Merge pull request #6 from xianyi/develop

update to current develop

Author: martin-frbg

appveyor.yml

@@ -38,7 +38,8 @@ environment:
     - COMPILER: MinGW64-gcc-7.2.0-mingw
       DYNAMIC_ARCH: OFF
       WITH_FORTRAN: ignore
-    - COMPILER: MinGW64-gcc-7.2.0
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
+      COMPILER: MinGW-gcc-6.3.0-32   
     - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
       COMPILER: MinGW-gcc-5.3.0
       WITH_FORTRAN: ignore
@@ -62,10 +63,10 @@ before_build:
   - set PATH=%PATH:C:\Program Files\Git\usr\bin;=%
   - if [%COMPILER%]==[MinGW-gcc-5.3.0] set PATH=C:\MinGW\bin;C:\msys64\usr\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
   - if [%COMPILER%]==[MinGW64-gcc-7.2.0-mingw] set PATH=C:\MinGW\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
-  - if [%COMPILER%]==[MinGW64-gcc-7.2.0] set PATH=C:\msys64\usr\bin;C:\mingw-w64\x86_64-7.2.0-posix-seh-rt_v5-rev1\mingw64\bin;%PATH%
+  - if [%COMPILER%]==[MinGW-gcc-6.3.0-32] set PATH=C:\msys64\usr\bin;C:\mingw-w64\i686-6.3.0-posix-dwarf-rt_v5-rev1\mingw64\bin;%PATH%
   - if [%COMPILER%]==[cl] cmake -G "Visual Studio 15 2017 Win64" ..
   - if [%COMPILER%]==[MinGW64-gcc-7.2.0-mingw] cmake -G "MinGW Makefiles" -DNOFORTRAN=1 ..
-  - if [%COMPILER%]==[MinGW64-gcc-7.2.0] cmake -G "MSYS Makefiles"  -DBINARY=32 -DNOFORTRAN=1 ..
+  - if [%COMPILER%]==[MinGW-gcc-6.3.0-32] cmake -G "MSYS Makefiles" -DNOFORTRAN=1 ..
   - if [%COMPILER%]==[MinGW-gcc-5.3.0] cmake -G "MSYS Makefiles" -DNOFORTRAN=1 ..
   - if [%WITH_FORTRAN%]==[no] cmake -G "Ninja" -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER=clang-cl -DMSVC_STATIC_CRT=ON ..
   - if [%WITH_FORTRAN%]==[yes] cmake -G "Ninja" -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_C_COMPILER=clang-cl -DCMAKE_Fortran_COMPILER=flang -DBUILD_WITHOUT_LAPACK=no -DNOFORTRAN=0 ..

cmake/cc.cmake

@@ -3,7 +3,7 @@
 ## Description: Ported from portion of OpenBLAS/Makefile.system
 ##              Sets C related variables.
 
-if (${CMAKE_C_COMPILER} STREQUAL "GNU" OR ${CMAKE_C_COMPILER} STREQUAL "LSB" OR ${CMAKE_C_COMPILER} STREQUAL "Clang")
+if (${CMAKE_C_COMPILER_ID} STREQUAL "GNU" OR ${CMAKE_C_COMPILER_ID} STREQUAL "LSB" OR ${CMAKE_C_COMPILER_ID} MATCHES "Clang")
 
   set(CCOMMON_OPT "${CCOMMON_OPT} -Wall")
   set(COMMON_PROF "${COMMON_PROF} -fno-inline")
@@ -43,23 +43,23 @@ if (${CMAKE_C_COMPILER} STREQUAL "GNU" OR ${CMAKE_C_COMPILER} STREQUAL "LSB" OR
   endif ()
 endif ()
 
-if (${CMAKE_C_COMPILER} STREQUAL "PGI")
+if (${CMAKE_C_COMPILER_ID} STREQUAL "PGI")
   if (BINARY64)
     set(CCOMMON_OPT "${CCOMMON_OPT} -tp p7-64")
   else ()
     set(CCOMMON_OPT "${CCOMMON_OPT} -tp p7")
   endif ()
 endif ()
 
-if (${CMAKE_C_COMPILER} STREQUAL "PATHSCALE")
+if (${CMAKE_C_COMPILER_ID} STREQUAL "PATHSCALE")
   if (BINARY64)
     set(CCOMMON_OPT "${CCOMMON_OPT} -m64")
   else ()
     set(CCOMMON_OPT "${CCOMMON_OPT} -m32")
   endif ()
 endif ()
 
-if (${CMAKE_C_COMPILER} STREQUAL "OPEN64")
+if (${CMAKE_C_COMPILER_ID} STREQUAL "OPEN64")
 
   if (MIPS64)
 
@@ -87,7 +87,7 @@ if (${CMAKE_C_COMPILER} STREQUAL "OPEN64")
   endif ()
 endif ()
 
