From 07e59c7bc7f375f792ec9734be42fe4fa391a8bb Mon Sep 17 00:00:00 2001 From: Felix Fietkau Date: Thu, 11 Mar 2010 18:49:01 +0000 Subject: [PATCH] mips: allow the compiler to optimize memset, memcmp, memcpy for better performance and (in some instances) smaller code SVN-Revision: 20151 --- .../202-mips_mem_functions_performance.patch | 83 +++++++++++++++++++ .../202-mips_mem_functions_performance.patch | 83 +++++++++++++++++++ .../202-mips_mem_functions_performance.patch | 83 +++++++++++++++++++ .../202-mips_mem_functions_performance.patch | 83 +++++++++++++++++++ .../202-mips_mem_functions_performance.patch | 83 +++++++++++++++++++ 5 files changed, 415 insertions(+) create mode 100644 target/linux/generic-2.6/patches-2.6.30/202-mips_mem_functions_performance.patch create mode 100644 target/linux/generic-2.6/patches-2.6.31/202-mips_mem_functions_performance.patch create mode 100644 target/linux/generic-2.6/patches-2.6.32/202-mips_mem_functions_performance.patch create mode 100644 target/linux/generic-2.6/patches-2.6.33/202-mips_mem_functions_performance.patch create mode 100644 target/linux/generic-2.6/patches-2.6.34/202-mips_mem_functions_performance.patch diff --git a/target/linux/generic-2.6/patches-2.6.30/202-mips_mem_functions_performance.patch b/target/linux/generic-2.6/patches-2.6.30/202-mips_mem_functions_performance.patch new file mode 100644 index 00000000000..6ea9ba51ee0 --- /dev/null +++ b/target/linux/generic-2.6/patches-2.6.30/202-mips_mem_functions_performance.patch @@ -0,0 +1,83 @@ +--- a/arch/mips/include/asm/string.h ++++ b/arch/mips/include/asm/string.h +@@ -133,11 +133,44 @@ strncmp(__const__ char *__cs, __const__ + + #define __HAVE_ARCH_MEMSET + extern void *memset(void *__s, int __c, size_t __count); ++#define memset(__s, __c, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memset((__s), (__c), __len); \ ++ else \ ++ __ret = __builtin_memset((__s), (__c), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMCPY + extern void *memcpy(void *__to, __const__ void *__from, size_t __n); ++#define memcpy(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memcpy((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memcpy((dst), (src), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMMOVE + extern void *memmove(void *__dest, __const__ void *__src, size_t __n); ++#define memmove(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memmove((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memmove((dst), (src), __len); \ ++ __ret; \ ++}) ++ ++#define __HAVE_ARCH_MEMCMP ++#define memcmp(src1, src2, len) __builtin_memcmp((src1), (src2), (len)) + + #endif /* _ASM_STRING_H */ +--- a/arch/mips/lib/Makefile ++++ b/arch/mips/lib/Makefile +@@ -3,7 +3,7 @@ + # + + lib-y += csum_partial.o delay.o memcpy.o memcpy-inatomic.o memset.o \ +- strlen_user.o strncpy_user.o strnlen_user.o uncached.o ++ strlen_user.o strncpy_user.o strnlen_user.o uncached.o memcmp.o + + obj-y += iomap.o + obj-$(CONFIG_PCI) += iomap-pci.o +--- /dev/null ++++ b/arch/mips/lib/memcmp.c +@@ -0,0 +1,22 @@ ++/* ++ * copied from linux/lib/string.c ++ * ++ * Copyright (C) 1991, 1992 Linus Torvalds ++ */ ++ ++#include ++#include ++ ++#undef memcmp ++int memcmp(const void *cs, const void *ct, size_t count) ++{ ++ const unsigned char *su1, *su2; ++ int res = 0; ++ ++ for (su1 = cs, su2 = ct; 0 < count; ++su1, ++su2, count--) ++ if ((res = *su1 - *su2) != 0) ++ break; ++ return res; ++} ++EXPORT_SYMBOL(memcmp); ++ diff --git a/target/linux/generic-2.6/patches-2.6.31/202-mips_mem_functions_performance.patch b/target/linux/generic-2.6/patches-2.6.31/202-mips_mem_functions_performance.patch new file mode 100644 index 00000000000..6ea9ba51ee0 --- /dev/null +++ b/target/linux/generic-2.6/patches-2.6.31/202-mips_mem_functions_performance.patch @@ -0,0 +1,83 @@ +--- a/arch/mips/include/asm/string.h ++++ b/arch/mips/include/asm/string.h +@@ -133,11 +133,44 @@ strncmp(__const__ char *__cs, __const__ + + #define __HAVE_ARCH_MEMSET + extern void *memset(void *__s, int __c, size_t __count); ++#define memset(__s, __c, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memset((__s), (__c), __len); \ ++ else \ ++ __ret = __builtin_memset((__s), (__c), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMCPY + extern void *memcpy(void *__to, __const__ void *__from, size_t __n); ++#define memcpy(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memcpy((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memcpy((dst), (src), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMMOVE + extern void *memmove(void *__dest, __const__ void *__src, size_t __n); ++#define memmove(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memmove((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memmove((dst), (src), __len); \ ++ __ret; \ ++}) ++ ++#define __HAVE_ARCH_MEMCMP ++#define memcmp(src1, src2, len) __builtin_memcmp((src1), (src2), (len)) + + #endif /* _ASM_STRING_H */ +--- a/arch/mips/lib/Makefile ++++ b/arch/mips/lib/Makefile +@@ -3,7 +3,7 @@ + # + + lib-y += csum_partial.o delay.o memcpy.o memcpy-inatomic.o memset.o \ +- strlen_user.o strncpy_user.o strnlen_user.o uncached.o ++ strlen_user.o strncpy_user.o strnlen_user.o uncached.o memcmp.o + + obj-y += iomap.o + obj-$(CONFIG_PCI) += iomap-pci.o +--- /dev/null ++++ b/arch/mips/lib/memcmp.c +@@ -0,0 +1,22 @@ ++/* ++ * copied from linux/lib/string.c ++ * ++ * Copyright (C) 1991, 1992 Linus Torvalds ++ */ ++ ++#include ++#include ++ ++#undef memcmp ++int memcmp(const void *cs, const void *ct, size_t count) ++{ ++ const unsigned char *su1, *su2; ++ int res = 0; ++ ++ for (su1 = cs, su2 = ct; 0 < count; ++su1, ++su2, count--) ++ if ((res = *su1 - *su2) != 0) ++ break; ++ return res; ++} ++EXPORT_SYMBOL(memcmp); ++ diff --git a/target/linux/generic-2.6/patches-2.6.32/202-mips_mem_functions_performance.patch b/target/linux/generic-2.6/patches-2.6.32/202-mips_mem_functions_performance.patch new file mode 100644 index 00000000000..6ea9ba51ee0 --- /dev/null +++ b/target/linux/generic-2.6/patches-2.6.32/202-mips_mem_functions_performance.patch @@ -0,0 +1,83 @@ +--- a/arch/mips/include/asm/string.h ++++ b/arch/mips/include/asm/string.h +@@ -133,11 +133,44 @@ strncmp(__const__ char *__cs, __const__ + + #define __HAVE_ARCH_MEMSET + extern void *memset(void *__s, int __c, size_t __count); ++#define memset(__s, __c, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memset((__s), (__c), __len); \ ++ else \ ++ __ret = __builtin_memset((__s), (__c), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMCPY + extern void *memcpy(void *__to, __const__ void *__from, size_t __n); ++#define memcpy(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memcpy((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memcpy((dst), (src), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMMOVE + extern void *memmove(void *__dest, __const__ void *__src, size_t __n); ++#define memmove(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memmove((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memmove((dst), (src), __len); \ ++ __ret; \ ++}) ++ ++#define __HAVE_ARCH_MEMCMP ++#define memcmp(src1, src2, len) __builtin_memcmp((src1), (src2), (len)) + + #endif /* _ASM_STRING_H */ +--- a/arch/mips/lib/Makefile ++++ b/arch/mips/lib/Makefile +@@ -3,7 +3,7 @@ + # + + lib-y += csum_partial.o delay.o memcpy.o memcpy-inatomic.o memset.o \ +- strlen_user.o strncpy_user.o strnlen_user.o uncached.o ++ strlen_user.o strncpy_user.o strnlen_user.o uncached.o memcmp.o + + obj-y += iomap.o + obj-$(CONFIG_PCI) += iomap-pci.o +--- /dev/null ++++ b/arch/mips/lib/memcmp.c +@@ -0,0 +1,22 @@ ++/* ++ * copied from linux/lib/string.c ++ * ++ * Copyright (C) 1991, 1992 Linus Torvalds ++ */ ++ ++#include ++#include ++ ++#undef memcmp ++int memcmp(const void *cs, const void *ct, size_t count) ++{ ++ const unsigned char *su1, *su2; ++ int res = 0; ++ ++ for (su1 = cs, su2 = ct; 0 < count; ++su1, ++su2, count--) ++ if ((res = *su1 - *su2) != 0) ++ break; ++ return res; ++} ++EXPORT_SYMBOL(memcmp); ++ diff --git a/target/linux/generic-2.6/patches-2.6.33/202-mips_mem_functions_performance.patch b/target/linux/generic-2.6/patches-2.6.33/202-mips_mem_functions_performance.patch new file mode 100644 index 00000000000..6ea9ba51ee0 --- /dev/null +++ b/target/linux/generic-2.6/patches-2.6.33/202-mips_mem_functions_performance.patch @@ -0,0 +1,83 @@ +--- a/arch/mips/include/asm/string.h ++++ b/arch/mips/include/asm/string.h +@@ -133,11 +133,44 @@ strncmp(__const__ char *__cs, __const__ + + #define __HAVE_ARCH_MEMSET + extern void *memset(void *__s, int __c, size_t __count); ++#define memset(__s, __c, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memset((__s), (__c), __len); \ ++ else \ ++ __ret = __builtin_memset((__s), (__c), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMCPY + extern void *memcpy(void *__to, __const__ void *__from, size_t __n); ++#define memcpy(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memcpy((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memcpy((dst), (src), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMMOVE + extern void *memmove(void *__dest, __const__ void *__src, size_t __n); ++#define memmove(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memmove((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memmove((dst), (src), __len); \ ++ __ret; \ ++}) ++ ++#define __HAVE_ARCH_MEMCMP ++#define memcmp(src1, src2, len) __builtin_memcmp((src1), (src2), (len)) + + #endif /* _ASM_STRING_H */ +--- a/arch/mips/lib/Makefile ++++ b/arch/mips/lib/Makefile +@@ -3,7 +3,7 @@ + # + + lib-y += csum_partial.o delay.o memcpy.o memcpy-inatomic.o memset.o \ +- strlen_user.o strncpy_user.o strnlen_user.o uncached.o ++ strlen_user.o strncpy_user.o strnlen_user.o uncached.o memcmp.o + + obj-y += iomap.o + obj-$(CONFIG_PCI) += iomap-pci.o +--- /dev/null ++++ b/arch/mips/lib/memcmp.c +@@ -0,0 +1,22 @@ ++/* ++ * copied from linux/lib/string.c ++ * ++ * Copyright (C) 1991, 1992 Linus Torvalds ++ */ ++ ++#include ++#include ++ ++#undef memcmp ++int memcmp(const void *cs, const void *ct, size_t count) ++{ ++ const unsigned char *su1, *su2; ++ int res = 0; ++ ++ for (su1 = cs, su2 = ct; 0 < count; ++su1, ++su2, count--) ++ if ((res = *su1 - *su2) != 0) ++ break; ++ return res; ++} ++EXPORT_SYMBOL(memcmp); ++ diff --git a/target/linux/generic-2.6/patches-2.6.34/202-mips_mem_functions_performance.patch b/target/linux/generic-2.6/patches-2.6.34/202-mips_mem_functions_performance.patch new file mode 100644 index 00000000000..6ea9ba51ee0 --- /dev/null +++ b/target/linux/generic-2.6/patches-2.6.34/202-mips_mem_functions_performance.patch @@ -0,0 +1,83 @@ +--- a/arch/mips/include/asm/string.h ++++ b/arch/mips/include/asm/string.h +@@ -133,11 +133,44 @@ strncmp(__const__ char *__cs, __const__ + + #define __HAVE_ARCH_MEMSET + extern void *memset(void *__s, int __c, size_t __count); ++#define memset(__s, __c, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memset((__s), (__c), __len); \ ++ else \ ++ __ret = __builtin_memset((__s), (__c), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMCPY + extern void *memcpy(void *__to, __const__ void *__from, size_t __n); ++#define memcpy(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memcpy((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memcpy((dst), (src), __len); \ ++ __ret; \ ++}) + + #define __HAVE_ARCH_MEMMOVE + extern void *memmove(void *__dest, __const__ void *__src, size_t __n); ++#define memmove(dst, src, len) \ ++({ \ ++ size_t __len = (len); \ ++ void *__ret; \ ++ if (__builtin_constant_p(len) && __len >= 64) \ ++ __ret = memmove((dst), (src), __len); \ ++ else \ ++ __ret = __builtin_memmove((dst), (src), __len); \ ++ __ret; \ ++}) ++ ++#define __HAVE_ARCH_MEMCMP ++#define memcmp(src1, src2, len) __builtin_memcmp((src1), (src2), (len)) + + #endif /* _ASM_STRING_H */ +--- a/arch/mips/lib/Makefile ++++ b/arch/mips/lib/Makefile +@@ -3,7 +3,7 @@ + # + + lib-y += csum_partial.o delay.o memcpy.o memcpy-inatomic.o memset.o \ +- strlen_user.o strncpy_user.o strnlen_user.o uncached.o ++ strlen_user.o strncpy_user.o strnlen_user.o uncached.o memcmp.o + + obj-y += iomap.o + obj-$(CONFIG_PCI) += iomap-pci.o +--- /dev/null ++++ b/arch/mips/lib/memcmp.c +@@ -0,0 +1,22 @@ ++/* ++ * copied from linux/lib/string.c ++ * ++ * Copyright (C) 1991, 1992 Linus Torvalds ++ */ ++ ++#include ++#include ++ ++#undef memcmp ++int memcmp(const void *cs, const void *ct, size_t count) ++{ ++ const unsigned char *su1, *su2; ++ int res = 0; ++ ++ for (su1 = cs, su2 = ct; 0 < count; ++su1, ++su2, count--) ++ if ((res = *su1 - *su2) != 0) ++ break; ++ return res; ++} ++EXPORT_SYMBOL(memcmp); ++