Merge pull request #5396 from abhishek-iitmadras/abhishekk_bfloat16

ARM64: Enable bfloat16 kernels by default

.github/workflows/apple_m.yml

@@ -87,10 +87,16 @@ jobs:
           echo "max_size = 300M" > ~/.ccache/ccache.conf
           echo "compression = true" >> ~/.ccache/ccache.conf
           ccache -s
+          
+      - name: Add gfortran runtime to link path
+        if: matrix.build == 'make' && runner.os == 'macOS'
+        run: |
+          GFORTRAN_LIBDIR=$(gfortran -print-file-name=libgfortran.dylib | xargs dirname)
+          echo "Using gfortran runtime in $GFORTRAN_LIBDIR"
+          echo "LDFLAGS=-L/opt/homebrew/opt/llvm/lib -L$GFORTRAN_LIBDIR" >> $GITHUB_ENV
 
       - name: Build OpenBLAS
         run: |
-          export LDFLAGS="-L/opt/homebrew/opt/llvm/lib"
           export CPPFLAGS="-I/opt/homebrew/opt/llvm/include"
           export CC="/opt/homebrew/opt/llvm/bin/clang"
           case "${{ matrix.build }}" in

.github/workflows/dynamic_arch.yml

@@ -89,6 +89,14 @@ jobs:
           echo "max_size = 300M" > ~/.ccache/ccache.conf
           echo "compression = true" >> ~/.ccache/ccache.conf
           ccache -s
+          
+      - name: Add gfortran runtime to link path
+        if: matrix.build == 'make' && runner.os == 'macOS'
+        run: |
+          GFORTRAN_LIBDIR=$(gfortran -print-file-name=libgfortran.dylib | xargs dirname)
+          echo "Using gfortran runtime in $GFORTRAN_LIBDIR"
+          # Preserve whatever LDFLAGS may already contain
+          echo "LDFLAGS=${LDFLAGS:+$LDFLAGS }-L$GFORTRAN_LIBDIR" >> "$GITHUB_ENV"
 
       - name: Build OpenBLAS
         run: |

CONTRIBUTORS.md

@@ -255,6 +255,7 @@ In chronological order:
 
 * Abhishek Kumar <https://github.com/abhishek-iitmadras>
   * [2025-04-22] Optimise dot kernel for NEOVERSE V1
+  * [2025-07-23] ARM64-Enable bfloat16 kernels by default 
 
 * Sharif Inamdar <sharif.inamdar@arm.com>
   * [2025-06-05] Optimize gemv_n_sve_v1x3 kernel

Makefile.system

@@ -270,6 +270,7 @@ SMALL_MATRIX_OPT = 1
 BUILD_BFLOAT16 = 1
 else ifeq ($(ARCH), arm64)
 SMALL_MATRIX_OPT = 1
+BUILD_BFLOAT16 = 1
 endif
 ifeq ($(ARCH), loongarch64)
 SMALL_MATRIX_OPT = 1
@@ -425,10 +426,8 @@ ifeq ($(OSNAME), Darwin)
 ifndef MACOSX_DEPLOYMENT_TARGET
 ifeq ($(ARCH), arm64)
 export MACOSX_DEPLOYMENT_TARGET=11.0
-ifeq ($(C_COMPILER), GCC)
 export NO_SVE = 1
 export NO_SME = 1
-endif
 else
 export MACOSX_DEPLOYMENT_TARGET=10.8
 endif

kernel/arm64/bgemm_kernel_4x4_neoversev1.c

@@ -27,6 +27,7 @@
  * *****************************************************************************/
 
 #include <arm_sve.h>
+#include <arm_neon.h>
 
 #include "common.h"
 

kernel/arm64/bgemm_kernel_4x4_neoversev1_impl.c

@@ -27,6 +27,7 @@
  * *****************************************************************************/
 
 #include <arm_sve.h>
+#include <arm_neon.h>
 
 #include "common.h"
 

kernel/arm64/bgemv_n_sve_v3x4.c

@@ -28,6 +28,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 
 #include <arm_sve.h>
+#include <arm_neon.h>
 
 #define UPDATE_PTRSx2 \
     a_ptr1 = a_ptr0 + lda;