-if (${CMAKE_C_COMPILER} STREQUAL "SUN")
+if (${CMAKE_C_COMPILER_ID} STREQUAL "SUN")
   set(CCOMMON_OPT "${CCOMMON_OPT} -w")
   if (X86)
     set(CCOMMON_OPT "${CCOMMON_OPT} -m32")

cmake/prebuild.cmake

@@ -105,6 +105,7 @@ if (DEFINED CORE AND CMAKE_CROSSCOMPILING AND NOT (${HOST_OS} STREQUAL "WINDOWSS
   # Perhaps this should be inside a different file as it grows larger
   file(APPEND ${TARGET_CONF_TEMP}
     "#define ${TCORE}\n"
+    "#define CORE_${TCORE}\n"
     "#define CHAR_CORENAME \"${TCORE}\"\n")
   if ("${TCORE}" STREQUAL "CORE2")
     file(APPEND ${TARGET_CONF_TEMP}
@@ -119,15 +120,23 @@ if (DEFINED CORE AND CMAKE_CROSSCOMPILING AND NOT (${HOST_OS} STREQUAL "WINDOWSS
       "#define HAVE_SSE\n"
       "#define HAVE_SSE2\n"
       "#define HAVE_SSE3\n"
-      "#define HAVE_SSSE3\n")
+      "#define HAVE_SSSE3\n"
+      "#define SLOCAL_BUFFER_SIZE\t16384\n"
+      "#define DLOCAL_BUFFER_SIZE\t16384\n"
+      "#define CLOCAL_BUFFER_SIZE\t16384\n"
+      "#define ZLOCAL_BUFFER_SIZE\t16384\n")
       set(SGEMM_UNROLL_M 8)
       set(SGEMM_UNROLL_N 4)
       set(DGEMM_UNROLL_M 4)
       set(DGEMM_UNROLL_N 4)
-      set(CGEMM_DEFAULT_UNROLL_M 4)
-      set(CGEMM_DEFAULT_UNROLL_N 2)
-      set(ZGEMM_DEFAULT_UNROLL_M 2)
-      set(ZGEMM_DEFAULT_UNROLL_N 2)      
+      set(CGEMM_UNROLL_M 4)
+      set(CGEMM_UNROLL_N 2)
+      set(ZGEMM_UNROLL_M 2)
+      set(ZGEMM_UNROLL_N 2)
+      set(CGEMM3M_UNROLL_M 8)
+      set(CGEMM3M_UNROLL_N 4)
+      set(ZGEMM3M_UNROLL_M 4)
+      set(ZGEMM3M_UNROLL_N 4)
   elseif ("${TCORE}" STREQUAL "ARMV7")
     file(APPEND ${TARGET_CONF_TEMP}
       "#define L1_DATA_SIZE\t65536\n"
@@ -143,6 +152,10 @@ if (DEFINED CORE AND CMAKE_CROSSCOMPILING AND NOT (${HOST_OS} STREQUAL "WINDOWSS
     set(SGEMM_UNROLL_N 4)
     set(DGEMM_UNROLL_M 4)
     set(DGEMM_UNROLL_N 4)
+    set(CGEMM_UNROLL_M 2)
+    set(CGEMM_UNROLL_N 2)
+    set(ZGEMM_UNROLL_M 2)
+    set(ZGEMM_UNROLL_N 2)
   elseif ("${TCORE}" STREQUAL "ARMV8")
     file(APPEND ${TARGET_CONF_TEMP}
       "#define L1_DATA_SIZE\t32768\n"
@@ -331,6 +344,9 @@ else(NOT CMAKE_CROSSCOMPILING)
     set(GETARCH_FLAGS ${GETARCH_FLAGS} -DFORCE_GENERIC)
   else()
     list(APPEND GETARCH_SRC ${PROJECT_SOURCE_DIR}/cpuid.S)
+    if (DEFINED TARGET_CORE)
+    set(GETARCH_FLAGS ${GETARCH_FLAGS} -DFORCE_${TARGET_CORE})
+  endif ()
   endif ()
 
   if ("${CMAKE_SYSTEM_NAME}" STREQUAL "WindowsStore")

cmake/system_check.cmake

@@ -39,10 +39,18 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc.*|power.*|Power.*")
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "mips64.*")
   set(MIPS64 1)
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "amd64.*|x86_64.*|AMD64.*")
-  if("${CMAKE_SIZEOF_VOID_P}" EQUAL "8")
-    set(X86_64 1)
+  if (NOT BINARY)
+    if("${CMAKE_SIZEOF_VOID_P}" EQUAL "8")
+      set(X86_64 1)
+    else()
+      set(X86 1)
+    endif()
   else()
-    set(X86 1)
+    if (${BINARY} EQUAL "64")
+       set(X86_64 1)
+    else ()
+       set(X86 1)
+    endif()
   endif()
 elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "i686.*|i386.*|x86.*|amd64.*|AMD64.*")
   set(X86 1)
@@ -54,6 +62,22 @@ elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(aarch64.*|AARCH64.*)")
   else()
     set(ARM 1)
   endif()
+elseif (${CMAKE_CROSSCOMPILING})
+   if (${TARGET} STREQUAL "CORE2")
+    if (NOT BINARY)
+       set(X86 1)
+    elseif (${BINARY} EQUAL "64")
+       set(X86_64 1)
+    else ()
+       set(X86 1)
+    endif()
+   elseif (${TARGET} STREQUAL "ARMV7")
+       set(ARM 1)
+   else()
+       set(ARM64 1)
+   endif ()
+else ()
+   message(WARNING "Target ARCH could not be determined, got \"${CMAKE_SYSTEM_PROCESSOR}\"")
 endif()
 
 if (X86_64)
@@ -92,4 +116,3 @@ set (CCOMMON_OPT "${CCOMMON_OPT} -DNO_AVX512")
 endif()
   file(REMOVE "avx512.tmp" "avx512.o")
 endif()
-