import files from glibc for an ia64/static port

MAINTAINERS

@@ -62,6 +62,10 @@ HPPA
 S:	Unmaintained
 
 
+IA64
+S:	Unmaintained
+
+
 I386
 N:	Erik Andersen
 E:	andersen@codepoet.org

extra/Configs/Config.ia64

@@ -0,0 +1,28 @@
+#
+# For a description of the syntax of this configuration file,
+# see extra/config/Kconfig-language.txt
+#
+
+config TARGET_ARCH
+	default "ia64"
+
+config ARCH_SUPPORTS_BIG_ENDIAN
+	bool
+	default n
+
+config ARCH_SUPPORTS_LITTLE_ENDIAN
+	bool
+	default y
+
+config FORCE_OPTIONS_FOR_ARCH
+	bool
+	default y
+	select ARCH_HAS_MMU
+	select HAS_NO_THREADS
+	select ARCH_HAS_NO_LDSO
+
+config ARCH_CFLAGS
+	string
+
+config LIBGCC_CFLAGS
+	string

extra/Configs/Config.in

@@ -41,6 +41,9 @@ config TARGET_i386
 config TARGET_i960
 	bool "i960 (BROKEN)"
 
+config TARGET_ia64
+	bool "ia64"
+
 config TARGET_m68k
 	bool "m68k"
 
@@ -122,6 +125,10 @@ if TARGET_i960
 source "extra/Configs/Config.i960"
 endif
 
+if TARGET_ia64
+source "extra/Configs/Config.ia64"
+endif
+
 if TARGET_m68k
 source "extra/Configs/Config.m68k"
 endif

libc/string/ia64/Makefile

@@ -0,0 +1,13 @@
+# Makefile for uClibc
+#
+# Copyright (C) 2000-2005 Erik Andersen <andersen@uclibc.org>
+#
+# Licensed under the LGPL v2.1, see the file COPYING.LIB in this tarball.
+#
+
+top_srcdir:=../../../
+top_builddir:=../../../
+all: objs
+include $(top_builddir)Rules.mak
+include ../Makefile.in
+include $(top_srcdir)Makerules

libc/string/ia64/bcopy.S

@@ -0,0 +1,10 @@
+#include "sysdep.h"
+
+ENTRY(bcopy)
+	.regstk 3, 0, 0, 0
+	mov r8 = in0
+	mov in0 = in1
+	;;
+	mov in1 = r8
+	br.cond.sptk.many HIDDEN_JUMPTARGET(memmove)
+END(bcopy)

libc/string/ia64/bzero.S

@@ -0,0 +1,315 @@
+/* Optimized version of the standard bzero() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2002 Free Software Foundation, Inc.
+   Contributed by Dan Pop for Itanium <Dan.Pop@cern.ch>.
+   Rewritten for McKinley by Sverre Jarp, HP Labs/CERN <Sverre.Jarp@cern.ch>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+        in0:    dest
+        in1:    count
+
+   The algorithm is fairly straightforward: set byte by byte until we
+   we get to a 16B-aligned address, then loop on 128 B chunks using an
+   early store as prefetching, then loop on 32B chucks, then clear remaining
+   words, finally clear remaining bytes.
+   Since a stf.spill f0 can store 16B in one go, we use this instruction
+   to get peak speed.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define dest		in0
+#define	cnt		in1
+
+#define tmp		r31
+#define save_lc		r30
+#define ptr0		r29
+#define ptr1		r28
+#define ptr2		r27
+#define ptr3		r26
+#define ptr9 		r24
+#define	loopcnt		r23
+#define linecnt		r22
+#define bytecnt		r21
+
+// This routine uses only scratch predicate registers (p6 - p15)
+#define p_scr		p6	// default register for same-cycle branches
+#define p_unalgn	p9
+#define p_y		p11
+#define p_n		p12
+#define p_yy		p13
+#define p_nn		p14
+
+#define movi0		mov
+
+#define MIN1		15
+#define MIN1P1HALF	8
+#define LINE_SIZE	128
+#define LSIZE_SH        7			// shift amount
+#define PREF_AHEAD	8
+
+#define USE_FLP
+#if defined(USE_INT)
+#define store		st8
+#define myval		r0
+#elif defined(USE_FLP)
+#define store		stf8
+#define myval		f0
+#endif
+
+.align	64
+ENTRY(bzero)
+{ .mmi
+	.prologue
+	alloc	tmp = ar.pfs, 2, 0, 0, 0
+	lfetch.nt1 [dest]
+	.save   ar.lc, save_lc
+	movi0	save_lc = ar.lc
+} { .mmi
+	.body
+	mov	ret0 = dest		// return value
+	nop.m	0
+	cmp.eq	p_scr, p0 = cnt, r0
+;; }
+{ .mmi
+	and	ptr2 = -(MIN1+1), dest	// aligned address
+	and	tmp = MIN1, dest	// prepare to check for alignment
+	tbit.nz p_y, p_n = dest, 0	// Do we have an odd address? (M_B_U)
+} { .mib
+	mov	ptr1 = dest
+	nop.i	0
+(p_scr)	br.ret.dpnt.many rp		// return immediately if count = 0
+;; }
+{ .mib
+	cmp.ne	p_unalgn, p0 = tmp, r0
+} { .mib					// NB: # of bytes to move is 1
+	sub	bytecnt = (MIN1+1), tmp		//     higher than loopcnt
+	cmp.gt	p_scr, p0 = 16, cnt		// is it a minimalistic task?
+(p_scr)	br.cond.dptk.many .move_bytes_unaligned	// go move just a few (M_B_U)
+;; }
+{ .mmi
+(p_unalgn) add	ptr1 = (MIN1+1), ptr2		// after alignment
+(p_unalgn) add	ptr2 = MIN1P1HALF, ptr2		// after alignment
+(p_unalgn) tbit.nz.unc p_y, p_n = bytecnt, 3	// should we do a st8 ?
+;; }
+{ .mib
+(p_y)	add	cnt = -8, cnt
+(p_unalgn) tbit.nz.unc p_yy, p_nn = bytecnt, 2	// should we do a st4 ?
+} { .mib
+(p_y)	st8	[ptr2] = r0,-4
+(p_n)	add	ptr2 = 4, ptr2
+;; }
+{ .mib
+(p_yy)	add	cnt = -4, cnt
+(p_unalgn) tbit.nz.unc p_y, p_n = bytecnt, 1	// should we do a st2 ?
+} { .mib
+(p_yy)	st4	[ptr2] = r0,-2
+(p_nn)	add	ptr2 = 2, ptr2
+;; }
+{ .mmi
+	mov	tmp = LINE_SIZE+1		// for compare
+(p_y)	add	cnt = -2, cnt
+(p_unalgn) tbit.nz.unc p_yy, p_nn = bytecnt, 0	// should we do a st1 ?
+} { .mmi
+	nop.m	0
+(p_y)	st2	[ptr2] = r0,-1
+(p_n)	add	ptr2 = 1, ptr2
+;; }
+
+{ .mmi
+(p_yy)	st1	[ptr2] = r0
+  	cmp.gt	p_scr, p0 = tmp, cnt		// is it a minimalistic task?
+} { .mbb
+(p_yy)	add	cnt = -1, cnt
+(p_scr)	br.cond.dpnt.many .fraction_of_line	// go move just a few
+;; }
+{ .mib
+	nop.m 	0
+	shr.u	linecnt = cnt, LSIZE_SH
+	nop.b	0
+;; }
+
+	.align 32
+.l1b:	// ------------------//  L1B: store ahead into cache lines; fill later
+{ .mmi
+	and	tmp = -(LINE_SIZE), cnt		// compute end of range
+	mov	ptr9 = ptr1			// used for prefetching
+	and	cnt = (LINE_SIZE-1), cnt	// remainder
+} { .mmi
+	mov	loopcnt = PREF_AHEAD-1		// default prefetch loop
+	cmp.gt	p_scr, p0 = PREF_AHEAD, linecnt	// check against actual value
+;; }
+{ .mmi
+(p_scr)	add	loopcnt = -1, linecnt
+	add	ptr2 = 16, ptr1	// start of stores (beyond prefetch stores)
+	add	ptr1 = tmp, ptr1	// first address beyond total range
+;; }
+{ .mmi
+	add	tmp = -1, linecnt	// next loop count
+	movi0	ar.lc = loopcnt
+;; }
+.pref_l1b:
+{ .mib
+	stf.spill [ptr9] = f0, 128	// Do stores one cache line apart
+	nop.i   0
+	br.cloop.dptk.few .pref_l1b
+;; }
+{ .mmi
+	add	ptr0 = 16, ptr2		// Two stores in parallel
+	movi0	ar.lc = tmp
+;; }
+.l1bx:
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 32
+ ;; }
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 32
+ ;; }
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 64
+ 	cmp.lt	p_scr, p0 = ptr9, ptr1	// do we need more prefetching?
+ ;; }
+{ .mmb
+	stf.spill [ptr2] = f0, 32
+(p_scr)	stf.spill [ptr9] = f0, 128
+	br.cloop.dptk.few .l1bx
+;; }
+{ .mib
+	cmp.gt  p_scr, p0 = 8, cnt	// just a few bytes left ?
+(p_scr)	br.cond.dpnt.many  .move_bytes_from_alignment
+;; }
+
+.fraction_of_line:
+{ .mib
+	add	ptr2 = 16, ptr1
+	shr.u	loopcnt = cnt, 5   	// loopcnt = cnt / 32
+;; }
+{ .mib
+	cmp.eq	p_scr, p0 = loopcnt, r0
+	add	loopcnt = -1, loopcnt
+(p_scr)	br.cond.dpnt.many .store_words
+;; }
+{ .mib
+	and	cnt = 0x1f, cnt		// compute the remaining cnt
+	movi0   ar.lc = loopcnt
+;; }
+	.align 32
+.l2:	// -----------------------------//  L2A:  store 32B in 2 cycles
+{ .mmb
+	store	[ptr1] = myval, 8
+	store	[ptr2] = myval, 8
+;; } { .mmb
+	store	[ptr1] = myval, 24
+	store	[ptr2] = myval, 24
+	br.cloop.dptk.many .l2
+;; }
+.store_words:
+{ .mib
+	cmp.gt	p_scr, p0 = 8, cnt	// just a few bytes left ?
+(p_scr)	br.cond.dpnt.many .move_bytes_from_alignment	// Branch
+;; }
+
+{ .mmi
+	store	[ptr1] = myval, 8	// store
+	cmp.le	p_y, p_n = 16, cnt	//
+	add	cnt = -8, cnt		// subtract
+;; }
+{ .mmi
+(p_y)	store	[ptr1] = myval, 8	// store
+(p_y)	cmp.le.unc p_yy, p_nn = 16, cnt
+(p_y)	add	cnt = -8, cnt		// subtract
+;; }
+{ .mmi					// store
+(p_yy)	store	[ptr1] = myval, 8
+(p_yy)	add	cnt = -8, cnt		// subtract
+;; }
+
+.move_bytes_from_alignment:
+{ .mib
+	cmp.eq	p_scr, p0 = cnt, r0
+	tbit.nz.unc p_y, p0 = cnt, 2	// should we terminate with a st4 ?
+(p_scr)	br.cond.dpnt.few .restore_and_exit
+;; }
+{ .mib
+(p_y)	st4	[ptr1] = r0,4
+	tbit.nz.unc p_yy, p0 = cnt, 1	// should we terminate with a st2 ?
+;; }
+{ .mib
+(p_yy)	st2	[ptr1] = r0,2
+	tbit.nz.unc p_y, p0 = cnt, 0	// should we terminate with a st1 ?
+;; }
+
+{ .mib
+(p_y)	st1	[ptr1] = r0
+;; }
+.restore_and_exit:
+{ .mib
+	nop.m	0
+	movi0	ar.lc = save_lc
+	br.ret.sptk.many rp
+;; }
+
+.move_bytes_unaligned:
+{ .mmi
+       .pred.rel "mutex",p_y, p_n
+       .pred.rel "mutex",p_yy, p_nn
+(p_n)	cmp.le  p_yy, p_nn = 4, cnt
+(p_y)	cmp.le  p_yy, p_nn = 5, cnt
+(p_n)	add	ptr2 = 2, ptr1
+} { .mmi
+(p_y)	add	ptr2 = 3, ptr1
+(p_y)	st1	[ptr1] = r0, 1		// fill 1 (odd-aligned) byte
+(p_y)	add	cnt = -1, cnt		// [15, 14 (or less) left]
+;; }
+{ .mmi
+(p_yy)	cmp.le.unc p_y, p0 = 8, cnt
+	add	ptr3 = ptr1, cnt	// prepare last store
+	movi0	ar.lc = save_lc
+} { .mmi
+(p_yy)	st2	[ptr1] = r0, 4		// fill 2 (aligned) bytes
+(p_yy)	st2	[ptr2] = r0, 4		// fill 2 (aligned) bytes
+(p_yy)	add	cnt = -4, cnt		// [11, 10 (o less) left]
+;; }
+{ .mmi
+(p_y)	cmp.le.unc p_yy, p0 = 8, cnt
+	add	ptr3 = -1, ptr3		// last store
+	tbit.nz p_scr, p0 = cnt, 1	// will there be a st2 at the end ?
+} { .mmi
+(p_y)	st2	[ptr1] = r0, 4		// fill 2 (aligned) bytes
+(p_y)	st2	[ptr2] = r0, 4		// fill 2 (aligned) bytes
+(p_y)	add	cnt = -4, cnt		// [7, 6 (or less) left]
+;; }
+{ .mmi
+(p_yy)	st2	[ptr1] = r0, 4		// fill 2 (aligned) bytes
+(p_yy)	st2	[ptr2] = r0, 4		// fill 2 (aligned) bytes
+					// [3, 2 (or less) left]
+	tbit.nz p_y, p0 = cnt, 0	// will there be a st1 at the end ?
+} { .mmi
+(p_yy)	add	cnt = -4, cnt
+;; }
+{ .mmb
+(p_scr)	st2	[ptr1] = r0		// fill 2 (aligned) bytes
+(p_y)	st1	[ptr3] = r0		// fill last byte (using ptr3)
+	br.ret.sptk.many rp
+;; }
+END(bzero)

libc/string/ia64/memccpy.S

@@ -0,0 +1,213 @@
+/* Optimized version of the memccpy() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: a pointer to the next byte after char in dest or NULL
+
+   Inputs:
+        in0:    dest
+        in1:    src
+  	in2:	char
+        in3:    byte count
+
+   This implementation assumes little endian mode (UM.be = 0).
+
+   This implementation assumes that it is safe to do read ahead
+   in the src block, without getting beyond its limit.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define OP_T_THRES 	16
+#define OPSIZ 		8
+
+#define saved_pr	r17
+#define saved_lc	r18
+#define dest		r19
+#define src		r20
+#define len		r21
+#define asrc		r22
+#define tmp		r23
+#define char		r24
+#define charx8		r25
+#define saved_ec	r26
+#define sh2		r28
+#define	sh1		r29
+#define loopcnt		r30
+#define	value		r31
+
+#ifdef GAS_ALIGN_BREAKS_UNWIND_INFO
+/* Manually force proper loop-alignment.  Note: be sure to
+   double-check the code-layout after making any changes to
+   this routine! */
+# define ALIGN(n)	{ nop 0 }
+#else
+# define ALIGN(n)	.align n
+#endif
+
+ENTRY(memccpy)
+	.prologue
+	alloc 	r2 = ar.pfs, 4, 40 - 4, 0, 40
+
+#include "softpipe.h"
+	.rotr	r[MEMLAT + 7], tmp1[4], tmp2[4], val[4], tmp3[2], pos0[2]
+	.rotp	p[MEMLAT + 6 + 1]
+
+	mov	ret0 = r0		// return NULL if no match
+	.save pr, saved_pr
+	mov	saved_pr = pr		// save the predicate registers
+	mov 	dest = in0		// dest
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc	// save the loop counter
+        mov 	saved_ec = ar.ec	// save the loop counter
+	.body
+	mov 	src = in1		// src
+	extr.u	char = in2, 0, 8	// char
+	mov	len = in3		// len
+	sub	tmp = r0, in0		// tmp = -dest
+	cmp.ne	p7, p0 = r0, r0		// clear p7
+	;;
+	and	loopcnt = 7, tmp	// loopcnt = -dest % 8
+	cmp.ge	p6, p0 = OP_T_THRES, len	// is len <= OP_T_THRES
+	mov	ar.ec = 0		// ec not guaranteed zero on entry
+(p6)	br.cond.spnt	.cpyfew		// copy byte by byte
+	;;
+	cmp.eq	p6, p0 = loopcnt, r0
+	mux1	charx8 = char, @brcst
+(p6)	br.cond.sptk .dest_aligned
+	sub	len = len, loopcnt	// len -= -dest % 8
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	;;
+	mov	ar.lc = loopcnt
+.l1:					// copy -dest % 8 bytes
+	ld1	value = [src], 1	// value = *src++
+	;;
+	st1	[dest] = value, 1	// *dest++ = value
+	cmp.eq	p6, p0 = value, char
+(p6)	br.cond.spnt .foundit
+	br.cloop.dptk .l1
+.dest_aligned:
+	and	sh1 = 7, src 		// sh1 = src % 8
+	and	tmp = -8, len   	// tmp = len & -OPSIZ
+	and	asrc = -8, src		// asrc = src & -OPSIZ  -- align src
+	shr.u	loopcnt = len, 3	// loopcnt = len / 8
+	and	len = 7, len ;;		// len = len % 8
+	shl	sh1 = sh1, 3		// sh1 = 8 * (src % 8)
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	mov     pr.rot = 1 << 16 ;;	// set rotating predicates
+	sub	sh2 = 64, sh1		// sh2 = 64 - sh1
+	mov	ar.lc = loopcnt		// set LC
+	cmp.eq  p6, p0 = sh1, r0 	// is the src aligned?
+(p6)    br.cond.sptk .src_aligned ;;
+	add	src = src, tmp		// src += len & -OPSIZ
+	mov	ar.ec = MEMLAT + 6 + 1 	// six more passes needed
+	ld8	r[1] = [asrc], 8 	// r[1] = w0
+	cmp.ne	p6, p0 = r0, r0	;;	// clear p6
+	ALIGN(32)
+.l2:
+(p[0])		ld8.s	r[0] = [asrc], 8		// r[0] = w1
+(p[MEMLAT])	shr.u	tmp1[0] = r[1 + MEMLAT], sh1	// tmp1 = w0 >> sh1
+(p[MEMLAT])	shl	tmp2[0] = r[0 + MEMLAT], sh2  	// tmp2 = w1 << sh2
+(p[MEMLAT+4])	xor	tmp3[0] = val[1], charx8
+(p[MEMLAT+5])	czx1.r	pos0[0] = tmp3[1]
+(p[MEMLAT+6])	chk.s	r[6 + MEMLAT], .recovery1	// our data isn't
+							// valid - rollback!
+(p[MEMLAT+6])	cmp.ne	p6, p0 = 8, pos0[1]
+(p6)		br.cond.spnt	.gotit
+(p[MEMLAT+6])	st8	[dest] = val[3], 8		// store val to dest
+(p[MEMLAT+3])	or	val[0] = tmp1[3], tmp2[3] 	// val = tmp1 | tmp2
+		br.ctop.sptk    .l2
+		br.cond.sptk .cpyfew
+
+.src_aligned:
+		cmp.ne  p6, p0 = r0, r0			// clear p6
+		mov     ar.ec = MEMLAT + 2 + 1 ;;	// set EC
+.l3:
+(p[0])		ld8.s	r[0] = [src], 8
+(p[MEMLAT])	xor	tmp3[0] = r[MEMLAT], charx8
+(p[MEMLAT+1])	czx1.r	pos0[0] = tmp3[1]
+(p[MEMLAT+2])	cmp.ne	p7, p0 = 8, pos0[1]
+(p[MEMLAT+2])	chk.s	r[MEMLAT+2], .recovery2
+(p7)		br.cond.spnt	.gotit
+.back2:
+(p[MEMLAT+2])	st8	[dest] = r[MEMLAT+2], 8
+		br.ctop.dptk .l3
+.cpyfew:
+	cmp.eq	p6, p0 = len, r0	// is len == 0 ?
+	adds	len = -1, len		// --len;
+(p6)	br.cond.spnt	.restore_and_exit ;;
+	mov	ar.lc = len
+.l4:
+	ld1	value = [src], 1
+	;;
+	st1	[dest] = value, 1
+	cmp.eq	p6, p0 = value, char
+(p6)	br.cond.spnt .foundit
+	br.cloop.dptk	.l4 ;;
+.foundit:
+(p6)	mov	ret0 = dest
+.restore_and_exit:
+	mov     pr = saved_pr, -1    	// restore the predicate registers
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	mov 	ar.ec = saved_ec ;;	// restore the epilog counter
+	br.ret.sptk.many b0
+.gotit:
+	.pred.rel "mutex" p6, p7
+(p6)	mov	value = val[3]		// if coming from l2
+(p7)	mov	value = r[MEMLAT+2]	// if coming from l3
+	mov	ar.lc = pos0[1] ;;
+.l5:
+	extr.u	tmp = value, 0, 8 ;;
+	st1	[dest] = tmp, 1
+	shr.u	value = value, 8
+	br.cloop.sptk .l5 ;;
+	mov 	ret0 = dest
+	mov	pr = saved_pr, -1
+	mov	ar.lc = saved_lc
+	br.ret.sptk.many b0
+
+.recovery1:
+	adds	src = -(MEMLAT + 6 + 1) * 8, asrc
+	mov	loopcnt = ar.lc
+	mov	tmp = ar.ec ;;
+	sub	sh1 = (MEMLAT + 6 + 1), tmp
+	shr.u	sh2 = sh2, 3
+	;; 
+	shl	loopcnt = loopcnt, 3
+	sub	src = src, sh2
+	shl	sh1 = sh1, 3
+	shl	tmp = tmp, 3
+	;;
+	add	len = len, loopcnt
+	add	src = sh1, src ;;
+	add	len = tmp, len
+.back1:
+	br.cond.sptk .cpyfew
+
+.recovery2:
+	add	tmp = -(MEMLAT + 3) * 8, src
+(p7)	br.cond.spnt .gotit
+	;;
+	ld8	r[MEMLAT+2] = [tmp] ;;
+	xor	pos0[1] = r[MEMLAT+2], charx8 ;;
+	czx1.r	pos0[1] = pos0[1] ;;
+	cmp.ne	p7, p6 = 8, pos0[1]
+(p7)	br.cond.spnt .gotit
+	br.cond.sptk .back2
+END(memccpy)

