svn commit: r209231 - in head/lib/libc/mips: . gen string sys
Jayachandran C.
jchandra at FreeBSD.org
Wed Jun 16 12:55:15 UTC 2010
Author: jchandra
Date: Wed Jun 16 12:55:14 2010
New Revision: 209231
URL: http://svn.freebsd.org/changeset/base/209231
Log:
Merge jmallett@'s n64 work into HEAD - changeset 1.
Update libc assembly code to use macros that work on both o32 and n64.
Merge string functions from NetBSD.
The changes are from http://svn.freebsd.org/base/user/jmallett/octeon
Approved by: rrs (mentor), jmallett
Modified:
head/lib/libc/mips/SYS.h
head/lib/libc/mips/gen/_setjmp.S
head/lib/libc/mips/gen/setjmp.S
head/lib/libc/mips/gen/sigsetjmp.S
head/lib/libc/mips/string/bcmp.S
head/lib/libc/mips/string/bcopy.S
head/lib/libc/mips/string/bzero.S
head/lib/libc/mips/string/ffs.S
head/lib/libc/mips/string/index.S
head/lib/libc/mips/string/rindex.S
head/lib/libc/mips/string/strcmp.S
head/lib/libc/mips/string/strlen.S
head/lib/libc/mips/sys/Ovfork.S
head/lib/libc/mips/sys/brk.S
head/lib/libc/mips/sys/cerror.S
head/lib/libc/mips/sys/exect.S
head/lib/libc/mips/sys/fork.S
head/lib/libc/mips/sys/pipe.S
head/lib/libc/mips/sys/ptrace.S
head/lib/libc/mips/sys/sbrk.S
Modified: head/lib/libc/mips/SYS.h
==============================================================================
--- head/lib/libc/mips/SYS.h Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/SYS.h Wed Jun 16 12:55:14 2010 (r209231)
@@ -1,4 +1,4 @@
-/* $NetBSD: SYS.h,v 1.18 2003/10/29 12:28:33 pooka Exp $ */
+/* $NetBSD: SYS.h,v 1.19 2009/12/14 01:07:41 matt Exp $ */
/* $FreeBSD$ */
/*-
@@ -79,14 +79,22 @@
*/
#ifdef __ABICALLS__
.abicalls
-# define PIC_PROLOGUE(x,sr) .set noreorder; .cpload sr; .set reorder
-# define PIC_CALL(l,sr) la sr, _C_LABEL(l); jr sr
+# if defined(__mips_o32) || defined(__mips_o64)
+# define PIC_PROLOGUE(x) SETUP_GP
+# define PIC_TAILCALL(l) PTR_LA t9, _C_LABEL(l); jr t9
+# define PIC_RETURN() j ra
+# else
+# define PIC_PROLOGUE(x) SETUP_GP64(t3, x)
+# define PIC_TAILCALL(l) PTR_LA t9, _C_LABEL(l); RESTORE_GP64; jr t9
+# define PIC_RETURN() RESTORE_GP64; j ra
+# endif
#else
-# define PIC_PROLOGUE(x,sr)
-# define PIC_CALL(l,sr) j _C_LABEL(l)
-#endif
+# define PIC_PROLOGUE(x)
+# define PIC_TAILCALL(l) j _C_LABEL(l)
+# define PIC_RETURN()
+#endif /* __ABICALLS__ */
-# define SYSTRAP(x) li v0, SYS_ ## x; syscall;
+# define SYSTRAP(x) li v0,SYS_ ## x; syscall;
/*
* Do a syscall that cannot fail (sync, get{p,u,g,eu,eg)id)
@@ -106,7 +114,7 @@
*/
#define PSEUDO_NOERROR(x) \
LEAF(__sys_ ## x); \
- .weak _C_LABEL(x); \
+ .weak _C_LABEL(x); \
_C_LABEL(x) = _C_LABEL(__CONCAT(__sys_,x)); \
.weak _C_LABEL(__CONCAT(_,x)); \
_C_LABEL(__CONCAT(_,x)) = _C_LABEL(__CONCAT(__sys_,x)); \
@@ -116,14 +124,14 @@ LEAF(__sys_ ## x); \
#define PSEUDO(x) \
LEAF(__sys_ ## x); \
- .weak _C_LABEL(x); \
+ .weak _C_LABEL(x); \
_C_LABEL(x) = _C_LABEL(__CONCAT(__sys_,x)); \
.weak _C_LABEL(__CONCAT(_,x)); \
_C_LABEL(__CONCAT(_,x)) = _C_LABEL(__CONCAT(__sys_,x)); \
- PIC_PROLOGUE(x,t9); \
+ PIC_PROLOGUE(__sys_ ## x); \
SYSTRAP(x); \
bne a3,zero,err; \
- j ra; \
+ PIC_RETURN(); \
err: \
- PIC_CALL(__cerror,t9); \
- END(__sys_ ## x)
+ PIC_TAILCALL(__cerror); \
+END(__sys_ ## x)
Modified: head/lib/libc/mips/gen/_setjmp.S
==============================================================================
--- head/lib/libc/mips/gen/_setjmp.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/gen/_setjmp.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -1,4 +1,4 @@
-/* $NetBSD: _setjmp.S,v 1.20 2005/10/07 17:16:40 tsutsui Exp $ */
+/* $NetBSD: _setjmp.S,v 1.20.34.5 2010/02/03 23:46:47 matt Exp $ */
/*-
* Copyright (c) 1991, 1993
@@ -36,14 +36,15 @@
__FBSDID("$FreeBSD$");
#include <machine/regnum.h>
-#if defined(LIBC_SCCS) && !defined(lint)
- ASMSTR("from: @(#)_setjmp.s 8.1 (Berkeley) 6/4/93")
- ASMSTR("$NetBSD: _setjmp.S,v 1.20 2005/10/07 17:16:40 tsutsui Exp $")
-#endif /* LIBC_SCCS and not lint */
+#include "SYS.h"
-#ifdef __ABICALLS__
- .abicalls
+#if defined(LIBC_SCCS) && !defined(lint)
+#if 0
+ RCSID("from: @(#)_setjmp.s 8.1 (Berkeley) 6/4/93")
+#else
+ RCSID("$NetBSD: _setjmp.S,v 1.20.34.5 2010/02/03 23:46:47 matt Exp $")
#endif
+#endif /* LIBC_SCCS and not lint */
/*
* C library -- _setjmp, _longjmp
@@ -56,62 +57,70 @@ __FBSDID("$FreeBSD$");
* The previous signal state is NOT restored.
