Rebase with AOCL5.1

.appveyor.yml

@@ -23,6 +23,12 @@ environment:
       CC: clang
       THREADING: openmp
 
+    - LIB_TYPE: static
+      CONFIG: auto
+      CC: clang
+      THREADING: openmp
+      SANDBOX: yes
+
 install:
 - set "PATH=C:\msys64\mingw64\bin;C:\msys64\bin;%PATH%"
 - if [%CC%]==[clang] set "PATH=C:\Program Files\LLVM\bin;%PATH%"
@@ -34,6 +40,7 @@ build_script:
 - if [%LIB_TYPE%]==[shared] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --enable-shared --disable-static"
 - if [%LIB_TYPE%]==[static] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --disable-shared --enable-static"
 - if not [%CBLAS%]==[no] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% --enable-cblas"
+- if [%SANDBOX%]==[yes] set "CONFIGURE_OPTS=%CONFIGURE_OPTS% -s gemmlike"
 - set RANLIB=echo
 - set LIBPTHREAD=
 - set "PATH=%PATH%;C:\blis\lib"

.circleci/config.yml

@@ -0,0 +1,264 @@
+version: 2.1
+
+branches:
+  only:
+    - master
+    - dev
+    - amd
+
+executors:
+  linux: # Docker using the Base Convenience Image
+    docker:
+      - image: cimg/base:2024.10
+  linuxnew: # Docker using the Base Convenience Image
+    docker:
+      - image: cimg/base:current-22.04
+  macos: &macos-executor # macos executor running Xcode
+    macos:
+      xcode: 14.2.0
+  linuxvm: # executor type
+    machine:
+      image: ubuntu-2204:current
+
+workflows:
+  build:
+    jobs:
+      # Default:
+      # - build:
+      #     os: linux
+      #     CC: gcc
+      #     OOT: 0
+      #     TEST: FAST
+      #     SDE: 0
+      #     LEVEL0: 0
+      #     THR: none
+      #     CONF: auto
+      #     BLD: ''
+      #     LDFLAGS: ''
+      #     TESTSUITE_WRAPPER: ''
+      #     PACKAGES: ''
+
+      # full testsuite (all tests + mixed datatype (gemm_nn only) + salt + OOT)
+      - build:
+          OOT: 1
+          TEST: ALL
+          THR: openmp,pthreads
+          CONF: x86_64
+
+      # SDE testing for x86_64
+      # Also test LEVEL0 here because g++ uses tons of memory for test_taxpbys.cxx
+      - build:
+          # linuxvm must be used because it provides 8G RAM and SDE fails with 4G RAM
+          os: linuxvm
+          SDE: 1
+          LEVEL0: 1
+          CONF: x86_64
+
+      # test generic kernels
+      - build:
+          CONF: generic_broadcast
+
+      # clang build
+      - build:
+          CC: clang
+          THR: openmp,pthreads
+          CXX: clang++
+          PACKAGES: clang libomp-dev
+
+      # macOS with system compiler (clang)
+      - build:
+          os: macos
+          THR: pthreads
+          CC: clang
+          CXX: clang++
+
+      # cortexa15 build and fast testsuite (qemu)
+      - build:
+          CC: arm-linux-gnueabihf-gcc
+          CXX: arm-linux-gnueabihf-g++
+          CONF: cortexa15
+          PACKAGES: 'gcc-arm-linux-gnueabihf g++-arm-linux-gnueabihf libc6-dev-armhf-cross qemu-system-arm qemu-user'
+          TESTSUITE_WRAPPER: 'qemu-arm -cpu cortex-a15 -L /usr/arm-linux-gnueabihf/'
+
+      # cortexa57 build and fast testsuite (qemu)
+      - build:
+          CC: aarch64-linux-gnu-gcc
+          CXX: aarch64-linux-gnu-g++
+          CONF: cortexa57
+          PACKAGES: 'gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user'
+          TESTSUITE_WRAPPER: 'qemu-aarch64 -L /usr/aarch64-linux-gnu/'
+
+      # Apple M1 (firestorm) build and fast testsuite (qemu)
+      - build:
+          CC: aarch64-linux-gnu-gcc
+          CXX: aarch64-linux-gnu-g++
+          CONF: firestorm
+          PACKAGES: 'gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user'
+          TESTSUITE_WRAPPER: 'qemu-aarch64 -L /usr/aarch64-linux-gnu/'
+
+      # armsve build and fast testsuite (qemu)
+      - build:
+          CC: aarch64-linux-gnu-gcc-10
+          CXX: aarch64-linux-gnu-g++-10
+          CONF: armsve
+          PACKAGES: 'gcc-10-aarch64-linux-gnu g++-10-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user'
+          TESTSUITE_WRAPPER: 'qemu-aarch64 -cpu max,sve=true,sve512=true -L /usr/aarch64-linux-gnu/'
+
+      # arm64 build and fast testsuite (qemu)
+      # NOTE: This entry omits the -cpu flag so that while both NEON and SVE kernels
+      # are compiled, only NEON kernels will be tested. (h/t to RuQing Xu)
+      - build:
+          CC: aarch64-linux-gnu-gcc-10
+          CXX: aarch64-linux-gnu-g++-10
+          CONF: arm64
+          PACKAGES: 'gcc-10-aarch64-linux-gnu g++-10-aarch64-linux-gnu libc6-dev-arm64-cross qemu-system-arm qemu-user'
+          TESTSUITE_WRAPPER: 'qemu-aarch64 -L /usr/aarch64-linux-gnu/'
+
+      # The RISC-V targets require the qemu version available in jammy or newer.
+      # When CI is upgraded, the packages should be activated and do_script.sh
+      # cleaned up.
+      # PACKAGES="qemu-user qemu-user-binfmt"
+      - build:
+          CONF: rv64iv
+          BLD: --disable-shared
+          LDFLAGS: -static
+      - build:
+          CONF: rv32iv
+          BLD: --disable-shared
+          LDFLAGS: -static
+      - build:
+          CONF: sifive_x280
+          BLD: --disable-shared