libc/string/ia64/memchr.S

@@ -0,0 +1,133 @@
+/* Optimized version of the standard memchr() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the address of the first occurence of chr in str or NULL
+
+   Inputs:
+  	in0:	str
+  	in1:	chr
+  	in2:	byte count
+
+   This implementation assumes little endian mode.  For big endian mode,
+   the instruction czx1.r should be replaced by czx1.l.
+
+   The algorithm is fairly straightforward: search byte by byte until we
+   we get to a word aligned address, then search word by word as much as
+   possible; the remaining few bytes are searched one at a time.
+
+   The word by word search is performed by xor-ing the word with a word
+   containing chr in every byte.  If there is a hit, the result will
+   contain a zero byte in the corresponding position.  The presence and
+   position of that zero byte is detected with a czx instruction.
+
+   All the loops in this function could have had the internal branch removed
+   if br.ctop and br.cloop could be predicated :-(.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define saved_pr	r15
+#define saved_lc	r16
+#define	chr		r17
+#define len		r18
+#define pos0		r20
+#define val		r21
+#define tmp		r24
+#define chrx8		r25
+#define loopcnt		r30
+
+#define str		in0
+
+ENTRY(__memchr)
+	.prologue
+	alloc r2 = ar.pfs, 3, 0, 29, 32
+#include "softpipe.h"
+	.rotr	value[MEMLAT+1], addr[MEMLAT+3], aux[2], poschr[2]
+	.rotp	p[MEMLAT+3]
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc 	// save the loop counter
+	.save pr, saved_pr
+	mov	saved_pr = pr		// save the predicates
+	.body
+	mov 	ret0 = str
+	and 	tmp = 7, str		// tmp = str % 8
+	cmp.ne	p7, p0 = r0, r0		// clear p7
+	extr.u	chr = in1, 0, 8		// chr = (unsigned char) in1
+	mov	len = in2
+	cmp.gtu	p6, p0 = 16, in2	// use a simple loop for short
+(p6)	br.cond.spnt .srchfew ;;	// searches
+	sub	loopcnt = 8, tmp	// loopcnt = 8 - tmp
+	cmp.eq	p6, p0 = tmp, r0
+(p6)	br.cond.sptk	.str_aligned;;
+	sub	len = len, loopcnt
+	adds	loopcnt = -1, loopcnt;;
+	mov	ar.lc = loopcnt
+.l1:
+	ld1	val = [ret0], 1
+	;;
+	cmp.eq	p6, p0 = val, chr
+(p6)	br.cond.spnt	.foundit
+	br.cloop.sptk	.l1 ;;
+.str_aligned:
+	cmp.ne	p6, p0 = r0, r0		// clear p6
+	shr.u	loopcnt = len, 3	// loopcnt = len / 8
+	and 	len = 7, len ;;		// remaining len = len & 7
+	adds	loopcnt = -1, loopcnt
+	mov	ar.ec = MEMLAT + 3
+	mux1	chrx8 = chr, @brcst ;;	// get a word full of chr
+	mov	ar.lc = loopcnt
+	mov	pr.rot = 1 << 16 ;;
+.l2:
+(p[0])		mov	addr[0] = ret0
+(p[0])		ld8	value[0] = [ret0], 8
+(p[MEMLAT])	xor	aux[0] = value[MEMLAT], chrx8
+(p[MEMLAT+1])	czx1.r	poschr[0] = aux[1]
+(p[MEMLAT+2])	cmp.ne	p7, p0 = 8, poschr[1]
+(p7)		br.cond.dpnt .foundit
+		br.ctop.dptk .l2
+.srchfew:
+	adds	loopcnt = -1, len
+	cmp.eq	p6, p0 = len, r0
+(p6)	br.cond.spnt .notfound ;;
+	mov	ar.lc = loopcnt
+.l3:
+	ld1	val = [ret0], 1
+	;;
+	cmp.eq	p6, p0 = val, chr
+(p6)	br.cond.dpnt	.foundit
+	br.cloop.sptk	.l3 ;;
+.notfound:
+	cmp.ne	p6, p0 = r0, r0	// clear p6 (p7 was already 0 when we got here)
+	mov	ret0 = r0 ;;	// return NULL
+.foundit:
+	.pred.rel "mutex" p6, p7
+(p6)	adds	ret0 = -1, ret0 		   // if we got here from l1 or l3
+(p7)	add	ret0 = addr[MEMLAT+2], poschr[1]   // if we got here from l2
+	mov	pr = saved_pr, -1
+	mov	ar.lc = saved_lc
+	br.ret.sptk.many b0
+
+END(__memchr)
+
+weak_alias (__memchr, memchr)
+#if !__BOUNDED_POINTERS__
+weak_alias (__memchr, __ubp_memchr)
+#endif
+libc_hidden_def (memchr)

libc/string/ia64/memcmp.S

@@ -0,0 +1,165 @@
+/* Optimized version of the standard memcmp() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2004 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the result of the comparison
+
+   Inputs:
+        in0:    dest (aka s1)
+        in1:    src  (aka s2)
+        in2:    byte count
+
+   In this form, it assumes little endian mode.  For big endian mode, the
+   the two shifts in .l2 must be inverted:
+
+	shl   	tmp1[0] = r[1 + MEMLAT], sh1   // tmp1 = w0 << sh1
+	shr.u   tmp2[0] = r[0 + MEMLAT], sh2   // tmp2 = w1 >> sh2
+
+   and all the mux1 instructions should be replaced by plain mov's.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define OP_T_THRES 	16
+#define OPSIZ 		8
+#define MEMLAT		2
+
+#define start		r15
+#define saved_pr	r17
+#define saved_lc	r18
+#define dest		r19
+#define src		r20
+#define len		r21
+#define asrc		r22
+#define tmp		r23
+#define value1		r24
+#define value2		r25
+#define sh2		r28
+#define	sh1		r29
+#define loopcnt		r30
+
+ENTRY(memcmp)
+	.prologue
+	alloc 	r2 = ar.pfs, 3, 37, 0, 40
+
+	.rotr	r[MEMLAT + 2], q[MEMLAT + 5], tmp1[4], tmp2[4], val[2]
+	.rotp	p[MEMLAT + 4 + 1]
+
+	mov	ret0 = r0		// by default return value = 0
+	.save pr, saved_pr
+	mov	saved_pr = pr		// save the predicate registers
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc	// save the loop counter
+	.body
+	mov 	dest = in0		// dest
+	mov 	src = in1		// src
+	mov	len = in2		// len
+	sub	tmp = r0, in0		// tmp = -dest
+	;;
+	and	loopcnt = 7, tmp		// loopcnt = -dest % 8
+	cmp.ge	p6, p0 = OP_T_THRES, len	// is len <= OP_T_THRES
+(p6)	br.cond.spnt	.cmpfew			// compare byte by byte
+	;;
+	cmp.eq	p6, p0 = loopcnt, r0
+(p6)	br.cond.sptk .dest_aligned
+	sub	len = len, loopcnt	// len -= -dest % 8
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	;;
+	mov	ar.lc = loopcnt
+.l1:					// copy -dest % 8 bytes
+	ld1	value1 = [src], 1	// value = *src++
+	ld1	value2 = [dest], 1
+	;;
+	cmp.ne	p6, p0 = value1, value2
+(p6)	br.cond.spnt .done
+	br.cloop.dptk .l1
+.dest_aligned:
+	and	sh1 = 7, src 		// sh1 = src % 8
+	and	tmp = -8, len   	// tmp = len & -OPSIZ
+	and	asrc = -8, src		// asrc = src & -OPSIZ  -- align src
+	shr.u	loopcnt = len, 3	// loopcnt = len / 8
+	and	len = 7, len ;;		// len = len % 8
+	shl	sh1 = sh1, 3		// sh1 = 8 * (src % 8)
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	mov     pr.rot = 1 << 16 ;;	// set rotating predicates
+	sub	sh2 = 64, sh1		// sh2 = 64 - sh1
+	mov	ar.lc = loopcnt		// set LC
+	cmp.eq  p6, p0 = sh1, r0 	// is the src aligned?
+(p6)    br.cond.sptk .src_aligned
+	add	src = src, tmp		// src += len & -OPSIZ
+	mov	ar.ec = MEMLAT + 4 + 1 	// four more passes needed
+	ld8	r[1] = [asrc], 8 ;;	// r[1] = w0
+	.align	32
+
+// We enter this loop with p6 cleared by the above comparison
+
+.l2:
+(p[0])		ld8	r[0] = [asrc], 8		// r[0] = w1
+(p[0])		ld8	q[0] = [dest], 8
+(p[MEMLAT])	shr.u	tmp1[0] = r[1 + MEMLAT], sh1	// tmp1 = w0 >> sh1
+(p[MEMLAT])	shl	tmp2[0] = r[0 + MEMLAT], sh2  	// tmp2 = w1 << sh2
+(p[MEMLAT+4])	cmp.ne	p6, p0 = q[MEMLAT + 4], val[1]
+(p[MEMLAT+3])	or	val[0] = tmp1[3], tmp2[3] 	// val = tmp1 | tmp2
+(p6)		br.cond.spnt .l2exit
+		br.ctop.sptk    .l2
+		br.cond.sptk .cmpfew
+.l3exit:
+	mux1	value1 = r[MEMLAT], @rev
+	mux1	value2 = q[MEMLAT], @rev
+	cmp.ne	p6, p0 = r0, r0	;;	// clear p6
+.l2exit:
+(p6)	mux1	value1 = val[1], @rev
+(p6)	mux1	value2 = q[MEMLAT + 4], @rev ;;
+	cmp.ltu	p6, p7 = value2, value1 ;;
+(p6)	mov	ret0 = -1
+(p7)	mov	ret0 = 1
+	mov     pr = saved_pr, -1    	// restore the predicate registers
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+.src_aligned:
+	cmp.ne	p6, p0 = r0, r0		// clear p6
+	mov     ar.ec = MEMLAT + 1 ;;	// set EC
+.l3:
+(p[0])		ld8	r[0] = [src], 8
+(p[0])		ld8	q[0] = [dest], 8
+(p[MEMLAT])	cmp.ne	p6, p0 = r[MEMLAT], q[MEMLAT]
+(p6)		br.cond.spnt .l3exit
+		br.ctop.dptk .l3 ;;
+.cmpfew:
+	cmp.eq	p6, p0 = len, r0	// is len == 0 ?
+	adds	len = -1, len		// --len;
+(p6)	br.cond.spnt	.restore_and_exit ;;
+	mov	ar.lc = len
+.l4:
+	ld1	value1 = [src], 1
+	ld1	value2 = [dest], 1
+	;;
+	cmp.ne	p6, p0 = value1, value2
+(p6)	br.cond.spnt	.done
+	br.cloop.dptk	.l4 ;;
+.done:
+(p6)	sub	ret0 = value2, value1	// don't execute it if falling thru
+.restore_and_exit:
+	mov     pr = saved_pr, -1    	// restore the predicate registers
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+END(memcmp)
+
+weak_alias (memcmp, bcmp)
+libc_hidden_def (memcmp)

libc/string/ia64/memcpy.S

@@ -0,0 +1,436 @@
+/* Optimized version of the standard memcpy() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop for Itanium <Dan.Pop@cern.ch>.
+   Rewritten for McKinley by Sverre Jarp, HP Labs/CERN <Sverre.Jarp@cern.ch>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+        in0:    dest
+        in1:    src
+        in2:    byte count
+
+   An assembly implementation of the algorithm used by the generic C
+   version from glibc.  The case when source and sest are aligned is
+   treated separately, for extra performance.
+
+   In this form, memcpy assumes little endian mode.  For big endian mode,
+   sh1 must be computed using an extra instruction: sub sh1 = 64, sh1
+   and the order of r[MEMLAT] and r[MEMLAT+1] must be reverted in the
+   shrp instruction.  */
+
+#define USE_LFETCH
+#define USE_FLP
+#include <sysdep.h>
+#undef ret
+
+#define LFETCH_DIST     500
+
+#define ALIGN_UNROLL_no   4 // no. of elements
+#define ALIGN_UNROLL_sh	  2 // (shift amount)
+
+#define MEMLAT	8
+#define Nrot	((4*(MEMLAT+2) + 7) & ~7)
+
+#define OP_T_THRES 	16
+#define OPSIZ 		8
+
+#define loopcnt		r14
+#define elemcnt		r15
+#define saved_pr	r16
+#define saved_lc	r17
+#define adest		r18
+#define dest		r19
+#define asrc		r20
+#define src		r21
+#define len		r22
+#define tmp2		r23
+#define tmp3		r24
+#define	tmp4		r25
+#define ptable		r26
+#define ploop56		r27
+#define	loopaddr	r28
+#define	sh1		r29
+#define ptr1		r30
+#define ptr2		r31
+
+#define movi0 		mov
+
+#define p_scr		p6
+#define p_xtr		p7
+#define p_nxtr		p8
+#define p_few		p9
+
+#if defined(USE_FLP)
+#define load		ldf8
+#define store		stf8
+#define tempreg		f6
+#define the_r		fr
+#define the_s		fs
+#define the_t		ft
+#define the_q		fq
+#define the_w		fw
+#define the_x		fx
+#define the_y		fy
+#define the_z		fz
+#elif defined(USE_INT)
+#define load		ld8
+#define store		st8
+#define tempreg		tmp2
+#define the_r		r
+#define the_s		s
+#define the_t		t
+#define the_q		q
+#define the_w		w
+#define the_x		x
+#define the_y		y
+#define the_z		z
+#endif
+
+#ifdef GAS_ALIGN_BREAKS_UNWIND_INFO
+/* Manually force proper loop-alignment.  Note: be sure to
+   double-check the code-layout after making any changes to
+   this routine! */
+# define ALIGN(n)	{ nop 0 }
+#else
+# define ALIGN(n)	.align n
+#endif
+
+#if defined(USE_LFETCH)
+#define LOOP(shift)						\
+		ALIGN(32);					\
+.loop##shift##:							\
+{ .mmb								\
+(p[0])	ld8.nt1	r[0] = [asrc], 8 ;				\
+(p[0])	lfetch.nt1 [ptr1], 16 ;					\
+	nop.b 0 ;						\
+} { .mib							\
+(p[MEMLAT+1]) st8 [dest] = tmp3, 8 ;				\
+(p[MEMLAT]) shrp tmp3 = r[MEMLAT], s[MEMLAT+1], shift ;		\
+ 	nop.b 0 ;;						\
+ } { .mmb							\
+(p[0])	ld8.nt1	s[0] = [asrc], 8 ;				\
+(p[0])	lfetch.nt1	[ptr2], 16 ;				\
+	nop.b 0 ;						\
+} { .mib							\
+(p[MEMLAT+1]) st8 [dest] = tmp4, 8 ;				\
+(p[MEMLAT]) shrp tmp4 = s[MEMLAT], r[MEMLAT], shift ;		\
+	br.ctop.sptk.many .loop##shift 				\
+;; }								\
+{ .mib								\
+	br.cond.sptk.many .copy_bytes ; /* deal with the remaining bytes */  \
+}
+#else
+#define LOOP(shift)						\
+		ALIGN(32);					\
+.loop##shift##:							\
+{ .mmb								\
+(p[0])	ld8.nt1	r[0] = [asrc], 8 ;				\
+	nop.b 0 ;						\
+} { .mib							\
+(p[MEMLAT+1]) st8 [dest] = tmp3, 8 ;				\
+(p[MEMLAT]) shrp tmp3 = r[MEMLAT], s[MEMLAT+1], shift ;		\
+ 	nop.b 0 ;;						\
+ } { .mmb							\
+(p[0])	ld8.nt1	s[0] = [asrc], 8 ;				\
+	nop.b 0 ;						\
+} { .mib							\
+(p[MEMLAT+1]) st8 [dest] = tmp4, 8 ;				\
+(p[MEMLAT]) shrp tmp4 = s[MEMLAT], r[MEMLAT], shift ;		\
+	br.ctop.sptk.many .loop##shift 				\
+;; }								\
+{ .mib								\
+	br.cond.sptk.many .copy_bytes ; /* deal with the remaining bytes */  \
+}
+#endif
+
+
+ENTRY(memcpy)
+{ .mmi
+	.prologue
+	alloc 	r2 = ar.pfs, 3, Nrot - 3, 0, Nrot
+	.rotr	r[MEMLAT+1], s[MEMLAT+2], q[MEMLAT+1], t[MEMLAT+1]
+	.rotp	p[MEMLAT+2]
+	.rotf	fr[MEMLAT+1], fq[MEMLAT+1], fs[MEMLAT+1], ft[MEMLAT+1]
+	mov	ret0 = in0		// return tmp2 = dest
+	.save   pr, saved_pr
+	movi0	saved_pr = pr		// save the predicate registers
+} { .mmi
+	and	tmp4 = 7, in0 		// check if destination is aligned
+	mov 	dest = in0		// dest
+	mov 	src = in1		// src
+;; }
+{ .mii
+	cmp.eq	p_scr, p0 = in2, r0	// if (len == 0)
+	.save   ar.lc, saved_lc
+        movi0 	saved_lc = ar.lc	// save the loop counter
+	.body
+	cmp.ge	p_few, p0 = OP_T_THRES, in2 // is len <= OP_T_THRESH
+} { .mbb
+	mov	len = in2		// len
+(p_scr)	br.cond.dpnt.few .restore_and_exit // 	Branch no. 1: return dest
+(p_few) br.cond.dpnt.many .copy_bytes	// Branch no. 2: copy byte by byte
+;; }
+{ .mmi
+#if defined(USE_LFETCH)
+	lfetch.nt1 [dest]		//
+	lfetch.nt1 [src]		//
+#endif
+	shr.u	elemcnt = len, 3	// elemcnt = len / 8
+} { .mib
+	cmp.eq	p_scr, p0 = tmp4, r0	// is destination aligned?
+	sub	loopcnt = 7, tmp4	//
+(p_scr) br.cond.dptk.many .dest_aligned
+;; }
+{ .mmi
+	ld1	tmp2 = [src], 1		//
+	sub	len = len, loopcnt, 1	// reduce len
+	movi0	ar.lc = loopcnt		//
+} { .mib
+	cmp.ne  p_scr, p0 = 0, loopcnt	// avoid loading beyond end-point
+;; }
+
+.l0:	// ---------------------------- // L0: Align src on 8-byte boundary
+{ .mmi
+	st1	[dest] = tmp2, 1	//
+(p_scr)	ld1	tmp2 = [src], 1		//
+} { .mib
+	cmp.lt	p_scr, p0 = 1, loopcnt	// avoid load beyond end-point
+	add	loopcnt = -1, loopcnt
+	br.cloop.dptk.few .l0		//
+;; }
+
+.dest_aligned:
+{ .mmi
+	and	tmp4 = 7, src		// ready for alignment check
+	shr.u	elemcnt = len, 3	// elemcnt = len / 8
+;; }
+{ .mib
+	cmp.ne	p_scr, p0 = tmp4, r0	// is source also aligned
+	tbit.nz p_xtr, p_nxtr = src, 3	// prepare a separate move if src
+} { .mib				// is not 16B aligned
+	add	ptr2 = LFETCH_DIST, dest	// prefetch address
+	add	ptr1 = LFETCH_DIST, src
+(p_scr) br.cond.dptk.many .src_not_aligned
+;; }
+
+// The optimal case, when dest, and src are aligned
+
+.both_aligned:
+{ .mmi
+	.pred.rel "mutex",p_xtr,p_nxtr
+(p_xtr)	cmp.gt  p_scr, p0 = ALIGN_UNROLL_no+1, elemcnt // Need N + 1 to qualify
+(p_nxtr) cmp.gt p_scr, p0 = ALIGN_UNROLL_no, elemcnt  // Need only N to qualify
+	movi0	pr.rot = 1 << 16	// set rotating predicates
+} { .mib
+(p_scr) br.cond.dpnt.many .copy_full_words
+;; }
+
+{ .mmi
+(p_xtr)	load	tempreg = [src], 8
+(p_xtr) add 	elemcnt = -1, elemcnt
+	movi0	ar.ec = MEMLAT + 1	// set the epilog counter
+;; }
+{ .mmi
+(p_xtr) add	len = -8, len		//
+	add 	asrc = 16, src 		// one bank apart (for USE_INT)
+	shr.u	loopcnt = elemcnt, ALIGN_UNROLL_sh  // cater for unrolling
+;;}
+{ .mmi
+	add	loopcnt = -1, loopcnt
+(p_xtr)	store	[dest] = tempreg, 8	// copy the "extra" word
+	nop.i	0
+;; }
+{ .mib
+	add	adest = 16, dest
+	movi0	ar.lc = loopcnt 	// set the loop counter
+;; }
+
+#ifdef  GAS_ALIGN_BREAKS_UNWIND_INFO
+	{ nop 0 }
+#else
+	.align	32
+#endif
+#if defined(USE_FLP)
+.l1: // ------------------------------- // L1: Everything a multiple of 8
+{ .mmi
+#if defined(USE_LFETCH)
+(p[0])	lfetch.nt1 [ptr2],32
+#endif
+(p[0])	ldfp8	the_r[0],the_q[0] = [src], 16
+(p[0])	add	len = -32, len
+} {.mmb
+(p[MEMLAT]) store [dest] = the_r[MEMLAT], 8
+(p[MEMLAT]) store [adest] = the_s[MEMLAT], 8
+;; }
+{ .mmi
+#if defined(USE_LFETCH)
+(p[0])	lfetch.nt1 [ptr1],32
+#endif
+(p[0])	ldfp8	the_s[0], the_t[0] = [src], 16
+} {.mmb
+(p[MEMLAT]) store [dest] = the_q[MEMLAT], 24
+(p[MEMLAT]) store [adest] = the_t[MEMLAT], 24
+	br.ctop.dptk.many .l1
+;; }
+#elif defined(USE_INT)
+.l1: // ------------------------------- // L1: Everything a multiple of 8
+{ .mmi
+(p[0])	load	the_r[0] = [src], 8
+(p[0])	load	the_q[0] = [asrc], 8
+(p[0])	add	len = -32, len
+} {.mmb
+(p[MEMLAT]) store [dest] = the_r[MEMLAT], 8
+(p[MEMLAT]) store [adest] = the_q[MEMLAT], 8
+;; }
+{ .mmi
+(p[0])	load	the_s[0]  = [src], 24
+(p[0])	load	the_t[0] = [asrc], 24
+} {.mmb
+(p[MEMLAT]) store [dest] = the_s[MEMLAT], 24
+(p[MEMLAT]) store [adest] = the_t[MEMLAT], 24
+#if defined(USE_LFETCH)
+;; }
+{ .mmb
+(p[0])	lfetch.nt1 [ptr2],32
+(p[0])	lfetch.nt1 [ptr1],32
+#endif
+	br.ctop.dptk.many .l1
+;; }
+#endif
+
+.copy_full_words:
+{ .mib
+	cmp.gt	p_scr, p0 = 8, len	//
+	shr.u	elemcnt = len, 3	//
+(p_scr) br.cond.dpnt.many .copy_bytes
+;; }
+{ .mii
+	load	tempreg = [src], 8
+	add	loopcnt = -1, elemcnt	//
+;; }
+{ .mii
+	cmp.ne	p_scr, p0 = 0, loopcnt	//
+	mov	ar.lc = loopcnt		//
+;; }
+
+.l2: // ------------------------------- // L2: Max 4 words copied separately
+{ .mmi
+	store	[dest] = tempreg, 8
+(p_scr)	load	tempreg = [src], 8	//
+	add	len = -8, len
+} { .mib
+	cmp.lt	p_scr, p0 = 1, loopcnt	// avoid load beyond end-point
+	add	loopcnt = -1, loopcnt
+	br.cloop.dptk.few  .l2
+;; }
+
+.copy_bytes:
+{ .mib
+	cmp.eq	p_scr, p0 = len, r0	// is len == 0 ?
+	add	loopcnt = -1, len	// len--;
+(p_scr)	br.cond.spnt	.restore_and_exit
+;; }
+{ .mii
+	ld1	tmp2 = [src], 1
+	movi0	ar.lc = loopcnt
+	cmp.ne	p_scr, p0 = 0, loopcnt	// avoid load beyond end-point
+;; }
+
+.l3: // ------------------------------- // L3: Final byte move
+{ .mmi
+	st1	[dest] = tmp2, 1
+(p_scr)	ld1	tmp2 = [src], 1
+} { .mib
+	cmp.lt	p_scr, p0 = 1, loopcnt	// avoid load beyond end-point
+	add	loopcnt = -1, loopcnt
+	br.cloop.dptk.few  .l3
+;; }
+
+.restore_and_exit:
+{ .mmi
+	movi0	pr = saved_pr, -1	// restore the predicate registers
+;; }
+{ .mib
+	movi0	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+;; }
+
+
+.src_not_aligned:
+{ .mmi
+	cmp.gt	p_scr, p0 = 16, len
+	and	sh1 = 7, src 		// sh1 = src % 8
+	shr.u	loopcnt = len, 4	// element-cnt = len / 16
+} { .mib
+	add	tmp4 = @ltoff(.table), gp
+	add 	tmp3 = @ltoff(.loop56), gp
+(p_scr)	br.cond.dpnt.many .copy_bytes	// do byte by byte if too few
+;; }
+{ .mmi
+	and	asrc = -8, src		// asrc = (-8) -- align src for loop
+	add 	loopcnt = -1, loopcnt	// loopcnt--
+	shl	sh1 = sh1, 3		// sh1 = 8 * (src % 8)
+} { .mmi
+	ld8	ptable = [tmp4]		// ptable = &table
+	ld8	ploop56 = [tmp3]	// ploop56 = &loop56
+	and	tmp2 = -16, len		// tmp2 = len & -OPSIZ
+;; }
+{ .mmi
+	add	tmp3 = ptable, sh1	// tmp3 = &table + sh1
+	add	src = src, tmp2		// src += len & (-16)
+	movi0	ar.lc = loopcnt		// set LC
+;; }
+{ .mmi
+	ld8	tmp4 = [tmp3]		// tmp4 = loop offset
+	sub	len = len, tmp2		// len -= len & (-16)
+	movi0	ar.ec = MEMLAT + 2 	// one more pass needed
+;; }
+{ .mmi
+	ld8	s[1] = [asrc], 8	// preload
+	sub	loopaddr = ploop56,tmp4	// loopadd = &loop56 - loop offset
+	movi0   pr.rot = 1 << 16	// set rotating predicates
+;; }
+{ .mib
+	nop.m	0
+	movi0	b6 = loopaddr
+	br	b6			// jump to the appropriate loop
+;; }
+
+	LOOP(8)
+	LOOP(16)
+	LOOP(24)
+	LOOP(32)
+	LOOP(40)
+	LOOP(48)
+	LOOP(56)
+END(memcpy)
+libc_hidden_def (memcpy)
+
+	.rodata
+	.align 8
+.table:
+	data8	0			// dummy entry
+	data8 	.loop56 - .loop8
+	data8 	.loop56 - .loop16
+	data8 	.loop56 - .loop24
+	data8	.loop56 - .loop32
+	data8	.loop56 - .loop40
+	data8	.loop56 - .loop48
+	data8	.loop56 - .loop56

