Merge pull request #2476 from xianyi/develop

Update from develop in preparation for 0.3.9

Author: martin-frbg

.drone.yml

@@ -92,7 +92,7 @@ steps:
     - mkdir build && cd build
     - cmake $CMAKE_FLAGS ..
     - make -j
-    - ctest
+    - ctest -V
 
 ---
 kind: pipeline
@@ -116,7 +116,7 @@ steps:
     - mkdir build && cd build
     - cmake $CMAKE_FLAGS ..
     - make -j
-    - ctest
+    - ctest -V
 
 ---
 kind: pipeline
@@ -140,4 +140,4 @@ steps:
     - mkdir build && cd build
     - cmake $CMAKE_FLAGS ..
     - make -j
-    - ctest
+    - ctest -V

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

@@ -0,0 +1,78 @@
+# Only the "head" branch of the OpenBLAS package is tested
+
+on:
+  push:
+    paths:
+      - '**/nightly-Homebrew-build.yml'
+  pull_request:
+    branches:
+      - develop
+    paths:
+      - '**/nightly-Homebrew-build.yml'
+  schedule:
+    - cron: 45 7 * * *
+# This is 7:45 AM UTC daily, late at night in the USA
+
+# Since push and pull_request will still always be building and testing the `develop` branch,
+# it only makes sense to test if this file has been changed
+
+name: Nightly-Homebrew-Build
+jobs:
+  build-OpenBLAS-with-Homebrew:
+    runs-on: macos-latest
+    env:
+      HOMEBREW_DEVELOPER: "ON"
+      HOMEBREW_DISPLAY_INSTALL_TIMES: "ON"
+      HOMEBREW_NO_ANALYTICS: "ON"
+      HOMEBREW_NO_AUTO_UPDATE: "ON"
+      HOMEBREW_NO_BOTTLE_SOURCE_FALLBACK: "ON"
+      HOMEBREW_NO_INSTALL_CLEANUP: "ON"
+
+    steps:
+      - name: Random delay for cron job
+        run: |
+          delay=$(( RANDOM % 600 ))
+          printf 'Delaying for %s seconds on event %s' ${delay} "${{ github.event_name }}"
+          sleep ${delay}
+        if: github.event_name == 'schedule'
+
+      - uses: actions/checkout@v2
+        # This isn't even needed, technically. Homebrew will get `develop` via git
+
+      - name: Update Homebrew
+        if: github.event_name != 'pull_request'
+        run: brew update || true
+
+      - name: Install prerequisites
+        run: brew install --fetch-HEAD --HEAD --only-dependencies --keep-tmp openblas
+
+      - name: Install and bottle OpenBLAS
+        run: brew install --fetch-HEAD --HEAD --build-bottle --keep-tmp openblas
+        # the HEAD flags tell Homebrew to build the develop branch fetch via git
+
+      - name: Create bottle
+        run: |
+          brew bottle -v openblas
+          mkdir bottles
+          mv *.bottle.tar.gz bottles
+
+      - name: Upload bottle
+        uses: actions/upload-artifact@v1
+        with:
+          name: openblas--HEAD.catalina.bottle.tar.gz
+          path: bottles
+
+      - name: Show linkage
+        run: brew linkage -v openblas
+
+      - name: Test openblas
+        run: brew test --HEAD --verbose openblas
+
+      - name: Audit openblas formula
+        run: |
+          brew audit --strict openblas
+          brew cat openblas
+
+      - name: Post logs on failure
+        if: failure()
+        run: brew gist-logs --with-hostname -v openblas

.gitignore

@@ -87,4 +87,5 @@ build.*
 *.swp
 benchmark/*.goto
 benchmark/smallscaling
-
+CMakeCache.txt
+CMakeFiles/*

.travis.yml

@@ -176,7 +176,7 @@ matrix:
     - <<: *test-macos
       osx_image: xcode10.1
       env:
-        - CC="/Applications/Xcode-10.1.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang -isysroot /Applications/Xcode-10.1.app/Contents/Developer/Platforms/iPhoneOS.platform/Developer/SDKs/iPhoneOS12.1.sdk"
+        - CC="/Applications/Xcode-10.1.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang"
         - 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"
 

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 8)
+set(OpenBLAS_PATCH_VERSION 9.dev)
 set(OpenBLAS_VERSION "${OpenBLAS_MAJOR_VERSION}.${OpenBLAS_MINOR_VERSION}.${OpenBLAS_PATCH_VERSION}")
 
