Merge pull request #62 from xianyi/develop

rebase

.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)

CMakeLists.txt

@@ -23,6 +23,7 @@ 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()

cmake/os.cmake

@@ -8,7 +8,7 @@ if (${CMAKE_SYSTEM_NAME} STREQUAL "Linux")
   set(NO_EXPRECISION 1)
 endif ()
 
-if (${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD|OpenBSD|NetBSD|DragonFly")
+if (${CMAKE_SYSTEM_NAME} MATCHES "FreeBSD|OpenBSD|NetBSD|DragonFly|Darwin")
   set(EXTRALIB "${EXTRALIB} -lm")
   set(NO_EXPRECISION 1)
 endif ()

cmake/system.cmake

@@ -297,6 +297,16 @@ if (USE_SIMPLE_THREADED_LEVEL3)
   set(CCOMMON_OPT "${CCOMMON_OPT} -DUSE_SIMPLE_THREADED_LEVEL3")
 endif ()
 
+if (NOT ${CMAKE_SYSTEM_NAME} STREQUAL "Windows")
+if (DEFINED MAX_STACK_ALLOC)
+if (NOT ${MAX_STACK_ALLOC} EQUAL 0)
+set(CCOMMON_OPT "${CCOMMON_OPT} -DMAX_STACK_ALLOC=${MAX_STACK_ALLOC}")
+endif ()
+else ()
+set(CCOMMON_OPT "${CCOMMON_OPT} -DMAX_STACK_ALLOC=2048")
+endif ()
+endif ()
+
 if (DEFINED LIBNAMESUFFIX)
   set(LIBPREFIX "libopenblas_${LIBNAMESUFFIX}")
 else ()

kernel/arm64/KERNEL.CORTEXA53

@@ -126,16 +126,9 @@ endif
 SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)
 SGEMMITCOPYOBJ =  sgemm_itcopy$(TSUFFIX).$(SUFFIX)
 endif
-ifeq ($(SGEMM_UNROLL_N), 16)
+
 SGEMMOTCOPY    =  sgemm_tcopy_$(SGEMM_UNROLL_N).S
-else
-SGEMMOTCOPY    =  ../generic/gemm_tcopy_$(SGEMM_UNROLL_N).c
-endif
-ifeq ($(SGEMM_UNROLL_N), 4)
 SGEMMONCOPY    =  sgemm_ncopy_$(SGEMM_UNROLL_N).S
-else
-SGEMMONCOPY    =  ../generic/gemm_ncopy_$(SGEMM_UNROLL_N).c
-endif
 SGEMMONCOPYOBJ =  sgemm_oncopy$(TSUFFIX).$(SUFFIX)
 SGEMMOTCOPYOBJ =  sgemm_otcopy$(TSUFFIX).$(SUFFIX)