libc/string/ia64/memmove.S

@@ -0,0 +1,251 @@
+/* Optimized version of the standard memmove() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+        in0:    dest
+        in1:    src
+        in2:    byte count
+
+   The core of the function is the memcpy implementation used in memcpy.S.
+   When bytes have to be copied backwards, only the easy case, when
+   all arguments are multiples of 8, is optimised.
+
+   In this form, it assumes little endian mode.  For big endian mode,
+   sh1 must be computed using an extra instruction: sub sh1 = 64, sh1
+   or the UM.be bit should be cleared at the beginning and set at the end.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define OP_T_THRES 	16
+#define OPSIZ 		 8
+
+#define adest		r15
+#define saved_pr	r17
+#define saved_lc	r18
+#define dest		r19
+#define src		r20
+#define len		r21
+#define asrc		r22
+#define tmp2		r23
+#define tmp3		r24
+#define	tmp4		r25
+#define ptable		r26
+#define ploop56		r27
+#define	loopaddr	r28
+#define	sh1		r29
+#define loopcnt		r30
+#define	value		r31
+
+#ifdef GAS_ALIGN_BREAKS_UNWIND_INFO
+# define ALIGN(n)	{ nop 0 }
+#else
+# define ALIGN(n)	.align n
+#endif
+
+#define LOOP(shift)							\
+		ALIGN(32);						\
+.loop##shift##:								\
+(p[0])		ld8	r[0] = [asrc], 8 ;	/* w1 */		\
+(p[MEMLAT+1])	st8	[dest] = value, 8 ;				\
+(p[MEMLAT])	shrp	value = r[MEMLAT], r[MEMLAT+1], shift ;		\
+		nop.b	0 ;						\
+		nop.b	0 ;						\
+		br.ctop.sptk .loop##shift ;				\
+		br.cond.sptk .cpyfew ; /* deal with the remaining bytes */
+
+#define MEMLAT	21
+#define Nrot	(((2*MEMLAT+3) + 7) & ~7)
+
+ENTRY(memmove)
+	.prologue
+	alloc 	r2 = ar.pfs, 3, Nrot - 3, 0, Nrot
+	.rotr	r[MEMLAT + 2], q[MEMLAT + 1]
+	.rotp	p[MEMLAT + 2]
+	mov	ret0 = in0		// return value = dest
+	.save pr, saved_pr
+	mov	saved_pr = pr		// save the predicate registers
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc	// save the loop counter
+	.body
+	or	tmp3 = in0, in1 ;;	// tmp3 = dest | src
+	or	tmp3 = tmp3, in2	// tmp3 = dest | src | len
+	mov 	dest = in0		// dest
+	mov 	src = in1		// src
+	mov	len = in2		// len
+	sub	tmp2 = r0, in0		// tmp2 = -dest
+	cmp.eq	p6, p0 = in2, r0	// if (len == 0)
+(p6)	br.cond.spnt .restore_and_exit;;// 	return dest;
+	and	tmp4 = 7, tmp3 		// tmp4 = (dest | src | len) & 7
+	cmp.le	p6, p0 = dest, src	// if dest <= src it's always safe
+(p6)	br.cond.spnt .forward		// to copy forward
+	add	tmp3 = src, len;;
+	cmp.lt	p6, p0 = dest, tmp3	// if dest > src && dest < src + len
+(p6)	br.cond.spnt .backward		// we have to copy backward
+
+.forward:
+	shr.u	loopcnt = len, 4 ;;	// loopcnt = len / 16
+	cmp.ne	p6, p0 = tmp4, r0	// if ((dest | src | len) & 7 != 0)
+(p6)	br.cond.sptk .next		//	goto next;
+
+// The optimal case, when dest, src and len are all multiples of 8
+
+	and	tmp3 = 0xf, len
+	mov	pr.rot = 1 << 16	// set rotating predicates
+	mov	ar.ec = MEMLAT + 1 ;;	// set the epilog counter
+	cmp.ne	p6, p0 = tmp3, r0	// do we have to copy an extra word?
+	adds	loopcnt = -1, loopcnt;;	// --loopcnt
+(p6)	ld8	value = [src], 8;;
+(p6)	st8	[dest] = value, 8	// copy the "odd" word
+	mov	ar.lc = loopcnt 	// set the loop counter
+	cmp.eq	p6, p0 = 8, len
+(p6)	br.cond.spnt .restore_and_exit;;// the one-word special case
+	adds	adest = 8, dest		// set adest one word ahead of dest
+	adds	asrc = 8, src ;;	// set asrc one word ahead of src
+	nop.b	0			// get the "golden" alignment for
+	nop.b	0			// the next loop
+.l0:
+(p[0])		ld8	r[0] = [src], 16
+(p[0])		ld8	q[0] = [asrc], 16
+(p[MEMLAT])	st8	[dest] = r[MEMLAT], 16
+(p[MEMLAT])	st8	[adest] = q[MEMLAT], 16
+		br.ctop.dptk .l0 ;;
+
+	mov	pr = saved_pr, -1	// restore the predicate registers
+	mov	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+.next:
+	cmp.ge	p6, p0 = OP_T_THRES, len	// is len <= OP_T_THRES
+	and	loopcnt = 7, tmp2 		// loopcnt = -dest % 8
+(p6)	br.cond.spnt	.cpyfew			// copy byte by byte
+	;;
+	cmp.eq	p6, p0 = loopcnt, r0
+(p6)	br.cond.sptk	.dest_aligned
+	sub	len = len, loopcnt	// len -= -dest % 8
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	;;
+	mov	ar.lc = loopcnt
+.l1:					// copy -dest % 8 bytes
+	ld1	value = [src], 1	// value = *src++
+	;;
+	st1	[dest] = value, 1	// *dest++ = value
+	br.cloop.dptk .l1
+.dest_aligned:
+	and	sh1 = 7, src 		// sh1 = src % 8
+	and	tmp2 = -8, len   	// tmp2 = len & -OPSIZ
+	and	asrc = -8, src		// asrc = src & -OPSIZ  -- align src
+	shr.u	loopcnt = len, 3	// loopcnt = len / 8
+	and	len = 7, len;;		// len = len % 8
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	addl	tmp4 = @ltoff(.table), gp
+	addl	tmp3 = @ltoff(.loop56), gp
+	mov     ar.ec = MEMLAT + 1	// set EC
+	mov     pr.rot = 1 << 16;;	// set rotating predicates
+	mov	ar.lc = loopcnt		// set LC
+	cmp.eq  p6, p0 = sh1, r0 	// is the src aligned?
+(p6)    br.cond.sptk .src_aligned
+	add	src = src, tmp2		// src += len & -OPSIZ
+	shl	sh1 = sh1, 3		// sh1 = 8 * (src % 8)
+	ld8	ploop56 = [tmp3]	// ploop56 = &loop56
+	ld8	ptable = [tmp4];;	// ptable = &table
+	add	tmp3 = ptable, sh1;;	// tmp3 = &table + sh1
+	mov	ar.ec = MEMLAT + 1 + 1 // one more pass needed
+	ld8	tmp4 = [tmp3];;		// tmp4 = loop offset
+	sub	loopaddr = ploop56,tmp4	// loopadd = &loop56 - loop offset
+	ld8	r[1] = [asrc], 8;;	// w0
+	mov	b6 = loopaddr;;
+	br	b6			// jump to the appropriate loop
+
+	LOOP(8)
+	LOOP(16)
+	LOOP(24)
+	LOOP(32)
+	LOOP(40)
+	LOOP(48)
+	LOOP(56)
+
+.src_aligned:
+.l3:
+(p[0])		ld8	r[0] = [src], 8
+(p[MEMLAT])	st8	[dest] = r[MEMLAT], 8
+		br.ctop.dptk .l3
+.cpyfew:
+	cmp.eq	p6, p0 = len, r0	// is len == 0 ?
+	adds	len = -1, len		// --len;
+(p6)	br.cond.spnt	.restore_and_exit ;;
+	mov	ar.lc = len
+.l4:
+	ld1	value = [src], 1
+	;;
+	st1	[dest] = value, 1
+	br.cloop.dptk	.l4 ;;
+.restore_and_exit:
+	mov     pr = saved_pr, -1    	// restore the predicate registers
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+
+// In the case of a backward copy, optimise only the case when everything
+// is a multiple of 8, otherwise copy byte by byte.  The backward copy is
+// used only when the blocks are overlapping and dest > src.
+
+.backward:
+	shr.u	loopcnt = len, 3	// loopcnt = len / 8
+	add	src = src, len		// src points one byte past the end
+	add	dest = dest, len ;; 	// dest points one byte past the end
+	mov	ar.ec = MEMLAT + 1	// set the epilog counter
+	mov	pr.rot = 1 << 16	// set rotating predicates
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+	cmp.ne	p6, p0 = tmp4, r0	// if ((dest | src | len) & 7 != 0)
+(p6)	br.cond.sptk .bytecopy ;;	// copy byte by byte backward
+	adds	src = -8, src		// src points to the last word
+	adds	dest = -8, dest 	// dest points to the last word
+	mov	ar.lc = loopcnt;;	// set the loop counter
+.l5:
+(p[0])		ld8	r[0] = [src], -8
+(p[MEMLAT])	st8	[dest] = r[MEMLAT], -8
+		br.ctop.dptk .l5
+		br.cond.sptk .restore_and_exit
+.bytecopy:
+	adds	src = -1, src		// src points to the last byte
+	adds	dest = -1, dest		// dest points to the last byte
+	adds	loopcnt = -1, len;;	// loopcnt = len - 1
+	mov	ar.lc = loopcnt;;	// set the loop counter
+.l6:
+(p[0])		ld1	r[0] = [src], -1
+(p[MEMLAT])	st1	[dest] = r[MEMLAT], -1
+		br.ctop.dptk .l6
+		br.cond.sptk .restore_and_exit
+END(memmove)
+
+	.rodata
+	.align 8
+.table:
+	data8	0			// dummy entry
+	data8 	.loop56 - .loop8
+	data8 	.loop56 - .loop16
+	data8 	.loop56 - .loop24
+	data8	.loop56 - .loop32
+	data8	.loop56 - .loop40
+	data8	.loop56 - .loop48
+	data8	.loop56 - .loop56
+
+libc_hidden_def (memmove)

libc/string/ia64/memset.S

@@ -0,0 +1,400 @@
+/* Optimized version of the standard memset() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2002, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop for Itanium <Dan.Pop@cern.ch>.
+   Rewritten for McKinley by Sverre Jarp, HP Labs/CERN <Sverre.Jarp@cern.ch>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+        in0:    dest
+        in1:    value
+        in2:    count
+
+   The algorithm is fairly straightforward: set byte by byte until we
+   we get to a 16B-aligned address, then loop on 128 B chunks using an
+   early store as prefetching, then loop on 32B chucks, then clear remaining
+   words, finally clear remaining bytes.
+   Since a stf.spill f0 can store 16B in one go, we use this instruction
+   to get peak speed when value = 0.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define dest		in0
+#define value		in1
+#define	cnt		in2
+
+#define tmp		r31
+#define save_lc		r30
+#define ptr0		r29
+#define ptr1		r28
+#define ptr2		r27
+#define ptr3		r26
+#define ptr9 		r24
+#define	loopcnt		r23
+#define linecnt		r22
+#define bytecnt		r21
+
+#define fvalue		f6
+
+// This routine uses only scratch predicate registers (p6 - p15)
+#define p_scr		p6			// default register for same-cycle branches
+#define p_nz		p7
+#define p_zr		p8
+#define p_unalgn	p9
+#define p_y		p11
+#define p_n		p12
+#define p_yy		p13
+#define p_nn		p14
+
+#define movi0		mov
+
+#define MIN1		15
+#define MIN1P1HALF	8
+#define LINE_SIZE	128
+#define LSIZE_SH        7			// shift amount
+#define PREF_AHEAD	8
+
+#define USE_FLP
+#if defined(USE_INT)
+#define store		st8
+#define myval           value
+#elif defined(USE_FLP)
+#define store		stf8
+#define myval		fvalue
+#endif
+
+.align	64
+ENTRY(memset)
+{ .mmi
+	.prologue
+	alloc	tmp = ar.pfs, 3, 0, 0, 0
+	lfetch.nt1 [dest]
+	.save   ar.lc, save_lc
+	movi0	save_lc = ar.lc
+} { .mmi
+	.body
+	mov	ret0 = dest		// return value
+	cmp.ne	p_nz, p_zr = value, r0	// use stf.spill if value is zero
+	cmp.eq	p_scr, p0 = cnt, r0
+;; }
+{ .mmi
+	and	ptr2 = -(MIN1+1), dest	// aligned address
+	and	tmp = MIN1, dest	// prepare to check for alignment
+	tbit.nz p_y, p_n = dest, 0	// Do we have an odd address? (M_B_U)
+} { .mib
+	mov	ptr1 = dest
+	mux1	value = value, @brcst	// create 8 identical bytes in word
+(p_scr)	br.ret.dpnt.many rp		// return immediately if count = 0
+;; }
+{ .mib
+	cmp.ne	p_unalgn, p0 = tmp, r0
+} { .mib				// NB: # of bytes to move is 1 higher
+	sub	bytecnt = (MIN1+1), tmp	//     than loopcnt
+	cmp.gt	p_scr, p0 = 16, cnt		// is it a minimalistic task?
+(p_scr)	br.cond.dptk.many .move_bytes_unaligned	// go move just a few (M_B_U)
+;; }
+{ .mmi
+(p_unalgn) add	ptr1 = (MIN1+1), ptr2		// after alignment
+(p_unalgn) add	ptr2 = MIN1P1HALF, ptr2		// after alignment
+(p_unalgn) tbit.nz.unc p_y, p_n = bytecnt, 3	// should we do a st8 ?
+;; }
+{ .mib
+(p_y)	add	cnt = -8, cnt
+(p_unalgn) tbit.nz.unc p_yy, p_nn = bytecnt, 2	// should we do a st4 ?
+} { .mib
+(p_y)	st8	[ptr2] = value, -4
+(p_n)	add	ptr2 = 4, ptr2
+;; }
+{ .mib
+(p_yy)	add	cnt = -4, cnt
+(p_unalgn) tbit.nz.unc p_y, p_n = bytecnt, 1	// should we do a st2 ?
+} { .mib
+(p_yy)	st4	[ptr2] = value, -2
+(p_nn)	add	ptr2 = 2, ptr2
+;; }
+{ .mmi
+	mov	tmp = LINE_SIZE+1		// for compare
+(p_y)	add	cnt = -2, cnt
+(p_unalgn) tbit.nz.unc p_yy, p_nn = bytecnt, 0	// should we do a st1 ?
+} { .mmi
+	setf.sig fvalue=value			// transfer value to FLP side
+(p_y)	st2	[ptr2] = value, -1
+(p_n)	add	ptr2 = 1, ptr2
+;; }
+
+{ .mmi
+(p_yy)	st1	[ptr2] = value
+  	cmp.gt	p_scr, p0 = tmp, cnt		// is it a minimalistic task?
+} { .mbb
+(p_yy)	add	cnt = -1, cnt
+(p_scr)	br.cond.dpnt.many .fraction_of_line	// go move just a few
+;; }
+
+{ .mib
+	nop.m 0
+	shr.u	linecnt = cnt, LSIZE_SH
+(p_zr)	br.cond.dptk.many .l1b			// Jump to use stf.spill
+;; }
+
+#ifndef GAS_ALIGN_BREAKS_UNWIND_INFO
+	.align 32 // -------- //  L1A: store ahead into cache lines; fill later
+#endif
+{ .mmi
+	and	tmp = -(LINE_SIZE), cnt		// compute end of range
+	mov	ptr9 = ptr1			// used for prefetching
+	and	cnt = (LINE_SIZE-1), cnt	// remainder
+} { .mmi
+	mov	loopcnt = PREF_AHEAD-1		// default prefetch loop
+	cmp.gt	p_scr, p0 = PREF_AHEAD, linecnt	// check against actual value
+;; }
+{ .mmi
+(p_scr)	add	loopcnt = -1, linecnt		// start of stores
+	add	ptr2 = 8, ptr1			// (beyond prefetch stores)
+	add	ptr1 = tmp, ptr1		// first address beyond total
+;; }						// range
+{ .mmi
+	add	tmp = -1, linecnt		// next loop count
+	movi0	ar.lc = loopcnt
+;; }
+.pref_l1a:
+{ .mib
+	store [ptr9] = myval, 128	// Do stores one cache line apart
+	nop.i	0
+	br.cloop.dptk.few .pref_l1a
+;; }
+{ .mmi
+	add	ptr0 = 16, ptr2		// Two stores in parallel
+	movi0	ar.lc = tmp
+;; }
+.l1ax:
+ { .mmi
+	store [ptr2] = myval, 8
+	store [ptr0] = myval, 8
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 24
+	store [ptr0] = myval, 24
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 8
+	store [ptr0] = myval, 8
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 24
+	store [ptr0] = myval, 24
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 8
+	store [ptr0] = myval, 8
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 24
+	store [ptr0] = myval, 24
+ ;; }
+ { .mmi
+	store [ptr2] = myval, 8
+	store [ptr0] = myval, 32
+ 	cmp.lt	p_scr, p0 = ptr9, ptr1		// do we need more prefetching?
+ ;; }
+{ .mmb
+	store [ptr2] = myval, 24
+(p_scr)	store [ptr9] = myval, 128
+	br.cloop.dptk.few .l1ax
+;; }
+{ .mbb
+	cmp.le  p_scr, p0 = 8, cnt		// just a few bytes left ?
+(p_scr) br.cond.dpnt.many  .fraction_of_line	// Branch no. 2
+	br.cond.dpnt.many  .move_bytes_from_alignment	// Branch no. 3
+;; }
+
+#ifdef GAS_ALIGN_BREAKS_UNWIND_INFO
+	{ nop 0 }
+#else
+	.align 32
+#endif
+.l1b:	// ------------------ //  L1B: store ahead into cache lines; fill later
+{ .mmi
+	and	tmp = -(LINE_SIZE), cnt		// compute end of range
+	mov	ptr9 = ptr1			// used for prefetching
+	and	cnt = (LINE_SIZE-1), cnt	// remainder
+} { .mmi
+	mov	loopcnt = PREF_AHEAD-1		// default prefetch loop
+	cmp.gt	p_scr, p0 = PREF_AHEAD, linecnt	// check against actual value
+;; }
+{ .mmi
+(p_scr)	add	loopcnt = -1, linecnt
+	add	ptr2 = 16, ptr1	// start of stores (beyond prefetch stores)
+	add	ptr1 = tmp, ptr1	// first address beyond total range
+;; }
+{ .mmi
+	add	tmp = -1, linecnt	// next loop count
+	movi0	ar.lc = loopcnt
+;; }
+.pref_l1b:
+{ .mib
+	stf.spill [ptr9] = f0, 128	// Do stores one cache line apart
+	nop.i   0
+	br.cloop.dptk.few .pref_l1b
+;; }
+{ .mmi
+	add	ptr0 = 16, ptr2		// Two stores in parallel
+	movi0	ar.lc = tmp
+;; }
+.l1bx:
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 32
+ ;; }
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 32
+ ;; }
+ { .mmi
+	stf.spill [ptr2] = f0, 32
+	stf.spill [ptr0] = f0, 64
+ 	cmp.lt	p_scr, p0 = ptr9, ptr1	// do we need more prefetching?
+ ;; }
+{ .mmb
+	stf.spill [ptr2] = f0, 32
+(p_scr)	stf.spill [ptr9] = f0, 128
+	br.cloop.dptk.few .l1bx
+;; }
+{ .mib
+	cmp.gt  p_scr, p0 = 8, cnt	// just a few bytes left ?
+(p_scr)	br.cond.dpnt.many  .move_bytes_from_alignment
+;; }
+
+.fraction_of_line:
+{ .mib
+	add	ptr2 = 16, ptr1
+	shr.u	loopcnt = cnt, 5   	// loopcnt = cnt / 32
+;; }
+{ .mib
+	cmp.eq	p_scr, p0 = loopcnt, r0
+	add	loopcnt = -1, loopcnt
+(p_scr)	br.cond.dpnt.many store_words
+;; }
+{ .mib
+	and	cnt = 0x1f, cnt		// compute the remaining cnt
+	movi0   ar.lc = loopcnt
+;; }
+#ifndef GAS_ALIGN_BREAKS_UNWIND_INFO
+	.align 32
+#endif
+.l2:	// ---------------------------- //  L2A:  store 32B in 2 cycles
+{ .mmb
+	store	[ptr1] = myval, 8
+	store	[ptr2] = myval, 8
+;; } { .mmb
+	store	[ptr1] = myval, 24
+	store	[ptr2] = myval, 24
+	br.cloop.dptk.many .l2
+;; }
+store_words:
+{ .mib
+	cmp.gt	p_scr, p0 = 8, cnt		// just a few bytes left ?
+(p_scr)	br.cond.dpnt.many .move_bytes_from_alignment	// Branch
+;; }
+
+{ .mmi
+	store	[ptr1] = myval, 8		// store
+	cmp.le	p_y, p_n = 16, cnt		//
+	add	cnt = -8, cnt			// subtract
+;; }
+{ .mmi
+(p_y)	store	[ptr1] = myval, 8		// store
+(p_y)	cmp.le.unc p_yy, p_nn = 16, cnt		//
+(p_y)	add	cnt = -8, cnt			// subtract
+;; }
+{ .mmi						// store
+(p_yy)	store	[ptr1] = myval, 8		//
+(p_yy)	add	cnt = -8, cnt			// subtract
+;; }
+
+.move_bytes_from_alignment:
+{ .mib
+	cmp.eq	p_scr, p0 = cnt, r0
+	tbit.nz.unc p_y, p0 = cnt, 2	// should we terminate with a st4 ?
+(p_scr)	br.cond.dpnt.few .restore_and_exit
+;; }
+{ .mib
+(p_y)	st4	[ptr1] = value, 4
+	tbit.nz.unc p_yy, p0 = cnt, 1	// should we terminate with a st2 ?
+;; }
+{ .mib
+(p_yy)	st2	[ptr1] = value, 2
+	tbit.nz.unc p_y, p0 = cnt, 0
+;; }
+
+{ .mib
+(p_y)	st1	[ptr1] = value
+;; }
+.restore_and_exit:
+{ .mib
+	nop.m	0
+	movi0	ar.lc = save_lc
+	br.ret.sptk.many rp
+;; }
+
+.move_bytes_unaligned:
+{ .mmi
+       .pred.rel "mutex",p_y, p_n
+       .pred.rel "mutex",p_yy, p_nn
+(p_n)	cmp.le  p_yy, p_nn = 4, cnt
+(p_y)	cmp.le  p_yy, p_nn = 5, cnt
+(p_n)	add	ptr2 = 2, ptr1
+} { .mmi
+(p_y)	add	ptr2 = 3, ptr1
+(p_y)	st1	[ptr1] = value, 1	// fill 1 (odd-aligned) byte
+(p_y)	add	cnt = -1, cnt		// [15, 14 (or less) left]
+;; }
+{ .mmi
+(p_yy)	cmp.le.unc p_y, p0 = 8, cnt
+	add	ptr3 = ptr1, cnt	// prepare last store
+	movi0	ar.lc = save_lc
+} { .mmi
+(p_yy)	st2	[ptr1] = value, 4	// fill 2 (aligned) bytes
+(p_yy)	st2	[ptr2] = value, 4	// fill 2 (aligned) bytes
+(p_yy)	add	cnt = -4, cnt		// [11, 10 (o less) left]
+;; }
+{ .mmi
+(p_y)	cmp.le.unc p_yy, p0 = 8, cnt
+	add	ptr3 = -1, ptr3		// last store
+	tbit.nz p_scr, p0 = cnt, 1	// will there be a st2 at the end ?
+} { .mmi
+(p_y)	st2	[ptr1] = value, 4	// fill 2 (aligned) bytes
+(p_y)	st2	[ptr2] = value, 4	// fill 2 (aligned) bytes
+(p_y)	add	cnt = -4, cnt		// [7, 6 (or less) left]
+;; }
+{ .mmi
+(p_yy)	st2	[ptr1] = value, 4	// fill 2 (aligned) bytes
+(p_yy)	st2	[ptr2] = value, 4	// fill 2 (aligned) bytes
+					// [3, 2 (or less) left]
+	tbit.nz p_y, p0 = cnt, 0	// will there be a st1 at the end ?
+} { .mmi
+(p_yy)	add	cnt = -4, cnt
+;; }
+{ .mmb
+(p_scr)	st2	[ptr1] = value		// fill 2 (aligned) bytes
+(p_y)	st1	[ptr3] = value		// fill last byte (using ptr3)
+	br.ret.sptk.many rp
+;; }
+END(memset)
+libc_hidden_def (memset)