+          LDFLAGS: -static
+
+jobs:
+  build:
+    parameters:
+      os:
+        type: executor
+        default: linux
+      CC:
+        type: string
+        default: gcc
+      CXX:
+        type: string
+        default: g++
+      OOT:
+        type: integer
+        default: 0
+      TEST:
+        type: string
+        default: FAST
+      SDE:
+        type: integer
+        default: 0
+      LEVEL0:
+        type: integer
+        default: 0
+      THR:
+        type: string
+        default: none
+      CONF:
+        type: string
+        default: auto
+      BLD:
+        type: string
+        default: ''
+      LDFLAGS:
+        type: string
+        default: ''
+      TESTSUITE_WRAPPER:
+        type: string
+        default: ''
+      PACKAGES:
+        type: string
+        default: ''
+    executor: << parameters.os >>
+    steps:
+      - checkout
+
+      - when:
+          condition:
+            not:
+              equal: [ *macos-executor, << parameters.os >> ]
+          steps:
+            - run:
+                name: Installing Dependencies
+                command:
+                  sudo apt-get update && sudo NEEDRESTART_MODE=a apt-get install -y make python3 << parameters.PACKAGES >>
+
+      - run:
+          name: Configuring, Building, Testing
+          command: |
+            export DIST_PATH=.
+            export CC="<< parameters.CC >>"
+            export CXX="<< parameters.CXX >>"
+            export OOT="<< parameters.OOT >>"
+            export CONF="<< parameters.CONF >>"
+            export TEST="<< parameters.TEST >>"
+            export BLD="<< parameters.BLD >>"
+            export LDFLAGS="<< parameters.LDFLAGS >>"
+            export SDE="<< parameters.SDE >>"
+            export LEVEL0="<< parameters.LEVEL0 >>"
+            export THR="<< parameters.THR >>"
+            export TESTSUITE_WRAPPER="<< parameters.TESTSUITE_WRAPPER >>"
+
+            pwd
+            if [ $OOT -eq 1 ]; then export DIST_PATH=`pwd`; mkdir ../oot; cd ../oot; chmod -R a-w $DIST_PATH; fi
+            pwd
+
+            if [ "$CONF" = "rv64iv" ]; then
+              $DIST_PATH/ci/do_riscv.sh "$CONF";
+              export CC=$DIST_PATH/../toolchain/riscv/bin/riscv64-unknown-linux-gnu-gcc;
+              export CXX=$DIST_PATH/../toolchain/riscv/bin/riscv64-unknown-linux-gnu-g++;
+              export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv64 -cpu rv64,vext_spec=v1.0,v=true,vlen=128 -B 0x100000";
+            fi
+            if [ "$CONF" = "rv32iv" ]; then
+              $DIST_PATH/ci/do_riscv.sh "$CONF";
+              export CC=$DIST_PATH/../toolchain/riscv/bin/riscv32-unknown-linux-gnu-gcc;
+              export CXX=$DIST_PATH/../toolchain/riscv/bin/riscv32-unknown-linux-gnu-g++;
+              export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv32 -cpu rv32,vext_spec=v1.0,v=true,vlen=128 -B 0x100000";
+            fi
+            if [ "$CONF" = "sifive_x280" ]; then
+              $DIST_PATH/ci/do_riscv.sh "$CONF";
+              export CC=$DIST_PATH/../toolchain/riscv/bin/clang;
+              export CXX=$DIST_PATH/../toolchain/riscv/bin/clang++;
+              export TESTSUITE_WRAPPER="$DIST_PATH/../toolchain/qemu-riscv64 -cpu rv64,vext_spec=v1.0,v=true,vlen=512 -B 0x100000";
+            fi
+
+            if [ "$CONF" = "generic_broadcast" ]; then
+              export CONF=generic
+              export CFLAGS="-DBLIS_BBM_s=2 -DBLIS_BBM_d=2 -DBLIS_BBM_c=2 -DBLIS_BBM_z=2 -DBLIS_BBN_s=4 -DBLIS_BBN_d=4 -DBLIS_BBN_c=4 -DBLIS_BBN_z=4"
+            fi
+
+            echo "Configuration:"
+            echo "CC                = $CC"
+            echo "CXX               = $CXX"
+            echo "OOT               = $OOT"
+            echo "CONF              = $CONF"
+            echo "THR               = $THR"
+            echo "TEST              = $TEST"
+            echo "BLD               = $BLD"
+            echo "SDE               = $SDE"
+            echo "LEVEL0            = $LEVEL0"
+            echo "DIST_PATH         = $DIST_PATH"
+            echo "CFLAGS            = $CFLAGS"
+            echo "LDFLAGS           = $LDFLAGS"
+            echo "TESTSUITE_WRAPPER = $TESTSUITE_WRAPPER"
+
+            $DIST_PATH/configure -p `pwd`/../install -t $THR $BLD CC=$CC $CONF
+            pwd
+            ls -l
+            $CC --version
+            $CC -v
+
+            make V=1 -j2
+            make install
+
+            if [ "$BLD" = "" ] && [ "$TESTSUITE_WRAPPER" = "" ] ; then $DIST_PATH/ci/cxx/cxx-test.sh $DIST_PATH $(ls -1 include); fi
+            # Qemu SVE is failing sgemmt in some cases. Skip as this issue is not observed
+            # on real chip (A64fx).
+            if [ "$CONF" = "armsve" ]; then sed -i 's/.*\<gemmt\>.*/0/' $DIST_PATH/testsuite/input.operations.fast; fi
+            if [ "$TEST" != "0" ]; then $DIST_PATH/ci/do_testsuite.sh; fi
+            if [ "$SDE" = "1" ]; then $DIST_PATH/ci/do_sde.sh; fi
+            if [ "$LEVEL0" = "1" ]; then $DIST_PATH/ci/do_level0.sh; fi

