AIX changes for Power8

Author: kavanabhat

common_power.h

@@ -39,6 +39,35 @@
 #ifndef COMMON_POWER
 #define COMMON_POWER
 
+#define str(x)	#x
+
+#ifdef OS_AIX
+#define XXSPLTD(T,A,z)	xxpermdi	T, A, A, 0b##z##z
+#define XXMRGHD(T,A,B)	xxpermdi	T, A, B, 0b00
+#define XXMRGLD(T,A,B)	xxpermdi	T, A, B, 0b11
+#define XXSWAPD(T,A)	xxpermdi	T, A, A, 0b10
+#define XVMOVDP(T,A)	xvcpsgndp	T, A, A
+
+#define XXSPLTD_S(T,A,z)	"xxpermdi	" str(T) ", " str(A) ", " str(A) ", 0b" str(z ## z) "	\n\t"
+#define XXMRGHD_S(T,A,B)	"xxpermdi	" str(T) ", " str(A) ", " str(B) ", 0b00	\n\t"
+#define XXMRGLD_S(T,A,B)	"xxpermdi	" str(T) ", " str(A) ", " str(B) ", 0b11	\n\t"
+#define XXSWAPD_S(T,A)	"xxpermdi	" str(T) ", " str(A) ", " str(A) ", 0b10	\n\t"
+
+#else
+#define XXSPLTD(T,A,z)	xxspltd	T, A, z
+#define XXMRGHD(T,A,B)	xxmrghd	T, A, B
+#define XXMRGLD(T,A,B)	xxmrgld	T, A, B
+#define XXSWAPD(T,A)	xxswapd	T, A
+#define XVMOVDP(T,A)	xvmovdp	T, A
+
+#define XXSPLTD_S(T,A,z)	"xxspltd	T, A, z	\n\t"
+#define XXMRGHD_S(T,A,B)	"xxmrghd	T, A, B	\n\t"
+#define XXMRGLD_S(T,A,B)	"xxmrgld	T, A, B	\n\t"
+#define XXSWAPD_S(T,A)	"xxswapd	T, A"
+
+#endif
+
+
 #if defined(POWER8) || defined(POWER9)
 #define MB		__asm__ __volatile__ ("eieio":::"memory")
 #define WMB		__asm__ __volatile__ ("eieio":::"memory")

kernel/Makefile.L3

@@ -68,10 +68,10 @@ static float casum_kernel_16 (long n, float *x)
        "addi		%2, %2, 128	\n\t"
 
        "addic.		%1, %1, -16	\n\t"
-       "ble		2f		\n\t"
+       "ble		two%=		\n\t"
 
-       ".p2align	5		\n"
-     "1:				\n\t"
+       ".align	5		\n"
+     "one%=:				\n\t"
 
        "xvabssp		48, 40		\n\t"
        "xvabssp		49, 41		\n\t"
@@ -108,9 +108,9 @@ static float casum_kernel_16 (long n, float *x)
        "xvaddsp		38, 38, %x5	\n\t"
        "xvaddsp		39, 39, %x6	\n\t"
 
-       "bgt		1b		\n"
+       "bgt		one%=		\n"
 
-     "2:				\n\t"
+     "two%=:				\n\t"
 
        "xvabssp		48, 40		\n\t"
        "xvabssp		49, 41		\n\t"

kernel/power/casum_microk_power8.c

@@ -62,10 +62,10 @@ static void ccopy_kernel_32 (long n, float *x, float *y)
        "addi		%2, %2, 128	\n\t"
 
        "addic.		%1, %1, -32	\n\t"
-       "ble		2f		\n\t"
+       "ble		two%=		\n\t"
 
-       ".p2align	5		\n"
-     "1:				\n\t"
+       ".align	5		\n"
+     "one%=:				\n\t"
 
        "stxvd2x		32, 0, %3	\n\t"
        "stxvd2x		33, %5, %3	\n\t"
@@ -108,9 +108,9 @@ static void ccopy_kernel_32 (long n, float *x, float *y)
        "addi		%2, %2, 128	\n\t"
 
        "addic.		%1, %1, -32	\n\t"
-       "bgt		1b		\n"
+       "bgt		one%=		\n"
 
-     "2:				\n\t"
+     "two%=:				\n\t"
 
        "stxvd2x		32, 0, %3	\n\t"
        "stxvd2x		33, %5, %3	\n\t"