*/
+ .set noreorder
LEAF(_setjmp)
-#ifdef __ABICALLS__
- .set noreorder
- .cpload t9
- subu sp, sp, CALLFRAME_SIZ # allocate stack frame
- .cprestore 16
+ REG_PROLOGUE
+ REG_LI v0, _JB_MAGIC__SETJMP
+ REG_S v0, (_JB_MAGIC * SZREG)(a0)
+ REG_S ra, (_JB_REG_RA * SZREG)(a0)
+ REG_S s0, (_JB_REG_S0 * SZREG)(a0)
+ REG_S s1, (_JB_REG_S1 * SZREG)(a0)
+ REG_S s2, (_JB_REG_S2 * SZREG)(a0)
+ REG_S s3, (_JB_REG_S3 * SZREG)(a0)
+ REG_S s4, (_JB_REG_S4 * SZREG)(a0)
+ REG_S s5, (_JB_REG_S5 * SZREG)(a0)
+ REG_S s6, (_JB_REG_S6 * SZREG)(a0)
+ REG_S s7, (_JB_REG_S7 * SZREG)(a0)
+ REG_S s8, (_JB_REG_S8 * SZREG)(a0)
+#if defined(__mips_n32) || defined(__mips_n64)
+ REG_S gp, (_JB_REG_GP * SZREG)(a0) # newabi gp is callee-saved
#endif
- li v0, _JB_MAGIC__SETJMP
- sw v0, (_JB_MAGIC * SZREG)(a0)
- sw ra, (_JB_REG_RA * SZREG)(a0)
- sw s0, (_JB_REG_S0 * SZREG)(a0)
- sw s1, (_JB_REG_S1 * SZREG)(a0)
- sw s2, (_JB_REG_S2 * SZREG)(a0)
- sw s3, (_JB_REG_S3 * SZREG)(a0)
- sw s4, (_JB_REG_S4 * SZREG)(a0)
- sw s5, (_JB_REG_S5 * SZREG)(a0)
- sw s6, (_JB_REG_S6 * SZREG)(a0)
- sw s7, (_JB_REG_S7 * SZREG)(a0)
- sw s8, (_JB_REG_S8 * SZREG)(a0)
-#ifdef __ABICALLS__
- addu sp, sp, CALLFRAME_SIZ # un-allocate the stack frame
-#endif
- sw sp, (_JB_REG_SP * SZREG)(a0)
+ REG_S sp, (_JB_REG_SP * SZREG)(a0)
+ REG_EPILOGUE
+
j ra
move v0, zero
END(_setjmp)
LEAF(_longjmp)
-#ifdef __ABICALLS__
- .set noreorder
- .cpload t9
- subu sp, sp, CALLFRAME_SIZ # allocate stack frame
- .cprestore 16
-#endif
- lw v0, (_JB_MAGIC * SZREG)(a0)
- lw ra, (_JB_REG_RA * SZREG)(a0)
- li t0, _JB_MAGIC__SETJMP
- bne v0, t0, botch # jump if error
- lw s0, (_JB_REG_S0 * SZREG)(a0)
- lw s1, (_JB_REG_S1 * SZREG)(a0)
- lw s2, (_JB_REG_S2 * SZREG)(a0)
- lw s3, (_JB_REG_S3 * SZREG)(a0)
- lw s4, (_JB_REG_S4 * SZREG)(a0)
- lw s5, (_JB_REG_S5 * SZREG)(a0)
- lw s6, (_JB_REG_S6 * SZREG)(a0)
- lw s7, (_JB_REG_S7 * SZREG)(a0)
- lw sp, (_JB_REG_SP * SZREG)(a0)
- lw s8, (_JB_REG_S8 * SZREG)(a0)
+ PIC_PROLOGUE(_longjmp)
+ PTR_SUBU sp, sp, CALLFRAME_SIZ
+ SAVE_GP(CALLFRAME_GP)
+
+ REG_PROLOGUE
+ REG_L v0, (_JB_MAGIC * SZREG)(a0) # get magic number
+ REG_L ra, (_JB_REG_RA * SZREG)(a0)
+ REG_LI t0, _JB_MAGIC__SETJMP
+ bne v0, t0, botch # jump if error
+ PTR_ADDU sp, sp, CALLFRAME_SIZ # does not matter, sanity
+ REG_L s0, (_JB_REG_S0 * SZREG)(a0)
+ REG_L s1, (_JB_REG_S1 * SZREG)(a0)
+ REG_L s2, (_JB_REG_S2 * SZREG)(a0)
+ REG_L s3, (_JB_REG_S3 * SZREG)(a0)
+ REG_L s4, (_JB_REG_S4 * SZREG)(a0)
+ REG_L s5, (_JB_REG_S5 * SZREG)(a0)
+ REG_L s6, (_JB_REG_S6 * SZREG)(a0)
+ REG_L s7, (_JB_REG_S7 * SZREG)(a0)
+#if defined(__mips_n32) || defined(__mips_n64)
+ REG_L gp, (_JB_REG_GP * SZREG)(a0)
+#endif
+ REG_L sp, (_JB_REG_SP * SZREG)(a0)
+ REG_L s8, (_JB_REG_S8 * SZREG)(a0)
+ REG_EPILOGUE
+ move v0, a1 # get return value in 1st arg
j ra
- move v0, a1
+ nop
botch:
- jal _C_LABEL(longjmperror)
- nop
- jal _C_LABEL(abort)
+ /*
+ * We know we aren't returning so we don't care about restoring
+ * our caller's GP.
