Doug / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit cf66bb93e0f75e0a4ba1ec070692618fa028e994

Authored by David Woodhouse
1 parent 27d7c2a006
Exists in smarc-l5.0.0_1.0.0-ga and in 5 other branches imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga

byteorder: allow arch to opt to use GCC intrinsics for byteswapping 

Since GCC 4.4, there have been __builtin_bswap32() and __builtin_bswap16()
intrinsics. A __builtin_bswap16() came a little later (4.6 for PowerPC,
48 for other platforms).

By using these instead of the inline assembler that most architectures
have in their __arch_swabXX() macros, we let the compiler see what's
actually happening. The resulting code should be at least as good, and
much *better* in the cases where it can be combined with a nearby load
or store, using a load-and-byteswap or store-and-byteswap instruction
(e.g. lwbrx/stwbrx on PowerPC, movbe on Atom).

When GCC is sufficiently recent *and* the architecture opts in to using
the intrinsics by setting CONFIG_ARCH_USE_BUILTIN_BSWAP, they will be
used in preference to the __arch_swabXX() macros. An architecture which
does not set ARCH_USE_BUILTIN_BSWAP will continue to use its own
hand-crafted macros.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Acked-by: H. Peter Anvin <hpa@linux.intel.com>

Showing 4 changed files with 45 additions and 3 deletions Side-by-side Diff

... ... @@ -112,6 +112,25 @@
112 112 See Documentation/unaligned-memory-access.txt for more
113 113 information on the topic of unaligned memory accesses.
114 114  
  115 +config ARCH_USE_BUILTIN_BSWAP
  116 + bool
  117 + help
  118 + Modern versions of GCC (since 4.4) have builtin functions
  119 + for handling byte-swapping. Using these, instead of the old
  120 + inline assembler that the architecture code provides in the
  121 + __arch_bswapXX() macros, allows the compiler to see what's
  122 + happening and offers more opportunity for optimisation. In
  123 + particular, the compiler will be able to combine the byteswap
  124 + with a nearby load or store and use load-and-swap or
  125 + store-and-swap instructions if the architecture has them. It
  126 + should almost *never* result in code which is worse than the
  127 + hand-coded assembler in <asm/swab.h>. But just in case it
  128 + does, the use of the builtins is optional.
  129 +
  130 + Any architecture with load-and-swap or store-and-swap
  131 + instructions should set this. And it shouldn't hurt to set it
  132 + on architectures that don't have such instructions.
  133 +
115 134 config HAVE_SYSCALL_WRAPPERS
116 135 bool
117 136  
include/linux/compiler-gcc4.h
Diff comments   View file @ cf66bb9
... ... @@ -63,4 +63,14 @@
63 63 #define __compiletime_warning(message) __attribute__((warning(message)))
64 64 #define __compiletime_error(message) __attribute__((error(message)))
65 65 #endif
  66 +
  67 +#ifdef CONFIG_ARCH_USE_BUILTIN_BSWAP
  68 +#if __GNUC_MINOR__ >= 4
  69 +#define __HAVE_BUILTIN_BSWAP32__
  70 +#define __HAVE_BUILTIN_BSWAP64__
  71 +#endif
  72 +#if __GNUC_MINOR__ >= 8 || (defined(__powerpc__) && __GNUC_MINOR__ >= 6)
  73 +#define __HAVE_BUILTIN_BSWAP16__
  74 +#endif
  75 +#endif
include/linux/compiler-intel.h
Diff comments   View file @ cf66bb9
... ... @@ -29,4 +29,10 @@
29 29 #endif
30 30  
31 31 #define uninitialized_var(x) x
  32 +
  33 +#ifndef __HAVE_BUILTIN_BSWAP16__
  34 +/* icc has this, but it's called _bswap16 */
  35 +#define __HAVE_BUILTIN_BSWAP16__
  36 +#define __builtin_bswap16 _bswap16
  37 +#endif
include/uapi/linux/swab.h
Diff comments   View file @ cf66bb9
... ... @@ -45,7 +45,9 @@
45 45  
46 46 static inline __attribute_const__ __u16 __fswab16(__u16 val)
47 47 {
48   -#ifdef __arch_swab16
  48 +#ifdef __HAVE_BUILTIN_BSWAP16__
  49 + return __builtin_bswap16(val);
  50 +#elif defined (__arch_swab16)
49 51 return __arch_swab16(val);
50 52 #else
51 53 return ___constant_swab16(val);
... ... @@ -54,7 +56,9 @@
54 56  
55 57 static inline __attribute_const__ __u32 __fswab32(__u32 val)
56 58 {
57   -#ifdef __arch_swab32
  59 +#ifdef __HAVE_BUILTIN_BSWAP32__
  60 + return __builtin_bswap32(val);
  61 +#elif defined(__arch_swab32)
58 62 return __arch_swab32(val);
59 63 #else
60 64 return ___constant_swab32(val);
... ... @@ -63,7 +67,9 @@
63 67  
64 68 static inline __attribute_const__ __u64 __fswab64(__u64 val)
65 69 {
66   -#ifdef __arch_swab64
  70 +#ifdef __HAVE_BUILTIN_BSWAP64__
  71 + return __builtin_bswap64(val);
  72 +#elif defined (__arch_swab64)
67 73 return __arch_swab64(val);
68 74 #elif defined(__SWAB_64_THRU_32__)
69 75 __u32 h = val >> 32;