_DIR})
-    set(LAPACK_LIBRARIES ${CLAPACK_F2C_LIB} ${CLAPACK_CBLASWR_LIB} ${CLAPACK_CLAPACK_LIB})
-    add_definitions(-DPADDLE_USE_CLAPACK)
-    message(STATUS "Found CLAPACK (include: ${LAPACK_INC_DIR}, library: ${LAPACK_LIBRARIES})")
-    set(LAPACK_FOUND ON)
-  endif()
-endmacro()
 
 ## Find MKL First.
 set(INTEL_ROOT "/opt/intel" CACHE PATH "Folder contains intel libs")
@@ -69,11 +44,8 @@ if(MKL_INC_DIR AND MKL_CORE_LIB AND MKL_SEQUENTIAL_LIB AND MKL_INTEL_LP64)
   message(STATUS "Found MKL (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
   set(CBLAS_FOUND ON)
   if(${MKL_LAPACK_INC_DIR})
-    set(LAPACK_INC_DIR ${MKL_LAPACK_INC_DIR})
+    add_definitions(-DPADDLE_USE_LAPACK)
     message(STATUS "Found lapack in MKL (include: ${MKL_LAPACK_INC_DIR})")
-    set(LAPACK_FOUND ON)
-  else()
-    find_lapack()
   endif()
   return() # return file.
 endif()
@@ -108,11 +80,8 @@ if(ATLAS_INC_DIR AND ATLAS_CBLAS_LIB AND ATLAS_LIB AND NOT CBLAS_FOUND)
   message(STATUS "Found ATLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
   set(CBLAS_FOUND ON)
   if(ATLAS_CLAPACK_INC_DIR)
-    set(LAPACK_INC_DIR ${ATLAS_CLAPACK_INC_DIR})
+    add_definitions(-DPADDLE_USE_LAPACK)
     message(STATUS "Found lapack in ATLAS (include: ${ATLAS_CLAPACK_INC_DIR})")
-    set(LAPACK_FOUND ON)
-  else()
-    find_lapack()
   endif()
   return()
 endif()
@@ -145,11 +114,8 @@ if(OPENBLAS_INC_DIR AND OPENBLAS_LIB)
   message(STATUS "Found OpenBLAS (include: ${CBLAS_INC_DIR}, library: ${CBLAS_LIBRARIES})")
   set(CBLAS_FOUND ON)
   if(OPENBLAS_LAPACKE_INC_DIR)
-    set(LAPACK_INC_DIR ${OPENBLAS_LAPACKE_INC_DIR})
+    add_definitions(-DPADDLE_USE_LAPACK)
     message(STATUS "Found lapack in OpenBLAS (include: ${OPENBLAS_LAPACKE_INC_DIR})")
-    set(LAPACK_FOUND ON)
-  else()
-    find_lapack()
   endif()
   return()
 endif()

cmake/external/openblas.cmake

@@ -20,7 +20,6 @@ IF(NOT ${CBLAS_FOUND})
     SET(CBLAS_SOURCES_DIR ${THIRD_PARTY_PATH}/openblas)
     SET(CBLAS_INSTALL_DIR ${THIRD_PARTY_PATH}/install/openblas)
     SET(CBLAS_INC_DIR "${CBLAS_INSTALL_DIR}/include" CACHE PATH "openblas include directory." FORCE)
-    SET(LAPACK_INC_DIR "${CBLAS_INSTALL_DIR}/include" CACHE PATH "lapack (in openblas) include directory." FORCE)
 
     IF(WIN32)
         SET(CBLAS_LIBRARIES "${CBLAS_INSTALL_DIR}/lib/openblas.lib" CACHE FILEPATH "openblas library." FORCE)
@@ -38,6 +37,8 @@ IF(NOT ${CBLAS_FOUND})
                 "you need to set gfortran compiler: cmake .. -DCMAKE_Fortran_COMPILER=...")
     ENDIF(NOT CMAKE_Fortran_COMPILER)
 
+    ADD_DEFINITIONS(-DPADDLE_USE_LAPACK)
+
     ExternalProject_Add(
         openblas
         ${EXTERNAL_PROJECT_LOG_ARGS}
@@ -65,4 +66,3 @@ IF(NOT ${CBLAS_FOUND})
 ENDIF(NOT ${CBLAS_FOUND})
 
 INCLUDE_DIRECTORIES(${CBLAS_INC_DIR})
-INCLUDE_DIRECTORIES(${LAPACK_INC_DIR})

cmake/external/zlib.cmake

@@ -22,7 +22,7 @@ SET(ZLIB_INCLUDE_DIR "${ZLIB_INSTALL_DIR}/include" CACHE PATH "zlib include dire
 IF(WIN32)
   SET(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/zlibstatic.lib" CACHE FILEPATH "zlib library." FORCE)
 ELSE(WIN32)
-  set(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/libz.a" CACHE FILEPATH "zlib library." FORCE)
+  SET(ZLIB_LIBRARIES "${ZLIB_INSTALL_DIR}/lib/libz.a" CACHE FILEPATH "zlib library." FORCE)
 ENDIF(WIN32)
 
