changed configuration stuff, replaced some Makefile.am by distfiles.

14 files changed, 689 insertions, 111 deletions

diff --git a/mpi/Makefile.am b/mpi/Makefile.am
index 33e1ac456..0daf86042 100644
--- a/mpi/Makefile.am
+++ b/mpi/Makefile.am
@@ -5,12 +5,11 @@ CFLAGS += -O2
 
 SUFFIXES = .S .s
 
-SUBDIRS = generic i386
 EXTRA_DIST = config.links
 
 
 noinst_LIBRARIES = mpi
-noinst_HEADERS	 = sysdep.h
+# noinst_HEADERS   =
 
 mpi_SOURCES = longlong.h     \
 	      mpi-add.c      \
diff --git a/mpi/Makefile.in b/mpi/Makefile.in
index 382a222cd..c179f53f9 100644
--- a/mpi/Makefile.in
+++ b/mpi/Makefile.in
@@ -42,11 +42,10 @@ INCLUDES =  -I$(top_srcdir)/include
 
 SUFFIXES = .S .s
 
-SUBDIRS = generic i386
 EXTRA_DIST = config.links
 
 noinst_LIBRARIES = mpi
-noinst_HEADERS	 = sysdep.h
+# noinst_HEADERS   =
 
 mpi_SOURCES = longlong.h     \
 	      mpi-add.c      \
@@ -99,8 +98,6 @@ EXTRA_mpi_SOURCES =
 LIBFILES =  libmpi.a
 AR = ar
 RANLIB = @RANLIB@
-HEADERS = $(noinst_HEADERS)
-
 DIST_COMMON = Makefile.am Makefile.in
 
 
@@ -161,45 +158,13 @@ libmpi.a: $(mpi_OBJECTS) $(mpi_LIBADD)
 	$(AR) cru libmpi.a $(mpi_OBJECTS) $(mpi_LIBADD)
 	$(RANLIB) libmpi.a
 
-# This directory's subdirectories are mostly independent; you can cd
-# into them and run `make' without going through this Makefile.
-# To change the values of `make' variables: instead of editing Makefiles,
-# (1) if the variable is set in `config.status', edit `config.status'
-#     (which will cause the Makefiles to be regenerated when you run `make');
-# (2) otherwise, pass the desired values on the `make' command line.
-
-@SET_MAKE@
-
-all-recursive install-data-recursive install-exec-recursive \
-installdirs-recursive install-recursive uninstall-recursive \
-check-recursive installcheck-recursive info-recursive dvi-recursive \
-mostlyclean-recursive clean-recursive distclean-recursive \
-maintainer-clean-recursive:
-	for subdir in $(SUBDIRS); do		\
-	  target=`echo $@ | sed s/-recursive//`; \
-	  echo making $$target in $$subdir;	\
-	  (cd $$subdir && $(MAKE) $$target)	\
-	   || case "$(MFLAGS)" in *k*) fail=yes;; *) exit 1;; esac; \
-	done && test -z "$$fail"
+ID: $(HEADERS) $(SOURCES)
+	here=`pwd` && cd $(srcdir) && mkid -f$$here/ID $(SOURCES) $(HEADERS)
 
 tags: TAGS
 
-tags-recursive:
-	list="$(SUBDIRS)"; for subdir in $$list; do \
-	  (cd $$subdir && $(MAKE) tags); \
-	done
-
-TAGS: tags-recursive $(HEADERS) $(SOURCES) $(CONFIG_HEADER) \
-		$(TAGS_DEPENDENCIES)
-	tags=; \
-	here=`pwd`; \
-	for subdir in $(SUBDIRS); do \
-	  test -f $$subdir/TAGS && { \
-	    tags="$$tags -i $$here/$$subdir/TAGS"; \
-	  }; \
-	done; \
-	test -z "$(ETAGS_ARGS)$(CONFIG_HEADER)$(SOURCES)$(HEADERS)$$tags" \
-	  || etags $(ETAGS_ARGS) $$tags $(CONFIG_HEADER) $(SOURCES) $(HEADERS)
+TAGS: $(HEADERS) $(SOURCES) $(TAGS_DEPENDENCIES)
+	here=`pwd` && cd $(srcdir) && etags $(ETAGS_ARGS) $(SOURCES) $(HEADERS) -o $$here/TAGS
 
 mostlyclean-tags:
 
