Merge pull request #2213 from xianyi/develop

Update from develop in preparation of the 0.3.7 release

Author: martin-frbg

.drone.yml

@@ -0,0 +1,143 @@
+---
+kind: pipeline
+name: arm64_gcc_make
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:19.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV8 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm32_gcc_make
+
+platform:
+  os: linux
+  arch: arm
+
+steps:
+- name: Build and Test
+  image: ubuntu:19.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV6 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm64_clang_make
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV8 NUM_THREADS=32'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+
+---
+kind: pipeline
+name: arm32_clang_cmake
+
+platform:
+  os: linux
+  arch: arm
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV6 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest
+
+---
+kind: pipeline
+name: arm64_gcc_cmake
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: gcc
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV8 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest
+
+---
+kind: pipeline
+name: arm64_clang_cmake
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: clang
+    CMAKE_FLAGS: '-DDYNAMIC_ARCH=1 -DTARGET=ARMV8 -DNUM_THREADS=32 -DNOFORTRAN=ON -DBUILD_WITHOUT_LAPACK=ON'
+  commands:
+    - echo "CMAKE_FLAGS:= $CMAKE_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC g++ perl cmake
+    - $CC --version
+    - mkdir build && cd build
+    - cmake $CMAKE_FLAGS ..
+    - make -j
+    - ctest

.travis.yml

@@ -25,6 +25,15 @@ matrix:
         - TARGET_BOX=LINUX64
         - BTYPE="BINARY=64"
 
+    - <<: *test-ubuntu
+      os: linux-ppc64le
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=POWER8 NUM_THREADS=32"
+      env:
+        # for matrix annotation only
+        - TARGET_BOX=PPC64LE_LINUX
+        - BTYPE="BINARY=64 USE_OPENMP=1"
+
     - <<: *test-ubuntu
       env:
         - TARGET_BOX=LINUX64
@@ -164,42 +173,6 @@ matrix:
       env:
         - BTYPE="BINARY=32"
 
-    - &emulated-arm
-      dist: trusty
-      sudo: required
-      services: docker
-      env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=gcc
-      name: "Emulated Build for ARMV6 with gcc"
-      before_install: sudo docker run --rm --privileged multiarch/qemu-user-static:register --reset
-      script: |
-        echo "FROM openblas/alpine:${IMAGE_ARCH}
-        COPY . /tmp/openblas
-        RUN mkdir /tmp/openblas/build                             &&  \
-            cd /tmp/openblas/build                                &&  \
-            CC=${COMPILER} cmake -D DYNAMIC_ARCH=OFF                  \
-                                 -D TARGET=${TARGET_ARCH}             \
-                                 -D BUILD_SHARED_LIBS=ON              \
-                                 -D BUILD_WITHOUT_LAPACK=ON           \
-                                 -D BUILD_WITHOUT_CBLAS=ON            \
-                                 -D CMAKE_BUILD_TYPE=Release ../  &&  \
-            cmake --build ." > Dockerfile
-        docker build .
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=clang
-      name: "Emulated Build for ARMV6 with clang"
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=gcc
-      name: "Emulated Build for ARMV8 with gcc"
-    - <<: *emulated-arm
-      env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=clang
-      name: "Emulated Build for ARMV8 with clang"
-
-  allow_failures:
-    - env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=gcc
-    - env: IMAGE_ARCH=arm32 TARGET_ARCH=ARMV6 COMPILER=clang
-    - env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=gcc
-    - env: IMAGE_ARCH=arm64 TARGET_ARCH=ARMV8 COMPILER=clang
-
 # whitelist
 branches:
   only:

CMakeLists.txt

@@ -6,7 +6,7 @@ cmake_minimum_required(VERSION 2.8.5)
 project(OpenBLAS C ASM)
 set(OpenBLAS_MAJOR_VERSION 0)
 set(OpenBLAS_MINOR_VERSION 3)
-set(OpenBLAS_PATCH_VERSION 6)
+set(OpenBLAS_PATCH_VERSION 7.dev)
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 
 # Adhere to GNU filesystem layout conventions
@@ -20,9 +20,14 @@ if(MSVC)
 option(BUILD_WITHOUT_LAPACK "Do not build LAPACK and LAPACKE (Only BLAS or CBLAS)" ON)
 endif()
 option(BUILD_WITHOUT_CBLAS "Do not build the C interface (CBLAS) to the BLAS functions" OFF)
-option(DYNAMIC_ARCH "Include support for multiple CPU targets, with automatic selection at runtime (x86/x86_64 only)" OFF)
-option(DYNAMIC_OLDER "Include specific support for older cpu models (Penryn,Dunnington,Atom,Nano,Opteron) with DYNAMIC_ARCH" OFF)
+option(DYNAMIC_ARCH "Include support for multiple CPU targets, with automatic selection at runtime (x86/x86_64, aarch64 or ppc only)" OFF)
+option(DYNAMIC_OLDER "Include specific support for older x86 cpu models (Penryn,Dunnington,Atom,Nano,Opteron) with DYNAMIC_ARCH" OFF)
 option(BUILD_RELAPACK "Build with ReLAPACK (recursive implementation of several LAPACK functions on top of standard LAPACK)" OFF)
+if(${CMAKE_SYSTEM_NAME} MATCHES "Linux")
+option(NO_AFFINITY "Disable support for CPU affinity masks to avoid binding processes from e.g. R or numpy/scipy to a single core" ON)
+else()
+set(NO_AFFINITY 1)
+endif()
 