libc/string/ia64/softpipe.h

@@ -0,0 +1,29 @@
+/* This file is part of the GNU C Library.
+   Copyright (C) 2000 Free Software Foundation, Inc.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* The latency of a memory load assumed by the assembly implementation
+   of the mem and str functions.  Since we don't have any clue about
+   where the data might be, let's assume it's in the L2 cache.  
+   Assuming L3 would be too pessimistic :-)
+
+   Some functions define MEMLAT as 2, because they expect their data
+   to be in the L1D cache.  */
+
+#ifndef MEMLAT
+# define MEMLAT 6
+#endif

libc/string/ia64/strchr.S

@@ -0,0 +1,112 @@
+/* Optimized version of the standard strchr() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the address of the first occurence of chr in str or NULL
+
+   Inputs:
+        in0:    str
+        in1:    chr
+
+   A modified version of memchr.S, the search ends when the character is
+   found or the terminating null character is encountered.
+
+   This implementation assumes little endian mode.  For big endian mode,
+   the instruction czx1.r should be replaced by czx1.l.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define saved_lc	r18
+#define poschr		r19
+#define pos0		r20
+#define val1		r21
+#define val2		r22
+#define tmp		r24
+#define chrx8		r25
+#define loopcnt		r30
+
+#define str		in0
+#define chr		in1
+
+ENTRY(strchr)
+	.prologue
+	alloc r2 = ar.pfs, 2, 0, 0, 0
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc 	// save the loop counter
+	.body
+	mov 	ret0 = str	
+	and 	tmp = 7, str		// tmp = str % 8
+	mux1	chrx8 = chr, @brcst
+	extr.u	chr = chr, 0, 8		// retain only the last byte
+	cmp.ne	p8, p0 = r0, r0		// clear p8
+	;;
+	sub	loopcnt = 8, tmp	// loopcnt = 8 - tmp
+	cmp.eq	p6, p0 = tmp, r0
+(p6)	br.cond.sptk	.str_aligned;;
+	adds	loopcnt = -1, loopcnt;;
+	mov	ar.lc = loopcnt
+.l1:
+	ld1	val2 = [ret0], 1
+	;;
+	cmp.eq	p6, p0 = val2, chr
+	cmp.eq	p7, p0 = val2, r0
+(p6)	br.cond.spnt	.restore_and_exit
+(p7)	br.cond.spnt	.notfound
+	br.cloop.sptk	.l1
+.str_aligned:
+	ld8	val1 = [ret0], 8;;
+	nop.b	0
+	nop.b 	0
+.l2:	
+	ld8.s	val2 = [ret0], 8	// don't bomb out here
+	czx1.r	pos0 = val1	
+	xor	tmp = val1, chrx8	// if val1 contains chr, tmp will
+	;;				// contain a zero in its position
+	czx1.r	poschr = tmp
+	cmp.ne	p6, p0 = 8, pos0
+	;;
+	cmp.ne	p7, p0 = 8, poschr
+(p7)	br.cond.spnt .foundit
+(p6)	br.cond.spnt .notfound
+	chk.s	val2, .recovery
+.back:
+	mov	val1 = val2	
+	br.cond.dptk .l2
+.foundit:
+(p6)	cmp.lt	p8, p0 = pos0, poschr	// we found chr and null in the word
+(p8)	br.cond.spnt .notfound		// null was found before chr
+	add	ret0 = ret0, poschr ;;
+	adds	ret0 = -15, ret0 ;;	// should be -16, but we decrement
+.restore_and_exit:			// ret0 in the next instruction
+	adds	ret0 = -1, ret0		// ret0 was pointing 1 char too far
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	br.ret.sptk.many b0
+.notfound:
+	mov	ret0 = r0		// return NULL if null was found
+	mov 	ar.lc = saved_lc
+	br.ret.sptk.many b0
+.recovery:
+	adds	ret0 = -8, ret0;;
+	ld8	val2 = [ret0], 8	// bomb out here
+	br.cond.sptk	.back
+END(strchr)
+
+weak_alias (strchr, index)
+libc_hidden_def (strchr)

libc/string/ia64/strcmp.S

@@ -0,0 +1,54 @@
+/* Optimized version of the standard strcmp() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the result of the comparison
+
+   Inputs:
+        in0:    s1
+        in1:    s2
+
+   Unlike memcmp(), this function is optimized for mismatches within the
+   first few characters.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define s1		in0
+#define s2		in1
+
+#define val1		r15
+#define val2		r16
+
+
+ENTRY(strcmp)
+	alloc	r2 = ar.pfs, 2, 0, 0, 0
+.loop:
+	ld1	val1 = [s1], 1
+	ld1	val2 = [s2], 1
+	cmp.eq	p6, p0 = r0, r0		// set p6
+	;;
+	cmp.ne.and p6, p0 = val1, r0
+	cmp.ne.and p6, p0 = val2, r0
+	cmp.eq.and p6, p0 = val1, val2
+(p6)	br.cond.sptk .loop
+	sub	ret0 = val1, val2
+	br.ret.sptk.many b0
+END(strcmp)
+libc_hidden_def (strcmp)

libc/string/ia64/strcpy.S

@@ -0,0 +1,145 @@
+/* Optimized version of the standard strcpy() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+        in0:    dest
+        in1:    src
+
+   In this form, it assumes little endian mode.  For big endian mode, the
+   the two shifts in .l2 must be inverted:
+
+	shl   	value = r[1], sh1   	// value = w0 << sh1
+	shr.u   tmp = r[0], sh2   	// tmp = w1 >> sh2
+ */
+
+#include <sysdep.h>
+#undef ret
+
+#define saved_lc	r15
+#define saved_pr	r16
+#define thresh		r17
+#define dest		r19
+#define src		r20
+#define len		r21
+#define asrc		r22
+#define tmp		r23
+#define pos		r24
+#define w0		r25
+#define w1		r26
+#define c		r27
+#define sh2		r28
+#define	sh1		r29
+#define loopcnt		r30
+#define	value		r31
+
+ENTRY(strcpy)
+	.prologue
+	alloc 	r2 = ar.pfs, 2, 0, 30, 32
+
+#define MEMLAT 2
+	.rotr	r[MEMLAT + 2]
+	.rotp	p[MEMLAT + 1]
+
+	mov	ret0 = in0		// return value = dest
+	.save pr, saved_pr
+	mov	saved_pr = pr           // save the predicate registers
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc	// save the loop counter
+	.body
+	sub	tmp = r0, in0 ;;	// tmp = -dest
+	mov 	dest = in0		// dest
+	mov 	src = in1		// src
+	and	loopcnt = 7, tmp ;;	// loopcnt = -dest % 8
+	cmp.eq	p6, p0 = loopcnt, r0
+	adds	loopcnt = -1, loopcnt	// --loopcnt
+(p6)	br.cond.sptk .dest_aligned ;;
+	mov	ar.lc = loopcnt
+.l1:					// copy -dest % 8 bytes
+	ld1	c = [src], 1		// c = *src++
+	;;
+	st1	[dest] = c, 1		// *dest++ = c
+	cmp.eq	p6, p0 = c, r0
+(p6)	br.cond.dpnt .restore_and_exit
+	br.cloop.dptk .l1 ;;
+.dest_aligned:
+	and	sh1 = 7, src 		// sh1 = src % 8
+	mov	ar.lc = -1		// "infinite" loop
+	and	asrc = -8, src ;;	// asrc = src & -OPSIZ  -- align src
+	sub	thresh = 8, sh1
+	mov	pr.rot = 1 << 16	// set rotating predicates
+	cmp.ne	p7, p0 = r0, r0		// clear p7
+	shl	sh1 = sh1, 3 ;;		// sh1 = 8 * (src % 8)
+	sub	sh2 = 64, sh1		// sh2 = 64 - sh1
+	cmp.eq  p6, p0 = sh1, r0 	// is the src aligned?
+(p6)    br.cond.sptk .src_aligned ;;
+	ld8	r[1] = [asrc],8 ;;
+
+	.align	32
+.l2:
+	ld8.s	r[0] = [asrc], 8
+	shr.u	value = r[1], sh1 ;; 	// value = w0 >> sh1
+	czx1.r	pos = value ;;		// do we have an "early" zero
+	cmp.lt	p7, p0 = pos, thresh	// in w0 >> sh1?
+(p7)	br.cond.dpnt .found0
+	chk.s	r[0], .recovery2	// it is safe to do that only
+.back2:					// after the previous test
+	shl	tmp = r[0], sh2  	// tmp = w1 << sh2
+	;;
+	or	value = value, tmp ;;	// value |= tmp
+	czx1.r	pos = value ;;
+	cmp.ne	p7, p0 = 8, pos
+(p7)	br.cond.dpnt .found0
+	st8	[dest] = value, 8	// store val to dest
+	br.ctop.dptk    .l2 ;;
+.src_aligned:
+.l3:
+(p[0])		ld8.s	r[0] = [src], 8
+(p[MEMLAT])	chk.s	r[MEMLAT], .recovery3
+.back3:
+(p[MEMLAT])	mov	value = r[MEMLAT]
+(p[MEMLAT])	czx1.r	pos = r[MEMLAT] ;;
+(p[MEMLAT])	cmp.ne	p7, p0 = 8, pos
+(p7)		br.cond.dpnt .found0
+(p[MEMLAT])	st8	[dest] = r[MEMLAT], 8
+		br.ctop.dptk .l3 ;;
+.found0:
+	mov	ar.lc = pos
+.l4:
+	extr.u	c = value, 0, 8		// c = value & 0xff
+	shr.u	value = value, 8
+	;;
+	st1	[dest] = c, 1
+	br.cloop.dptk	.l4 ;;
+.restore_and_exit:
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	mov	pr = saved_pr, -1	// restore the predicate registers
+	br.ret.sptk.many b0
+.recovery2:
+	add	tmp = -8, asrc ;;
+	ld8	r[0] = [tmp]
+	br.cond.sptk .back2
+.recovery3:
+	add	tmp = -(MEMLAT + 1) * 8, src ;;
+	ld8	r[MEMLAT] = [tmp]
+	br.cond.sptk .back3
+END(strcpy)
+libc_hidden_def (strcpy)

libc/string/ia64/strlen.S

@@ -0,0 +1,98 @@
+/* Optimized version of the standard strlen() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003, 2005 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the length of the input string
+
+   Input:
+        in0:    str
+
+   Look for the null character byte by byte, until we reach a word aligned
+   address, then search word by word, using the czx instruction.  We're
+   also doing one word of read ahead, which could cause problems if the
+   null character is on the last word of a page and the next page is not
+   mapped in the process address space.  Hence the use of the speculative
+   load.
+
+   This implementation assumes little endian mode.  For big endian mode,
+   the instruction czx1.r should be replaced by czx1.l.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define saved_lc	r18
+#define str		r19
+#define pos0		r20
+#define val1		r21
+#define val2		r22
+#define origadd		r23
+#define tmp		r24
+#define loopcnt		r30
+#define len		ret0
+
+ENTRY(strlen)
+	.prologue
+	alloc r2 = ar.pfs, 1, 0, 0, 0
+	.save ar.lc, saved_lc
+        mov 	saved_lc = ar.lc 	// save the loop counter
+	.body
+	mov 	str = in0	
+	mov 	len = r0		// len = 0
+	and 	tmp = 7, in0		// tmp = str % 8
+	;;
+	sub	loopcnt = 8, tmp	// loopcnt = 8 - tmp
+	cmp.eq	p6, p0 = tmp, r0
+(p6)	br.cond.sptk	.str_aligned;;
+	adds	loopcnt = -1, loopcnt;;
+	mov	ar.lc = loopcnt
+.l1:
+	ld1	val2 = [str], 1
+	;;
+	cmp.eq	p6, p0 = val2, r0
+(p6)	br.cond.spnt	.restore_and_exit
+	adds	len = 1, len
+	br.cloop.dptk	.l1
+.str_aligned:
+	mov	origadd = str		// origadd = orig
+	ld8	val1 = [str], 8;;
+	nop.b	0
+	nop.b 	0
+.l2:	ld8.s	val2 = [str], 8		// don't bomb out here
+	czx1.r	pos0 = val1	
+	;;
+	cmp.ne	p6, p0 = 8, pos0
+(p6)	br.cond.spnt .foundit
+	chk.s	val2, .recovery
+.back:
+	mov	val1 = val2	
+	br.cond.dptk	.l2
+.foundit:
+	sub	tmp = str, origadd	// tmp = crt address - orig
+	add	len = len, pos0;;
+	add	len = len, tmp;;
+	adds	len = -16, len
+.restore_and_exit:
+	mov ar.lc = saved_lc		// restore the loop counter
+	br.ret.sptk.many b0
+.recovery:
+	adds	str = -8, str;;
+	ld8	val2 = [str], 8		// bomb out here
+	br.cond.sptk	.back
+END(strlen)
+libc_hidden_def (strlen)

libc/string/ia64/strncmp.S

@@ -0,0 +1,62 @@
+/* Optimized version of the standard strncmp() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: the result of the comparison
+
+   Inputs:
+        in0:    s1
+        in1:    s2
+  	in2:	n
+
+   Unlike memcmp(), this function is optimized for mismatches within the
+   first few characters.  */
+
+#include <sysdep.h>
+#undef ret
+
+#define s1		in0
+#define s2		in1
+#define n		in2
+
+#define val1		r15
+#define val2		r16
+
+
+ENTRY(strncmp)
+	alloc	r2 = ar.pfs, 3, 0, 0, 0
+	mov	ret0 = r0
+	cmp.eq  p6, p0 = r0, r0		// set p6
+	cmp.eq	p7, p0 = n, r0		// return immediately if n == 0
+(p7)	br.cond.spnt .restore_and_exit ;;
+.loop:
+	ld1	val1 = [s1], 1
+	ld1	val2 = [s2], 1
+	adds	n = -1, n		// n--
+	;;
+	cmp.ne.and p6, p0 = val1, r0
+	cmp.ne.and p6, p0 = val2, r0
+	cmp.ne.and p6, p0 = n, r0
+	cmp.eq.and p6, p0 = val1, val2
+(p6)	br.cond.sptk .loop
+	sub	ret0 = val1, val2
+.restore_and_exit:
+	br.ret.sptk.many b0
+END(strncmp)	
+libc_hidden_def (strncmp)

libc/string/ia64/strncpy.S

@@ -0,0 +1,232 @@
+/* Optimized version of the standard strncpy() function.
+   This file is part of the GNU C Library.
+   Copyright (C) 2000, 2001, 2002, 2003 Free Software Foundation, Inc.
+   Contributed by Dan Pop <Dan.Pop@cern.ch>
+	      and Jakub Jelinek <jakub@redhat.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* Return: dest
+
+   Inputs:
+	in0:    dest
+	in1:    src
+	in2:	len
+
+   In this form, it assumes little endian mode.
+ */
+
+#include <sysdep.h>
+#undef ret
+
+#define saved_lc	r15
+#define saved_pr	r16
+#define thresh		r17
+#define dest		r18
+#define dest2		r19
+#define src		r20
+#define len		r21
+#define asrc		r22
+#define tmp		r23
+#define pos		r24
+#define w0		r25
+#define w1		r26
+#define c		r27
+#define sh2		r28
+#define	sh1		r29
+#define loopcnt		r30
+#define	value		r31
+
+ENTRY(strncpy)
+	.prologue
+	alloc 	r2 = ar.pfs, 3, 0, 29, 32
+
+#define MEMLAT 2
+	.rotr	r[MEMLAT + 2]
+	.rotp	p[MEMLAT + 1]
+
+	mov	ret0 = in0		// return value = dest
+	.save pr, saved_pr
+	mov	saved_pr = pr           // save the predicate registers
+	.save ar.lc, saved_lc
+	mov 	saved_lc = ar.lc	// save the loop counter
+	mov	ar.ec = 0		// ec is not guaranteed to
+					// be zero upon function entry
+	.body
+	cmp.geu p6, p5 = 24, in2
+(p6)	br.cond.spnt .short_len
+	sub	tmp = r0, in0 ;;	// tmp = -dest
+	mov	len = in2		// len
+	mov 	dest = in0		// dest
+	mov 	src = in1		// src
+	and	tmp = 7, tmp ;;		// loopcnt = -dest % 8
+	cmp.eq	p6, p7 = tmp, r0
+	adds	loopcnt = -1, tmp	// --loopcnt
+(p6)	br.cond.sptk .dest_aligned ;;
+	sub	len = len, tmp		// len -= -dest % 8
+	mov	ar.lc = loopcnt
+.l1:					// copy -dest % 8 bytes
+(p5)	ld1	c = [src], 1		// c = *src++
+	;;
+	st1	[dest] = c, 1		// *dest++ = c
+	cmp.ne	p5, p7 = c, r0
+	br.cloop.dptk .l1 ;;
+(p7)	br.cond.dpnt	.found0_align
+
+.dest_aligned:				// p7 should be cleared here
+	shr.u	c = len, 3		// c = len / 8
+	and	sh1 = 7, src 		// sh1 = src % 8
+	and	asrc = -8, src ;;	// asrc = src & -OPSIZ  -- align src
+	adds	c = (MEMLAT-1), c	// c = (len / 8) + MEMLAT - 1
+	sub	thresh = 8, sh1
+	mov	pr.rot = 1 << 16	// set rotating predicates
+	shl	sh1 = sh1, 3 ;;		// sh1 = 8 * (src % 8)
+	mov	ar.lc = c		// "infinite" loop
+	sub	sh2 = 64, sh1		// sh2 = 64 - sh1
+	cmp.eq  p6, p0 = sh1, r0 	// is the src aligned?
+(p6)    br.cond.sptk .src_aligned
+	adds	c = -(MEMLAT-1), c ;;	// c = (len / 8)
+	ld8	r[1] = [asrc],8
+	mov	ar.lc = c ;;
+
+	.align	32
+.l2:
+(p6)	st8	[dest] = value, 8	// store val to dest
+	ld8.s	r[0] = [asrc], 8
+	shr.u	value = r[1], sh1 ;; 	// value = w0 >> sh1
+	czx1.r	pos = value ;;		// do we have an "early" zero
+	cmp.lt	p7, p0 = pos, thresh	// in w0 >> sh1?
+	adds	len = -8, len		// len -= 8
+(p7)	br.cond.dpnt .nonalign_found0
+	chk.s	r[0], .recovery2	// it is safe to do that only
+.back2:					// after the previous test
+	shl	tmp = r[0], sh2  	// tmp = w1 << sh2
+	;;
+	or	value = value, tmp ;;	// value |= tmp
+	czx1.r	pos = value ;;
+	cmp.ne	p7, p6 = 8, pos
+(p7)	br.cond.dpnt .nonalign_found0
+	br.ctop.dptk    .l2 ;;
+	adds	len = 8, len
+	br.cond.sptk	.not_found0 ;;
+.nonalign_found0:
+	cmp.gtu	p6, p0 = -8, len
+(p6)	br.cond.dptk .found0
+	adds	len = 8, len
+	br.cond.sptk	.not_found0 ;;
+
+	.align	32
+.src_aligned:
+.l3:
+(p[0])		ld8.s	r[0] = [src], 8
+(p[MEMLAT])	chk.s	r[MEMLAT], .recovery3
+.back3:
+(p[MEMLAT])	mov	value = r[MEMLAT]
+(p[MEMLAT])	czx1.r	pos = r[MEMLAT] ;;
+(p[MEMLAT])	cmp.ne	p7, p0 = 8, pos
+(p[MEMLAT])	adds	len = -8, len	// len -= 8
+(p7)		br.cond.dpnt .found0
+(p[MEMLAT])	st8	[dest] = r[MEMLAT], 8
+		br.ctop.dptk .l3 ;;
+
+	chk.s	r[MEMLAT-1], .recovery4
+.back4:
+	mov	value = r[MEMLAT-1]
+
+.not_found0:
+	cmp.eq	p5, p6 = len, r0
+	adds	len = -1, len
+(p5)	br.cond.dptk	.restore_and_exit ;;
+	mov	ar.lc = len
+.l4:
+(p6)	extr.u	c = value, 0, 8		// c = value & 0xff
+(p6)	shr.u	value = value, 8 ;;
+	st1	[dest] = c, 1
+	cmp.ne	p6, p0 = c, r0
+	br.cloop.dptk	.l4
+	br.cond.sptk	.restore_and_exit
+
+.found0_align:
+	mov	pos = 0
+	adds	len = -8, len
+	mov	value = 0 ;;
+.found0:
+	shl	tmp = pos, 3
+	shr.u	loopcnt = len, 4	// loopcnt = len / 16
+	mov	c = -1 ;;
+	cmp.eq	p6, p0 = loopcnt, r0
+	adds	loopcnt = -1, loopcnt
+	shl	c = c, tmp ;;
+	and	len = 0xf, len
+	andcm	value = value, c
+	mov	ar.lc = loopcnt ;;
+	cmp.le	p7, p0 = 8, len
+	adds	dest2 = 16, dest
+	st8	[dest] = value, 8
+	and	len = 0x7, len
+(p6)	br.cond.dpnt	.l6 ;;
+.l5:
+	st8	[dest] = r0, 16
+	st8	[dest2] = r0, 16
+	br.cloop.dptk	.l5 ;;
+.l6:
+(p7)	st8	[dest] = r0, 8
+	cmp.eq	p5, p0 = len, r0
+	adds	len = -1, len
+(p5)	br.cond.dptk .restore_and_exit ;;
+	mov	ar.lc = len ;;
+.l7:
+	st1	[dest] = r0, 1
+	br.cloop.dptk	.l7 ;;
+.restore_and_exit:
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	mov	pr = saved_pr, -1	// restore the predicate registers
+	br.ret.sptk.many b0
+
+.short_len:
+	cmp.eq	p5, p0 = in2, r0
+	adds	loopcnt = -1, in2
+(p5)	br.cond.spnt .restore_and_exit ;;
+	mov	ar.lc = loopcnt		// p6 should be set when we get here
+.l8:
+(p6)	ld1	c = [in1], 1		// c = *src++
+	;;
+	st1	[in0] = c, 1		// *dest++ = c
+(p6)	cmp.ne	p6, p0 = c, r0
+	br.cloop.dptk .l8
+	;;
+	mov 	ar.lc = saved_lc	// restore the loop counter
+	mov	pr = saved_pr, -1	// restore the predicate registers
+	br.ret.sptk.many b0
+.recovery2:
+	add	c = 8, len
+	add	tmp = -8, asrc ;;
+	cmp.gtu	p8, p5 = c, thresh ;;
+(p8)	ld8	r[0] = [tmp]
+(p5)	mov	r[0] = r0
+	br.cond.sptk .back2
+.recovery3:
+	add	tmp = -(MEMLAT + 1) * 8, src ;;
+	ld8	r[MEMLAT] = [tmp]
+	br.cond.sptk .back3
+.recovery4:
+	cmp.eq	p5, p6 = len, r0
+	add	tmp = -MEMLAT * 8, src ;;
+(p6)	ld8	r[MEMLAT - 1] = [tmp]
+(p5)	mov	r[MEMLAT - 1] = r0
+	br.cond.sptk .back4
+END(strncpy)
+libc_hidden_def (strncpy)

libc/string/ia64/sysdep.h

@@ -0,0 +1,168 @@
+/* Copyright (C) 1999, 2000, 2002, 2003, 2004 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Written by Jes Sorensen, <Jes.Sorensen@cern.ch>, April 1999.
+   Based on code originally written by David Mosberger-Tang
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _LINUX_IA64_SYSDEP_H
+#define _LINUX_IA64_SYSDEP_H 1
+
+#include <features.h>
+#include <asm/unistd.h>
+
+#ifdef __ASSEMBLER__
+
+/* Macros to help writing .prologue directives in assembly code.  */
+#define ASM_UNW_PRLG_RP			0x8
+#define ASM_UNW_PRLG_PFS		0x4
+#define ASM_UNW_PRLG_PSP		0x2
+#define ASM_UNW_PRLG_PR			0x1
+#define ASM_UNW_PRLG_GRSAVE(ninputs)	(32+(ninputs))
+
+#ifdef	__STDC__
+#define C_LABEL(name)		name##:
+#else
+#define C_LABEL(name)		name/**/:
+#endif
+
+#define CALL_MCOUNT
+
+#define ENTRY(name)				\
+	.text;					\
+	.align 32;				\
+	.proc C_SYMBOL_NAME(name);		\
+	.global C_SYMBOL_NAME(name);		\
+	C_LABEL(name)				\
+	CALL_MCOUNT
+
+#define LEAF(name)				\
+  .text;					\
+  .align 32;					\
+  .proc C_SYMBOL_NAME(name);			\
+  .global name;					\
+  C_LABEL(name)
+
+/* Mark the end of function SYM.  */
+#undef END
+#define END(sym)	.endp C_SYMBOL_NAME(sym)
+
+/* For Linux we can use the system call table in the header file
+	/usr/include/asm/unistd.h
+   of the kernel.  But these symbols do not follow the SYS_* syntax
+   so we have to redefine the `SYS_ify' macro here.  */
+#undef SYS_ify
+#ifdef __STDC__
+# define SYS_ify(syscall_name)	__NR_##syscall_name
+#else
+# define SYS_ify(syscall_name)	__NR_/**/syscall_name
+#endif
+
+/* Linux uses a negative return value to indicate syscall errors, unlike
+   most Unices, which use the condition codes' carry flag.
+
+   Since version 2.1 the return value of a system call might be negative
+   even if the call succeeded.  E.g., the `lseek' system call might return
+   a large offset.  Therefore we must not anymore test for < 0, but test
+   for a real error by making sure the value in %d0 is a real error
+   number.  Linus said he will make sure the no syscall returns a value
+   in -1 .. -4095 as a valid result so we can savely test with -4095.  */
+
+/* We don't want the label for the error handler to be visible in the symbol
+   table when we define it here.  */
+#define SYSCALL_ERROR_LABEL __syscall_error
+
+#undef PSEUDO
+#define	PSEUDO(name, syscall_name, args)	\
+  ENTRY(name)					\
+    DO_CALL (SYS_ify(syscall_name));		\
+	cmp.eq p6,p0=-1,r10;			\
+(p6)	br.cond.spnt.few __syscall_error;
+
+#define DO_CALL_VIA_BREAK(num)			\
+	mov r15=num;				\
+	break __BREAK_SYSCALL
+
+#ifdef IA64_USE_NEW_STUB
+# ifdef SHARED
+#  define DO_CALL(num)				\
+	.prologue;				\
+	adds r2 = SYSINFO_OFFSET, r13;;		\
+	ld8 r2 = [r2];				\
+	.save ar.pfs, r11;			\
+	mov r11 = ar.pfs;;			\
+	.body;					\
+	mov r15 = num;				\
+	mov b7 = r2;				\
+	br.call.sptk.many b6 = b7;;		\
+	.restore sp;				\
+	mov ar.pfs = r11;			\
+	.prologue;				\
+	.body
+# else /* !SHARED */
+#  define DO_CALL(num)				\
+	.prologue;				\
+	mov r15 = num;				\
+	movl r2 = _dl_sysinfo;;			\
+	ld8 r2 = [r2];				\
+	.save ar.pfs, r11;			\
+	mov r11 = ar.pfs;;			\
+	.body;					\
+	mov b7 = r2;				\
+	br.call.sptk.many b6 = b7;;		\
+	.restore sp;				\
+	mov ar.pfs = r11;			\
+	.prologue;				\
+	.body
+# endif
+#else
+# define DO_CALL(num)				DO_CALL_VIA_BREAK(num)
+#endif
+
+#undef PSEUDO_END
+#define PSEUDO_END(name)	.endp C_SYMBOL_NAME(name);
+
+#undef PSEUDO_NOERRNO
+#define	PSEUDO_NOERRNO(name, syscall_name, args)	\
+  ENTRY(name)						\
+    DO_CALL (SYS_ify(syscall_name));
+
+#undef PSEUDO_END_NOERRNO
+#define PSEUDO_END_NOERRNO(name)	.endp C_SYMBOL_NAME(name);
+
+#undef PSEUDO_ERRVAL
+#define	PSEUDO_ERRVAL(name, syscall_name, args)	\
+  ENTRY(name)					\
+    DO_CALL (SYS_ify(syscall_name));		\
+	cmp.eq p6,p0=-1,r10;			\
+(p6)	mov r10=r8;
+
+
+#undef PSEUDO_END_ERRVAL
+#define PSEUDO_END_ERRVAL(name)	.endp C_SYMBOL_NAME(name);
+
+#undef END
+#define END(name)						\
+	.size	C_SYMBOL_NAME(name), . - C_SYMBOL_NAME(name) ;	\
+	.endp	C_SYMBOL_NAME(name)
+
+#define ret			br.ret.sptk.few b0
+#define ret_NOERRNO		ret
+#define ret_ERRVAL		ret
+
+#endif /* not __ASSEMBLER__ */
+
+#endif /* linux/ia64/sysdep.h */

libc/sysdeps/linux/ia64/Makefile

@@ -0,0 +1,15 @@
+# Makefile for uClibc
+#
+# Copyright (C) 2000-2005 Erik Andersen <andersen@uclibc.org>
+#
+# Licensed under the LGPL v2.1, see the file COPYING.LIB in this tarball.
+#
+
+TOPDIR=../../../../
+
+top_srcdir=$(TOPDIR)
+top_builddir=../../../../
+all: objs
+include $(top_builddir)Rules.mak
+include Makefile.arch
+include $(top_srcdir)Makerules

libc/sysdeps/linux/ia64/Makefile.arch

@@ -0,0 +1,14 @@
+# Makefile for uClibc
+#
+# Copyright (C) 2000-2005 Erik Andersen <andersen@uclibc.org>
+#
+# Licensed under the LGPL v2.1, see the file COPYING.LIB in this tarball.
+#
+
+CSRC := __syscall_error.c
+
+SSRC := \
+	__longjmp.S brk.S bsd-setjmp.S bsd-_setjmp.S clone2.S fork.S \
+	pipe.S setjmp.S syscall.S vfork.S
+
+include $(top_srcdir)libc/sysdeps/linux/Makefile.commonarch

libc/sysdeps/linux/ia64/__longjmp.S

@@ -0,0 +1,160 @@
+/* Copyright (C) 1999, 2000, 2001 Free Software Foundation, Inc.
+   Contributed by David Mosberger-Tang <davidm@hpl.hp.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.
+
+   Note that __sigsetjmp() did NOT flush the register stack.  Instead,
+   we do it here since __longjmp() is usually much less frequently
+   invoked than __sigsetjmp(). The only difficulty is that __sigsetjmp()
+   didn't (and wouldn't be able to) save ar.rnat either.  This is a problem
+   because if we're not careful, we could end up loading random NaT bits.
+   There are two cases:
+
+	(i)  ar.bsp < ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+		ar.rnat contains the desired bits---preserve ar.rnat
+		across loadrs and write to ar.bspstore
+
+	(ii) ar.bsp >= ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+		The desired ar.rnat is stored in
+		ia64_rse_rnat_addr(jmpbuf.ar_bsp).  Load those
+		bits into ar.rnat after setting ar.bspstore. */
+
+#include "sysdep.h"
+#include <features.h>
+
+#	define	pPos	p6	/* is rotate count positive? */
+#	define	pNeg	p7	/* is rotate count negative? */
+
+
+	/* __longjmp(__jmp_buf buf, int val) */
+
+LEAF(__longjmp)
+	alloc r8=ar.pfs,2,1,0,0
+	mov r27=ar.rsc
+	add r2=0x98,in0		// r2 <- &jmpbuf.orig_jmp_buf_addr
+	;;
+	ld8 r8=[r2],-16		// r8 <- orig_jmp_buf_addr
+	mov r10=ar.bsp
+	and r11=~0x3,r27	// clear ar.rsc.mode
+	;;
+	flushrs			// flush dirty regs to backing store (must be first in insn grp)
+	ld8 r23=[r2],8		// r23 <- jmpbuf.ar_bsp
+	sub r8=r8,in0		// r8 <- &orig_jmpbuf - &jmpbuf
+	;;
+	ld8 r25=[r2]		// r25 <- jmpbuf.ar_unat
+	extr.u r8=r8,3,6	// r8 <- (&orig_jmpbuf - &jmpbuf)/8 & 0x3f
+	;;
+	cmp.lt pNeg,pPos=r8,r0
+	mov r2=in0
+	;;
+(pPos)	mov r16=r8
+(pNeg)	add r16=64,r8
+(pPos)	sub r17=64,r8
+(pNeg)	sub r17=r0,r8
+	;;
+	mov ar.rsc=r11		// put RSE in enforced lazy mode
+	shr.u r8=r25,r16
+	add r3=8,in0		// r3 <- &jmpbuf.r1
+	shl r9=r25,r17
+	;;
+	or r25=r8,r9
+	;;
+	mov r26=ar.rnat
+	mov ar.unat=r25		// setup ar.unat (NaT bits for r1, r4-r7, and r12)
+	;;
+	ld8.fill.nta sp=[r2],16	// r12 (sp)
+	ld8.fill.nta gp=[r3],16		// r1 (gp)
+	dep r11=-1,r23,3,6	// r11 <- ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+	;;
+	ld8.nta r16=[r2],16		// caller's unat
+	ld8.nta r17=[r3],16		// fpsr
+	;;
+	ld8.fill.nta r4=[r2],16	// r4
+	ld8.fill.nta r5=[r3],16		// r5 (gp)
+	cmp.geu p8,p0=r10,r11	// p8 <- (ar.bsp >= jmpbuf.ar_bsp)
+	;;
+	ld8.fill.nta r6=[r2],16	// r6
+	ld8.fill.nta r7=[r3],16		// r7
+	;;
+	mov ar.unat=r16			// restore caller's unat
+	mov ar.fpsr=r17			// restore fpsr
+	;;
+	ld8.nta r16=[r2],16		// b0
+	ld8.nta r17=[r3],16		// b1
+	;;
+(p8)	ld8 r26=[r11]		// r26 <- *ia64_rse_rnat_addr(jmpbuf.ar_bsp)
+	mov ar.bspstore=r23	// restore ar.bspstore
+	;;
+	ld8.nta r18=[r2],16		// b2
+	ld8.nta r19=[r3],16		// b3
+	;;
+	ld8.nta r20=[r2],16		// b4
+	ld8.nta r21=[r3],16		// b5
+	;;
+	ld8.nta r11=[r2],16		// ar.pfs
+	ld8.nta r22=[r3],56		// ar.lc
+	;;
+	ld8.nta r24=[r2],32		// pr
+	mov b0=r16
+	;;
+	ldf.fill.nta f2=[r2],32
+	ldf.fill.nta f3=[r3],32
+	mov b1=r17
+	;;
+	ldf.fill.nta f4=[r2],32
+	ldf.fill.nta f5=[r3],32
+	mov b2=r18
+	;;
+	ldf.fill.nta f16=[r2],32
+	ldf.fill.nta f17=[r3],32
+	mov b3=r19
+	;;
+	ldf.fill.nta f18=[r2],32
+	ldf.fill.nta f19=[r3],32
+	mov b4=r20
+	;;
+	ldf.fill.nta f20=[r2],32
+	ldf.fill.nta f21=[r3],32
+	mov b5=r21
+	;;
+	ldf.fill.nta f22=[r2],32
+	ldf.fill.nta f23=[r3],32
+	mov ar.lc=r22
+	;;
+	ldf.fill.nta f24=[r2],32
+	ldf.fill.nta f25=[r3],32
+	cmp.eq p8,p9=0,in1
+	;;
+	ldf.fill.nta f26=[r2],32
+	ldf.fill.nta f27=[r3],32
+	mov ar.pfs=r11
+	;;
+	ldf.fill.nta f28=[r2],32
+	ldf.fill.nta f29=[r3],32
+	;;
+	ldf.fill.nta f30=[r2]
+	ldf.fill.nta f31=[r3]
+(p8)	mov r8=1
+
+	mov ar.rnat=r26		// restore ar.rnat
+	;;
+	mov ar.rsc=r27		// restore ar.rsc
+(p9)	mov r8=in1
+
+	invala			// virt. -> phys. regnum mapping may change
+	mov pr=r24,-1
+	ret
+END(__longjmp)

libc/sysdeps/linux/ia64/__syscall_error.c

@@ -0,0 +1,19 @@
+/* Wrapper for setting errno.
+ *
+ * Copyright (C) 2000-2006 Erik Andersen <andersen@uclibc.org>
+ *
+ * Licensed under the LGPL v2.1, see the file COPYING.LIB in this tarball.
+ */
+
+#include <errno.h>
+#include <features.h>
+
+/* This routine is jumped to by all the syscall handlers, to stash
+ * an error number into errno.  */
+int __syscall_error(void) attribute_hidden;
+int __syscall_error(void)
+{
+	register int err_no asm("%r8");
+	__set_errno(-err_no);
+	return -1;
+}

libc/sysdeps/linux/ia64/bits/endian.h

@@ -0,0 +1,7 @@
+/* Linux/ia64 is little-endian.  */
+
+#ifndef _ENDIAN_H
+# error "Never use <bits/endian.h> directly; include <endian.h> instead."
+#endif
+
+#define __BYTE_ORDER __LITTLE_ENDIAN

libc/sysdeps/linux/ia64/bits/fcntl.h

@@ -0,0 +1,183 @@
+/* O_*, F_*, FD_* bit values for Linux/IA64.
+   Copyright (C) 1999, 2000, 2004 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef	_FCNTL_H
+# error "Never use <bits/fcntl.h> directly; include <fcntl.h> instead."
+#endif
+
+
+#include <sys/types.h>
+
+/* open/fcntl - O_SYNC is only implemented on blocks devices and on files
+   located on an ext2 file system */
+#define O_ACCMODE	  0003
+#define O_RDONLY	    00
+#define O_WRONLY	    01
+#define O_RDWR		    02
+#define O_CREAT		  0100	/* not fcntl */
+#define O_EXCL		  0200	/* not fcntl */
+#define O_NOCTTY	  0400	/* not fcntl */
+#define O_TRUNC		 01000	/* not fcntl */
+#define O_APPEND	 02000
+#define O_NONBLOCK	 04000
+#define O_NDELAY	O_NONBLOCK
+#define O_SYNC		010000
+#define O_FSYNC		O_SYNC
+#define O_ASYNC		020000
+#define O_DIRECT	040000
+
+#ifdef __USE_GNU
+# define O_DIRECTORY	0200000	/* must be a directory */
+# define O_NOFOLLOW	0400000 /* don't follow links */
+# define O_NOATIME	01000000 /* Do not set atime.  */
+#endif
+
+#ifdef __USE_LARGEFILE64
+/* Not necessary, files are always with 64bit off_t.  */
+# define O_LARGEFILE	0
+#endif
+
+/* For now Linux has synchronisity options for data and read operations.
+   We define the symbols here but let them do the same as O_SYNC since
+   this is a superset.	*/
+#if defined __USE_POSIX199309 || defined __USE_UNIX98
+# define O_DSYNC	O_SYNC	/* Synchronize data.  */
+# define O_RSYNC	O_SYNC	/* Synchronize read operations.	 */
+#endif
+
+/* Values for the second argument to `fcntl'.  */
+#define F_DUPFD		0	/* Duplicate file descriptor.  */
+#define F_GETFD		1	/* Get file descriptor flags.  */
+#define F_SETFD		2	/* Set file descriptor flags.  */
+#define F_GETFL		3	/* Get file status flags.  */
+#define F_SETFL		4	/* Set file status flags.  */
+#define F_GETLK		5	/* Get record locking info.  */
+#define F_SETLK		6	/* Set record locking info (non-blocking).  */
+#define F_SETLKW	7	/* Set record locking info (blocking).	*/
+
+/* Not necessary, we always have 64-bit offsets.  */
+#define F_GETLK64	5	/* Get record locking info.  */
+#define F_SETLK64	6	/* Set record locking info (non-blocking).  */
+#define F_SETLKW64	7	/* Set record locking info (blocking).	*/
+
+#if defined __USE_BSD || defined __USE_UNIX98
+# define F_SETOWN	8	/* Get owner of socket (receiver of SIGIO).  */
+# define F_GETOWN	9	/* Set owner of socket (receiver of SIGIO).  */
+#endif
+
+#ifdef __USE_GNU
+# define F_SETSIG	10	/* Set number of signal to be sent.  */
+# define F_GETSIG	11	/* Get number of signal to be sent.  */
+#endif
+
+#ifdef __USE_GNU
+# define F_SETLEASE	1024	/* Set a lease.	 */
+# define F_GETLEASE	1025	/* Enquire what lease is active.  */
+# define F_NOTIFY	1026	/* Request notfications on a directory.	 */
+#endif
+
+/* For F_[GET|SET]FL.  */
+#define FD_CLOEXEC	1	/* actually anything with low bit set goes */
+
+/* For posix fcntl() and `l_type' field of a `struct flock' for lockf().  */
+#define F_RDLCK		0	/* Read lock.  */
+#define F_WRLCK		1	/* Write lock.	*/
+#define F_UNLCK		2	/* Remove lock.	 */
+
+/* for old implementation of bsd flock () */
+#define F_EXLCK		4	/* or 3 */
+#define F_SHLCK		8	/* or 4 */
+
+#ifdef __USE_BSD
+/* Operations for bsd flock(), also used by the kernel implementation */
+# define LOCK_SH	1	/* shared lock */
+# define LOCK_EX	2	/* exclusive lock */
+# define LOCK_NB	4	/* or'd with one of the above to prevent
+				   blocking */
+# define LOCK_UN	8	/* remove lock */
+#endif
+
+#ifdef __USE_GNU
+# define LOCK_MAND	32	/* This is a mandatory flock:	*/
+# define LOCK_READ	64	/* ... which allows concurrent read operations.	 */
+# define LOCK_WRITE	128	/* ... which allows concurrent write operations.  */
+# define LOCK_RW	192	/* ... Which allows concurrent read & write operations.	 */
+#endif
+
+
+#ifdef __USE_GNU
+/* Types of directory notifications that may be requested with F_NOTIFY.  */
+# define DN_ACCESS	0x00000001	/* File accessed.  */
+# define DN_MODIFY	0x00000002	/* File modified.  */
+# define DN_CREATE	0x00000004	/* File created.  */
+# define DN_DELETE	0x00000008	/* File removed.  */
+# define DN_RENAME	0x00000010	/* File renamed.  */
+# define DN_ATTRIB	0x00000020	/* File changed attibutes.  */
+# define DN_MULTISHOT	0x80000000	/* Don't remove notifier.  */
+#endif
+
+/* We don't need to support __USE_FILE_OFFSET64.  */
+struct flock
+  {
+    short int l_type;	/* Type of lock: F_RDLCK, F_WRLCK, or F_UNLCK.	*/
+    short int l_whence;	/* Where `l_start' is relative to (like `lseek').  */
+    __off_t l_start;	/* Offset where the lock begins.  */
+    __off_t l_len;	/* Size of the locked area; zero means until EOF.  */
+    __pid_t l_pid;	/* Process holding the lock.  */
+  };
+
+#ifdef __USE_LARGEFILE64
+struct flock64
+  {
+    short int l_type;	/* Type of lock: F_RDLCK, F_WRLCK, or F_UNLCK.	*/
+    short int l_whence;	/* Where `l_start' is relative to (like `lseek').  */
+    __off64_t l_start;	/* Offset where the lock begins.  */
+    __off64_t l_len;	/* Size of the locked area; zero means until EOF.  */
+    __pid_t l_pid;	/* Process holding the lock.  */
+  };
+#endif
+
+
+/* Define some more compatibility macros to be backward compatible with
+   BSD systems which did not managed to hide these kernel macros.  */
+#ifdef	__USE_BSD
+# define FAPPEND	O_APPEND
+# define FFSYNC		O_FSYNC
+# define FASYNC		O_ASYNC
+# define FNONBLOCK	O_NONBLOCK
+# define FNDELAY	O_NDELAY
+#endif /* Use BSD.  */
+
+/* Advise to `posix_fadvise'.  */
+#ifdef __USE_XOPEN2K
+# define POSIX_FADV_NORMAL	0 /* No further special treatment.  */
+# define POSIX_FADV_RANDOM	1 /* Expect random page references.  */
+# define POSIX_FADV_SEQUENTIAL	2 /* Expect sequential page references.	 */
+# define POSIX_FADV_WILLNEED	3 /* Will need these pages.  */
+# define POSIX_FADV_DONTNEED	4 /* Don't need these pages.  */
+# define POSIX_FADV_NOREUSE	5 /* Data will be accessed once.  */
+#endif
+
+__BEGIN_DECLS
+
+/* Provide kernel hint to read ahead.  */
+extern ssize_t readahead (int __fd, __off64_t __offset, size_t __count)
+    __THROW;
+
+__END_DECLS

libc/sysdeps/linux/ia64/bits/kernel_stat.h

@@ -0,0 +1,57 @@
+/* Ripped from linux/include/asm-ia64/stat.h
+ * and renamed 'struct stat' to 'struct kernel_stat' */
+
+#ifndef _ASM_IA64_STAT_H
+#define _ASM_IA64_STAT_H
+
+/*
+ * Modified 1998, 1999
+ *	David Mosberger-Tang <davidm@hpl.hp.com>, Hewlett-Packard Co
+ */
+
+struct kernel_stat {
+	unsigned long	st_dev;
+	unsigned long	st_ino;
+	unsigned long	st_nlink;
+	unsigned int	st_mode;
+	unsigned int	st_uid;
+	unsigned int	st_gid;
+	unsigned int	__pad0;
+	unsigned long	st_rdev;
+	unsigned long	st_size;
+	unsigned long	st_atime;
+	unsigned long	st_atime_nsec;
+	unsigned long	st_mtime;
+	unsigned long	st_mtime_nsec;
+	unsigned long	st_ctime;
+	unsigned long	st_ctime_nsec;
+	unsigned long	st_blksize;
+	long		st_blocks;
+	unsigned long	__unused[3];
+};
+
+#define STAT_HAVE_NSEC 1
+
+struct __old_kernel_stat {
+	unsigned int	st_dev;
+	unsigned int	st_ino;
+	unsigned int	st_mode;
+	unsigned int	st_nlink;
+	unsigned int	st_uid;
+	unsigned int	st_gid;
+	unsigned int	st_rdev;
+	unsigned int	__pad1;
+	unsigned long	st_size;
+	unsigned long	st_atime;
+	unsigned long	st_mtime;
+	unsigned long	st_ctime;
+	unsigned int	st_blksize;
+	int		st_blocks;
+	unsigned int	__unused1;
+	unsigned int	__unused2;
+};
+
+/* ia64 stat64 is same as stat */
+#define kernel_stat64 kernel_stat
+
+#endif /* _ASM_IA64_STAT_H */

libc/sysdeps/linux/ia64/bits/kernel_types.h

@@ -0,0 +1,56 @@
+/* Note that we use the exact same include guard #define names
+ * as asm/posix_types.h.  This will avoid gratuitous conflicts 
+ * with the posix_types.h kernel header, and will ensure that 
+ * our private content, and not the kernel header, will win.
+ *  -Erik
+ */
+#ifndef _ASM_IA64_POSIX_TYPES_H
+#define _ASM_IA64_POSIX_TYPES_H
+
+/*
+ * This file is generally used by user-level software, so you need to
+ * be a little careful about namespace pollution etc.  Also, we cannot
+ * assume GCC is being used.
+ *
+ * Based on <asm-alpha/posix_types.h>.
+ *
+ * Modified 1998-2000, 2003
+ *	David Mosberger-Tang <davidm@hpl.hp.com>, Hewlett-Packard Co
+ */
+
+typedef unsigned long	__kernel_ino_t;
+typedef unsigned int	__kernel_mode_t;
+typedef unsigned int	__kernel_nlink_t;
+typedef long		__kernel_off_t;
+typedef long long	__kernel_loff_t;
+typedef int		__kernel_pid_t;
+typedef int		__kernel_ipc_pid_t;
+typedef unsigned int	__kernel_uid_t;
+typedef unsigned int	__kernel_gid_t;
+typedef unsigned long	__kernel_size_t;
+typedef long		__kernel_ssize_t;
+typedef long		__kernel_ptrdiff_t;
+typedef long		__kernel_time_t;
+typedef long		__kernel_suseconds_t;
+typedef long		__kernel_clock_t;
+typedef int		__kernel_timer_t;
+typedef int		__kernel_clockid_t;
+typedef int		__kernel_daddr_t;
+typedef char *		__kernel_caddr_t;
+typedef unsigned long	__kernel_sigset_t;	/* at least 32 bits */
+typedef unsigned short	__kernel_uid16_t;
+typedef unsigned short	__kernel_gid16_t;
+
+typedef struct {
+	int	val[2];
+} __kernel_fsid_t;
+
+typedef __kernel_uid_t __kernel_old_uid_t;
+typedef __kernel_gid_t __kernel_old_gid_t;
+typedef __kernel_uid_t __kernel_uid32_t;
+typedef __kernel_gid_t __kernel_gid32_t;
+
+typedef unsigned int	__kernel_dev_t;
+typedef unsigned int	__kernel_old_dev_t;
+
+#endif /* _ASM_IA64_POSIX_TYPES_H */

libc/sysdeps/linux/ia64/bits/setjmp.h

@@ -0,0 +1,40 @@
+/* Define the machine-dependent type `jmp_buf'.  Linux/IA-64 version.
+   Copyright (C) 1999, 2000, 2003 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David Mosberger-Tang <davidm@hpl.hp.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _BITS_SETJMP_H
+#define _BITS_SETJMP_H  1
+
+#if !defined _SETJMP_H && !defined _PTHREAD_H
+# error "Never include <bits/setjmp.h> directly; use <setjmp.h> instead."
+#endif
+
+/* User code must not depend on the internal representation of jmp_buf. */
+
+#define _JBLEN	70
+
+/* the __jmp_buf element type should be __float80 per ABI... */
+typedef long __jmp_buf[_JBLEN] __attribute__ ((aligned (16))); /* guarantees 128-bit alignment! */
+
+/* Test if longjmp to JMPBUF would unwind the frame containing a local
+   variable at ADDRESS.  */
+#define _JMPBUF_UNWINDS(_jmpbuf, _address)		\
+     ((void *)(_address) < (void *)(((long *)_jmpbuf)[0]))
+
+#endif  /* bits/setjmp.h */

libc/sysdeps/linux/ia64/bits/sigaction.h

@@ -0,0 +1,73 @@
+/* Definitions for Linux/ia64 sigaction.
+   Copyright (C) 1996, 1997, 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SIGNAL_H
+# error "Never include <bits/sigaction.h> directly; use <signal.h> instead."
+#endif
+
+/* Structure describing the action to be taken when a signal arrives.  */
+struct sigaction
+  {
+    /* Signal handler.  */
+#ifdef __USE_POSIX199309
+    union
+      {
+	/* Used if SA_SIGINFO is not set.  */
+	__sighandler_t sa_handler;
+	/* Used if SA_SIGINFO is set.  */
+	void (*sa_sigaction) (int, siginfo_t *, void *);
+      }
+    __sigaction_handler;
+# define sa_handler	__sigaction_handler.sa_handler
+# define sa_sigaction	__sigaction_handler.sa_sigaction
+#else
+    __sighandler_t sa_handler;
+#endif
+
+    /* Special flags.  */
+    unsigned long int sa_flags;
+
+    /* Additional set of signals to be blocked.  */
+    __sigset_t sa_mask;
+  };
+
+/* Bits in `sa_flags'.  */
+#define SA_NOCLDSTOP  0x00000001 /* Don't send SIGCHLD when children stop.  */
+#define SA_NOCLDWAIT  0x00000002 /* Don't create zombie on child death.  */
+#define SA_SIGINFO    0x00000004
+#if defined __USE_UNIX98 || defined __USE_MISC
+# define SA_ONSTACK   0x08000000 /* Use signal stack by using `sa_restorer'. */
+# define SA_RESTART   0x10000000 /* Restart syscall on signal return.  */
+# define SA_NODEFER   0x40000000 /* Don't automatically block the signal
+				    when its handler is being executed.  */
+# define SA_RESETHAND 0x80000000 /* Reset to SIG_DFL on entry to handler.  */
+#endif
+#ifdef __USE_MISC
+# define SA_INTERRUPT 0x20000000 /* Historic no-op.  */
+
+/* Some aliases for the SA_ constants.  */
+# define SA_NOMASK    SA_NODEFER
+# define SA_ONESHOT   SA_RESETHAND
+# define SA_STACK     SA_ONSTACK
+#endif
+
+/* Values for the HOW argument to `sigprocmask'.  */
+#define SIG_BLOCK          0	/* for blocking signals */
+#define SIG_UNBLOCK        1	/* for unblocking signals */
+#define SIG_SETMASK        2	/* for setting the signal mask */

libc/sysdeps/linux/ia64/bits/sigcontext.h

@@ -0,0 +1,79 @@
+/* Copyright (C) 1996, 1997, 1998, 2000, 2001, 2003, 2004
+   Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Jes Sorensen <jes@linuxcare.com>, July 2000
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#if !defined _SIGNAL_H && !defined _SYS_UCONTEXT_H
+# error "Never use <bits/sigcontext.h> directly; include <signal.h> instead."
+#endif
+
+#ifndef _BITS_SIGCONTEXT_H
+#define _BITS_SIGCONTEXT_H 1
+
+#include <bits/sigstack.h>
+
+struct ia64_fpreg
+  {
+    union
+      {
+	unsigned long bits[2];
+      } u;
+  } __attribute__ ((aligned (16)));
+
+struct sigcontext
+{
+  unsigned long int sc_flags;	/* see manifest constants below */
+  unsigned long int sc_nat;	/* bit i == 1 iff scratch reg gr[i] is a NaT */
+  stack_t sc_stack;		/* previously active stack */
+
+  unsigned long int sc_ip;	/* instruction pointer */
+  unsigned long int sc_cfm;	/* current frame marker */
+  unsigned long int sc_um;	/* user mask bits */
+  unsigned long int sc_ar_rsc;	/* register stack configuration register */
+  unsigned long int sc_ar_bsp;	/* backing store pointer */
+  unsigned long int sc_ar_rnat;	/* RSE NaT collection register */
+  unsigned long int sc_ar_ccv;	/* compare & exchange compare value register */
+  unsigned long int sc_ar_unat;	/* ar.unat of interrupted context */
+  unsigned long int sc_ar_fpsr;	/* floating-point status register */
+  unsigned long int sc_ar_pfs;	/* previous function state */
+  unsigned long int sc_ar_lc;	/* loop count register */
+  unsigned long int sc_pr;	/* predicate registers */
+  unsigned long int sc_br[8];	/* branch registers */
+  unsigned long int sc_gr[32];	/* general registers (static partition) */
+  struct ia64_fpreg sc_fr[128];	/* floating-point registers */
+  unsigned long int sc_rbs_base;/* NULL or new base of sighandler's rbs */
+  unsigned long int sc_loadrs;	/* see description above */
+  unsigned long int sc_ar25;	/* cmp8xchg16 uses this */
+  unsigned long int sc_ar26;	/* rsvd for scratch use */
+  unsigned long int sc_rsvd[12];/* reserved for future use */
+
+  /* sc_mask is actually an sigset_t but we don't want to
+   * include the kernel headers here. */
+  unsigned long int sc_mask;	/* signal mask to restore after handler returns */
+};
+
+/* sc_flag bit definitions. */
+#define IA64_SC_FLAG_ONSTACK_BIT	0	/* is handler running on signal stack? */
+#define IA64_SC_FLAG_IN_SYSCALL_BIT	1	/* did signal interrupt a syscall? */
+#define IA64_SC_FLAG_FPH_VALID_BIT	2	/* is state in f[32]-f[127] valid? */
+
+#define IA64_SC_FLAG_ONSTACK		(1 << IA64_SC_FLAG_ONSTACK_BIT)
+#define IA64_SC_FLAG_IN_SYSCALL		(1 << IA64_SC_FLAG_IN_SYSCALL_BIT)
+#define IA64_SC_FLAG_FPH_VALID		(1 << IA64_SC_FLAG_FPH_VALID_BIT)
+
+#endif /* _BITS_SIGCONTEXT_H */

libc/sysdeps/linux/ia64/bits/sigstack.h

@@ -0,0 +1,63 @@
+/* sigstack, sigaltstack definitions.
+   Copyright (C) 1998, 2000, 2002, 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SIGNAL_H
+# error "Never include this file directly.  Use <signal.h> instead"
+#endif
+
+#ifndef _SIGSTACK_H
+#define _SIGSTACK_H	1
+
+/* Structure describing a signal stack (obsolete).  */
+struct sigstack
+  {
+    __ptr_t ss_sp;		/* Signal stack pointer.  */
+    int ss_onstack;		/* Nonzero if executing on this stack.  */
+  };
+
+
+/* Possible values for `ss_flags.'.  */
+enum
+{
+  SS_ONSTACK = 1,
+#define SS_ONSTACK	SS_ONSTACK
+  SS_DISABLE
+#define SS_DISABLE	SS_DISABLE
+};
+
+/* Minimum stack size for a signal handler.
+
+   Yes, this should be 131072 but the constant got defined incorrectly
+   in the kernel and we have to live with it.  Users should in any case
+   use SIGSTKSZ as the size user-supplied buffers should have.  */
+#define MINSIGSTKSZ	131027
+
+/* System default stack size.  */
+#define SIGSTKSZ	262144
+
+
+/* Alternate, preferred interface.  */
+typedef struct sigaltstack
+  {
+    __ptr_t ss_sp;
+    int ss_flags;
+    size_t ss_size;
+  } stack_t;
+
+#endif	/* bits/sigstack.h */

libc/sysdeps/linux/ia64/bits/stackinfo.h

@@ -0,0 +1,29 @@
+/* Copyright (C) 2001 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* This file contains a bit of information about the stack allocation
+   of the processor.  */
+
+#ifndef _STACKINFO_H
+#define _STACKINFO_H	1
+
+/* On IA-64 the stack grows down.  The register stack is of no concern
+   here.  */
+#define _STACK_GROWS_DOWN	1
+
+#endif	/* stackinfo.h */

libc/sysdeps/linux/ia64/bits/syscalls.h

@@ -0,0 +1,198 @@
+/* Copyright (C) 1999, 2000, 2002, 2003, 2004 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Written by Jes Sorensen, <Jes.Sorensen@cern.ch>, April 1999.
+   Based on code originally written by David Mosberger-Tang
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _BITS_SYSCALLS_H
+#define _BITS_SYSCALLS_H
+
+#ifndef _SYSCALL_H
+# error "Never use <bits/syscalls.h> directly; include <sys/syscall.h> instead."
+#endif
+
+/* This includes the `__NR_<name>' syscall numbers taken from the Linux kernel
+ * header files.  It also defines the traditional `SYS_<name>' macros for older
+ * programs.  */
+#include <bits/sysnum.h>
+
+#ifndef __set_errno
+# define __set_errno(val) (*__errno_location ()) = (val)
+#endif
+
+
+#ifndef __ASSEMBLER__
+
+#undef IA64_USE_NEW_STUB
+
+#define SYS_ify(syscall_name)   __NR_##syscall_name
+
+/* taken from asm-ia64/break.h */
+#define __IA64_BREAK_SYSCALL	"0x100000"
+
+#define _DO_SYSCALL(name, nr, args...) \
+    LOAD_ARGS_##nr (args) \
+    register long _r8 asm ("r8"); \
+    register long _r10 asm ("r10"); \
+    register long _r15 asm ("r15") = SYS_ify(name); \
+    long _retval; \
+    LOAD_REGS_##nr \
+    __asm __volatile ("break " __IA64_BREAK_SYSCALL ";;\n\t" \
+		: "=r" (_r8), "=r" (_r10), "=r" (_r15) ASM_OUTARGS_##nr \
+		: "2" (_r15) ASM_ARGS_##nr \
+		: "memory" ASM_CLOBBERS_##nr); \
+    _retval = _r8; \
+	if (_r10 == -1) { \
+		__set_errno (_retval); \
+		_retval = -1; \
+	}
+
+#define LOAD_ARGS_0()
+#define LOAD_REGS_0
+#define LOAD_ARGS_1(a1)					\
+  long _arg1 = (long) (a1);				\
+  LOAD_ARGS_0 ()
+#define LOAD_REGS_1					\
+  register long _out0 asm ("out0") = _arg1;		\
+  LOAD_REGS_0
+#define LOAD_ARGS_2(a1, a2)				\
+  long _arg2 = (long) (a2);				\
+  LOAD_ARGS_1 (a1)
+#define LOAD_REGS_2					\
+  register long _out1 asm ("out1") = _arg2;		\
+  LOAD_REGS_1
+#define LOAD_ARGS_3(a1, a2, a3)				\
+  long _arg3 = (long) (a3);				\
+  LOAD_ARGS_2 (a1, a2)
+#define LOAD_REGS_3					\
+  register long _out2 asm ("out2") = _arg3;		\
+  LOAD_REGS_2
+#define LOAD_ARGS_4(a1, a2, a3, a4)			\
+  long _arg4 = (long) (a4);				\
+  LOAD_ARGS_3 (a1, a2, a3)
+#define LOAD_REGS_4					\
+  register long _out3 asm ("out3") = _arg4;		\
+  LOAD_REGS_3
+#define LOAD_ARGS_5(a1, a2, a3, a4, a5)			\
+  long _arg5 = (long) (a5);				\
+  LOAD_ARGS_4 (a1, a2, a3, a4)
+#define LOAD_REGS_5					\
+  register long _out4 asm ("out4") = _arg5;		\
+  LOAD_REGS_4
+#define LOAD_ARGS_6(a1, a2, a3, a4, a5, a6)		\
+  long _arg6 = (long) (a6);	    			\
+  LOAD_ARGS_5 (a1, a2, a3, a4, a5)
+#define LOAD_REGS_6					\
+  register long _out5 asm ("out5") = _arg6;		\
+  LOAD_REGS_5
+
+#define ASM_OUTARGS_0
+#define ASM_OUTARGS_1	ASM_OUTARGS_0, "=r" (_out0)
+#define ASM_OUTARGS_2	ASM_OUTARGS_1, "=r" (_out1)
+#define ASM_OUTARGS_3	ASM_OUTARGS_2, "=r" (_out2)
+#define ASM_OUTARGS_4	ASM_OUTARGS_3, "=r" (_out3)
+#define ASM_OUTARGS_5	ASM_OUTARGS_4, "=r" (_out4)
+#define ASM_OUTARGS_6	ASM_OUTARGS_5, "=r" (_out5)
+
+#ifdef IA64_USE_NEW_STUB
+#define ASM_ARGS_0
+#define ASM_ARGS_1	ASM_ARGS_0, "4" (_out0)
+#define ASM_ARGS_2	ASM_ARGS_1, "5" (_out1)
+#define ASM_ARGS_3	ASM_ARGS_2, "6" (_out2)
+#define ASM_ARGS_4	ASM_ARGS_3, "7" (_out3)
+#define ASM_ARGS_5	ASM_ARGS_4, "8" (_out4)
+#define ASM_ARGS_6	ASM_ARGS_5, "9" (_out5)
+#else
+#define ASM_ARGS_0
+#define ASM_ARGS_1	ASM_ARGS_0, "3" (_out0)
+#define ASM_ARGS_2	ASM_ARGS_1, "4" (_out1)
+#define ASM_ARGS_3	ASM_ARGS_2, "5" (_out2)
+#define ASM_ARGS_4	ASM_ARGS_3, "6" (_out3)
+#define ASM_ARGS_5	ASM_ARGS_4, "7" (_out4)
+#define ASM_ARGS_6	ASM_ARGS_5, "8" (_out5)
+#endif
+
+#define ASM_CLOBBERS_0	ASM_CLOBBERS_1, "out0"
+#define ASM_CLOBBERS_1	ASM_CLOBBERS_2, "out1"
+#define ASM_CLOBBERS_2	ASM_CLOBBERS_3, "out2"
+#define ASM_CLOBBERS_3	ASM_CLOBBERS_4, "out3"
+#define ASM_CLOBBERS_4	ASM_CLOBBERS_5, "out4"
+#define ASM_CLOBBERS_5	ASM_CLOBBERS_6, "out5"
+#define ASM_CLOBBERS_6_COMMON	, "out6", "out7",			\
+  /* Non-stacked integer registers, minus r8, r10, r15.  */		\
+  "r2", "r3", "r9", "r11", "r12", "r13", "r14", "r16", "r17", "r18",	\
+  "r19", "r20", "r21", "r22", "r23", "r24", "r25", "r26", "r27",	\
+  "r28", "r29", "r30", "r31",						\
+  /* Predicate registers.  */						\
+  "p6", "p7", "p8", "p9", "p10", "p11", "p12", "p13", "p14", "p15",	\
+  /* Non-rotating fp registers.  */					\
+  "f6", "f7", "f8", "f9", "f10", "f11", "f12", "f13", "f14", "f15",	\
+  /* Branch registers.  */						\
+  "b6"
+
+#ifdef IA64_USE_NEW_STUB
+# define ASM_CLOBBERS_6	ASM_CLOBBERS_6_COMMON
+#else
+# define ASM_CLOBBERS_6	ASM_CLOBBERS_6_COMMON , "b7"
+#endif
+
+
+
+#define _syscall0(type,name) \
+type name(void) \
+{ \
+	_DO_SYSCALL(name, 0); return (type) _retval; \
+}
+
+#define _syscall1(type,name,type1,arg1) \
+type name(type1 arg1) \
+{ \
+	_DO_SYSCALL(name, 1, arg1); return (type) _retval; \
+}
+
+#define _syscall2(type,name,type1,arg1,type2,arg2) \
+type name(type1 arg1, type2 arg2) \
+{ \
+	_DO_SYSCALL(name, 2, arg1, arg2); return (type) _retval; \
+}
+
+#define _syscall3(type,name,type1,arg1,type2,arg2,type3,arg3) \
+type name(type1 arg1, type2 arg2, type3 arg3) \
+{ \
+	_DO_SYSCALL(name, 3, arg1, arg2, arg3); return (type) _retval; \
+}
+
+#define _syscall4(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
+{ \
+	_DO_SYSCALL(name, 4, arg1, arg2, arg3, arg4); return (type) _retval; \
+}
+
+#define _syscall5(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5) \
+type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5) \
+{ \
+	_DO_SYSCALL(name, 5, arg1, arg2, arg3, arg4, arg5); return (type) _retval; \
+}
+
+#define _syscall6(type,name,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5,type6,arg6) \
+type name(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5, type6 arg6) \
+{ \
+	_DO_SYSCALL(name, 6, arg1, arg2, arg3, arg4, arg5, arg6); return (type) _retval; \
+}
+
+#endif /* __ASSEMBLER__ */
+#endif /* _BITS_SYSCALLS_H */

