alpha: introduce asm/swab.h
authorHarvey Harrison <harvey.harrison@gmail.com>
Tue, 6 Jan 2009 22:56:29 +0000 (14:56 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 7 Jan 2009 02:10:27 +0000 (18:10 -0800)
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/alpha/include/asm/Kbuild
arch/alpha/include/asm/byteorder.h
arch/alpha/include/asm/swab.h [new file with mode: 0644]

index b7c8f18..4dad273 100644 (file)
@@ -9,3 +9,4 @@ unifdef-y += console.h
 unifdef-y += fpu.h
 unifdef-y += sysinfo.h
 unifdef-y += compiler.h
+unifdef-y += swab.h
index 58e958f..6772f31 100644 (file)
@@ -1,47 +1,7 @@
 #ifndef _ALPHA_BYTEORDER_H
 #define _ALPHA_BYTEORDER_H
 
-#include <asm/types.h>
-#include <linux/compiler.h>
-#include <asm/compiler.h>
-
-#ifdef __GNUC__
-
-static inline __attribute_const__ __u32 __arch__swab32(__u32 x)
-{
-       /*
-        * Unfortunately, we can't use the 6 instruction sequence
-        * on ev6 since the latency of the UNPKBW is 3, which is
-        * pretty hard to hide.  Just in case a future implementation
-        * has a lower latency, here's the sequence (also by Mike Burrows)
-        *
-        * UNPKBW a0, v0       v0: 00AA00BB00CC00DD
-        * SLL v0, 24, a0      a0: BB00CC00DD000000
-        * BIS v0, a0, a0      a0: BBAACCBBDDCC00DD
-        * EXTWL a0, 6, v0     v0: 000000000000BBAA
-        * ZAP a0, 0xf3, a0    a0: 00000000DDCC0000
-        * ADDL a0, v0, v0     v0: ssssssssDDCCBBAA
-        */
-
-       __u64 t0, t1, t2, t3;
-
-       t0 = __kernel_inslh(x, 7);      /* t0 : 0000000000AABBCC */
-       t1 = __kernel_inswl(x, 3);      /* t1 : 000000CCDD000000 */
-       t1 |= t0;                       /* t1 : 000000CCDDAABBCC */
-       t2 = t1 >> 16;                  /* t2 : 0000000000CCDDAA */
-       t0 = t1 & 0xFF00FF00;           /* t0 : 00000000DD00BB00 */
-       t3 = t2 & 0x00FF00FF;           /* t3 : 0000000000CC00AA */
-       t1 = t0 + t3;                   /* t1 : ssssssssDDCCBBAA */
-
-       return t1;
-}
-
-#define __arch__swab32 __arch__swab32
-
-#endif /* __GNUC__ */
-
-#define __BYTEORDER_HAS_U64__
-
+#include <asm/swab.h>
 #include <linux/byteorder/little_endian.h>
 
 #endif /* _ALPHA_BYTEORDER_H */
diff --git a/arch/alpha/include/asm/swab.h b/arch/alpha/include/asm/swab.h
new file mode 100644 (file)
index 0000000..68e7089
--- /dev/null
@@ -0,0 +1,42 @@
+#ifndef _ALPHA_SWAB_H
+#define _ALPHA_SWAB_H
+
+#include <asm/types.h>
+#include <linux/compiler.h>
+#include <asm/compiler.h>
+
+#ifdef __GNUC__
+
+static inline __attribute_const__ __u32 __arch_swab32(__u32 x)
+{
+       /*
+        * Unfortunately, we can't use the 6 instruction sequence
+        * on ev6 since the latency of the UNPKBW is 3, which is
+        * pretty hard to hide.  Just in case a future implementation
+        * has a lower latency, here's the sequence (also by Mike Burrows)
+        *
+        * UNPKBW a0, v0       v0: 00AA00BB00CC00DD
+        * SLL v0, 24, a0      a0: BB00CC00DD000000
+        * BIS v0, a0, a0      a0: BBAACCBBDDCC00DD
+        * EXTWL a0, 6, v0     v0: 000000000000BBAA
+        * ZAP a0, 0xf3, a0    a0: 00000000DDCC0000
+        * ADDL a0, v0, v0     v0: ssssssssDDCCBBAA
+        */
+
+       __u64 t0, t1, t2, t3;
+
+       t0 = __kernel_inslh(x, 7);      /* t0 : 0000000000AABBCC */
+       t1 = __kernel_inswl(x, 3);      /* t1 : 000000CCDD000000 */
+       t1 |= t0;                       /* t1 : 000000CCDDAABBCC */
+       t2 = t1 >> 16;                  /* t2 : 0000000000CCDDAA */
+       t0 = t1 & 0xFF00FF00;           /* t0 : 00000000DD00BB00 */
+       t3 = t2 & 0x00FF00FF;           /* t3 : 0000000000CC00AA */
+       t1 = t0 + t3;                   /* t1 : ssssssssDDCCBBAA */
+
+       return t1;
+}
+#define __arch_swab32 __arch_swab32
+
+#endif /* __GNUC__ */
+
+#endif /* _ALPHA_SWAB_H */