From 93e2e85139509338c68279c7260ebb68177b23a9 Mon Sep 17 00:00:00 2001 From: Michal Simek Date: Sat, 9 Oct 2010 13:58:24 +1000 Subject: microblaze: Separate library optimized functions memcpy/memmove/memset Signed-off-by: Michal Simek --- arch/microblaze/lib/memcpy.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) (limited to 'arch/microblaze/lib/memcpy.c') diff --git a/arch/microblaze/lib/memcpy.c b/arch/microblaze/lib/memcpy.c index 014bac92bdff..ab2d115f9ee5 100644 --- a/arch/microblaze/lib/memcpy.c +++ b/arch/microblaze/lib/memcpy.c @@ -33,17 +33,24 @@ #include #ifdef __HAVE_ARCH_MEMCPY +#ifndef CONFIG_OPT_LIB_FUNCTION void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) { const char *src = v_src; char *dst = v_dst; -#ifndef CONFIG_OPT_LIB_FUNCTION + /* Simple, byte oriented memcpy. */ while (c--) *dst++ = *src++; return v_dst; -#else +} +#else /* CONFIG_OPT_LIB_FUNCTION */ +void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) +{ + const char *src = v_src; + char *dst = v_dst; + /* The following code tries to optimize the copy by using unsigned * alignment. This will work fine if both source and destination are * aligned on the same boundary. However, if they are aligned on @@ -150,7 +157,7 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) } return v_dst; -#endif } +#endif /* CONFIG_OPT_LIB_FUNCTION */ EXPORT_SYMBOL(memcpy); #endif /* __HAVE_ARCH_MEMCPY */ -- cgit v1.2.2 From 1180b28ca82c529972bfd438467d5cd71cca5372 Mon Sep 17 00:00:00 2001 From: Michal Simek Date: Sat, 9 Oct 2010 14:05:58 +1000 Subject: microblaze: Support C optimized lib functions for little-endian Optimized C library functions can rapidly speedup the kernel. memset doesn't need to be optimized because there is no difference in behavior on little/big endian cpu. Signed-off-by: Michal Simek --- arch/microblaze/lib/memcpy.c | 33 ++++++++++++++++++++++++++++++--- 1 file changed, 30 insertions(+), 3 deletions(-) (limited to 'arch/microblaze/lib/memcpy.c') diff --git a/arch/microblaze/lib/memcpy.c b/arch/microblaze/lib/memcpy.c index ab2d115f9ee5..cc495d7d99cc 100644 --- a/arch/microblaze/lib/memcpy.c +++ b/arch/microblaze/lib/memcpy.c @@ -93,7 +93,7 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) case 0x1: /* Unaligned - Off by 1 */ /* Word align the source */ i_src = (const void *) ((unsigned)src & ~3); - +#ifndef __MICROBLAZEEL__ /* Load the holding buffer */ buf_hold = *i_src++ << 8; @@ -102,7 +102,16 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) *i_dst++ = buf_hold | value >> 24; buf_hold = value << 8; } +#else + /* Load the holding buffer */ + buf_hold = (*i_src++ & 0xFFFFFF00) >>8; + for (; c >= 4; c -= 4) { + value = *i_src++; + *i_dst++ = buf_hold | ((value & 0xFF) << 24); + buf_hold = (value & 0xFFFFFF00) >>8; + } +#endif /* Realign the source */ src = (const void *)i_src; src -= 3; @@ -110,7 +119,7 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) case 0x2: /* Unaligned - Off by 2 */ /* Word align the source */ i_src = (const void *) ((unsigned)src & ~3); - +#ifndef __MICROBLAZEEL__ /* Load the holding buffer */ buf_hold = *i_src++ << 16; @@ -119,7 +128,16 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) *i_dst++ = buf_hold | value >> 16; buf_hold = value << 16; } +#else + /* Load the holding buffer */ + buf_hold = (*i_src++ & 0xFFFF0000 )>>16; + for (; c >= 4; c -= 4) { + value = *i_src++; + *i_dst++ = buf_hold | ((value & 0xFFFF)<<16); + buf_hold = (value & 0xFFFF0000) >>16; + } +#endif /* Realign the source */ src = (const void *)i_src; src -= 2; @@ -127,7 +145,7 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) case 0x3: /* Unaligned - Off by 3 */ /* Word align the source */ i_src = (const void *) ((unsigned)src & ~3); - +#ifndef __MICROBLAZEEL__ /* Load the holding buffer */ buf_hold = *i_src++ << 24; @@ -136,7 +154,16 @@ void *memcpy(void *v_dst, const void *v_src, __kernel_size_t c) *i_dst++ = buf_hold | value >> 8; buf_hold = value << 24; } +#else + /* Load the holding buffer */ + buf_hold = (*i_src++ & 0xFF000000) >> 24; + for (; c >= 4; c -= 4) { + value = *i_src++; + *i_dst++ = buf_hold | ((value & 0xFFFFFF) << 8); + buf_hold = (value & 0xFF000000) >> 24; + } +#endif /* Realign the source */ src = (const void *)i_src; src -= 1; -- cgit v1.2.2