 # Adhere to GNU filesystem layout conventions

Changelog.txt

@@ -1,4 +1,48 @@
 OpenBLAS ChangeLog
+====================================================================
+Version 0.3.9
+ 1-Mar-2020
+ 
+ common:
+ 	* Fixed a miscompilation of the GETRF functions with CMAKE
+	* Imported bugfix 390 from LAPACK (missing NaN propagation in xCOMBSSQ)
+	* The size of the memory buffer used for splitting GEMM tasks across
+          multiple threads can now be configured in the build system.
+
+POWER:
+	* Fixed several compilation problems related to endianness
+	  and ELF version on POWER8 and POWER9
+	* Fixed use of the absolute value IAMIN/IAMAX instead of IMIN/IMAX
+	* Fixed a race condition in the level3 blas code
+	
+MIPS64:
+	* Fixed use of the absoltute value IAMIN/IAMAX instead of IMIN/IMAX
+	
+ARMV7:
+	* Fixed a race condition in the level3 blas code
+	* Fixed compilation on Android
+ARMV8:
+	* Added support for Ampere EMAG8180
+	* Added support for Neoverse N1
+	* Improved performance of the blas_lock function
+	* Fixed a race condition in the level3 blas code
+	* Fixed a performance regression on TSV110-based servers
+
+x86_64:
+	* Fixed a long-standing error with undeclared register overwrites
+	  in the DSCAL microkernel for HASWELL,SKYLAKEX and ZEN
+	* Fixed a long-standing bug in the SSE implementation of IAMAX
+	* Fixed a CMAKE build failure with DYNAMIC_ARCH
+	* Fixed cpu autodetection of Goldmont+, Cannon Lake and Ice Lake
+	* Fixed a compilation failure on OSX with compiler name containing dash
+	* Fixed compilation with MinGW on SkylakeX
+	* Improved speed of the AVX512 GEMM3M kernel on SkylakeX
+	* Added an AVX512 STRMM kernel for SkylakeX
+	* Improved GEMM performance on Haswell and Zen
+
+zarch:
+	* fixed compilation of the DYNAMIC_ARCH code
+	
 ====================================================================
 Version 0.3.8
  9-Feb-2020

Makefile.arm

@@ -1,7 +1,7 @@
 ifeq ($(CORE), $(filter $(CORE),ARMV7 CORTEXA9 CORTEXA15))
 ifeq ($(OSNAME), Android)
-CCOMMON_OPT += -mfpu=neon
-FCOMMON_OPT += -mfpu=neon
+CCOMMON_OPT += -mfpu=neon -march=armv7-a
+FCOMMON_OPT += -mfpu=neon -march=armv7-a
 else
 CCOMMON_OPT += -mfpu=vfpv3 -march=armv7-a
 FCOMMON_OPT += -mfpu=vfpv3 -march=armv7-a

Makefile.arm64

@@ -24,6 +24,23 @@ CCOMMON_OPT += -march=armv8-a -mtune=cortex-a73
 FCOMMON_OPT += -march=armv8-a -mtune=cortex-a73
 endif
 
+# Use a72 tunings because Neoverse-N1 is only available
+# in GCC>=9
+ifeq ($(CORE), NEOVERSEN1)
+ifeq ($(GCCVERSIONGTEQ7), 1)
+ifeq ($(GCCVERSIONGTEQ9), 1)
+CCOMMON_OPT += -march=armv8.2-a -mtune=neoverse-n1
+FCOMMON_OPT += -march=armv8.2-a -mtune=neoverse-n1
+else
+CCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a72
+FCOMMON_OPT += -march=armv8.2-a -mtune=cortex-a72
+endif
+else
+CCOMMON_OPT += -march=armv8-a -mtune=cortex-a72
+FCOMMON_OPT += -march=armv8-a -mtune=cortex-a72
+endif
+endif
+
 ifeq ($(CORE), THUNDERX)
 CCOMMON_OPT += -march=armv8-a -mtune=thunderx
 FCOMMON_OPT += -march=armv8-a -mtune=thunderx

Makefile.install

@@ -82,7 +82,7 @@ ifeq ($(OSNAME), $(filter $(OSNAME),FreeBSD OpenBSD NetBSD DragonFly))
 endif
 ifeq ($(OSNAME), Darwin)
 	@-cp $(LIBDYNNAME) "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)"
-	@-install_name_tool -id "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)" "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)"
+	@-install_name_tool -id "$(OPENBLAS_LIBRARY_DIR)/$(LIBPREFIX).$(MAJOR_VERSION).dylib" "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)/$(LIBDYNNAME)"
 	@cd "$(DESTDIR)$(OPENBLAS_LIBRARY_DIR)" ; \
 	ln -fs $(LIBDYNNAME) $(LIBPREFIX).dylib ; \
 	ln -fs $(LIBDYNNAME) $(LIBPREFIX).$(MAJOR_VERSION).dylib

Makefile.prebuild

@@ -42,7 +42,7 @@ all: getarch_2nd
 	./getarch_2nd  1 >> $(TARGET_CONF)
 
 config.h : c_check f_check getarch
-	perl ./c_check $(TARGET_MAKE) $(TARGET_CONF) $(CC) $(TARGET_FLAGS)
+	perl ./c_check $(TARGET_MAKE) $(TARGET_CONF) $(CC) $(TARGET_FLAGS) $(CFLAGS)
 ifneq ($(ONLY_CBLAS), 1)
 	perl ./f_check $(TARGET_MAKE) $(TARGET_CONF) $(FC) $(TARGET_FLAGS)
 else
@@ -59,13 +59,13 @@ endif
 
 
 getarch : getarch.c cpuid.S dummy $(CPUIDEMU)
-	$(HOSTCC) $(CFLAGS) $(EXFLAGS) -o $(@F) getarch.c cpuid.S $(CPUIDEMU)
+	$(HOSTCC) $(HOST_CFLAGS) $(EXFLAGS) -o $(@F) getarch.c cpuid.S $(CPUIDEMU)
 
 getarch_2nd : getarch_2nd.c config.h dummy
 ifndef TARGET_CORE
-	$(HOSTCC) -I. $(CFLAGS) -o $(@F) getarch_2nd.c
+	$(HOSTCC) -I. $(HOST_CFLAGS) -o $(@F) getarch_2nd.c
 else
-	$(HOSTCC) -I. $(CFLAGS) -DBUILD_KERNEL -o $(@F) getarch_2nd.c
+	$(HOSTCC) -I. $(HOST_CFLAGS) -DBUILD_KERNEL -o $(@F) getarch_2nd.c
 endif
 
 dummy: