The generic implementation of strlen() reads strings byte per byte.
This patch implements strlen() in assembly for PPC32 based on
a read of entire words, in the same spirit as what some other
arches and glibc do.
For long strings, the time spent in strlen is reduced by 50-60%
Signed-off-by: Christophe Leroy <christophe.le...@c-s.fr>
---
Applies after the patch 'powerpc/lib: move PPC32 specific functions out of
string.S'
arch/powerpc/include/asm/string.h | 3 +++
arch/powerpc/lib/string_32.S | 40 +++++++++++++++++++++++++++++++++++++++
2 files changed, 43 insertions(+)
diff --git a/arch/powerpc/include/asm/string.h
b/arch/powerpc/include/asm/string.h
index 0f41686b6243..23ee2a0f2b21 100644
--- a/arch/powerpc/include/asm/string.h
+++ b/arch/powerpc/include/asm/string.h
@@ -15,6 +15,9 @@
#define __HAVE_ARCH_MEMCHR
#define __HAVE_ARCH_MEMSET16
#define __HAVE_ARCH_MEMCPY_FLUSHCACHE
+#ifdef CONFIG_PPC32
+#define __HAVE_ARCH_STRLEN
+#endif
extern char * strcpy(char *,const char *);
extern __kernel_size_t strlen(const char *);
diff --git a/arch/powerpc/lib/string_32.S b/arch/powerpc/lib/string_32.S
index c4e70123d245..31575a698c97 100644
--- a/arch/powerpc/lib/string_32.S
+++ b/arch/powerpc/lib/string_32.S
@@ -62,6 +62,46 @@ _GLOBAL(memcmp)
blr
EXPORT_SYMBOL(memcmp)
+_GLOBAL(strlen)
+ andi. r9, r3, 3
+ addi r10, r3, -4
+ beq+ 2f
+1: lbz r9, 4(r10)
+ addi r10, r10, 1
+ cmpwi cr0, r9, 0
+ beq 19f
+ andi. r9, r10, 3
+ bne 1b
+2: lis r6, 0x8080
+ ori r6, r6, 0x8080
+ rlwinm r7, r6, 1, 0xffffffff
+3: lwzu r9, 4(r10)
+ subf r8, r7, r9
+ andc r11, r6, r9
+ and. r8, r8, r11
+ beq+ 3b
+ rlwinm. r8, r9, 0, 0xff000000
+ beq 20f
+ rlwinm. r8, r9, 0, 0x00ff0000
+ beq 21f
+ rlwinm. r8, r9, 0, 0x0000ff00
+ beq 22f
+ rlwinm. r8, r9, 0, 0x000000ff
+ bne 3b
+23: subf r3, r3, r10
+ addi r3, r3, 3
+ blr
+22: subf r3, r3, r10
+ addi r3, r3, 2
+ blr
+21: subf r3, r3, r10
+ addi r3, r3, 1
+ blr
+19: addi r10, r10, 3
+20: subf r3, r3, r10
+ blr
+EXPORT_SYMBOL(strlen)
+
CACHELINE_BYTES = L1_CACHE_BYTES
LG_CACHELINE_BYTES = L1_CACHE_SHIFT
CACHELINE_MASK = (L1_CACHE_BYTES-1)
--
2.13.3