+ */
+ PTR_LA t9, _C_LABEL(longjmperror)
+ jalr t9
nop
+
+ PIC_TAILCALL(abort)
END(_longjmp)
Modified: head/lib/libc/mips/gen/setjmp.S
==============================================================================
--- head/lib/libc/mips/gen/setjmp.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/gen/setjmp.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -41,6 +41,8 @@ __FBSDID("$FreeBSD$");
ASMSTR("$NetBSD: setjmp.S,v 1.17 2005/09/17 11:49:39 tsutsui Exp $")
#endif /* LIBC_SCCS and not lint */
+#include "SYS.h"
+
#ifdef __ABICALLS__
.abicalls
#endif
@@ -61,100 +63,102 @@ __FBSDID("$FreeBSD$");
NESTED(setjmp, SETJMP_FRAME_SIZE, ra)
.mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ)
- .set noreorder
-#ifdef __ABICALLS__
- .cpload t9
-#endif
- subu sp, sp, SETJMP_FRAME_SIZE # allocate stack frame
-#ifdef __ABICALLS__
- .cprestore 16
-#endif
- sw ra, CALLFRAME_RA(sp) # save RA
- sw a0, CALLFRAME_SIZ(sp) # store env
+ SETUP_GP
+ PTR_SUBU sp, sp, SETJMP_FRAME_SIZE # allocate stack frame
+ SAVE_GP(CALLFRAME_GP)
+ SETUP_GP64(CALLFRAME_GP, setjmp)
+
+ REG_S ra, CALLFRAME_RA(sp) # save RA
+ REG_S a0, CALLFRAME_SIZ(sp) # store env
/* Get the signal mask. */
- addu a2, a0, _JB_SIGMASK * SZREG # &oenv
+ PTR_ADDU a2, a0, _JB_SIGMASK * SZREG # &oenv
li a0, 1 # SIG_SETBLOCK
move a1, zero # &env == 0
- la t9, _C_LABEL(sigprocmask) # get current signal mask
- jal t9
- nop
+ PTR_LA t9, _C_LABEL(sigprocmask) # get current signal mask
+ jalr t9
- lw a0, CALLFRAME_SIZ(sp) # restore env pointer
- lw ra, CALLFRAME_RA(sp) # restore RA
- addu sp, sp, SETJMP_FRAME_SIZE # pop stack frame
-
- li v0, _JB_MAGIC_SETJMP
- sw v0, (_JB_MAGIC * SZREG)(a0)
- sw ra, (_JB_REG_RA * SZREG)(a0)
- sw s0, (_JB_REG_S0 * SZREG)(a0)
- sw s1, (_JB_REG_S1 * SZREG)(a0)
- sw s2, (_JB_REG_S2 * SZREG)(a0)
- sw s3, (_JB_REG_S3 * SZREG)(a0)
- sw s4, (_JB_REG_S4 * SZREG)(a0)
- sw s5, (_JB_REG_S5 * SZREG)(a0)
- sw s6, (_JB_REG_S6 * SZREG)(a0)
- sw s7, (_JB_REG_S7 * SZREG)(a0)
- sw sp, (_JB_REG_SP * SZREG)(a0)
- sw s8, (_JB_REG_S8 * SZREG)(a0)
+ RESTORE_GP64
+ REG_L a0, CALLFRAME_SIZ(sp) # restore env pointer
+ REG_L ra, CALLFRAME_RA(sp) # restore RA
+ PTR_ADDU sp, sp, SETJMP_FRAME_SIZE # pop stack frame
+
+ REG_LI v0, _JB_MAGIC_SETJMP
+ REG_S v0, (_JB_MAGIC * SZREG)(a0)
+ REG_S ra, (_JB_REG_RA * SZREG)(a0)
+ REG_S s0, (_JB_REG_S0 * SZREG)(a0)
+ REG_S s1, (_JB_REG_S1 * SZREG)(a0)
+ REG_S s2, (_JB_REG_S2 * SZREG)(a0)
+ REG_S s3, (_JB_REG_S3 * SZREG)(a0)
+ REG_S s4, (_JB_REG_S4 * SZREG)(a0)
+ REG_S s5, (_JB_REG_S5 * SZREG)(a0)
+ REG_S s6, (_JB_REG_S6 * SZREG)(a0)
+ REG_S s7, (_JB_REG_S7 * SZREG)(a0)
+ REG_S sp, (_JB_REG_SP * SZREG)(a0)
+ REG_S s8, (_JB_REG_S8 * SZREG)(a0)
+#if defined(__mips_n32) || defined(__mips_n64)
+ REG_S gp, (_JB_REG_GP * SZREG)(a0)
+#endif
move v0, zero
- j ra
+ jr ra
END(setjmp)
#define LONGJMP_FRAME_SIZE (CALLFRAME_SIZ + (SZREG * 2))
NESTED(longjmp, LONGJMP_FRAME_SIZE, ra)
.mask 0x80000000, (CALLFRAME_RA - CALLFRAME_SIZ)
- .set noreorder
-#ifdef __ABICALLS__
- .cpload t9
-#endif
- subu sp, sp, LONGJMP_FRAME_SIZE # allocate stack frame
-#ifdef __ABICALLS__
- .cprestore 16
-#endif
- sw ra, CALLFRAME_RA(sp) # save RA
- lw v0, (_JB_MAGIC * SZREG)(a0)
- li t0, _JB_MAGIC_SETJMP