 # Add a prefix or suffix to all exported symbol names in the shared library.
 # Avoids conflicts with other BLAS libraries, especially when using
@@ -206,7 +211,8 @@ if (USE_THREAD)
   target_link_libraries(${OpenBLAS_LIBNAME} ${CMAKE_THREAD_LIBS_INIT})
 endif()
 
-if (MSVC OR NOT NOFORTRAN)
+#if (MSVC OR NOT NOFORTRAN)
+if (NOT NO_CBLAS)
   # Broken without fortran on unix
   add_subdirectory(utest)
 endif()

CONTRIBUTORS.md

@@ -167,4 +167,7 @@ In chronological order:
   * [2017-02-26] ztrmm kernel for IBM z13
   * [2017-03-13] strmm and ctrmm kernel for IBM z13
   * [2017-09-01] initial Blas Level-1,2 (double precision) for IBM z13
-
+  * [2018-03-07] added missing Blas Level 1-2  (double precision) simd codes
+  * [2019-02-01] added missing Blas Level-1,2 (single precision)  simd codes
+  * [2019-03-14] power9 dgemm/dtrmm kernel
+  * [2019-04-29] power9 sgemm/strmm kernel 

Makefile

@@ -34,7 +34,7 @@ endif
 
 LAPACK_NOOPT := $(filter-out -O0 -O1 -O2 -O3 -Ofast,$(LAPACK_FFLAGS))
 
-SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench
+SUBDIRS_ALL = $(SUBDIRS) test ctest utest exports benchmark ../laswp ../bench cpp_thread_test
 
 .PHONY : all libs netlib $(RELA) test ctest shared install
 .NOTPARALLEL : all libs $(RELA) prof lapack-test install blas-test
@@ -109,6 +109,7 @@ endif
 ifeq ($(OSNAME), Darwin)
 	@$(MAKE) -C exports dyn
 	@ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib
+	@ln -fs $(LIBDYNNAME) $(LIBPREFIX).$(MAJOR_VERSION).dylib
 endif
 ifeq ($(OSNAME), WINNT)
 	@$(MAKE) -C exports dll
@@ -123,10 +124,13 @@ ifeq ($(NOFORTRAN), $(filter 0,$(NOFORTRAN)))
 	touch $(LIBNAME)
 ifndef NO_FBLAS
 	$(MAKE) -C test all
-	$(MAKE) -C utest all
 endif
+	$(MAKE) -C utest all
 ifndef NO_CBLAS
 	$(MAKE) -C ctest all
+ifeq ($(CPP_THREAD_SAFETY_TEST), 1)
+	$(MAKE) -C cpp_thread_test all
+endif
 endif
 endif
 

Makefile.arm

@@ -1,19 +1,14 @@
 ifeq ($(CORE), $(filter $(CORE),ARMV7 CORTEXA9 CORTEXA15))
 ifeq ($(OSNAME), Android)
-CCOMMON_OPT += -mfpu=neon  -march=armv7-a
-FCOMMON_OPT += -mfpu=neon  -march=armv7-a
+CCOMMON_OPT += -mfpu=neon
+FCOMMON_OPT += -mfpu=neon
 else
 CCOMMON_OPT += -mfpu=vfpv3 -march=armv7-a
 FCOMMON_OPT += -mfpu=vfpv3 -march=armv7-a
 endif
 endif
 
 ifeq ($(CORE), ARMV6)
-CCOMMON_OPT += -mfpu=vfp -march=armv6
-FCOMMON_OPT += -mfpu=vfp -march=armv6
-endif
-
-ifeq ($(CORE), ARMV5)
-CCOMMON_OPT += -march=armv5
-FCOMMON_OPT += -march=armv5
+CCOMMON_OPT += -mfpu=vfp
+FCOMMON_OPT += -mfpu=vfp
 endif

Makefile.install

@@ -83,7 +83,8 @@ ifeq ($(OSNAME), Darwin)
 	@-cp $(LIBDYNNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
 	@-install_name_tool -id "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)" "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
-	ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib
+	ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib ; \
+	ln -fs $(LIBDYNNAME) $(LIBPREFIX).$(MAJOR_VERSION).dylib
 endif
 ifeq ($(OSNAME), WINNT)
 	@-cp $(LIBDLLNAME) "$(DESTDIR)$(OPENBLAS_BINARY_DIR)"

Makefile.power

@@ -29,6 +29,10 @@ FCOMMON_OPT += -O2 -frecursive -mcpu=power8 -mtune=power8 -malign-power -fno-fas
 endif
 endif
 
+# workaround for C->FORTRAN ABI violation in LAPACKE
+ifeq ($(F_COMPILER), GFORTRAN)
+FCOMMON_OPT += -fno-optimize-sibling-calls
+endif
 
 FLAMEPATH	= $(HOME)/flame/lib