.dir-locals.el

@@ -1,9 +1,32 @@
-;; First (minimal) attempt at configuring Emacs CC mode for the BLIS
-;; layout requirements.
-((c-mode . ((c-file-style . "stroustrup")
+;; Emacs formatting for the BLIS layout requirements.
+
+(
+ ;; Recognize *.mk files as Makefile fragments
+ (auto-mode-alist . (("\\.mk\\'" . makefile-mode)) )
+
+ ;; Makefiles require tabs and are almost always width 8
+ (makefile-mode . (
+                   (indent-tabs-mode . t)
+                   (tab-width . 8)
+                   )
+                )
+
+ ;; C code formatting roughly according to docs/CodingConventions.md
+ (c-mode . (
+            (c-file-style . "bsd")
             (c-basic-offset . 4)
             (comment-start . "// ")
             (comment-end . "")
-            (indent-tabs-mode . t)
-            (tab-width . 4)
-            (parens-require-spaces . nil))))
+            (parens-require-spaces . nil)
+            )
+         )
+
+ ;; Default formatting for all source files not overriden above
+ (prog-mode . (
+               (indent-tabs-mode . nil)
+               (tab-width . 4)
+               (require-final-newline . t)
+               (eval add-hook `before-save-hook `delete-trailing-whitespace)
+               )
+            )
+)

.gitignore

@@ -31,6 +31,7 @@
 
 config.mk
 bli_config.h
+bli_addon.h
 
 # -- monolithic headers --
 
@@ -43,6 +44,7 @@ include/*/*.h
 # -- misc. --
 
 # BLIS testsuite output file
+output.testsuite
 output.testsuite.*
 
 # BLAS test output files
@@ -52,3 +54,6 @@ out.*
 GPATH
 GRTAGS
 GTAGS
+
+# Mac DS.store files
+.DS_Store