Merge pull request #2667 from xianyi/develop

Merge develop into 0.3.0 for 0.3.10 release

Author: martin-frbg

.drone.yml

@@ -8,7 +8,7 @@ platform:
 
 steps:
 - name: Build and Test
-  image: ubuntu:19.04
+  image: ubuntu:18.04
   environment:
     CC: gcc
     COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV8 NUM_THREADS=32'
@@ -32,7 +32,7 @@ platform:
 
 steps:
 - name: Build and Test
-  image: ubuntu:19.04
+  image: ubuntu:18.04
   environment:
     CC: gcc
     COMMON_FLAGS: 'DYNAMIC_ARCH=1 TARGET=ARMV6 NUM_THREADS=32'
@@ -141,3 +141,52 @@ steps:
     - cmake $CMAKE_FLAGS ..
     - make -j
     - ctest -V
+
+---
+kind: pipeline
+name: arm64_native_test
+
+platform:
+  os: linux
+  arch: arm64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'USE_OPENMP=1'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl python g++
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+    - make -C cpp_thread_test dgemm_tester
+---
+kind: pipeline
+name: epyc_native_test
+
+platform:
+  os: linux
+  arch: amd64
+
+steps:
+- name: Build and Test
+  image: ubuntu:18.04
+  environment:
+    CC: gcc
+    COMMON_FLAGS: 'USE_OPENMP=1'
+  commands:
+    - echo "MAKE_FLAGS:= $COMMON_FLAGS"
+    - apt-get update -y
+    - apt-get install -y make $CC gfortran perl python g++
+    - $CC --version
+    - make QUIET_MAKE=1 $COMMON_FLAGS
+    - make -C test $COMMON_FLAGS
+    - make -C ctest $COMMON_FLAGS
+    - make -C utest $COMMON_FLAGS
+    - make -C cpp_thread_test dgemm_tester

.github/workflows/dynamic_arch.yml

@@ -0,0 +1,81 @@
+name: continuous build
+
+on: [push, pull_request]
+
+jobs:
+  build:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest]
+        build: [cmake, make]
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v2
+
+      - name: Compilation cache
+        uses: actions/cache@v2
+        with:
+          path: ~/.ccache
+          # We include the commit sha in the cache key, as new cache entries are
+          # only created if there is no existing entry for the key yet.
+          key: ${{ runner.os }}-ccache-${{ github.sha }}
+          # Restore any ccache cache entry, if none for
+          # ${{ runner.os }}-ccache-${{ github.sha }} exists
+          restore-keys: |
+            ${{ runner.os }}-ccache
+
+      - name: Print system information
+        run: |
+          if [ "$RUNNER_OS" == "Linux" ]; then
+            cat /proc/cpuinfo
+          elif [ "$RUNNER_OS" == "macOS" ]; then
+            sysctl -a | grep machdep.cpu
+          else
+            echo "$RUNNER_OS not supported"
+            exit 1
+          fi
+
+      - name: Install Dependencies
+        run: |
+          if [ "$RUNNER_OS" == "Linux" ]; then
+            sudo apt-get install -y gfortran cmake ccache
+          elif [ "$RUNNER_OS" == "macOS" ]; then
+            brew install coreutils cmake ccache
+          else
+            echo "$RUNNER_OS not supported"
+            exit 1
+          fi
+          ccache -M 300M  # Limit the ccache size; Github's overall cache limit is 5GB
+
+      - name: Build
+        if: matrix.build == 'make'
+        run: |
+          if [ "$RUNNER_OS" == "Linux" ]; then
+            export PATH="/usr/lib/ccache:${PATH}"
+          elif [ "$RUNNER_OS" == "macOS" ]; then
+            export PATH="$(brew --prefix)/opt/ccache/libexec:${PATH}"
+          else
+            echo "$RUNNER_OS not supported"
+            exit 1
+          fi
+
+          make -j$(nproc) DYNAMIC_ARCH=1 USE_OPENMP=0
+
+      - name: CMake build
+        if: matrix.build == 'cmake'
+        run: |
+          if [ "$RUNNER_OS" == "Linux" ]; then
+            export PATH="/usr/lib/ccache:${PATH}"
+          elif [ "$RUNNER_OS" == "macOS" ]; then
+            export PATH="$(brew --prefix)/opt/ccache/libexec:${PATH}"
+          else
+            echo "$RUNNER_OS not supported"
+            exit 1
+          fi
+
+          mkdir build
+          cd build
+          cmake -DDYNAMIC_ARCH=1 -DNOFORTRAN=0 -DBUILD_WITHOUT_LAPACK=0  -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_BUILD_TYPE=Release ..
+          make -j$(nproc)

.github/workflows/nightly-Homebrew-build.yml

@@ -21,6 +21,7 @@ jobs:
   build-OpenBLAS-with-Homebrew:
     runs-on: macos-latest
     env:
+      DEVELOPER_DIR: /Applications/Xcode_11.4.1.app/Contents/Developer
       HOMEBREW_DEVELOPER: "ON"
       HOMEBREW_DISPLAY_INSTALL_TIMES: "ON"
       HOMEBREW_NO_ANALYTICS: "ON"

.gitignore

@@ -70,6 +70,7 @@ test/SBLAT2.SUMM
 test/SBLAT3.SUMM
 test/ZBLAT2.SUMM
 test/ZBLAT3.SUMM
+test/SHBLAT3.SUMM
 test/cblat1
 test/cblat2
 test/cblat3
@@ -79,6 +80,7 @@ test/dblat3
 test/sblat1
 test/sblat2
 test/sblat3
+test/test_shgemm
 test/zblat1
 test/zblat2
 test/zblat3

.travis.yml

@@ -16,7 +16,6 @@ matrix:
       before_script: &common-before
         - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=NEHALEM NUM_THREADS=32"
       script:
-        - set -e
         - make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
         - make -C test $COMMON_FLAGS $BTYPE
         - make -C ctest $COMMON_FLAGS $BTYPE
@@ -34,6 +33,16 @@ matrix:
         - TARGET_BOX=PPC64LE_LINUX
         - BTYPE="BINARY=64 USE_OPENMP=1"
 
+    - <<: *test-ubuntu
+      os: linux
+      arch: s390x
+      before_script:
+        - COMMON_FLAGS="DYNAMIC_ARCH=1 TARGET=Z13 NUM_THREADS=32"
+      env:
+        # for matrix annotation only
+        - TARGET_BOX=IBMZ_LINUX
+        - BTYPE="BINARY=64 USE_OPENMP=1"
+
     - <<: *test-ubuntu
       env:
         - TARGET_BOX=LINUX64
@@ -98,7 +107,6 @@ matrix:
         - sudo sh alpine-chroot-install -p 'build-base gfortran perl linux-headers'
       before_script: *common-before
       script:
-        - set -e
         # XXX: Disable some warnings for now to avoid exceeding Travis limit for log size.
         - alpine make QUIET_MAKE=1 $COMMON_FLAGS $BTYPE
               CFLAGS="-Wno-misleading-indentation -Wno-sign-conversion -Wno-incompatible-pointer-types"
@@ -141,7 +149,6 @@ matrix:
       before_script:
         - COMMON_ARGS="-DTARGET=NEHALEM -DNUM_THREADS=32"
       script:
-        - set -e
         - mkdir build
         - CONFIG=Release
         - cmake -Bbuild -H. $CMAKE_ARGS $COMMON_ARGS -DCMAKE_BUILD_TYPE=$CONFIG
@@ -180,6 +187,12 @@ matrix:
         - CFLAGS="-O2 -Wno-macro-redefined -isysroot /Applications/Xcode-10.1.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS12.1.sdk -arch arm64 -miphoneos-version-min=10.0"
         - BTYPE="TARGET=ARMV8 BINARY=64 HOSTCC=clang NOFORTRAN=1"
 
