Optimise scopy for POWER9

Use lxvd2x instruction instead of lxvw4x.
lxvd2x performs far better on the new POWER architecture than lxvw4x.

kernel/power/scopy_microk_power8.c

@@ -39,14 +39,14 @@ static void scopy_kernel_32 (long n, float *x, float *y)
 {
   __asm__
     (
-       "lxvw4x		40, 0, %2	\n\t"
-       "lxvw4x		41, %5, %2	\n\t"
-       "lxvw4x		42, %6, %2	\n\t"
-       "lxvw4x		43, %7, %2	\n\t"
-       "lxvw4x		44, %8, %2	\n\t"
-       "lxvw4x		45, %9, %2	\n\t"
-       "lxvw4x		46, %10, %2	\n\t"
-       "lxvw4x		47, %11, %2	\n\t"
+       "lxvd2x		40, 0, %2	\n\t"
+       "lxvd2x		41, %5, %2	\n\t"
+       "lxvd2x		42, %6, %2	\n\t"
+       "lxvd2x		43, %7, %2	\n\t"
+       "lxvd2x		44, %8, %2	\n\t"
+       "lxvd2x		45, %9, %2	\n\t"
+       "lxvd2x		46, %10, %2	\n\t"
+       "lxvd2x		47, %11, %2	\n\t"
 
        "addi		%2, %2, 128	\n\t"
 
@@ -56,22 +56,22 @@ static void scopy_kernel_32 (long n, float *x, float *y)
        ".p2align	5		\n"
      "1:				\n\t"
 
-       "stxvw4x		40, 0, %3	\n\t"
-       "stxvw4x		41, %5, %3	\n\t"
-       "lxvw4x		40, 0, %2	\n\t"
-       "lxvw4x		41, %5, %2	\n\t"
-       "stxvw4x		42, %6, %3	\n\t"
-       "stxvw4x		43, %7, %3	\n\t"
-       "lxvw4x		42, %6, %2	\n\t"
-       "lxvw4x		43, %7, %2	\n\t"
-       "stxvw4x		44, %8, %3	\n\t"
-       "stxvw4x		45, %9, %3	\n\t"
-       "lxvw4x		44, %8, %2	\n\t"
-       "lxvw4x		45, %9, %2	\n\t"
-       "stxvw4x		46, %10, %3	\n\t"
-       "stxvw4x		47, %11, %3	\n\t"
-       "lxvw4x		46, %10, %2	\n\t"
-       "lxvw4x		47, %11, %2	\n\t"
+       "stxvd2x		40, 0, %3	\n\t"
+       "stxvd2x		41, %5, %3	\n\t"
+       "lxvd2x		40, 0, %2	\n\t"
+       "lxvd2x		41, %5, %2	\n\t"
+       "stxvd2x		42, %6, %3	\n\t"
+       "stxvd2x		43, %7, %3	\n\t"
+       "lxvd2x		42, %6, %2	\n\t"
+       "lxvd2x		43, %7, %2	\n\t"
+       "stxvd2x		44, %8, %3	\n\t"
+       "stxvd2x		45, %9, %3	\n\t"
+       "lxvd2x		44, %8, %2	\n\t"
+       "lxvd2x		45, %9, %2	\n\t"
+       "stxvd2x		46, %10, %3	\n\t"
+       "stxvd2x		47, %11, %3	\n\t"
+       "lxvd2x		46, %10, %2	\n\t"
+       "lxvd2x		47, %11, %2	\n\t"
 
        "addi		%3, %3, 128	\n\t"
        "addi		%2, %2, 128	\n\t"
@@ -81,14 +81,14 @@ static void scopy_kernel_32 (long n, float *x, float *y)
 
      "2:				\n\t"
 
-       "stxvw4x		40, 0, %3	\n\t"
-       "stxvw4x		41, %5, %3	\n\t"
-       "stxvw4x		42, %6, %3	\n\t"
-       "stxvw4x		43, %7, %3	\n\t"
-       "stxvw4x		44, %8, %3	\n\t"
-       "stxvw4x		45, %9, %3	\n\t"
-       "stxvw4x		46, %10, %3	\n\t"
-       "stxvw4x		47, %11, %3	\n"
+       "stxvd2x		40, 0, %3	\n\t"
+       "stxvd2x		41, %5, %3	\n\t"
+       "stxvd2x		42, %6, %3	\n\t"
+       "stxvd2x		43, %7, %3	\n\t"
+       "stxvd2x		44, %8, %3	\n\t"
+       "stxvd2x		45, %9, %3	\n\t"
+       "stxvd2x		46, %10, %3	\n\t"
+       "stxvd2x		47, %11, %3	\n"
 
      "#n=%1 x=%4=%2 y=%0=%3 o16=%5 o32=%6 o48=%7 o64=%8 o80=%9 o96=%10 o112=%11"
      :