@@ -218,14 +183,6 @@ distdir: $(DEP_DISTFILES)
 	  || ln $(srcdir)/$$file $(distdir)/$$file 2> /dev/null \
 	  || cp -p $(srcdir)/$$file $(distdir)/$$file; \
 	done
-	for subdir in $(SUBDIRS); do		\
-	  test -d $(distdir)/$$subdir		\
-	  || mkdir $(distdir)/$$subdir		\
-	  || exit 1;				\
-	  chmod 777 $(distdir)/$$subdir;	\
-	  (cd $$subdir && $(MAKE) distdir=../$(distdir)/$$subdir distdir) \
-	    || exit 1; \
-	done
 
 # This fragment is probably only useful for maintainers.  It relies on
 # GNU make and gcc.  It is only included in the generated Makefile.in
@@ -253,30 +210,28 @@ $(srcdir)/.deps/%.P: $(srcdir)/%.c
 	fi
 
 # End of maintainer-only section
-info: info-recursive
-
-dvi: dvi-recursive
+info:
 
-check: all check-recursive
+dvi:
 
-installcheck: installcheck-recursive
+check: all
 
-all-am: $(LIBFILES) $(HEADERS) Makefile
+installcheck:
 
-install-exec: install-exec-recursive
+install-exec: 
 
-install-data: install-data-recursive
+install-data: 
 
-install: install-recursive
+install: install-exec install-data all
 	@:
 
-uninstall: uninstall-recursive
+uninstall: 
 
-all: all-recursive all-am
+all: $(LIBFILES) Makefile
 
 install-strip:
 	$(MAKE) INSTALL_PROGRAM='$(INSTALL_PROGRAM) -s' install
-installdirs: installdirs-recursive
+installdirs:
 
 
 mostlyclean-generic:
@@ -292,42 +247,29 @@ distclean-generic:
 maintainer-clean-generic:
 	test -z "$(MAINTAINERCLEANFILES)" || rm -f $(MAINTAINERCLEANFILES)
 	test -z "$(BUILT_SOURCES)" || rm -f $(BUILT_SOURCES)
-mostlyclean-am:  mostlyclean-noinstLIBRARIES mostlyclean-compile \
+mostlyclean:  mostlyclean-noinstLIBRARIES mostlyclean-compile \
 		mostlyclean-tags mostlyclean-generic
 
-clean-am:  clean-noinstLIBRARIES clean-compile clean-tags clean-generic \
-		mostlyclean-am 
-
-distclean-am:  distclean-noinstLIBRARIES distclean-compile \
-		distclean-tags distclean-generic clean-am 
+clean:  clean-noinstLIBRARIES clean-compile clean-tags clean-generic \
+		mostlyclean 
 
-maintainer-clean-am:  maintainer-clean-noinstLIBRARIES \
-		maintainer-clean-compile maintainer-clean-tags \
-		maintainer-clean-generic distclean-am 
-
-mostlyclean:  mostlyclean-am mostlyclean-recursive
-
-clean:  clean-am clean-recursive
-
-distclean:  distclean-am distclean-recursive
+distclean:  distclean-noinstLIBRARIES distclean-compile distclean-tags \
+		distclean-generic clean 
 	rm -f config.status
 
-maintainer-clean:  maintainer-clean-am maintainer-clean-recursive
+maintainer-clean:  maintainer-clean-noinstLIBRARIES \
+		maintainer-clean-compile maintainer-clean-tags \
+		maintainer-clean-generic distclean 
 	@echo "This command is intended for maintainers to use;"
 	@echo "it deletes files that may require special tools to rebuild."
 