+ PIC_PROLOGUE(longjmp)
+ PTR_SUBU sp, sp, LONGJMP_FRAME_SIZE # allocate stack frame
+ SAVE_GP(CALLFRAME_GP)
+
+ REG_S ra, CALLFRAME_RA(sp) # save RA
+ REG_L v0, (_JB_MAGIC * SZREG)(a0)
+ REG_LI t0, _JB_MAGIC_SETJMP
bne v0, t0, botch # jump if error
nop
- sw a0, CALLFRAME_SIZ(sp) # save env
- sw a1, (CALLFRAME_SIZ + SZREG)(sp) # save return value
+ REG_S a0, CALLFRAME_SIZ(sp) # save env
+ REG_S a1, (CALLFRAME_SIZ + SZREG)(sp) # save return value
# set sigmask
- addu a1, a0, _JB_SIGMASK * SZREG # &set
+ PTR_ADDU a1, a0, _JB_SIGMASK * SZREG # &set
move a2, zero # &oset == NULL
li a0, 3 # SIG_SETMASK
- la t9,_C_LABEL(sigprocmask) # set current signal mask
+ PTR_LA t9,_C_LABEL(sigprocmask) # set current signal mask
jal t9
nop
- lw a0, CALLFRAME_SIZ(sp) # restore env
- lw a1, (CALLFRAME_SIZ + SZREG)(sp) # restore return value
+ REG_L a0, CALLFRAME_SIZ(sp) # restore env
+ REG_L a1, (CALLFRAME_SIZ + SZREG)(sp) # restore return value
+
+ REG_L ra, (_JB_REG_RA * SZREG)(a0)
+ REG_L s0, (_JB_REG_S0 * SZREG)(a0)
+ REG_L s1, (_JB_REG_S1 * SZREG)(a0)
+ REG_L s2, (_JB_REG_S2 * SZREG)(a0)
+ REG_L s3, (_JB_REG_S3 * SZREG)(a0)
+ REG_L s4, (_JB_REG_S4 * SZREG)(a0)
+ REG_L s5, (_JB_REG_S5 * SZREG)(a0)
+ REG_L s6, (_JB_REG_S6 * SZREG)(a0)
+ REG_L s7, (_JB_REG_S7 * SZREG)(a0)
+ REG_L sp, (_JB_REG_SP * SZREG)(a0)
+ REG_L s8, (_JB_REG_S8 * SZREG)(a0)
+#if defined(__mips_n32) || defined(__mips_n64)
+ REG_L gp, (_JB_REG_GP * SZREG)(a0)
+#endif
- lw ra, (_JB_REG_RA * SZREG)(a0)
- lw s0, (_JB_REG_S0 * SZREG)(a0)
- lw s1, (_JB_REG_S1 * SZREG)(a0)
- lw s2, (_JB_REG_S2 * SZREG)(a0)
- lw s3, (_JB_REG_S3 * SZREG)(a0)
- lw s4, (_JB_REG_S4 * SZREG)(a0)
- lw s5, (_JB_REG_S5 * SZREG)(a0)
- lw s6, (_JB_REG_S6 * SZREG)(a0)
- lw s7, (_JB_REG_S7 * SZREG)(a0)
- lw sp, (_JB_REG_SP * SZREG)(a0)
- lw s8, (_JB_REG_S8 * SZREG)(a0)
move v0, a1
j ra
nop
botch:
- la t9, _C_LABEL(longjmperror)
- jal t9
+ /*
+ * We know we aren't returning so we don't care about restoring
+ * our caller's GP.
+ */
+ PTR_LA t9, _C_LABEL(longjmperror)
+ jalr t9
nop
- la t9, _C_LABEL(abort)
- jal t9
- nop
+ PIC_TAILCALL(abort)
END(longjmp)
Modified: head/lib/libc/mips/gen/sigsetjmp.S
==============================================================================
--- head/lib/libc/mips/gen/sigsetjmp.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/gen/sigsetjmp.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -40,6 +40,8 @@ __FBSDID("$FreeBSD$");
ASMSTR("$NetBSD: sigsetjmp.S,v 1.8 2005/09/17 11:49:39 tsutsui Exp $")
#endif /* LIBC_SCCS and not lint */
+#include "SYS.h"
+
#ifdef __ABICALLS__
.abicalls
#endif
@@ -57,30 +59,19 @@ __FBSDID("$FreeBSD$");
*/
LEAF(sigsetjmp)
-#ifdef __ABICALLS__
- .set noreorder
- .cpload t9
- .set reorder
-#endif
+ PIC_PROLOGUE(sigsetjmp)
+
bne a1, 0x0, 1f # do saving of signal mask?
- la t9, _setjmp
- jr t9
+ PIC_TAILCALL(_setjmp)
-1: la t9, setjmp
- jr t9
+1: PIC_TAILCALL(setjmp)
END(sigsetjmp)
LEAF(siglongjmp)
-#ifdef __ABICALLS__
- .set noreorder
- .cpload t9
- .set reorder
-#endif
- lw t0, (_JB_MAGIC * SZREG)(a0)
- li t1, _JB_MAGIC__SETJMP
+ PIC_PROLOGUE(siglongjmp)
+ REG_L t0, (_JB_MAGIC * SZREG)(a0)
+ REG_LI t1, _JB_MAGIC__SETJMP
bne t0, t1, 1f # setjmp or _setjmp magic?
- la t9, _longjmp
- jr t9
-1: la t9, longjmp
- jr t9
+ PIC_TAILCALL(_longjmp)
+1: PIC_TAILCALL(longjmp)
END(siglongjmp)
Modified: head/lib/libc/mips/string/bcmp.S
==============================================================================
--- head/lib/libc/mips/string/bcmp.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/string/bcmp.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -1,4 +1,4 @@
-/* $NetBSD: bcmp.S,v 1.8 2003/08/07 16:42:16 agc Exp $ */
+/* $NetBSD: bcmp.S,v 1.9 2009/12/14 01:07:42 matt Exp $ */
/*-
* Copyright (c) 1991, 1993
@@ -35,9 +35,15 @@
#include <machine/asm.h>
__FBSDID("$FreeBSD$");
+#define _LOCORE /* XXX not really, just assembly-code source */
+#include <machine/endian.h> /* LWLO/LWHI, SWLO/SWHI */
+
#if defined(LIBC_SCCS) && !defined(lint)
+#if 0
ASMSTR("from: @(#)bcmp.s 8.1 (Berkeley) 6/4/93")
- ASMSTR("$NetBSD: bcmp.S,v 1.8 2003/08/07 16:42:16 agc Exp $")
+#else
+ ASMSTR("$NetBSD: bcmp.S,v 1.9 2009/12/14 01:07:42 matt Exp $")
+#endif
#endif /* LIBC_SCCS and not lint */
#ifdef __ABICALLS__
@@ -49,86 +55,76 @@ __FBSDID("$FreeBSD$");
LEAF(bcmp)
.set noreorder
- blt a2, 16, small # is it worth any trouble?
- xor v0, a0, a1 # compare low two bits of addresses
- and v0, v0, 3
- subu a3, zero, a1 # compute # bytes to word align address
- bne v0, zero, unaligned # not possible to align addresses
- and a3, a3, 3
-
- beq a3, zero, 1f
- subu a2, a2, a3 # subtract from remaining count
- move v0, v1 # init v0,v1 so unmodified bytes match
-#ifdef __MIPSEB__
- lwl v0, 0(a0) # read 1, 2, or 3 bytes
- lwl v1, 0(a1)
-#else
- lwr v0, 0(a0) # read 1, 2, or 3 bytes
- lwr v1, 0(a1)
-#endif
- addu a1, a1, a3
- bne v0, v1, nomatch
- addu a0, a0, a3
+ blt a2, 16, small # is it worth any trouble?
+ xor v0, a0, a1 # compare low two bits of addresses
+ and v0, v0, 3
+ PTR_SUBU a3, zero, a1 # compute # bytes to word align address
+ bne v0, zero, unaligned # not possible to align addresses
+ and a3, a3, 3
+
+ beq a3, zero, 1f
+ PTR_SUBU a2, a2, a3 # subtract from remaining count
+ move v0, v1 # init v0,v1 so unmodified bytes match
+ LWHI v0, 0(a0) # read 1, 2, or 3 bytes
+ LWHI v1, 0(a1)
+ PTR_ADDU a1, a1, a3
+ bne v0, v1, nomatch
+ PTR_ADDU a0, a0, a3
1:
- and a3, a2, ~3 # compute number of whole words left
- subu a2, a2, a3 # which has to be >= (16-3) & ~3
- addu a3, a3, a0 # compute ending address
+ and a3, a2, ~3 # compute number of whole words left
+ PTR_SUBU a2, a2, a3 # which has to be >= (16-3) & ~3
+ PTR_ADDU a3, a3, a0 # compute ending address
2:
- lw v0, 0(a0) # compare words
- lw v1, 0(a1)
- addu a0, a0, 4
- bne v0, v1, nomatch
- addu a1, a1, 4
- bne a0, a3, 2b
+ lw v0, 0(a0) # compare words
+ lw v1, 0(a1)
+ PTR_ADDU a0, a0, 4
+ bne v0, v1, nomatch
+ PTR_ADDU a1, a1, 4
+ bne a0, a3, 2b
nop
- b small # finish remainder
+ b small # finish remainder
nop
unaligned:
- beq a3, zero, 2f
- subu a2, a2, a3 # subtract from remaining count
- addu a3, a3, a0 # compute ending address
+ beq a3, zero, 2f
+ PTR_SUBU a2, a2, a3 # subtract from remaining count
+ PTR_ADDU a3, a3, a0 # compute ending address
1:
- lbu v0, 0(a0) # compare bytes until a1 word aligned
- lbu v1, 0(a1)
- addu a0, a0, 1
- bne v0, v1, nomatch
- addu a1, a1, 1
- bne a0, a3, 1b
+ lbu v0, 0(a0) # compare bytes until a1 word aligned
+ lbu v1, 0(a1)
+ PTR_ADDU a0, a0, 1
+ bne v0, v1, nomatch
+ PTR_ADDU a1, a1, 1
+ bne a0, a3, 1b
nop
2:
- and a3, a2, ~3 # compute number of whole words left
- subu a2, a2, a3 # which has to be >= (16-3) & ~3
- addu a3, a3, a0 # compute ending address
+ and a3, a2, ~3 # compute number of whole words left
+ PTR_SUBU a2, a2, a3 # which has to be >= (16-3) & ~3
+ PTR_ADDU a3, a3, a0 # compute ending address
3:
-#ifdef __MIPSEB__