libc/sysdeps/linux/ia64/bits/uClibc_arch_features.h

@@ -0,0 +1,41 @@
+/*
+ * Track misc arch-specific features that aren't config options
+ */
+
+#ifndef _BITS_UCLIBC_ARCH_FEATURES_H
+#define _BITS_UCLIBC_ARCH_FEATURES_H
+
+/* instruction used when calling abort() to kill yourself */
+#define __UCLIBC_ABORT_INSTRUCTION__ "break 0"
+
+/* can your target use syscall6() for mmap ? */
+#define __UCLIBC_MMAP_HAS_6_ARGS__
+
+/* does your target use syscall4() for truncate64 ? (32bit arches only) */
+#undef __UCLIBC_TRUNCATE64_HAS_4_ARGS__
+
+/* does your target have a broken create_module() ? */
+#undef __UCLIBC_BROKEN_CREATE_MODULE__
+
+/* does your target prefix all symbols with an _ ? */
+#define __UCLIBC_NO_UNDERSCORES__
+
+/* does your target have an asm .set ? */
+#define __UCLIBC_HAVE_ASM_SET_DIRECTIVE__
+
+/* define if target doesn't like .global */
+#undef __UCLIBC_ASM_GLOBAL_DIRECTIVE__
+
+/* define if target supports .weak */
+#define __UCLIBC_HAVE_ASM_WEAK_DIRECTIVE__
+
+/* define if target supports .weakext */
+#undef __UCLIBC_HAVE_ASM_WEAKEXT_DIRECTIVE__
+
+/* needed probably only for ppc64 */
+#undef __UCLIBC_HAVE_ASM_GLOBAL_DOT_NAME__
+
+/* define if target supports IEEE signed zero floats */
+#define __UCLIBC_HAVE_SIGNED_ZERO__
+
+#endif /* _BITS_UCLIBC_ARCH_FEATURES_H */

libc/sysdeps/linux/ia64/bits/wordsize.h

@@ -0,0 +1,19 @@
+/* Copyright (C) 1999 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#define __WORDSIZE	64

libc/sysdeps/linux/ia64/brk.S

@@ -0,0 +1,52 @@
+/* brk system call for Linux/ia64
+   Copyright (C) 1999,2000,2001,2003 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Written by Stephane Eranian <eranian@hpl.hp.com> and
+	      Jes Sorensen, <Jes.Sorensen@cern.ch>, April 1999.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "sysdep.h"
+
+#include <asm/unistd.h>
+#include <asm/errno.h>
+
+	.global __curbrk
+	.type __curbrk,@object
+	.size __curbrk,8
+	.data
+	.align	8
+__curbrk:
+	data8	0
+
+weak_alias (__curbrk, ___brk_addr)
+
+LEAF(__brk)
+	.regstk 1, 0, 0, 0
+	DO_CALL(__NR_brk)
+	cmp.ltu	p6, p0 = ret0, in0
+	addl r9 = @ltoff(__curbrk), gp
+	;;
+	ld8 r9 = [r9]
+(p6) 	mov ret0 = ENOMEM
+(p6)	br.cond.spnt.few __syscall_error
+	;;
+	st8 [r9] = ret0
+	mov ret0 = 0
+	ret
+END(__brk)
+
+weak_alias (__brk, brk)

libc/sysdeps/linux/ia64/bsd-_setjmp.S

@@ -0,0 +1 @@
+/* _setjmp is in setjmp.S  */

libc/sysdeps/linux/ia64/bsd-setjmp.S

@@ -0,0 +1 @@
+/* setjmp is in setjmp.S  */

libc/sysdeps/linux/ia64/clone2.S

@@ -0,0 +1,105 @@
+/* Copyright (C) 2000, 2001, 2003, 2004 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "sysdep.h"
+
+#include "sysdep.h"
+#include <asm/errno.h>
+
+
+/* int  __clone2(int (*fn) (void *arg), void *child_stack_base, 	*/
+/* 	         size_t child_stack_size, int flags, void *arg,		*/
+/*	         pid_t *parent_tid, void *tls, pid_t *child_tid)	*/
+
+#define CHILD	p8
+#define PARENT	p9
+
+ENTRY(__clone2)
+	.prologue
+	alloc r2=ar.pfs,8,1,6,0
+	cmp.eq p6,p0=0,in0
+	mov r8=EINVAL
+	mov out0=in3		/* Flags are first syscall argument.	*/
+	mov out1=in1		/* Stack address.			*/
+(p6)	br.cond.spnt.many __syscall_error
+	;;
+	mov out2=in2		/* Stack size.				*/
+	mov out3=in5		/* Parent TID Pointer			*/
+	mov out4=in7		/* Child TID Pointer			*/
+ 	mov out5=in6		/* TLS pointer				*/
+	/*
+	 * clone2() is special: the child cannot execute br.ret right
+	 * after the system call returns, because it starts out
+	 * executing on an empty stack.  Because of this, we can't use
+	 * the new (lightweight) syscall convention here.  Instead, we
+	 * just fall back on always using "break".
+	 *
+	 * Furthermore, since the child starts with an empty stack, we
+	 * need to avoid unwinding past invalid memory.  To that end,
+	 * we'll pretend now that __clone2() is the end of the
+	 * call-chain.  This is wrong for the parent, but only until
+	 * it returns from clone2() but it's better than the
+	 * alternative.
+	 */
+	mov r15=SYS_ify (clone2)
+	.save rp, r0
+	break __BREAK_SYSCALL
+	.body
+        cmp.eq p6,p0=-1,r10
+	cmp.eq CHILD,PARENT=0,r8 /* Are we the child?   */
+(p6)	br.cond.spnt.many __syscall_error
+	;;
+(CHILD)	mov loc0=gp
+(PARENT) ret
+	;;
+#ifdef RESET_PID
+	tbit.nz p6,p0=in3,16	/* CLONE_THREAD */
+	tbit.z p7,p10=in3,8	/* CLONE_VM */
+(p6)	br.cond.dptk 1f
+	;;
+	mov r15=SYS_ify (getpid)
+(p10)	addl r8=-1,r0
+(p7)	break __BREAK_SYSCALL
+	;;
+	add r9=PID,r13
+	add r10=TID,r13
+	;;
+	st4 [r9]=r8
+	st4 [r10]=r8
+	;;
+#endif
+1:	ld8 out1=[in0],8	/* Retrieve code pointer.	*/
+	mov out0=in4		/* Pass proper argument	to fn */
+	;;
+	ld8 gp=[in0]		/* Load function gp.		*/
+	mov b6=out1
+	br.call.dptk.many rp=b6	/* Call fn(arg) in the child 	*/
+	;;
+	mov out0=r8		/* Argument to _exit		*/
+	mov gp=loc0
+	.globl HIDDEN_JUMPTARGET(_exit)
+	br.call.dpnt.many rp=HIDDEN_JUMPTARGET(_exit)
+				/* call _exit with result from fn.	*/
+	ret			/* Not reached.		*/
+PSEUDO_END(__clone2)
+
+/* For now we leave __clone undefined.  This is unlikely to be a	*/
+/* problem, since at least the i386 __clone in glibc always failed	*/
+/* with a 0 sp (eventhough the kernel explicitly handled it).		*/
+/* Thus all such calls needed to pass an explicit sp, and as a result,	*/
+/* would be unlikely to work on ia64.					*/

libc/sysdeps/linux/ia64/crt1.S

@@ -0,0 +1,129 @@
+/* Copyright (C) 1999, 2000, 2001, 2002 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Written by Jes Sorensen, <Jes.Sorensen@cern.ch>, April 1999.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   In addition to the permissions in the GNU Lesser General Public
+   License, the Free Software Foundation gives you unlimited
+   permission to link the compiled version of this file with other
+   programs, and to distribute those programs without any restriction
+   coming from the use of this file. (The GNU Lesser General Public
+   License restrictions do apply in other respects; for example, they
+   cover modification of the file, and distribution when not linked
+   into another program.)
+
+   Note that people who make modified versions of this file are not
+   obligated to grant this special exception for their modified
+   versions; it is their choice whether to do so. The GNU Lesser
+   General Public License gives permission to release a modified
+   version without this exception; this exception also makes it
+   possible to release a modified version which carries forward this
+   exception.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#define __ASSEMBLY__
+#include "sysdep.h"
+
+#include <asm/unistd.h>
+#include <asm/fpu.h>
+
+/*
+ * Arguments for __uClibc_main:
+ *	out0:	main
+ *	out1:	argc
+ *	out2:	argv
+ *	out3:	init
+ *	out4:	fini
+ *	out5:	rtld_fini
+ *	out6:	stack_end
+ */
+
+#if defined(__UCLIBC_CTOR_DTOR__)
+.type   _init,%function
+.type   _fini,%function
+#else
+.weak   _init
+.weak   _fini
+#endif
+
+	.align 32
+	.global _start
+
+	.proc _start
+	.type _start,@function
+_start:
+	.prologue
+	.save rp, r0
+	.body
+	.prologue
+	{ .mlx
+	  alloc r2 = ar.pfs,0,0,7,0
+	  movl r3 = FPSR_DEFAULT
+	}
+	{ .mlx
+	  adds out2 = 16, sp	/* get address of argc value */
+	  movl gp = @gprel(0f)
+	  ;;
+	}
+0:	{ .mmi
+	  ld8 out1 = [out2], 8	/* load argc and move out2 to become argv */
+	  mov.m r10 = ar.bsp	/* fetch rbs base address */
+	  mov r9 = ip
+	  ;;
+	}
+	{ .mii
+	  mov ar.fpsr = r3
+	  sub gp = r9, gp	/* back-compute gp value */
+	  adds out6 = 16, sp	/* highest non-environment stack address */
+	  ;;
+	}
+	{
+	  addl r11 = @ltoff(__libc_ia64_register_backing_store_base), gp
+	  addl out0 = @ltoff(@fptr(main)), gp
+	  addl out3 = @ltoff(@fptr(_init)), gp
+	  ;;
+	}
+	{ .mmi
+	  ld8 r3 = [r11]	/* pointer to __libc_ia64_register_backing_store_base */
+	  ld8 out0 = [out0]	/* pointer to `main' function descriptor */
+	  addl out4 = @ltoff(@fptr(_fini)), gp
+	  ;;
+	}
+	{ .mmi
+	  ld8 out3 = [out3]	/* pointer to `init' function descriptor */
+	  ld8 out4 = [out4]	/* pointer to `fini' function descriptor */
+	  nop 0
+	}
+	.body
+	{ .mib
+	  st8 [r3] = r10
+	  mov out5 = ret0	/* dynamic linker destructor */
+	  br.call.sptk.few rp = __uClibc_main
+	}
+	{ .mib
+	  break 0	/* break miserably if we ever return */
+	}
+	.endp _start
+
+/* Define a symbol for the first piece of initialized data.  */
+	.data
+	.globl __data_start
+__data_start:
+	.long 0
+	.weak data_start
+	data_start = __data_start
+
+	.common __libc_ia64_register_backing_store_base, 8, 8

libc/sysdeps/linux/ia64/crti.S

@@ -0,0 +1,36 @@
+/* glibc's sysdeps/ia64/elf/initfini.c used for reference [PROLOG] */
+
+	.text
+	.section .init
+	.global _init#
+	.proc _init#
+_init:
+	.prologue
+	.save ar.pfs, r34
+	alloc r34 = ar.pfs, 0, 3, 0, 0
+	.vframe r32
+	mov r32 = r12
+	.save rp, r33
+	mov r33 = b0
+	.body
+	adds r12 = -16, r12
+	;;
+	.endp _init#
+
+
+
+	.section .fini
+	.global _fini#
+	.proc _fini#
+_fini:
+	.prologue
+	.save ar.pfs, r34
+	alloc r34 = ar.pfs, 0, 3, 0, 0
+	.vframe r32
+	mov r32 = r12
+	.save rp, r33
+	mov r33 = b0
+	.body
+	adds r12 = -16, r12
+	;;
+	.endp _fini#

libc/sysdeps/linux/ia64/crtn.S

@@ -0,0 +1,33 @@
+/* glibc's sysdeps/ia64/elf/initfini.c used for reference [EPILOG] */
+
+	.text
+	.section .init
+	.proc _init#
+_init:
+	.prologue
+	.save ar.pfs, r34
+	.vframe r32
+	.save rp, r33
+	.body
+	.regstk 0,2,0,0
+	mov r12 = r32
+	mov ar.pfs = r34
+	mov b0 = r33
+	br.ret.sptk.many b0
+	.endp _init#
+
+
+
+	.section .fini
+	.proc _fini#
+_fini:
+	.prologue
+	.save ar.pfs, r34
+	.vframe r32
+	.save rp, r33
+	.body
+	mov r12 = r32
+	mov ar.pfs = r34
+	mov b0 = r33
+	br.ret.sptk.many b0
+	.endp _fini#

libc/sysdeps/linux/ia64/fork.S

@@ -0,0 +1,41 @@
+/* Copyright (C) 2000, 2002 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+
+#include "sysdep.h"
+#define _SIGNAL_H
+#include <bits/signum.h>
+
+/* pid_t fork(void); */
+/* Implemented as a clone system call with parameters SIGCHLD and 0	*/
+
+ENTRY(__libc_fork)
+	alloc r2=ar.pfs,0,0,2,0
+	mov out0=SIGCHLD	/* Return SIGCHLD when child finishes	*/
+				/* no other clone flags; nothing shared	*/
+	mov out1=0		/* Standard sp value.			*/
+	;;
+	DO_CALL (SYS_ify (clone))
+	cmp.eq p6,p0=-1,r10
+(p6)	br.cond.spnt.few __syscall_error
+	ret
+PSEUDO_END(__libc_fork)
+
+weak_alias (__libc_fork, __fork)
+libc_hidden_def (__fork)
+weak_alias (__libc_fork, fork)

libc/sysdeps/linux/ia64/pipe.S

@@ -0,0 +1,38 @@
+/* Copyright (C) 1999, 2000, 2002 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David Mosberger <davidm@hpl.hp.com>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+/* __pipe is a special syscall since it returns two values.  */
+
+#include "sysdep.h"
+
+ENTRY(__pipe)
+       .regstk 1,0,0,0
+       DO_CALL (SYS_ify (pipe))
+       cmp.ne p6,p0=-1,r10
+       ;;
+(p6)   st4 [in0]=r8,4
+(p6)   mov ret0=0
+       ;;
+(p6)   st4 [in0]=r9
+(p6)   ret
+       br.cond.spnt.few __syscall_error
+PSEUDO_END(__pipe)
+
+libc_hidden_def (__pipe)
+weak_alias (__pipe, pipe)

libc/sysdeps/linux/ia64/setjmp.S

@@ -0,0 +1,189 @@
+/* Copyright (C) 1999, 2000, 2001, 2002, 2004, 2005
+   Free Software Foundation, Inc.
+   Contributed by David Mosberger-Tang <davidm@hpl.hp.com>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.
+
+   The layout of the jmp_buf is as follows.  This is subject to change
+   and user-code should never depend on the particular layout of
+   jmp_buf!
+
+
+  	offset:	description:
+	-------	------------
+  	0x000	stack pointer (r12)	; unchangeable (see _JMPBUF_UNWINDS)
+  	0x008	r1 (gp)
+	0x010	caller's unat
+	0x018	fpsr
+  	0x020	r4
+  	0x028	r5
+  	0x030	r6
+  	0x038	r7
+  	0x040	rp (b0)
+  	0x048	b1
+  	0x050	b2
+  	0x058	b3
+  	0x060	b4
+  	0x068	b5
+  	0x070	ar.pfs
+  	0x078	ar.lc
+  	0x080	pr
+  	0x088	ar.bsp			; unchangeable (see __longjmp.S)
+  	0x090	ar.unat
+	0x098	&__jmp_buf	; address of the jmpbuf (needed to locate NaT bits in unat)
+	0x0a0	 f2
+	0x0b0	 f3
+	0x0c0	 f4
+	0x0d0	 f5
+	0x0e0	f16
+  	0x0f0	f17
+  	0x100	f18
+  	0x110	f19
+  	0x120	f20
+  	0x130	f21
+  	0x130	f22
+  	0x140	f23
+  	0x150	f24
+  	0x160	f25
+  	0x170	f26
+  	0x180	f27
+  	0x190	f28
+  	0x1a0	f29
+  	0x1b0	f30
+  	0x1c0	f31 */
+
+#include "sysdep.h"
+#include <features.h>
+
+	/* The following two entry points are the traditional entry points: */
+
+LEAF(setjmp)
+	alloc r8=ar.pfs,2,0,0,0
+	mov in1=1
+	br.cond.sptk.many _GI___sigsetjmp
+END(setjmp)
+
+LEAF(_setjmp)
+	alloc r8=ar.pfs,2,0,0,0
+	mov in1=0
+	br.cond.sptk.many _GI___sigsetjmp
+END(_setjmp)
+libc_hidden_def (_setjmp)
+
+	/* __sigsetjmp(__jmp_buf buf, int savemask) */
+
+ENTRY(__sigsetjmp)
+	.prologue ASM_UNW_PRLG_RP|ASM_UNW_PRLG_PFS, ASM_UNW_PRLG_GRSAVE(2)
+	alloc loc1=ar.pfs,2,5,2,0
+	.save ar.unat, loc2
+	mov loc2=ar.unat
+	;;
+	mov r17=ar.fpsr
+	mov r2=in0
+	add r3=8,in0
+	;;
+.mem.offset 8,0;	st8.spill.nta [r2]=sp,16	// r12 (sp)
+.mem.offset 0,0;	st8.spill.nta [r3]=gp,16	// r1 (gp)
+	;;
+	st8.nta [r2]=loc2,16		// save caller's unat
+	st8.nta [r3]=r17,16		// save fpsr
+	add r8=0xa0,in0
+	;;
+.mem.offset 8,0;	st8.spill.nta [r2]=r4,16	// r4
+.mem.offset 0,0;	st8.spill.nta [r3]=r5,16	// r5
+	add r9=0xb0,in0
+	;;
+	stf.spill.nta [r8]=f2,32
+	stf.spill.nta [r9]=f3,32
+	mov loc0=rp
+	.body
+	;;
+	stf.spill.nta [r8]=f4,32
+	stf.spill.nta [r9]=f5,32
+	mov r17=b1
+	;;
+	stf.spill.nta [r8]=f16,32
+	stf.spill.nta [r9]=f17,32
+	mov r18=b2
+	;;
+	stf.spill.nta [r8]=f18,32
+	stf.spill.nta [r9]=f19,32
+	mov r19=b3
+	;;
+	stf.spill.nta [r8]=f20,32
+	stf.spill.nta [r9]=f21,32
+	mov r20=b4
+	;;
+	stf.spill.nta [r8]=f22,32
+	stf.spill.nta [r9]=f23,32
+	mov r21=b5
+	;;
+	stf.spill.nta [r8]=f24,32
+	stf.spill.nta [r9]=f25,32
+	mov r22=ar.lc
+	;;
+	stf.spill.nta [r8]=f26,32
+	stf.spill.nta [r9]=f27,32
+	mov r24=pr
+	;;
+	stf.spill.nta [r8]=f28,32
+	stf.spill.nta [r9]=f29,32
+	;;
+	stf.spill.nta [r8]=f30
+	stf.spill.nta [r9]=f31
+
+.mem.offset 8,0;	st8.spill.nta [r2]=r6,16	// r6
+.mem.offset 0,0;	st8.spill.nta [r3]=r7,16	// r7
+	;;
+	mov r23=ar.bsp
+	mov r25=ar.unat
+	mov out0=in0
+
+	st8.nta [r2]=loc0,16		// b0
+	st8.nta [r3]=r17,16		// b1
+	mov out1=in1
+	;;
+	st8.nta [r2]=r18,16		// b2
+	st8.nta [r3]=r19,16		// b3
+	;;
+	st8.nta [r2]=r20,16		// b4
+	st8.nta [r3]=r21,16		// b5
+	;;
+	st8.nta [r2]=loc1,16		// ar.pfs
+	st8.nta [r3]=r22,16		// ar.lc
+	;;
+	st8.nta [r2]=r24,16		// pr
+	st8.nta [r3]=r23,16		// ar.bsp
+	;;
+	st8.nta [r2]=r25		// ar.unat
+	st8.nta [r3]=in0		// &__jmp_buf
+#if defined NOT_IN_libc && defined IS_IN_rtld
+	/* In ld.so we never save the signal mask.  */
+	;;
+#else
+	br.call.dpnt.few rp=__sigjmp_save
+#endif
+.ret0:					// force a new bundle ::q
+	mov.m ar.unat=loc2		// restore caller's unat
+	mov rp=loc0
+	mov ar.pfs=loc1
+	mov r8=0
+	ret
+END(__sigsetjmp)
+strong_alias(__sigsetjmp, _GI___sigsetjmp)
+
+weak_extern(_setjmp)
+weak_extern(setjmp)

