Update.

2003-09-14  Ulrich Drepper  <drepper@redhat.com>

	* dirent/dirent.h: Remove __THROW from scandir.

2003-09-14  Philip Blundell  <philb@gnu.org>

	* sysdeps/unix/sysv/linux/arm/mmap.S: Use sys_mmap2 if it's known
	to be available.

	* sysdeps/unix/sysv/linux/arm/mmap64.S: Optimise code a little.

	* sysdeps/arm/memset.S: Rewrite.

5 files changed, 115 insertions, 62 deletions

diff --git a/ChangeLog b/ChangeLog
index 468a551..ac3576f 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,16 @@
+2003-09-14  Ulrich Drepper  <drepper@redhat.com>
+
+	* dirent/dirent.h: Remove __THROW from scandir.
+
+2003-09-14  Philip Blundell  <philb@gnu.org>
+
+	* sysdeps/unix/sysv/linux/arm/mmap.S: Use sys_mmap2 if it's known
+	to be available.
+
+	* sysdeps/unix/sysv/linux/arm/mmap64.S: Optimise code a little.
+
+	* sysdeps/arm/memset.S: Rewrite.
+
 2003-09-12  Jakub Jelinek  <jakub@redhat.com>
 
 	* sysdeps/unix/sysv/linux/s390/bits/typesizes.h: New.
diff --git a/dirent/dirent.h b/dirent/dirent.h
index 9f7c380..319fb74 100644
--- a/dirent/dirent.h
+++ b/dirent/dirent.h
@@ -239,14 +239,14 @@ extern int dirfd (DIR *__dirp) __THROW;
 extern int scandir (__const char *__restrict __dir,
 		    struct dirent ***__restrict __namelist,
 		    int (*__selector) (__const struct dirent *),
-		    int (*__cmp) (__const void *, __const void *)) __THROW;
+		    int (*__cmp) (__const void *, __const void *));
 # else
 #  ifdef __REDIRECT
 extern int __REDIRECT (scandir,
 		       (__const char *__restrict __dir,
 			struct dirent ***__restrict __namelist,
 			int (*__selector) (__const struct dirent *),
-			int (*__cmp) (__const void *, __const void *)) __THROW,
+			int (*__cmp) (__const void *, __const void *)),
 		       scandir64);
 #  else
 #   define scandir scandir64
@@ -259,7 +259,7 @@ extern int __REDIRECT (scandir,
 extern int scandir64 (__const char *__restrict __dir,
 		      struct dirent64 ***__restrict __namelist,
 		      int (*__selector) (__const struct dirent64 *),
-		      int (*__cmp) (__const void *, __const void *)) __THROW;
+		      int (*__cmp) (__const void *, __const void *));
 # endif
 
 /* Function to compare two `struct dirent's alphabetically.  */
@@ -269,8 +269,7 @@ extern int alphasort (__const void *__e1, __const void *__e2)
 # else
 #  ifdef __REDIRECT
 extern int __REDIRECT (alphasort,
-		       (__const void *__e1, __const void *__e2)
-		       __THROW,
+		       (__const void *__e1, __const void *__e2) __THROW,
 		       alphasort64) __attribute_pure__;
 #  else
 #   define alphasort alphasort64
@@ -290,8 +289,7 @@ extern int versionsort (__const void *__e1, __const void *__e2)
 #  else
 #   ifdef __REDIRECT
 extern int __REDIRECT (versionsort,
-		       (__const void *__e1, __const void *__e2)
-		       __THROW,
+		       (__const void *__e1, __const void *__e2) __THROW,
 		       versionsort64) __attribute_pure__;
 #   else
 #    define versionsort versionsort64
diff --git a/sysdeps/arm/memset.S b/sysdeps/arm/memset.S
index 0b62413..1e2699d 100644
--- a/sysdeps/arm/memset.S
+++ b/sysdeps/arm/memset.S
@@ -22,47 +22,46 @@
 /* void *memset (dstpp, c, len) */
 