 INCLUDE_DIRECTORIES(${ZLIB_INCLUDE_DIR})

cmake/simd.cmake

@@ -10,7 +10,7 @@ IF(CMAKE_COMPILER_IS_GNUCC OR CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID
     set(SSE3_FLAG "-msse3")
     SET(AVX_FLAG "-mavx")
     SET(AVX2_FLAG "-mavx2")
-    SET(NEON_FLAG "-mfloat-abi=softfp -mfpu=neon")
+    SET(NEON_FLAG "-pie -fPIE -mfloat-abi=softfp -mfpu=neon")
 ELSEIF(MSVC)
     set(MMX_FLAG "/arch:MMX")
     set(SSE2_FLAG "/arch:SSE2")

paddle/math/MathFunctions.cpp

@@ -85,11 +85,16 @@ int getrf<float>(const CBLAS_ORDER order,
                  float* A,
                  const int lda,
                  int* ipiv) {
+#ifdef PADDLE_USE_LAPACK
 #ifdef PADDLE_USE_ATLAS
   return clapack_sgetrf(order, M, N, A, lda, ipiv);
 #else
   return LAPACKE_sgetrf(order, M, N, A, lda, ipiv);
 #endif
+#else
+  LOG(FATAL) << "Not implemented";
+#endif
+  return 0;
 }
 
 template <>
@@ -99,11 +104,16 @@ int getrf<double>(const CBLAS_ORDER order,
                   double* A,
                   const int lda,
                   int* ipiv) {
+#ifdef PADDLE_USE_LAPACK
 #ifdef PADDLE_USE_ATLAS
   return clapack_dgetrf(order, M, N, A, lda, ipiv);
 #else
   return LAPACKE_dgetrf(order, M, N, A, lda, ipiv);
 #endif
+#else
+  LOG(FATAL) << "Not implemented".
+#endif
+  return 0;
 }
 
 template <>
@@ -112,11 +122,16 @@ int getri<float>(const CBLAS_ORDER order,
                  float* A,
                  const int lda,
                  const int* ipiv) {
+#ifdef PADDLE_USE_LAPACK
 #ifdef PADDLE_USE_ATLAS
   return clapack_sgetri(order, N, A, lda, ipiv);
 #else
   return LAPACKE_sgetri(order, N, A, lda, ipiv);
 #endif
+#else
+  LOG(FATAL) << "Not implemented".
+#endif
+  return 0;
 }
 
 template <>
@@ -125,11 +140,16 @@ int getri<double>(const CBLAS_ORDER order,
                   double* A,
                   const int lda,
                   const int* ipiv) {
+#ifdef PADDLE_USE_LAPACK
 #ifdef PADDLE_USE_ATLAS
   return clapack_dgetri(order, N, A, lda, ipiv);
 #else
   return LAPACKE_dgetri(order, N, A, lda, ipiv);
 #endif
+#else
+  LOG(FATAL) << "Not implemented".
+#endif
+  return 0;
 }
 
 template <>

paddle/math/MathFunctions.h

@@ -17,23 +17,19 @@ limitations under the License. */
 
 #ifdef PADDLE_USE_MKL
 #include <mkl.h>
-#ifdef PADDLE_USE_CLAPACK
-#include <clapack.h>
-#else
+#ifdef PADDLE_USE_LAPACK
 #include <mkl_lapacke.h>
 #endif
 #else
 extern "C" {
 #include <cblas.h>
 }
+#ifdef PADDLE_USE_LAPACK
 #ifdef PADDLE_USE_ATLAS
 extern "C" {
 #include <clapack.h>
 }
 #else
-#ifdef PADDLE_USE_CLAPACK
-#include <clapack.h>
-#else
 #include <lapacke.h>
 #endif
 #endif

paddle/utils/CpuId.h

@@ -18,7 +18,6 @@ namespace paddle {
 // clang-format off
 enum simd_t {
   SIMD_NONE   = 0,          ///< None
-#if !defined(__ANDROID__)
   SIMD_SSE    = 1 << 0,     ///< SSE
   SIMD_SSE2   = 1 << 1,     ///< SSE 2
   SIMD_SSE3   = 1 << 2,     ///< SSE 3
@@ -30,9 +29,7 @@ enum simd_t {
   SIMD_AVX    = 1 << 8,     ///< AVX
   SIMD_AVX2   = 1 << 9,     ///< AVX 2
   SIMD_AVX512 = 1 << 10,    ///< AVX 512
-#else
-  SIMD_NEON   = 1 << 0,     ///< NEON
-#endif
+  SIMD_NEON   = 1 << 11,    ///  NEON
 };
 // clang-format on
 
@@ -99,6 +96,7 @@ private:
 #define HAS_AVX     HAS_SIMD(SIMD_AVX)
 #define HAS_AVX2    HAS_SIMD(SIMD_AVX2)
 #define HAS_AVX512  HAS_SIMD(SIMD_AVX512)
+#define HAS_NEON    HAS_SIMD(SIMD_NEON)
 // clang-format on
 
 }  // namespace paddle