- lwl v0, 0(a0) # compare words a0 unaligned, a1 aligned
- lwr v0, 3(a0)
-#else
- lwr v0, 0(a0) # compare words a0 unaligned, a1 aligned
- lwl v0, 3(a0)
-#endif
- lw v1, 0(a1)
- addu a0, a0, 4
- bne v0, v1, nomatch
- addu a1, a1, 4
- bne a0, a3, 3b
+ LWHI v0, 0(a0) # compare words a0 unaligned, a1 aligned
+ LWLO v0, 3(a0)
+ lw v1, 0(a1)
+ PTR_ADDU a0, a0, 4
+ bne v0, v1, nomatch
+ PTR_ADDU a1, a1, 4
+ bne a0, a3, 3b
nop
small:
- ble a2, zero, match
- addu a3, a2, a0 # compute ending address
+ ble a2, zero, match
+ PTR_ADDU a3, a2, a0 # compute ending address
1:
- lbu v0, 0(a0)
- lbu v1, 0(a1)
- addu a0, a0, 1
- bne v0, v1, nomatch
- addu a1, a1, 1
- bne a0, a3, 1b
+ lbu v0, 0(a0)
+ lbu v1, 0(a1)
+ PTR_ADDU a0, a0, 1
+ bne v0, v1, nomatch
+ PTR_ADDU a1, a1, 1
+ bne a0, a3, 1b
nop
match:
- j ra
- move v0, zero
+ j ra
+ move v0, zero
nomatch:
- j ra
- li v0, 1
+ j ra
+ li v0, 1
.set reorder
END(bcmp)
Modified: head/lib/libc/mips/string/bcopy.S
==============================================================================
--- head/lib/libc/mips/string/bcopy.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/string/bcopy.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -1,4 +1,4 @@
-/* $NetBSD: bcopy.S,v 1.2 2005/12/27 11:23:53 tsutsui Exp $ */
+/* $NetBSD: bcopy.S,v 1.3 2009/12/14 00:39:00 matt Exp $ */
/*
* Mach Operating System
@@ -38,9 +38,15 @@
#include <machine/asm.h>
__FBSDID("$FreeBSD$");
+#define _LOCORE /* XXX not really, just assembly-code source */
+#include <machine/endian.h>
+
#if defined(LIBC_SCCS) && !defined(lint)
+#if 0
ASMSTR("from: @(#)mips_bcopy.s 2.2 CMU 18/06/93")
- ASMSTR("$NetBSD: bcopy.S,v 1.2 2005/12/27 11:23:53 tsutsui Exp $")
+#else
+ ASMSTR("$NetBSD: bcopy.S,v 1.3 2009/12/14 00:39:00 matt Exp $")
+#endif
#endif /* LIBC_SCCS and not lint */
#ifdef __ABICALLS__
@@ -99,71 +105,72 @@ LEAF(FUNCTION)
* copy is alignable. eg if src and dest are both
* on a halfword boundary.
*/
- andi t1,DSTREG,3 # get last 3 bits of dest
- bne t1,zero,3f
- andi t0,SRCREG,3 # get last 3 bits of src
- bne t0,zero,5f
+ andi t1,DSTREG,(SZREG-1) # get last bits of dest
+ bne t1,zero,3f # dest unaligned
+ andi t0,SRCREG,(SZREG-1) # get last bits of src
+ bne t0,zero,5f
/*
- * Forward aligned->aligned copy, 8*4 bytes at a time.
+ * Forward aligned->aligned copy, 8 words at a time.
*/
- li AT,-32
- and t0,SIZEREG,AT # count truncated to multiple of 32 */
- addu a3,SRCREG,t0 # run fast loop up to this address
- sltu AT,SRCREG,a3 # any work to do?
- beq AT,zero,2f
- subu SIZEREG,t0
+98:
+ li AT,-(SZREG*8)
+ and t0,SIZEREG,AT # count truncated to multiples
+ PTR_ADDU a3,SRCREG,t0 # run fast loop up to this addr
+ sltu AT,SRCREG,a3 # any work to do?
+ beq AT,zero,2f
+ PTR_SUBU SIZEREG,t0
/*
* loop body
*/
1: # cp
- lw t3,0(SRCREG)
- lw v1,4(SRCREG)
- lw t0,8(SRCREG)
- lw t1,12(SRCREG)
- addu SRCREG,32
- sw t3,0(DSTREG)
- sw v1,4(DSTREG)
- sw t0,8(DSTREG)
- sw t1,12(DSTREG)
- lw t1,-4(SRCREG)
- lw t0,-8(SRCREG)
- lw v1,-12(SRCREG)
- lw t3,-16(SRCREG)
- addu DSTREG,32
- sw t1,-4(DSTREG)
- sw t0,-8(DSTREG)
- sw v1,-12(DSTREG)
- bne SRCREG,a3,1b
- sw t3,-16(DSTREG)
+ REG_L t3,(0*SZREG)(SRCREG)
+ REG_L v1,(1*SZREG)(SRCREG)
+ REG_L t0,(2*SZREG)(SRCREG)
+ REG_L t1,(3*SZREG)(SRCREG)
+ PTR_ADDU SRCREG,SZREG*8
+ REG_S t3,(0*SZREG)(DSTREG)
+ REG_S v1,(1*SZREG)(DSTREG)
+ REG_S t0,(2*SZREG)(DSTREG)
+ REG_S t1,(3*SZREG)(DSTREG)
+ REG_L t1,(-1*SZREG)(SRCREG)
+ REG_L t0,(-2*SZREG)(SRCREG)
+ REG_L v1,(-3*SZREG)(SRCREG)
+ REG_L t3,(-4*SZREG)(SRCREG)
+ PTR_ADDU DSTREG,SZREG*8
+ REG_S t1,(-1*SZREG)(DSTREG)
+ REG_S t0,(-2*SZREG)(DSTREG)
+ REG_S v1,(-3*SZREG)(DSTREG)
+ bne SRCREG,a3,1b
+ REG_S t3,(-4*SZREG)(DSTREG)
/*
* Copy a word at a time, no loop unrolling.