libc/sysdeps/linux/ia64/sys/io.h

@@ -0,0 +1,68 @@
+/* Copyright (C) 1999, 2000 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+     Contributed by David Mosberger-Tang <davidm@hpl.hp.com>
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef	_SYS_IO_H
+#define	_SYS_IO_H	1
+
+#include <features.h>
+
+__BEGIN_DECLS
+
+/* If TURN_ON is TRUE, request for permission to do direct i/o on the
+   port numbers in the range [FROM,FROM+NUM-1].  Otherwise, turn I/O
+   permission off for that range.  This call requires root privileges.
+
+   Portability note: not all Linux platforms support this call.  Most
+   platforms based on the PC I/O architecture probably will, however.
+   E.g., Linux/Alpha for Alpha PCs supports this.  */
+extern int ioperm (unsigned long int __from, unsigned long int __num,
+		   int __turn_on);
+
+/* Set the I/O privilege level to LEVEL.  If LEVEL>3, permission to
+   access any I/O port is granted.  This call requires root
+   privileges. */
+extern int iopl (int __level);
+
+extern unsigned int _inb (unsigned long int __port);
+extern unsigned int _inb (unsigned long int __port);
+extern unsigned int _inw (unsigned long int __port);
+extern unsigned int _inl (unsigned long int __port);
+extern void _outb (unsigned char __val, unsigned long int __port);
+extern void _outw (unsigned short __val, unsigned long int __port);
+extern void _outl (unsigned int __val, unsigned long int __port);
+
+#define inb	_inb
+#define inw	_inw
+#define inl	_inl
+#define outb	_outb
+#define outw	_outw
+#define outl	_outl
+
+/* Access PCI space protected from machine checks.  */
+extern int pciconfig_read (unsigned long int __bus, unsigned long int __dfn,
+			   unsigned long int __off, unsigned long int __len,
+			   unsigned char *__buf);
+
+extern int pciconfig_write (unsigned long int __bus, unsigned long int __dfn,
+			    unsigned long int __off, unsigned long int __len,
+			    unsigned char *__buf);
+
+__END_DECLS
+
+#endif /* _SYS_IO_H */

libc/sysdeps/linux/ia64/sys/procfs.h

@@ -0,0 +1,130 @@
+/* Copyright (C) 1999, 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SYS_PROCFS_H
+#define _SYS_PROCFS_H	1
+
+/* This is somehow modelled after the file of the same name on SysVr4
+   systems.  It provides a definition of the core file format for ELF
+   used on Linux.  */
+
+#include <features.h>
+#include <signal.h>
+#include <sys/time.h>
+#include <sys/types.h>
+#include <sys/ucontext.h>
+#include <sys/user.h>
+
+__BEGIN_DECLS
+
+struct elf_siginfo
+  {
+    int si_signo;			/* Signal number.  */
+    int si_code;			/* Extra code.  */
+    int si_errno;			/* Errno.  */
+  };
+
+/* We really need just 72 but let's leave some headroom...  */
+#define ELF_NGREG	128
+/* f0 and f1 could be omitted, but so what...  */
+#define ELF_NFPREG	128
+
+typedef unsigned long elf_greg_t;
+typedef elf_greg_t elf_gregset_t[ELF_NGREG];
+
+typedef struct ia64_fpreg elf_fpreg_t;
+typedef elf_fpreg_t elf_fpregset_t[ELF_NFPREG];
+
+typedef elf_greg_t greg_t;
+typedef elf_gregset_t gregset_t;
+typedef elf_fpregset_t fpregset_t;
+#define NGREG ELF_NGREG
+
+/* Definitions to generate Intel SVR4-like core files.  These mostly
+   have the same names as the SVR4 types with "elf_" tacked on the
+   front to prevent clashes with linux definitions, and the typedef
+   forms have been avoided.  This is mostly like the SVR4 structure,
+   but more Linuxy, with things that Linux does not support and which
+   gdb doesn't really use excluded.  Fields present but not used are
+   marked with "XXX".  */
+struct elf_prstatus
+  {
+#if 0
+    long int pr_flags;			/* XXX Process flags.  */
+    short int pr_why;			/* XXX Reason for process halt.  */
+    short int pr_what;			/* XXX More detailed reason.  */
+#endif
+    struct elf_siginfo pr_info;		/* Info associated with signal.  */
+    short int pr_cursig;		/* Current signal.  */
+    unsigned long int pr_sigpend;	/* Set of pending signals.  */
+    unsigned long int pr_sighold;	/* Set of held signals.  */
+#if 0
+    struct sigaltstack pr_altstack;	/* Alternate stack info.  */
+    struct sigaction pr_action;		/* Signal action for current sig.  */
+#endif
+    __pid_t pr_pid;
+    __pid_t pr_ppid;
+    __pid_t pr_pgrp;
+    __pid_t pr_sid;
+    struct timeval pr_utime;		/* User time.  */
+    struct timeval pr_stime;		/* System time.  */
+    struct timeval pr_cutime;		/* Cumulative user time.  */
+    struct timeval pr_cstime;		/* Cumulative system time.  */
+#if 0
+    long int pr_instr;			/* Current instruction.  */
+#endif
+    elf_gregset_t pr_reg;		/* GP registers.  */
+    int pr_fpvalid;			/* True if math copro being used.  */
+  };
+
+
+#define ELF_PRARGSZ     (80)    /* Number of chars for args */
+
+struct elf_prpsinfo
+  {
+    char pr_state;			/* Numeric process state.  */
+    char pr_sname;			/* Char for pr_state.  */
+    char pr_zomb;			/* Zombie.  */
+    char pr_nice;			/* Nice val.  */
+    unsigned long int pr_flag;		/* Flags.  */
+    unsigned int pr_uid;
+    unsigned int pr_gid;
+    int pr_pid, pr_ppid, pr_pgrp, pr_sid;
+    /* Lots missing */
+    char pr_fname[16];			/* Filename of executable.  */
+    char pr_psargs[ELF_PRARGSZ];	/* Initial part of arg list.  */
+  };
+
+/* Addresses.  */
+typedef void *psaddr_t;
+
+/* Register sets.  Linux has different names.  */
+typedef gregset_t prgregset_t;
+typedef fpregset_t prfpregset_t;
+
+/* We don't have any differences between processes and threads,
+   therefore habe only ine PID type.  */
+typedef __pid_t lwpid_t;
+
+
+typedef struct elf_prstatus prstatus_t;
+typedef struct elf_prpsinfo prpsinfo_t;
+
+__END_DECLS
+
+#endif	/* sys/procfs.h */

libc/sysdeps/linux/ia64/sys/ptrace.h

@@ -0,0 +1,135 @@
+/* `ptrace' debugger support interface.  Linux/ia64 version.
+   Copyright (C) 2001 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SYS_PTRACE_H
+#define _SYS_PTRACE_H	1
+
+#include <features.h>
+#include <sys/ucontext.h>
+
+__BEGIN_DECLS
+
+/* Type of the REQUEST argument to `ptrace.'  */
+enum __ptrace_request
+{
+  /* Indicate that the process making this request should be traced.
+     All signals received by this process can be intercepted by its
+     parent, and its parent can use the other `ptrace' requests.  */
+  PTRACE_TRACEME = 0,
+#define PT_TRACE_ME PTRACE_TRACEME
+
+  /* Return the word in the process's text space at address ADDR.  */
+  PTRACE_PEEKTEXT = 1,
+#define PT_READ_I PTRACE_PEEKTEXT
+
+  /* Return the word in the process's data space at address ADDR.  */
+  PTRACE_PEEKDATA = 2,
+#define PT_READ_D PTRACE_PEEKDATA
+
+  /* Return the word in the process's user area at offset ADDR.  */
+  PTRACE_PEEKUSER = 3,
+#define PT_READ_U PTRACE_PEEKUSER
+
+  /* Write the word DATA into the process's text space at address ADDR.  */
+  PTRACE_POKETEXT = 4,
+#define PT_WRITE_I PTRACE_POKETEXT
+
+  /* Write the word DATA into the process's data space at address ADDR.  */
+  PTRACE_POKEDATA = 5,
+#define PT_WRITE_D PTRACE_POKEDATA
+
+  /* Write the word DATA into the process's user area at offset ADDR.  */
+  PTRACE_POKEUSER = 6,
+#define PT_WRITE_U PTRACE_POKEUSER
+
+  /* Continue the process.  */
+  PTRACE_CONT = 7,
+#define PT_CONTINUE PTRACE_CONT
+
+  /* Kill the process.  */
+  PTRACE_KILL = 8,
+#define PT_KILL PTRACE_KILL
+
+  /* Single step the process.
+     This is not supported on all machines.  */
+  PTRACE_SINGLESTEP = 9,
+#define PT_STEP PTRACE_SINGLESTEP
+
+  /* Execute process until next taken branch.  */
+  PTRACE_SINGLEBLOCK = 12,
+#define PT_STEPBLOCK PTRACE_SINGLEBLOCK
+
+  /* Get siginfo for process.  */
+  PTRACE_GETSIGINFO = 13,
+#define PT_GETSIGINFO PTRACE_GETSIGINFO
+
+  /* Set new siginfo for process.  */
+  PTRACE_SETSIGINFO = 14,
+#define PT_GETSIGINFO PTRACE_GETSIGINFO
+
+  /* Attach to a process that is already running. */
+  PTRACE_ATTACH = 16,
+#define PT_ATTACH PTRACE_ATTACH
+
+  /* Detach from a process attached to with PTRACE_ATTACH.  */
+  PTRACE_DETACH = 17,
+#define PT_DETACH PTRACE_DETACH
+
+  /* Get all registers (pt_all_user_regs) in one shot */
+  PTRACE_GETREGS = 18,
+#define PT_GETREGS PTRACE_GETREGS
+
+  /* Set all registers (pt_all_user_regs) in one shot */
+  PTRACE_SETREGS = 19,
+#define PT_SETREGS PTRACE_SETREGS
+
+  /* Continue and stop at the next (return from) syscall.  */
+  PTRACE_SYSCALL = 24
+#define PT_SYSCALL PTRACE_SYSCALL
+};
+
+/* pt_all_user_regs is used for PTRACE_GETREGS/PTRACE_SETREGS.  */
+struct pt_all_user_regs
+  {
+    unsigned long nat;
+    unsigned long cr_iip;
+    unsigned long cfm;
+    unsigned long cr_ipsr;
+    unsigned long pr;
+
+    unsigned long gr[32];
+    unsigned long br[8];
+    unsigned long ar[128];
+    struct ia64_fpreg fr[128];
+  };
+
+/* Perform process tracing functions.  REQUEST is one of the values
+   above, and determines the action to be taken.
+   For all requests except PTRACE_TRACEME, PID specifies the process to be
+   traced.
+
+   PID and the other arguments described above for the various requests should
+   appear (those that are used for the particular request) as:
+     pid_t PID, void *ADDR, int DATA, void *ADDR2
+   after REQUEST.  */
+extern long int ptrace (enum __ptrace_request __request, ...) __THROW;
+
+__END_DECLS
+
+#endif /* _SYS_PTRACE_H */

libc/sysdeps/linux/ia64/sys/ucontext.h

@@ -0,0 +1,66 @@
+/* Copyright (C) 1998, 2000, 2001, 2002, 2004 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SYS_UCONTEXT_H
+#define _SYS_UCONTEXT_H	1
+
+#include <features.h>
+#include <signal.h>
+
+#include <bits/sigcontext.h>
+
+/*
+ * These are here mostly for backwards compatibility with older Unices.
+ * IA-64 Linux does not distinguish between "struct sigcontext" and
+ * "ucontext_t" as all the necessary info is inside the former.
+ */
+
+typedef struct sigcontext mcontext_t;
+
+#if defined __cplusplus && __GNUC_PREREQ (3, 5)
+# define _SC_GR0_OFFSET	\
+	__builtin_offsetof (struct sigcontext, sc_gr[0])
+#elif defined __GNUC__
+# define _SC_GR0_OFFSET	\
+	(((char *) &((struct sigcontext *) 0)->sc_gr[0]) - (char *) 0)
+#else
+# define _SC_GR0_OFFSET	0xc8	/* pray that this is correct... */
+#endif
+
+typedef struct ucontext
+  {
+    union
+      {
+	mcontext_t _mc;
+	struct
+	  {
+	    unsigned long _pad[_SC_GR0_OFFSET/8];
+	    struct ucontext *_link;	/* this should overlay sc_gr[0] */
+	  }
+	_uc;
+      }
+    _u;
+  }
+ucontext_t;
+
+#define uc_mcontext	_u._mc
+#define uc_sigmask	_u._mc.sc_mask
+#define uc_stack	_u._mc.sc_stack
+#define uc_link		_u._uc._link
+
+#endif /* sys/ucontext.h */

libc/sysdeps/linux/ia64/sys/user.h

@@ -0,0 +1,54 @@
+/* Copyright (C) 2002 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _SYS_USER_H
+#define _SYS_USER_H	1
+
+#include <features.h>
+#include <sys/types.h>
+
+/* This definition comes directly from the kernel headers.  If
+   anything changes in them this header has to be changed, too.  */
+
+
+/* The definition in the kernel has the comment "XXX fix me".  */
+#define EF_SIZE		3072
+
+
+struct user
+{
+  unsigned long int regs[EF_SIZE / 8 + 32];	/* Integer and fp regs.  */
+  size_t u_tsize;				/* Text size (pages).  */
+  size_t u_dsize;				/* Data size (pages).  */
+  size_t u_ssize;				/* Stack size (pages).  */
+  unsigned long int start_code;			/* Text starting address.  */
+  unsigned long int start_data;			/* Data starting address.  */
+  unsigned long int start_stack;		/* Stack starting address.  */
+  long int signal;				/* Signal causing core dump. */
+  struct regs *u_ar0;				/* Help gdb find registers.  */
+  unsigned long int magic;			/* Identifies a core file.  */
+  char u_comm[32];				/* User command name.  */
+};
+
+#define NBPG			PAGE_SIZE
+#define UPAGES			1
+#define HOST_TEXT_START_ADDR	(u.start_code)
+#define HOST_DATA_START_ADDR	(u.start_data)
+#define HOST_STACK_END_ADDR	(u.start_stack + u.u_ssize * NBPG)
+
+#endif	/* sys/user.h */

libc/sysdeps/linux/ia64/syscall.S

@@ -0,0 +1,30 @@
+/* Copyright (C) 1999, 2000 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Jes Sorensen <Jes.Sorensen@cern.ch>.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "sysdep.h"
+
+ENTRY(syscall)
+	alloc r2=ar.pfs,1,0,8,0
+	mov r15=r32		/* syscall number */
+	break __BREAK_SYSCALL
+	;;
+	cmp.ne p6,p0=-1,r10	/* r10 = -1 on error */
+(p6)	ret
+	br.cond.spnt.few __syscall_error
+PSEUDO_END(syscall)

libc/sysdeps/linux/ia64/sysdep.h

@@ -0,0 +1,168 @@
+/* Copyright (C) 1999, 2000, 2002, 2003, 2004 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Written by Jes Sorensen, <Jes.Sorensen@cern.ch>, April 1999.
+   Based on code originally written by David Mosberger-Tang
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _LINUX_IA64_SYSDEP_H
+#define _LINUX_IA64_SYSDEP_H 1
+
+#include <features.h>
+#include <asm/unistd.h>
+
+#ifdef __ASSEMBLER__
+
+/* Macros to help writing .prologue directives in assembly code.  */
+#define ASM_UNW_PRLG_RP			0x8
+#define ASM_UNW_PRLG_PFS		0x4
+#define ASM_UNW_PRLG_PSP		0x2
+#define ASM_UNW_PRLG_PR			0x1
+#define ASM_UNW_PRLG_GRSAVE(ninputs)	(32+(ninputs))
+
+#ifdef	__STDC__
+#define C_LABEL(name)		name##:
+#else
+#define C_LABEL(name)		name/**/:
+#endif
+
+#define CALL_MCOUNT
+
+#define ENTRY(name)				\
+	.text;					\
+	.align 32;				\
+	.proc C_SYMBOL_NAME(name);		\
+	.global C_SYMBOL_NAME(name);		\
+	C_LABEL(name)				\
+	CALL_MCOUNT
+
+#define LEAF(name)				\
+  .text;					\
+  .align 32;					\
+  .proc C_SYMBOL_NAME(name);			\
+  .global name;					\
+  C_LABEL(name)
+
+/* Mark the end of function SYM.  */
+#undef END
+#define END(sym)	.endp C_SYMBOL_NAME(sym)
+
+/* For Linux we can use the system call table in the header file
+	/usr/include/asm/unistd.h
+   of the kernel.  But these symbols do not follow the SYS_* syntax
+   so we have to redefine the `SYS_ify' macro here.  */
+#undef SYS_ify
+#ifdef __STDC__
+# define SYS_ify(syscall_name)	__NR_##syscall_name
+#else
+# define SYS_ify(syscall_name)	__NR_/**/syscall_name
+#endif
+
+/* Linux uses a negative return value to indicate syscall errors, unlike
+   most Unices, which use the condition codes' carry flag.
+
+   Since version 2.1 the return value of a system call might be negative
+   even if the call succeeded.  E.g., the `lseek' system call might return
+   a large offset.  Therefore we must not anymore test for < 0, but test
+   for a real error by making sure the value in %d0 is a real error
+   number.  Linus said he will make sure the no syscall returns a value
+   in -1 .. -4095 as a valid result so we can savely test with -4095.  */
+
+/* We don't want the label for the error handler to be visible in the symbol
+   table when we define it here.  */
+#define SYSCALL_ERROR_LABEL __syscall_error
+
+#undef PSEUDO
+#define	PSEUDO(name, syscall_name, args)	\
+  ENTRY(name)					\
+    DO_CALL (SYS_ify(syscall_name));		\
+	cmp.eq p6,p0=-1,r10;			\
+(p6)	br.cond.spnt.few __syscall_error;
+
+#define DO_CALL_VIA_BREAK(num)			\
+	mov r15=num;				\
+	break __BREAK_SYSCALL
+
+#ifdef IA64_USE_NEW_STUB
+# ifdef SHARED
+#  define DO_CALL(num)				\
+	.prologue;				\
+	adds r2 = SYSINFO_OFFSET, r13;;		\
+	ld8 r2 = [r2];				\
+	.save ar.pfs, r11;			\
+	mov r11 = ar.pfs;;			\
+	.body;					\
+	mov r15 = num;				\
+	mov b7 = r2;				\
+	br.call.sptk.many b6 = b7;;		\
+	.restore sp;				\
+	mov ar.pfs = r11;			\
+	.prologue;				\
+	.body
+# else /* !SHARED */
+#  define DO_CALL(num)				\
+	.prologue;				\
+	mov r15 = num;				\
+	movl r2 = _dl_sysinfo;;			\
+	ld8 r2 = [r2];				\
+	.save ar.pfs, r11;			\
+	mov r11 = ar.pfs;;			\
+	.body;					\
+	mov b7 = r2;				\
+	br.call.sptk.many b6 = b7;;		\
+	.restore sp;				\
+	mov ar.pfs = r11;			\
+	.prologue;				\
+	.body
+# endif
+#else
+# define DO_CALL(num)				DO_CALL_VIA_BREAK(num)
+#endif
+
+#undef PSEUDO_END
+#define PSEUDO_END(name)	.endp C_SYMBOL_NAME(name);
+
+#undef PSEUDO_NOERRNO
+#define	PSEUDO_NOERRNO(name, syscall_name, args)	\
+  ENTRY(name)						\
+    DO_CALL (SYS_ify(syscall_name));
+
+#undef PSEUDO_END_NOERRNO
+#define PSEUDO_END_NOERRNO(name)	.endp C_SYMBOL_NAME(name);
+
+#undef PSEUDO_ERRVAL
+#define	PSEUDO_ERRVAL(name, syscall_name, args)	\
+  ENTRY(name)					\
+    DO_CALL (SYS_ify(syscall_name));		\
+	cmp.eq p6,p0=-1,r10;			\
+(p6)	mov r10=r8;
+
+
+#undef PSEUDO_END_ERRVAL
+#define PSEUDO_END_ERRVAL(name)	.endp C_SYMBOL_NAME(name);
+
+#undef END
+#define END(name)						\
+	.size	C_SYMBOL_NAME(name), . - C_SYMBOL_NAME(name) ;	\
+	.endp	C_SYMBOL_NAME(name)
+
+#define ret			br.ret.sptk.few b0
+#define ret_NOERRNO		ret
+#define ret_ERRVAL		ret
+
+#endif /* not __ASSEMBLER__ */
+
+#endif /* linux/ia64/sysdep.h */

libc/sysdeps/linux/ia64/vfork.S

@@ -0,0 +1,44 @@
+/* Copyright (C) 2000, 2002 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
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library 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.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+
+#include "sysdep.h"
+#define _SIGNAL_H
+#include <bits/signum.h>
+
+/* The following are defined in linux/sched.h, which unfortunately	*/
+/* is not safe for inclusion in an assembly file.			*/
+#define CLONE_VM        0x00000100      /* set if VM shared between processes */
+#define CLONE_VFORK     0x00004000      /* set if the parent wants the child to wake it up on mm_release */
+
+/* pid_t vfork(void); */
+/* Implemented as __clone_syscall(CLONE_VFORK | CLONE_VM | SIGCHLD, 0)	*/
+
+ENTRY(__vfork)
+	alloc r2=ar.pfs,0,0,2,0
+	mov out0=CLONE_VM+CLONE_VFORK+SIGCHLD
+	mov out1=0		/* Standard sp value.			*/
+	;;
+	DO_CALL_VIA_BREAK (SYS_ify (clone))
+	cmp.eq p6,p0=-1,r10
+(p6)	br.cond.spnt.few __syscall_error
+	ret
+PSEUDO_END(__vfork)
+libc_hidden_def (__vfork)
+
+weak_alias (__vfork, vfork)

test/unistd/clone.c

@@ -28,6 +28,7 @@
 #include <signal.h>
 #include <sched.h>
 #include <sys/wait.h>
+#include "clone_cruft.h"
 
 #define GOT1     (1 << 1)
 #define GOT2     (1 << 2)
@@ -56,15 +57,15 @@ int main(void)
 
 	signal(SIGCHLD, child_handler);
 
-	if ((clone1 = clone(clone_main, clone1_stack, 0, (void*)11)) == -1) {
+	if ((clone1 = do_clone(clone_main, clone1_stack, 0, (void*)11)) == -1) {
 		perror("Clone 1 failed");
 		exit(-1);
 	}
-	if ((clone2 = clone(clone_main, clone2_stack, 0, (void*)22)) == -1) {
+	if ((clone2 = do_clone(clone_main, clone2_stack, 0, (void*)22)) == -1) {
 		perror("Clone 2 failed");
 		exit(-2);
 	}
-	if ((clone3 = clone(clone_main, clone3_stack, 0, (void*)33)) == -1) {
+	if ((clone3 = do_clone(clone_main, clone3_stack, 0, (void*)33)) == -1) {
 		perror("Clone 3 failed");
 		exit(-3);
 	}

test/unistd/clone_cruft.h

@@ -0,0 +1,20 @@
+/* because people like to make things difficult */
+
+#undef do_clone
+
+#if defined __ia64__
+
+# define do_clone(fn, stack, flags, arg) \
+	__clone2(fn, stack, sizeof(stack), flags, arg, NULL, NULL, NULL)
+
+#elif defined __hppa__
+
+# define do_clone(fn, stack, flags, arg) \
+	clone(fn, stack, flags, arg)
+
+#else
+
+# define do_clone(fn, stack, flags, arg) \
+	clone(fn, stack+sizeof(stack), flags, arg)
+
+#endif

test/unistd/errno.c

@@ -4,6 +4,7 @@
 #include <stdlib.h>
 #include <errno.h>
 #include <sched.h>
+#include "clone_cruft.h"
 
 int child_fn(void *arg)
 {
@@ -15,7 +16,7 @@ int main(void)
 {
 	int r_clone, ret_errno;
 
-	r_clone = clone(child_fn, NULL, (int) NULL, NULL);
+	r_clone = do_clone(child_fn, NULL, (int) NULL, NULL);
 	ret_errno = errno;
 	if (ret_errno != EINVAL || r_clone != -1) {
 		fprintf(stderr, "clone: res=%d (wanted -1) errno=%d (wanted %d)\n",