openwrtv3/target/linux/generic/pending-4.9/306-mips_mem_functions_performance.patch

From: Felix Fietkau <nbd@nbd.name>
Subject: [PATCH] mips: allow the compiler to optimize memset, memcmp, memcpy  for better performance and (in some instances) smaller code

lede-commit: 07e59c7bc7f375f792ec9734be42fe4fa391a8bb
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
 arch/mips/boot/compressed/Makefile |  3 ++-
 arch/mips/include/asm/string.h     | 38 ++++++++++++++++++++++++++++++++++++++
 arch/mips/lib/Makefile             |  2 +-
 arch/mips/lib/memcmp.c             | 22 ++++++++++++++++++++++
 4 files changed, 63 insertions(+), 2 deletions(-)
 create mode 100644 arch/mips/lib/memcmp.c

diff --git a/arch/mips/boot/compressed/Makefile b/arch/mips/boot/compressed/Makefile
index 90aca95fe314..3cd3b391ef49 100644
--- a/arch/mips/boot/compressed/Makefile
+++ b/arch/mips/boot/compressed/Makefile
@@ -23,7 +23,8 @@ KBUILD_CFLAGS := $(shell echo $(KBUILD_CFLAGS) | sed -e "s/-pg//")
 KBUILD_CFLAGS := $(filter-out -fstack-protector, $(KBUILD_CFLAGS))
 
 KBUILD_CFLAGS := $(LINUXINCLUDE) $(KBUILD_CFLAGS) -D__KERNEL__ \
-	-DBOOT_HEAP_SIZE=$(BOOT_HEAP_SIZE) -D"VMLINUX_LOAD_ADDRESS_ULL=$(VMLINUX_LOAD_ADDRESS)ull"
+	-DBOOT_HEAP_SIZE=$(BOOT_HEAP_SIZE) -D"VMLINUX_LOAD_ADDRESS_ULL=$(VMLINUX_LOAD_ADDRESS)ull" \
+	-D__ZBOOT__
 
 KBUILD_AFLAGS := $(LINUXINCLUDE) $(KBUILD_AFLAGS) -D__ASSEMBLY__ \
 	-DBOOT_HEAP_SIZE=$(BOOT_HEAP_SIZE) \
diff --git a/arch/mips/include/asm/string.h b/arch/mips/include/asm/string.h
index 29030cb398ee..7b737f9b6d58 100644
--- a/arch/mips/include/asm/string.h
+++ b/arch/mips/include/asm/string.h
@@ -140,4 +140,42 @@ extern void *memcpy(void *__to, __const__ void *__from, size_t __n);
 #define __HAVE_ARCH_MEMMOVE
 extern void *memmove(void *__dest, __const__ void *__src, size_t __n);
 
+#ifndef __ZBOOT__
+#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 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 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
+
 #endif /* _ASM_STRING_H */
diff --git a/arch/mips/lib/Makefile b/arch/mips/lib/Makefile
index 0344e575f522..33a0211e954a 100644
--- a/arch/mips/lib/Makefile
+++ b/arch/mips/lib/Makefile
@@ -4,7 +4,7 @@
 
 lib-y	+= bitops.o csum_partial.o delay.o memcpy.o memset.o \
 	   mips-atomic.o strlen_user.o strncpy_user.o \
-	   strnlen_user.o uncached.o
+	   strnlen_user.o uncached.o memcmp.o
 
 obj-y			+= iomap.o
 obj-$(CONFIG_PCI)	+= iomap-pci.o
diff --git a/arch/mips/lib/memcmp.c b/arch/mips/lib/memcmp.c
new file mode 100644
index 000000000000..35ef1646286e
--- /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 <linux/module.h>
+#include <linux/string.h>
+
+#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);
+
-- 
2.11.0