*/
2: # wordcopy
- andi t2,SIZEREG,3 # get byte count / 4
- subu t2,SIZEREG,t2 # t2 = number of words to copy * 4
- beq t2,zero,3f
- addu t0,SRCREG,t2 # stop at t0
- subu SIZEREG,SIZEREG,t2
+ andi t2,SIZEREG,(SZREG-1) # get byte count / SZREG
+ PTR_SUBU t2,SIZEREG,t2 # t2 = words to copy * SZREG
+ beq t2,zero,3f
+ PTR_ADDU t0,SRCREG,t2 # stop at t0
+ PTR_SUBU SIZEREG,SIZEREG,t2
1:
- lw t3,0(SRCREG)
- addu SRCREG,4
- sw t3,0(DSTREG)
- bne SRCREG,t0,1b
- addu DSTREG,4
+ REG_L t3,0(SRCREG)
+ PTR_ADDU SRCREG,SZREG
+ REG_S t3,0(DSTREG)
+ bne SRCREG,t0,1b
+ PTR_ADDU DSTREG,SZREG
3: # bytecopy
- beq SIZEREG,zero,4f # nothing left to do?
+ beq SIZEREG,zero,4f # nothing left to do?
nop
1:
- lb t3,0(SRCREG)
- addu SRCREG,1
- sb t3,0(DSTREG)
- subu SIZEREG,1
- bgtz SIZEREG,1b
- addu DSTREG,1
+ lb t3,0(SRCREG)
+ PTR_ADDU SRCREG,1
+ sb t3,0(DSTREG)
+ PTR_SUBU SIZEREG,1
+ bgtz SIZEREG,1b
+ PTR_ADDU DSTREG,1
4: # copydone
j ra
@@ -173,96 +180,91 @@ LEAF(FUNCTION)
* Copy from unaligned source to aligned dest.
*/
5: # destaligned
- andi t0,SIZEREG,3 # t0 = bytecount mod 4
- subu a3,SIZEREG,t0 # number of words to transfer
- beq a3,zero,3b
+ andi t0,SIZEREG,(SZREG-1) # t0 = bytecount mod SZREG
+ PTR_SUBU a3,SIZEREG,t0 # number of words to transfer
+ beq a3,zero,3b
nop
- move SIZEREG,t0 # this many to do after we are done
- addu a3,SRCREG,a3 # stop point
+ move SIZEREG,t0 # this many to do after we are done
+ PTR_ADDU a3,SRCREG,a3 # stop point
1:
-#ifdef __MIPSEB__
- lwl t3,0(SRCREG)
- lwr t3,3(SRCREG)
-#else
- lwr t3,0(SRCREG)
- lwl t3,3(SRCREG)
-#endif
- addi SRCREG,4
- sw t3,0(DSTREG)
- bne SRCREG,a3,1b
- addi DSTREG,4
+ REG_LHI t3,0(SRCREG)
+ REG_LLO t3,SZREG-1(SRCREG)
+ PTR_ADDI SRCREG,SZREG
+ REG_S t3,0(DSTREG)
+ bne SRCREG,a3,1b
+ PTR_ADDI DSTREG,SZREG
- j 3b
+ b 3b
nop
6: # backcopy -- based on above
- addu SRCREG,SIZEREG
- addu DSTREG,SIZEREG
- andi t1,DSTREG,3 # get last 3 bits of dest
- bne t1,zero,3f
- andi t0,SRCREG,3 # get last 3 bits of src
- bne t0,zero,5f
+ PTR_ADDU SRCREG,SIZEREG
+ PTR_ADDU DSTREG,SIZEREG
+ andi t1,DSTREG,SZREG-1 # get last 3 bits of dest
+ bne t1,zero,3f
+ andi t0,SRCREG,SZREG-1 # get last 3 bits of src
+ bne t0,zero,5f
/*
* Forward aligned->aligned copy, 8*4 bytes at a time.
*/
- li AT,-32
- and t0,SIZEREG,AT # count truncated to multiple of 32
- beq t0,zero,2f # any work to do?
- subu SIZEREG,t0
- subu a3,SRCREG,t0
+ li AT,(-8*SZREG)
+ and t0,SIZEREG,AT # count truncated to multiple of 32
+ beq t0,zero,2f # any work to do?
+ PTR_SUBU SIZEREG,t0
+ PTR_SUBU a3,SRCREG,t0
/*
* loop body
*/
1: # cp
- lw t3,-16(SRCREG)
- lw v1,-12(SRCREG)
- lw t0,-8(SRCREG)
- lw t1,-4(SRCREG)
- subu SRCREG,32
- sw t3,-16(DSTREG)
- sw v1,-12(DSTREG)
- sw t0,-8(DSTREG)
- sw t1,-4(DSTREG)
- lw t1,12(SRCREG)
- lw t0,8(SRCREG)
- lw v1,4(SRCREG)
- lw t3,0(SRCREG)
- subu DSTREG,32
- sw t1,12(DSTREG)
- sw t0,8(DSTREG)
- sw v1,4(DSTREG)
- bne SRCREG,a3,1b
- sw t3,0(DSTREG)
+ REG_L t3,(-4*SZREG)(SRCREG)
+ REG_L v1,(-3*SZREG)(SRCREG)
+ REG_L t0,(-2*SZREG)(SRCREG)
+ REG_L t1,(-1*SZREG)(SRCREG)
+ PTR_SUBU SRCREG,8*SZREG
+ REG_S t3,(-4*SZREG)(DSTREG)
+ REG_S v1,(-3*SZREG)(DSTREG)
+ REG_S t0,(-2*SZREG)(DSTREG)
+ REG_S t1,(-1*SZREG)(DSTREG)
+ REG_L t1,(3*SZREG)(SRCREG)
+ REG_L t0,(2*SZREG)(SRCREG)
+ REG_L v1,(1*SZREG)(SRCREG)
+ REG_L t3,(0*SZREG)(SRCREG)
+ PTR_SUBU DSTREG,8*SZREG
+ REG_S t1,(3*SZREG)(DSTREG)
+ REG_S t0,(2*SZREG)(DSTREG)
+ REG_S v1,(1*SZREG)(DSTREG)
+ bne SRCREG,a3,1b
+ REG_S t3,(0*SZREG)(DSTREG)
/*
* Copy a word at a time, no loop unrolling.
