diff --git a/Makefile.system b/Makefile.system
index cc72c02e8..db651ef99 100644
--- a/Makefile.system
+++ b/Makefile.system
@@ -617,7 +617,6 @@ DYNAMIC_CORE += POWER8
 ifneq ($(C_COMPILER), GCC)
 DYNAMIC_CORE += POWER9
 DYNAMIC_CORE += POWER10
-override LDFLAGS += -Wl,-no-power10-stubs
 endif
 ifeq ($(C_COMPILER), GCC)
 ifeq ($(GCCVERSIONGT5), 1)
@@ -627,11 +626,9 @@ $(info, OpenBLAS: Your gcc version is too old to build the POWER9 kernels.)
 endif
 ifeq ($(GCCVERSIONGTEQ11), 1)
 DYNAMIC_CORE += POWER10
-override LDFLAGS += -Wl,-no-power10-stubs
 else ifeq ($(GCCVERSIONGTEQ10), 1)
 ifeq ($(GCCMINORVERSIONGTEQ2), 1)
 DYNAMIC_CORE += POWER10
-override LDFLAGS += -Wl,-no-power10-stubs
 endif
 else
 $(info, OpenBLAS: Your gcc version is too old to build the POWER10 kernels.)
diff --git a/kernel/power/dgemm_kernel_power10.c b/kernel/power/dgemm_kernel_power10.c
index b3ee301be..a0bc1a777 100644
--- a/kernel/power/dgemm_kernel_power10.c
+++ b/kernel/power/dgemm_kernel_power10.c
@@ -27,64 +27,64 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 #include <altivec.h>
 
-typedef unsigned char vec_t __attribute__ ((vector_size (16)));
+typedef __vector unsigned char  vec_t;
 typedef FLOAT v4sf_t __attribute__ ((vector_size (16)));
 typedef FLOAT v2sf_t __attribute__ ((vector_size (8)));
 
 #ifdef TRMMKERNEL
 #define SAVE_ACC(ACC, J)  \
-          __builtin_mma_disassemble_acc (result, ACC); \
+          __builtin_mma_disassemble_acc ((void *)result, ACC); \
           rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
+          rowC[0] = result[0] * alpha; \
           rowC = (v4sf_t *) &CO[1*ldc+J]; \
-          rowC[0] = result[2] * alpha; \
+          rowC[0] = result[1] * alpha; \
           rowC = (v4sf_t *) &CO[2*ldc+J]; \
-          rowC[0] = result[1] * alpha; \
-          rowC = (v4sf_t *) &CO[3*ldc+J]; \
-          rowC[0] = result[0] * alpha;
-#define SAVE_ACC1(ACC, J)  \
-          __builtin_mma_disassemble_acc (result, ACC); \
-          rowC = (v4sf_t *) &CO[4* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
-          rowC = (v4sf_t *) &CO[5*ldc+J]; \
           rowC[0] = result[2] * alpha; \
-          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC = (v4sf_t *) &CO[3*ldc+J]; \
+          rowC[0] = result[3] * alpha;
+#define SAVE_ACC1(ACC, J)  \
+          __builtin_mma_disassemble_acc ((void *)result, ACC); \
+          rowC = (v4sf_t *) &CO[4* ldc+J]; \
+          rowC[0] = result[0] * alpha; \
+          rowC = (v4sf_t *) &CO[5*ldc+J]; \
           rowC[0] = result[1] * alpha; \
+          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC[0] = result[2] * alpha; \
           rowC = (v4sf_t *) &CO[7*ldc+J]; \
-          rowC[0] = result[0] * alpha;
+          rowC[0] = result[3] * alpha;
 #define  SAVE2x4_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
+          rowC[0] = result[0] * alpha; \
 	  rowC = (v4sf_t *) &CO[1* ldc+J]; \
-          rowC[0] = result[2] * alpha;
+          rowC[0] = result[1] * alpha;
 #else
 #define SAVE_ACC(ACC, J)  \
-          __builtin_mma_disassemble_acc (result, ACC); \
+          __builtin_mma_disassemble_acc ((void *)result, ACC); \
           rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
           rowC = (v4sf_t *) &CO[1*ldc+J]; \
-          rowC[0] += result[2] * alpha; \
+          rowC[0] += result[1] * alpha; \
           rowC = (v4sf_t *) &CO[2*ldc+J]; \
-          rowC[0] += result[1] * alpha; \
-          rowC = (v4sf_t *) &CO[3*ldc+J]; \
-          rowC[0] += result[0] * alpha;
-#define SAVE_ACC1(ACC, J)  \
-          __builtin_mma_disassemble_acc (result, ACC); \
-          rowC = (v4sf_t *) &CO[4* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
-          rowC = (v4sf_t *) &CO[5*ldc+J]; \
           rowC[0] += result[2] * alpha; \
-          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC = (v4sf_t *) &CO[3*ldc+J]; \
+          rowC[0] += result[3] * alpha;
+#define SAVE_ACC1(ACC, J)  \
+          __builtin_mma_disassemble_acc ((void *)result, ACC); \
+          rowC = (v4sf_t *) &CO[4* ldc+J]; \
+          rowC[0] += result[0] * alpha; \
+          rowC = (v4sf_t *) &CO[5*ldc+J]; \
           rowC[0] += result[1] * alpha; \
+          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC[0] += result[2] * alpha; \
           rowC = (v4sf_t *) &CO[7*ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[3] * alpha;
 #define  SAVE2x4_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
 	  rowC = (v4sf_t *) &CO[1* ldc+J]; \
-          rowC[0] += result[2] * alpha;
+          rowC[0] += result[1] * alpha;
 #endif
 
