[A64FX]: add tt for a64fx dot

kernel/arm64/KERNEL.A64FX

@@ -5,8 +5,8 @@ DGEMVNKERNEL = gemv_n_sve_v4x3.c
 SGEMVTKERNEL = gemv_t_sve_v4x3.c
 DGEMVTKERNEL = gemv_t_sve_v4x3.c
 
-DDOTKERNEL   = dot_sve_v8.c
-SDOTKERNEL   = dot_sve_v8.c
+DDOTKERNEL   = dot.c
+SDOTKERNEL   = dot.c
 
 SAXPYKERNEL  = axpy_sve.c
 DAXPYKERNEL  = axpy_sve.c

kernel/arm64/dot.c

@@ -42,6 +42,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #ifdef USE_SVE
 #ifdef DOT_KERNEL_SVE
 #include DOT_KERNEL_SVE
+#elif defined(A64FX)
+#include "dot_kernel_sve_v8.c"
 #else
 #include "dot_kernel_sve.c"
 #endif
@@ -82,14 +84,43 @@ static inline int get_dot_optimal_nthreads_neoversev1(BLASLONG N, int ncpu) {
 }
 #endif
 
+#if defined(DYNAMIC_ARCH) || defined(A64FX)
+static inline int get_dot_optimal_nthreads_a64fx(BLASLONG N, int ncpu) {
+  #ifdef DOUBLE
+  return (N <= 11000L) ? 1 
+      : (N <= 20000L)  ? MIN(ncpu, 2) 
+      : (N <= 35000L)  ? MIN(ncpu, 4) 
+      : (N <= 50000L)  ? MIN(ncpu, 6) 
+      : (N <= 440000L)  ? MIN(ncpu, 8) 
+      : (N <= 880000L)  ? MIN(ncpu, 16) 
+      : (N <= 1020000L) ? MIN(ncpu, 24) 
+      : ncpu;
+  #else
+  return (N <= 22000L) ? 1 
+      : (N <= 39000L)  ? MIN(ncpu, 2) 
+      : (N <= 79000L)  ? MIN(ncpu, 4) 
+      : (N <= 120000L)  ? MIN(ncpu, 6) 
+      : (N <= 1020000L)  ? MIN(ncpu, 8) 
+      : ncpu;
+  #endif
+}
+#endif
+
 static inline int get_dot_optimal_nthreads(BLASLONG n) {
   int ncpu = num_cpu_avail(1);
 
-#if defined(NEOVERSEV1) && !defined(COMPLEX) && !defined(BFLOAT16)
+#if defined(A64FX) && !defined(COMPLEX) && !defined(BFLOAT16)
+  return get_dot_optimal_nthreads_a64fx(n, ncpu);
+#elif defined(NEOVERSEV1) && !defined(COMPLEX) && !defined(BFLOAT16)
   return get_dot_optimal_nthreads_neoversev1(n, ncpu);
 #elif defined(DYNAMIC_ARCH) && !defined(COMPLEX) && !defined(BFLOAT16)
-  if (strcmp(gotoblas_corename(), "neoversev1") == 0) {
-    return get_dot_optimal_nthreads_neoversev1(n, ncpu);
+  {
+    const char *core = gotoblas_corename();
+    if (strcmp(core, "a64fx") == 0) {
+      return get_dot_optimal_nthreads_a64fx(n, ncpu);
+    } else if (strcmp(core, "neoversev1") == 0) {
+      return get_dot_optimal_nthreads_neoversev1(n, ncpu);
+    }
   }
 #endif