*/
2: # wordcopy
- andi t2,SIZEREG,3 # get byte count / 4
- subu t2,SIZEREG,t2 # t2 = number of words to copy * 4
- beq t2,zero,3f
- subu t0,SRCREG,t2 # stop at t0
- subu SIZEREG,SIZEREG,t2
+ andi t2,SIZEREG,SZREG-1 # get byte count / 4
+ PTR_SUBU t2,SIZEREG,t2 # t2 = number of words to copy
+ beq t2,zero,3f
+ PTR_SUBU t0,SRCREG,t2 # stop at t0
+ PTR_SUBU SIZEREG,SIZEREG,t2
1:
- lw t3,-4(SRCREG)
- subu SRCREG,4
- sw t3,-4(DSTREG)
- bne SRCREG,t0,1b
- subu DSTREG,4
+ REG_L t3,-SZREG(SRCREG)
+ PTR_SUBU SRCREG,SZREG
+ REG_S t3,-SZREG(DSTREG)
+ bne SRCREG,t0,1b
+ PTR_SUBU DSTREG,SZREG
3: # bytecopy
- beq SIZEREG,zero,4f # nothing left to do?
+ beq SIZEREG,zero,4f # nothing left to do?
nop
1:
- lb t3,-1(SRCREG)
- subu SRCREG,1
- sb t3,-1(DSTREG)
- subu SIZEREG,1
- bgtz SIZEREG,1b
- subu DSTREG,1
+ lb t3,-1(SRCREG)
+ PTR_SUBU SRCREG,1
+ sb t3,-1(DSTREG)
+ PTR_SUBU SIZEREG,1
+ bgtz SIZEREG,1b
+ PTR_SUBU DSTREG,1
4: # copydone
j ra
@@ -272,27 +274,22 @@ LEAF(FUNCTION)
* Copy from unaligned source to aligned dest.
*/
5: # destaligned
- andi t0,SIZEREG,3 # t0 = bytecount mod 4
- subu a3,SIZEREG,t0 # number of words to transfer
- beq a3,zero,3b
+ andi t0,SIZEREG,SZREG-1 # t0 = bytecount mod 4
+ PTR_SUBU a3,SIZEREG,t0 # number of words to transfer
+ beq a3,zero,3b
nop
- move SIZEREG,t0 # this many to do after we are done
- subu a3,SRCREG,a3 # stop point
+ move SIZEREG,t0 # this many to do after we are done
+ PTR_SUBU a3,SRCREG,a3 # stop point
1:
-#ifdef __MIPSEB__
- lwl t3,-4(SRCREG)
- lwr t3,-1(SRCREG)
-#else
- lwr t3,-4(SRCREG)
- lwl t3,-1(SRCREG)
-#endif
- subu SRCREG,4
- sw t3,-4(DSTREG)
- bne SRCREG,a3,1b
- subu DSTREG,4
+ REG_LHI t3,-SZREG(SRCREG)
+ REG_LLO t3,-1(SRCREG)
+ PTR_SUBU SRCREG,SZREG
+ REG_S t3,-SZREG(DSTREG)
+ bne SRCREG,a3,1b
+ PTR_SUBU DSTREG,SZREG
- j 3b
+ b 3b
nop
.set reorder
Modified: head/lib/libc/mips/string/bzero.S
==============================================================================
--- head/lib/libc/mips/string/bzero.S Wed Jun 16 12:42:20 2010 (r209230)
+++ head/lib/libc/mips/string/bzero.S Wed Jun 16 12:55:14 2010 (r209231)
@@ -1,4 +1,4 @@
-/* $NetBSD: bzero.S,v 1.8 2003/08/07 16:42:16 agc Exp $ */
+/* $NetBSD: bzero.S,v 1.10 2009/12/14 02:53:52 matt Exp $ */
/*-
* Copyright (c) 1991, 1993
@@ -36,10 +36,15 @@
__FBSDID("$FreeBSD$");
#if defined(LIBC_SCCS) && !defined(lint)
+#if 0
ASMSTR("from: @(#)bzero.s 8.1 (Berkeley) 6/4/93")
- ASMSTR("$NetBSD: bzero.S,v 1.8 2003/08/07 16:42:16 agc Exp $")
+#else
+ ASMSTR("$NetBSD: bzero.S,v 1.10 2009/12/14 02:53:52 matt Exp $")
+#endif
#endif /* LIBC_SCCS and not lint */
+#define _LOCORE /* XXX not really, just assembly-code source */
+#include <machine/endian.h>
#ifdef __ABICALLS__
.abicalls
@@ -49,34 +54,48 @@ __FBSDID("$FreeBSD$");
LEAF(bzero)
.set noreorder
- blt a1, 12, smallclr # small amount to clear?
- subu a3, zero, a0 # compute # bytes to word align address
- and a3, a3, 3
- beq a3, zero, 1f # skip if word aligned
- subu a1, a1, a3 # subtract from remaining count
-#ifdef __MIPSEB__
- swl zero, 0(a0) # clear 1, 2, or 3 bytes to align
-#else
- swr zero, 0(a0) # clear 1, 2, or 3 bytes to align
+ blt a1, 3*SZREG, smallclr # small amount to clear?
+ PTR_SUBU a3, zero, a0 # compute # bytes to word align address
+ and a3, a3, SZREG-1
+ beq a3, zero, 1f # skip if word aligned
+#if SZREG == 4
+ PTR_SUBU a1, a1, a3 # subtract from remaining count
+ SWHI zero, 0(a0) # clear 1, 2, or 3 bytes to align
+ PTR_ADDU a0, a0, a3
+#endif
+#if SZREG == 8
+ PTR_SUBU a1, a1, a3 # subtract from remaining count
*** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
More information about the svn-src-head
mailing list