 #define SET_ACC_ZERO4() \
diff --git a/kernel/power/sgemm_kernel_power10.c b/kernel/power/sgemm_kernel_power10.c
index 01c122c6d..81a5ec76b 100644
--- a/kernel/power/sgemm_kernel_power10.c
+++ b/kernel/power/sgemm_kernel_power10.c
@@ -27,103 +27,103 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include "common.h"
 #include <altivec.h>
 
-typedef unsigned char vec_t __attribute__ ((vector_size (16)));
+typedef __vector unsigned char  vec_t;
 typedef FLOAT v4sf_t __attribute__ ((vector_size (16)));
 typedef FLOAT v2sf_t __attribute__ ((vector_size (8)));
 #if defined(TRMMKERNEL)
 #define SAVE_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
+          rowC[0] = result[0] * alpha; \
           rowC = (v4sf_t *) &CO[1*ldc+J]; \
-          rowC[0] = result[2] * alpha; \
+          rowC[0] = result[1] * alpha; \
           rowC = (v4sf_t *) &CO[2*ldc+J]; \
-          rowC[0] = result[1] * alpha; \
+          rowC[0] = result[2] * alpha; \
           rowC = (v4sf_t *) &CO[3*ldc+J]; \
-          rowC[0] = result[0] * alpha;
+          rowC[0] = result[3] * alpha;
 #define SAVE_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[4* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
+          rowC[0] = result[0] * alpha; \
           rowC = (v4sf_t *) &CO[5*ldc+J]; \
-          rowC[0] = result[2] * alpha; \
-          rowC = (v4sf_t *) &CO[6*ldc+J]; \
           rowC[0] = result[1] * alpha; \
+          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC[0] = result[2] * alpha; \
           rowC = (v4sf_t *) &CO[7*ldc+J]; \
-          rowC[0] = result[0] * alpha;
+          rowC[0] = result[3] * alpha;
 #define  SAVE4x2_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v2sf_t *) &CO[0* ldc+J]; \
-          rowC[0] = result[6] * alpha; \
+          rowC[0] = result[0] * alpha; \
 	  rowC = (v2sf_t *) &CO[1* ldc+J]; \
-          rowC[0] = result[4] * alpha; \
+          rowC[0] = result[2] * alpha; \
 	  rowC = (v2sf_t *) &CO[2* ldc+J]; \
-          rowC[0] = result[2] * alpha; \
-	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
-          rowC[0] = result[0] * alpha;
-#define  SAVE4x2_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
-	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
-          rowC[0] = result[6] * alpha; \
-	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] = result[4] * alpha; \
-	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
+          rowC[0] = result[6] * alpha;
+#define  SAVE4x2_ACC1(ACC, J)  \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
+	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
+          rowC[0] = result[0] * alpha; \
+	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] = result[2] * alpha; \
+	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+          rowC[0] = result[4] * alpha; \
 	  rowC = (v2sf_t *) &CO[7* ldc+J]; \
-          rowC[0] = result[0] * alpha;
+          rowC[0] = result[6] * alpha;
 #define  SAVE2x4_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] = result[3] * alpha; \
+          rowC[0] = result[0] * alpha; \
 	  rowC = (v4sf_t *) &CO[1* ldc+J]; \
-          rowC[0] = result[2] * alpha;
+          rowC[0] = result[1] * alpha;
 #else
 #define SAVE_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
           rowC = (v4sf_t *) &CO[1*ldc+J]; \
-          rowC[0] += result[2] * alpha; \
+          rowC[0] += result[1] * alpha; \
           rowC = (v4sf_t *) &CO[2*ldc+J]; \
-          rowC[0] += result[1] * alpha; \
+          rowC[0] += result[2] * alpha; \
           rowC = (v4sf_t *) &CO[3*ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[3] * alpha;
 #define SAVE_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[4* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
           rowC = (v4sf_t *) &CO[5*ldc+J]; \
-          rowC[0] += result[2] * alpha; \
-          rowC = (v4sf_t *) &CO[6*ldc+J]; \
           rowC[0] += result[1] * alpha; \
+          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC[0] += result[2] * alpha; \
           rowC = (v4sf_t *) &CO[7*ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[3] * alpha;
 #define  SAVE4x2_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v2sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[6] * alpha; \
+          rowC[0] += result[0] * alpha; \
 	  rowC = (v2sf_t *) &CO[1* ldc+J]; \
-          rowC[0] += result[4] * alpha; \
+          rowC[0] += result[2] * alpha; \
 	  rowC = (v2sf_t *) &CO[2* ldc+J]; \
-          rowC[0] += result[2] * alpha; \
-	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
-          rowC[0] += result[0] * alpha;
-#define  SAVE4x2_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
-	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
-          rowC[0] += result[6] * alpha; \
-	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] += result[4] * alpha; \
-	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
+          rowC[0] += result[6] * alpha;
+#define  SAVE4x2_ACC1(ACC, J)  \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
+	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
+          rowC[0] += result[0] * alpha; \
+	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] += result[2] * alpha; \
+	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+          rowC[0] += result[4] * alpha; \
 	  rowC = (v2sf_t *) &CO[7* ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[6] * alpha;
 #define  SAVE2x4_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
 	  rowC = (v4sf_t *) &CO[1* ldc+J]; \
-          rowC[0] += result[2] * alpha;
+          rowC[0] += result[1] * alpha;
 #endif
 #define KERNEL(i, j) \
           __builtin_mma_xvf32gerpp (&acc0, rowB[i], rowA[j]); \
diff --git a/kernel/power/shgemm_kernel_power10.c b/kernel/power/shgemm_kernel_power10.c
index 7455f925c..1ae9e04bf 100644
--- a/kernel/power/shgemm_kernel_power10.c
+++ b/kernel/power/shgemm_kernel_power10.c
@@ -45,7 +45,7 @@ bfloat16tof32 (bfloat16 f16)
 #define BF16TOF32(x) x
 #endif
 
