tile: optimize strnlen using SIMD instructions

Using strlen as a model, add length checking to create strnlen.

Signed-off-by: Ken Steele <ken@tilera.com>
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>

3 files changed, 96 insertions, 1 deletions

diff --git a/arch/tile/lib/Makefile b/arch/tile/lib/Makefile
index 985f598..5d84437 100644
--- a/arch/tile/lib/Makefile
+++ b/arch/tile/lib/Makefile
@@ -4,7 +4,7 @@
 
 lib-y = cacheflush.o checksum.o cpumask.o delay.o uaccess.o \
 	memmove.o memcpy_$(BITS).o memchr_$(BITS).o memset_$(BITS).o \
-	strchr_$(BITS).o strlen_$(BITS).o
+	strchr_$(BITS).o strlen_$(BITS).o strnlen_$(BITS).o
 
 ifeq ($(CONFIG_TILEGX),y)
 CFLAGS_REMOVE_memcpy_user_64.o = -fno-omit-frame-pointer
diff --git a/arch/tile/lib/strnlen_32.c b/arch/tile/lib/strnlen_32.c
new file mode 100644
index 0000000..1434141
--- /dev/null
+++ b/arch/tile/lib/strnlen_32.c
@@ -0,0 +1,47 @@
+/*
+ * Copyright 2013 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/module.h>
+
+size_t strnlen(const char *s, size_t count)
+{
+	/* Get an aligned pointer. */
+	const uintptr_t s_int = (uintptr_t) s;
+	const uint32_t *p = (const uint32_t *)(s_int & -4);
+	size_t bytes_read = sizeof(*p) - (s_int & (sizeof(*p) - 1));
+	size_t len;
+	uint32_t v, bits;
+
+	/* Avoid page fault risk by not reading any bytes when count is 0. */
+	if (count == 0)
+		return 0;
+
+	/* Read first word, but force bytes before the string to be nonzero. */
+	v = *p | ((1 << ((s_int << 3) & 31)) - 1);
+
+	while ((bits = __insn_seqb(v, 0)) == 0) {
+		if (bytes_read >= count) {
+			/* Read COUNT bytes and didn't find the terminator. */
+			return count;
+		}
+		v = *++p;
+		bytes_read += sizeof(v);
+	}
+
+	len = ((const char *) p) + (__insn_ctz(bits) >> 3) - s;
+	return (len < count ? len : count);
+}
+EXPORT_SYMBOL(strnlen);
diff --git a/arch/tile/lib/strnlen_64.c b/arch/tile/lib/strnlen_64.c
new file mode 100644
index 0000000..2e8de6a
--- /dev/null
+++ b/arch/tile/lib/strnlen_64.c
@@ -0,0 +1,48 @@
+/*
+ * Copyright 2013 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/module.h>
+#include "string-endian.h"
+
+size_t strnlen(const char *s, size_t count)
+{
+	/* Get an aligned pointer. */
+	const uintptr_t s_int = (uintptr_t) s;
+	const uint64_t *p = (const uint64_t *)(s_int & -8);
+	size_t bytes_read = sizeof(*p) - (s_int & (sizeof(*p) - 1));
+	size_t len;
+	uint64_t v, bits;
+
+	/* Avoid page fault risk by not reading any bytes when count is 0. */
+	if (count == 0)
+		return 0;
+
+	/* Read and MASK the first word. */
+	v = *p | MASK(s_int);
+
+	while ((bits = __insn_v1cmpeqi(v, 0)) == 0) {
+		if (bytes_read >= count) {
+			/* Read COUNT bytes and didn't find the terminator. */
+			return count;
+		}
+		v = *++p;
+		bytes_read += sizeof(v);
+	}
+
+	len = ((const char *) p) + (CFZ(bits) >> 3) - s;
+	return (len < count ? len : count);
+}
+EXPORT_SYMBOL(strnlen);