 ENTRY(memset)
-	mov	a4, a1
-	cmp	a3, $8		@ at least 8 bytes to do?
-	blt	2f
-	orr	a2, a2, a2, lsl $8
-	orr	a2, a2, a2, lsl $16
+	mov	r3, r0
+	cmp	r2, #8
+	bcc	2f		@ less than 8 bytes to move
+
 1:
-	tst	a4, $3		@ aligned yet?
-	strneb	a2, [a4], $1
-	subne	a3, a3, $1
+	tst	r3, #3		@ aligned yet?
+	strneb	r1, [r3], #1
+	subne	r2, r2, #1
 	bne	1b
-	mov	ip, a2
+
+	orr	r1, r1, r1, lsl $8
+	orr	r1, r1, r1, lsl $16
+
 1:
-	cmp	a3, $8		@ 8 bytes still to do?
-	blt	2f
-	stmia	a4!, {a2, ip}
-	sub	a3, a3, $8
-	cmp	a3, $8		@ 8 bytes still to do?
-	blt	2f
-	stmia	a4!, {a2, ip}
-	sub	a3, a3, $8
-	cmp	a3, $8		@ 8 bytes still to do?
-	blt	2f
-	stmia	a4!, {a2, ip}
-	sub	a3, a3, $8
-	cmp	a3, $8		@ 8 bytes still to do?
-	stmgeia	a4!, {a2, ip}
-	subge	a3, a3, $8
-	bge	1b
+	subs	r2, r2, #8
+	strcs	r1, [r3], #4	@ store up to 32 bytes per loop iteration
+	strcs	r1, [r3], #4
+	subcss	r2, r2, #8
+	strcs	r1, [r3], #4
+	strcs	r1, [r3], #4
+	subcss	r2, r2, #8
+	strcs	r1, [r3], #4
+	strcs	r1, [r3], #4
+	subcss	r2, r2, #8
+	strcs	r1, [r3], #4
+	strcs	r1, [r3], #4
+	bcs	1b
+
+	and	r2, r2, #7
 2:
-	movs	a3, a3		@ anything left?
-	RETINSTR(moveq,pc,lr)	@ nope
-	rsb	a3, a3, $7
-	add	pc, pc, a3, lsl $2
-	mov	r0, r0
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	strb	a2, [a4], $1
-	RETINSTR(mov,pc,lr)
+	subs	r2, r2, #1	@ store up to 4 bytes per loop iteration
+	strcsb	r1, [r3], #1
+	subcss	r2, r2, #1
+	strcsb	r1, [r3], #1
+	subcss	r2, r2, #1
+	strcsb	r1, [r3], #1
+	subcss	r2, r2, #1
+	strcsb	r1, [r3], #1
+	bcs	2b
+	
+	DO_RET(lr)
 END(memset)
 libc_hidden_builtin_def (memset)
diff --git a/sysdeps/unix/sysv/linux/arm/mmap.S b/sysdeps/unix/sysv/linux/arm/mmap.S
index af93c7b..7beba68 100644
--- a/sysdeps/unix/sysv/linux/arm/mmap.S
+++ b/sysdeps/unix/sysv/linux/arm/mmap.S
@@ -1,4 +1,4 @@
-/* Copyright (C) 1998, 2000 Free Software Foundation, Inc.
+/* Copyright (C) 1998, 2000, 2003 Free Software Foundation, Inc.
    This file is part of the GNU C Library.
 
    The GNU C Library is free software; you can redistribute it and/or
@@ -17,11 +17,47 @@
    02111-1307 USA.  */
 
 #include <sysdep.h>
+#include <kernel-features.h>
+
+#define	EINVAL		22
 
 	.text
 
 ENTRY (__mmap)
+# ifdef __ASSUME_MMAP2_SYSCALL
+	/* This code is actually a couple of cycles slower than the
+	   sys_mmap version below, so it might seem like a loss.  But the
+	   code path inside the kernel is sufficiently much shorter to
+	   make it a net gain to use mmap2 when it's known to be
+	   available.  */
+
+	/* shuffle args */
+	str	r5, [sp, #-4]!
+	ldr	r5, [sp, #8]
+	str	r4, [sp, #-4]!
+	ldr	r4, [sp, #8]
+
+	/* convert offset to pages */
+	movs	ip, r5, lsl #20
+	bne	.Linval
+	mov	r5, r5, lsr #12
+	
+	/* do the syscall */
+	swi	SYS_ify (mmap2)
+
+	/* restore registers */
+2:
+	ldr	r4, [sp], #4
+	ldr	r5, [sp], #4
+
+	cmn	r0, $4096
+	RETINSTR(movcc, pc, lr)
+	b	PLTJMP(syscall_error)
 