-typedef unsigned char vec_t __attribute__ ((vector_size (16)));
+typedef __vector unsigned char  vec_t;
 typedef FLOAT v4sf_t __attribute__ ((vector_size (16)));
 typedef FLOAT v2sf_t __attribute__ ((vector_size (8)));
 
@@ -64,54 +64,54 @@ vector char mask =
 #define MERGE_LOW(x, y) (vec_t) vec_mergel ((vector short)x, (vector short)y)
 
 #define SAVE_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
           rowC = (v4sf_t *) &CO[1*ldc+J]; \
-          rowC[0] += result[2] * alpha; \
+          rowC[0] += result[1] * alpha; \
           rowC = (v4sf_t *) &CO[2*ldc+J]; \
-          rowC[0] += result[1] * alpha; \
+          rowC[0] += result[2] * alpha; \
           rowC = (v4sf_t *) &CO[3*ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[3] * alpha;
 #define SAVE_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[4* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
           rowC = (v4sf_t *) &CO[5*ldc+J]; \
-          rowC[0] += result[2] * alpha; \
-          rowC = (v4sf_t *) &CO[6*ldc+J]; \
           rowC[0] += result[1] * alpha; \
+          rowC = (v4sf_t *) &CO[6*ldc+J]; \
+          rowC[0] += result[2] * alpha; \
           rowC = (v4sf_t *) &CO[7*ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[3] * alpha;
 #define  SAVE4x2_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v2sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[6] * alpha; \
+          rowC[0] += result[0] * alpha; \
 	  rowC = (v2sf_t *) &CO[1* ldc+J]; \
-          rowC[0] += result[4] * alpha; \
+          rowC[0] += result[2] * alpha; \
 	  rowC = (v2sf_t *) &CO[2* ldc+J]; \
-          rowC[0] += result[2] * alpha; \
-	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
-          rowC[0] += result[0] * alpha;
-#define  SAVE4x2_ACC1(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
-	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
-          rowC[0] += result[6] * alpha; \
-	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] += result[4] * alpha; \
-	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+	  rowC = (v2sf_t *) &CO[3* ldc+J]; \
+          rowC[0] += result[6] * alpha;
+#define  SAVE4x2_ACC1(ACC, J)  \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
+	  rowC = (v2sf_t *) &CO[4* ldc+J]; \
+          rowC[0] += result[0] * alpha; \
+	  rowC = (v2sf_t *) &CO[5* ldc+J]; \
           rowC[0] += result[2] * alpha; \
+	  rowC = (v2sf_t *) &CO[6* ldc+J]; \
+          rowC[0] += result[4] * alpha; \
 	  rowC = (v2sf_t *) &CO[7* ldc+J]; \
-          rowC[0] += result[0] * alpha;
+          rowC[0] += result[6] * alpha;
 
 #define MMA __builtin_mma_xvbf16ger2pp
 
 #define  SAVE2x4_ACC(ACC, J)  \
-	  __builtin_mma_disassemble_acc (result, ACC); \
+	  __builtin_mma_disassemble_acc ((void *)result, ACC); \
 	  rowC = (v4sf_t *) &CO[0* ldc+J]; \
-          rowC[0] += result[3] * alpha; \
+          rowC[0] += result[0] * alpha; \
 	  rowC = (v4sf_t *) &CO[1* ldc+J]; \
-          rowC[0] += result[2] * alpha;
+          rowC[0] += result[1] * alpha;
 
 #define SET_ACC_ZERO4() \
 	  __builtin_mma_xxsetaccz (&acc0); \