POWER: AIX: Make use of power10 optimization

POWER10 optimizations are disabled when using default AIX assembler.
As we have fixed many issues recently, enabling optimization path
for default assembler.

Author: Rajalakshmi Srinivasaraghavan

Makefile.system

@@ -277,10 +277,6 @@ endif
 ifndef GOTOBLAS_MAKEFILE
 export GOTOBLAS_MAKEFILE = 1
 
-# Determine if the assembler is GNU Assembler
-HAVE_GAS := $(shell $(AS) -v < /dev/null 2>&1 | grep GNU 2>&1 >/dev/null ; echo $$?)
-GETARCH_FLAGS += -DHAVE_GAS=$(HAVE_GAS)
-
 # Generating Makefile.conf and config.h
 DUMMY := $(shell $(MAKE) -C $(TOPDIR) -f Makefile.prebuild CC="$(CC)" FC="$(FC)" HOSTCC="$(HOSTCC)" HOST_CFLAGS="$(GETARCH_FLAGS)" CFLAGS="$(CFLAGS)" BINARY=$(BINARY) USE_OPENMP=$(USE_OPENMP) DYNAMIC_ARCH=$(DYNAMIC_ARCH) TARGET_CORE=$(TARGET_CORE) ONLY_CBLAS=$(ONLY_CBLAS) TARGET=$(TARGET) all)
 
@@ -760,7 +756,11 @@ DYNAMIC_CORE += POWER9
 else
 $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
 endif
+ifeq ($(OSNAME), AIX)
+LDVERSIONGTEQ35 := 1
+else
 LDVERSIONGTEQ35 := $(shell expr `$(CC) -Wl,--version 2> /dev/null | head -1 | cut -f2 -d "." | cut -f1 -d "-"` \>= 35)
+endif
 ifeq ($(GCCVERSIONGTEQ11)$(LDVERSIONGTEQ35), 11)
 DYNAMIC_CORE += POWER10
 CCOMMON_OPT += -DHAVE_P10_SUPPORT

kernel/power/KERNEL.POWER10

@@ -1,6 +1,3 @@
-ifeq ($(HAVE_GAS), 1)
-include $(KERNELDIR)/KERNEL.POWER8
-else
 #SGEMM_BETA = ../generic/gemm_beta.c
 #DGEMM_BETA = ../generic/gemm_beta.c
 #CGEMM_BETA = ../generic/zgemm_beta.c
@@ -265,4 +262,3 @@ QCABS_KERNEL	= ../generic/cabs.c
 #Dump kernel
 CGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
 ZGEMM3MKERNEL    = ../generic/zgemm3mkernel_dump.c
-endif

param.h

@@ -2600,13 +2600,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 #define SGEMM_DEFAULT_UNROLL_M 16
 #define SGEMM_DEFAULT_UNROLL_N 8
-#if defined(HAVE_GAS) && (HAVE_GAS == 1) 
-#define DGEMM_DEFAULT_UNROLL_M 16
-#define DGEMM_DEFAULT_UNROLL_N 4
-#else
 #define DGEMM_DEFAULT_UNROLL_M 8
 #define DGEMM_DEFAULT_UNROLL_N 8
-#endif
 #define CGEMM_DEFAULT_UNROLL_M 8
 #define CGEMM_DEFAULT_UNROLL_N 4
 #define ZGEMM_DEFAULT_UNROLL_M 8