+.Linval:
+	mov	r0, #-EINVAL
+	b	2b
+# else
 	/* Because we can only get five args through the syscall interface, and
 	   mmap() takes six, we need to build a parameter block and pass its
 	   address instead.  The 386 port does a similar trick.  */
@@ -49,6 +85,7 @@ ENTRY (__mmap)
 	cmn	r0, $4096
 	RETINSTR(movcc, pc, lr)
 	b	PLTJMP(syscall_error);
+#endif
 
 PSEUDO_END (__mmap)
 
diff --git a/sysdeps/unix/sysv/linux/arm/mmap64.S b/sysdeps/unix/sysv/linux/arm/mmap64.S
index 3936e25..f8361b5 100644
--- a/sysdeps/unix/sysv/linux/arm/mmap64.S
+++ b/sysdeps/unix/sysv/linux/arm/mmap64.S
@@ -27,38 +27,44 @@
 	.text
 ENTRY (__mmap64)
 #ifdef __NR_mmap2
-	stmfd	sp!, {r4, r5, lr}
-	ldr	r5, [sp, $16]
-	ldr	r4, [sp, $12]
-	movs	ip, r5, lsl $20		@ check that offset is page-aligned
+	ldr	ip, [sp, $4]		@ offset low part
+	str	r5, [sp, #-4]!
+	ldr	r5, [sp, $12]		@ offset high part
+	str	r4, [sp, #-4]!
+	movs	r4, ip, lsl $20		@ check that offset is page-aligned
+	mov	ip, ip, lsr $12
+	moveqs	r4, r5, lsr $12		@ check for overflow
 	bne	.Linval
-	ldr	ip, [sp, $20]
-	mov	r5, r5, lsr $12
-	orr	r5, r5, ip, lsl $20	@ compose page offset
-	movs	ip, ip, lsr $12
-	bne	.Linval			@ check for overflow
+	ldr	r4, [sp, $8]		@ load fd
+	orr	r5, ip, r5, lsl $20	@ compose page offset
 	mov	ip, r0
 	swi	SYS_ify (mmap2)
 	cmn	r0, $4096
-	LOADREGS(ccfd, sp!, {r4, r5, pc})
 # ifdef __ASSUME_MMAP2_SYSCALL
-	ldmfd	sp!, {r4, r5, lr}
+	ldr	r4, [sp], #4
+	ldr	r5, [sp], #4
+	RETINSTR(movcc, pc, lr)	
 	b	PLTJMP(syscall_error)
 # else
+	ldrcc	r4, [sp], #4
+	ldrcc	r5, [sp], #4
+	RETINSTR(movcc, pc, lr)
 	cmn	r0, $ENOSYS
-	ldmnefd	sp!, {r4, r5, lr}
-	bne	PLTJMP(syscall_error)
+	bne	.Lerror
 	/* The current kernel does not support mmap2.  Fall back to plain
 	   mmap if the offset is small enough.  */
-	ldr	r5, [sp, $20]
+	ldr	r5, [sp, $16]
 	mov	r0, ip			@ first arg was clobbered
 	teq	r5, $0
-	ldmeqfd	sp!, {r4, r5, lr}
+	ldreq	r4, [sp], #4
+	ldreq	r5, [sp], #4
 	beq	PLTJMP(__mmap)
 # endif
 .Linval:
 	mov	r0, $-EINVAL
-	ldmfd	sp!, {r4, r5, lr}
+.Lerror:
+	ldr	r4, [sp], #4
+	ldr	r5, [sp], #4
 	b	PLTJMP(syscall_error)
 #else
 	/* The kernel headers do not support mmap2.  Fall back to plain