+    - <<: *test-macos
+      osx_image: xcode10.1
+      env:
+        - CC="/Applications/Xcode-10.1.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang"
+        - CFLAGS="-O2 -mno-thumb -Wno-macro-redefined -isysroot /Applications/Xcode-10.1.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS12.1.sdk -arch armv7 -miphoneos-version-min=5.1"
+        - BTYPE="TARGET=ARMV7 HOSTCC=clang NOFORTRAN=1"
 # whitelist
 branches:
   only:

CMakeLists.txt

@@ -6,7 +6,8 @@ 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 9)
+set(OpenBLAS_PATCH_VERSION 10)
+
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 
 # Adhere to GNU filesystem layout conventions
@@ -23,6 +24,7 @@ option(BUILD_WITHOUT_CBLAS "Do not build the C interface (CBLAS) to the BLAS fun
 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)
+option(USE_LOCKING "Use locks even in single-threaded builds to make them callable from multiple threads" 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()
@@ -86,9 +88,13 @@ if (NOT NO_LAPACK)
   list(APPEND SUBDIRS lapack)
 endif ()
 
+if (NOT DEFINED BUILD_HALF)
+ set (BUILD_HALF false)
+endif ()
 # set which float types we want to build for
 if (NOT DEFINED BUILD_SINGLE AND NOT DEFINED BUILD_DOUBLE AND NOT DEFINED BUILD_COMPLEX AND NOT DEFINED BUILD_COMPLEX16)
   # if none are defined, build for all
+#  set(BUILD_HALF true)
   set(BUILD_SINGLE true)
   set(BUILD_DOUBLE true)
   set(BUILD_COMPLEX true)
@@ -120,6 +126,11 @@ if (BUILD_COMPLEX16)
   list(APPEND FLOAT_TYPES "ZCOMPLEX") # defines COMPLEX and DOUBLE
 endif ()
 
+if (BUILD_HALF)
+  message(STATUS "Building Half Precision")
+  list(APPEND FLOAT_TYPES "HALF") # defines nothing
+endif ()
+
 if (NOT DEFINED CORE OR "${CORE}" STREQUAL "UNKNOWN")
   message(FATAL_ERROR "Detecting CPU failed. Please set TARGET explicitly, e.g. make TARGET=your_cpu_target. Please read README for details.")
 endif ()
@@ -234,7 +245,7 @@ if (BUILD_SHARED_LIBS AND BUILD_RELAPACK)
   if (NOT MSVC)
     target_link_libraries(${OpenBLAS_LIBNAME} "-Wl,-allow-multiple-definition")
   else()
-    target_link_libraries(${OpenBLAS_LIBNAME} "/FORCE:MULTIPLE")
+   set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /FORCE:MULTIPLE")
   endif()
 endif()
 

CONTRIBUTORS.md

@@ -180,3 +180,13 @@ In chronological order:
   * [2019-12-23] optimize AVX2 CGEMM and ZGEMM
   * [2019-12-30] AVX2 CGEMM3M & ZGEMM3M kernels
   * [2020-01-07] optimize AVX2 SGEMM and STRMM
+
+* Rajalakshmi Srinivasaraghavan <https://github.com/RajalakshmiSR>
+  * [2020-04-15] Half-precision GEMM for bfloat16
+
+* Marius Hillenbrand <https://github.com/mhillenibm>
+  * [2020-05-12] Revise dynamic architecture detection for IBM z
+  * [2020-05-12] Add new sgemm and strmm kernel for IBM z14
+
+* Danfeng Zhang <https://github.com/craft-zhang>
+  * [2020-05-20] Improve performance of SGEMM and STRMM on Arm Cortex-A53

Changelog.txt

@@ -1,4 +1,77 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.3.10
+ 14-Jun-2020
+
+common:
+	* Improved thread locking behaviour in blas_server and parallel getrf                         
+	* Imported bugfix 394 from LAPACK (spurious reference to "XERBL"
+	  due to overlong lines)
+	* Imported bugfix 403 from LAPACK (compile option "recursive" required
+	  for correctness with Intel and PGI)
+	* Imported bugfix 408 from LAPACK (wrong scaling in ZHEEQUB)
+	* Imported bugfix 411 from LAPACK (infinite loop in LARGV/LARTG/LARTGP)
+	* Fixed mismatches between BUFFERSIZE and GEMM_UNROLL parameters that
+	  could lead to crashes at large matrix sizes  
+	* Restored internal soname in dynamic libraries on FreeBSD and Dragonfly 
+	* Added API (openblas_setaffinity) to set the thread affinity on Linux
+	* Added initial infrastructure for half-precision floating point 
+	  (bfloat16) support with a generic implementation of SHGEMM     
+	* Added CMAKE build system support for building the cblas_Xgemm3m
+	  functions
+	* Fixed CMAKE support for building in a path with embedded spaces
+	* Fixed CMAKE (non)handling of NO_EXPRECISION and MAX_STACK_ALLOC
+	* Fixed GCC version detection in the Makefiles
+	* Allowed overriding the names of AR, AS and LD in Makefile builds
+
+POWER:
+	* Fixed big-endian POWER8 ELFv2 builds on FreeBSD
+	* Fixed GCC version checks and DYNAMIC_ARCH builds on POWER9
+	* Fixed CMAKE build support for POWER9
+	* fixed a potential race condition in the thread buffer allocation
+	* Worked around LAPACK test failures on PPC G4
+
+MIPS:
+	* Fixed a potential race condition in the thread buffer allocation
+	* Added support for MIPS 24K/24KE family based on P5600 kernels
+
+MIPS64:
+	* fixed a potential race condition in the thread buffer allocation
+	* Added TARGET=GENERIC
+
+ARMV7:
+	* Fixed a race condition in the thread buffer allocation
+
+ARMV8:
+	* Fixed a race condition in the thread buffer allocation
+	* Fixed zero initialisation in the assembly for SGEMM and DGEMM BETA
+	* Improved performance of the ThunderX2 DAXPY kernel
+	* Added an optimized SGEMM kernel for Cortex A53    
+	* Fixed Makefile support for INTERFACE64 (8-byte integer)
+
+x86_64:
+	* Fixed a syntax error in the CMAKE setup for SkylakeX
+	* Improved performance of STRSM on Haswell, SkylakeX and Ryzen
+	* Improved SGEMM performance on SGEMM for workloads with ldc a
+	  multiple of 1024
+	* Improved DGEMM performance on Skylake X
+	* Fixed unwanted AVX512-dependency of SGEMM in DYNAMIC_ARCH
+	  builds created on SkylakeX
+	* Removed data alignment requirement in the SSE2 copy kernels
+	  that could cause spurious crashes
+	* Added a workaround for an optimizer bug in AppleClang 11.0.3
+	* Fixed LAPACK test failures due to wrong options for Intel Fortran
+	* Fixed compilation and LAPACK test results with recent Flang
+	  and AMD AOCC
+	* Fixed DYNAMIC_ARCH builds with CMAKE on OS X
+	* Fixed missing exports of cblas_i?amin, cblas_i?min, cblas_i?max,
+	  cblas_?sum, cblas_?gemm3m in the shared library on OS 
+	* Fixed reporting of cpu name in DYNAMIC_ARCH builds (would sometimes
+	  show the name of an older generation chip supported by the same kernels)
+
+IBM Z:
+	* Improved performance of SGEMM/STRMM and DGEMM/DTRMM on Z14
+
 ====================================================================
 Version 0.3.9
  1-Mar-2020