 .PHONY: default mostlyclean-noinstLIBRARIES distclean-noinstLIBRARIES \
 clean-noinstLIBRARIES maintainer-clean-noinstLIBRARIES \
 mostlyclean-compile distclean-compile clean-compile \
-maintainer-clean-compile install-data-recursive \
-uninstall-data-recursive install-exec-recursive \
-uninstall-exec-recursive installdirs-recursive uninstalldirs-recursive \
-all-recursive check-recursive installcheck-recursive info-recursive \
-dvi-recursive mostlyclean-recursive distclean-recursive clean-recursive \
-maintainer-clean-recursive tags tags-recursive mostlyclean-tags \
-distclean-tags clean-tags maintainer-clean-tags distdir info dvi check \
-installcheck all-am install-exec install-data install uninstall all \
-installdirs mostlyclean-generic distclean-generic clean-generic \
+maintainer-clean-compile tags mostlyclean-tags distclean-tags \
+clean-tags maintainer-clean-tags distdir info dvi check installcheck \
+install-exec install-data install uninstall all installdirs \
+mostlyclean-generic distclean-generic clean-generic \
 maintainer-clean-generic clean mostlyclean distclean maintainer-clean
 
 CFLAGS += -O2
diff --git a/mpi/alpha/distfiles b/mpi/alpha/distfiles
new file mode 100644
index 000000000..4dd0ffe3a
--- /dev/null
+++ b/mpi/alpha/distfiles
@@ -0,0 +1,3 @@
+
+udiv-qrnnd.S
+
diff --git a/mpi/alpha/udiv-qrnnd.S b/mpi/alpha/udiv-qrnnd.S
new file mode 100644
index 000000000..487991cd8
--- /dev/null
+++ b/mpi/alpha/udiv-qrnnd.S
@@ -0,0 +1,161 @@
+/* Alpha 21064 __udiv_qrnnd
+ *
+ * Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
+ * Copyright (c) 1997 by Werner Koch (dd9jn)
+ *
+ * This file is part of G10.
+ *
+ * G10 is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * G10 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
+ *
+ * Note: This code is heavily based on the GNU MP Library.
+ *	 Actually it's the same code with only minor changes in the
+ *	 way the data is stored; this is to support the abstraction
+ *	 of an optional secure memory allocation which may be used
+ *	 to avoid revealing of sensitive data due to paging etc.
+ *	 The GNU MP Library itself is published under the LGPL;
+ *	 however I decided to publish this code under the plain GPL.
+ */
+
+
+
+	.set noreorder
+	.set noat
+.text
+	.align	3
+	.globl	__udiv_qrnnd
+	.ent	__udiv_qrnnd
+__udiv_qrnnd:
+	.frame $30,0,$26,0
+	.prologue 0
+#define cnt	$2
+#define tmp	$3
+#define rem_ptr $16
+#define n1	$17
+#define n0	$18
+#define d	$19
+#define qb	$20
+
+	ldiq	cnt,16
+	blt	d,.Largedivisor
+
+.Loop1: cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	d,n1,qb
+	subq	n1,d,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	d,n1,qb
+	subq	n1,d,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	d,n1,qb
+	subq	n1,d,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	d,n1,qb
+	subq	n1,d,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	subq	cnt,1,cnt
+	bgt	cnt,.Loop1
+	stq	n1,0(rem_ptr)
+	bis	$31,n0,$0
+	ret	$31,($26),1
+
+.Largedivisor:
+	and	n0,1,$4
+
+	srl	n0,1,n0
+	sll	n1,63,tmp
+	or	tmp,n0,n0
+	srl	n1,1,n1
+
+	and	d,1,$6
+	srl	d,1,$5
+	addq	$5,$6,$5
+
+.Loop2: cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	$5,n1,qb
+	subq	n1,$5,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	$5,n1,qb
+	subq	n1,$5,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	$5,n1,qb
+	subq	n1,$5,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	cmplt	n0,0,tmp
+	addq	n1,n1,n1
+	bis	n1,tmp,n1
+	addq	n0,n0,n0
+	cmpule	$5,n1,qb
+	subq	n1,$5,tmp
+	cmovne	qb,tmp,n1
+	bis	n0,qb,n0
+	subq	cnt,1,cnt
+	bgt	cnt,.Loop2
+
+	addq	n1,n1,n1
+	addq	$4,n1,n1
+	bne	$6,.LOdd
+	stq	n1,0(rem_ptr)
+	bis	$31,n0,$0
+	ret	$31,($26),1
+
+.LOdd:
+	/* q' in n0. r' in n1 */
+	addq	n1,n0,n1
+	cmpult	n1,n0,tmp	# tmp := carry from addq
+	beq	tmp,.LLp6
+	addq	n0,1,n0
+	subq	n1,d,n1
+.LLp6:	cmpult	n1,d,tmp
+	bne	tmp,.LLp7
+	addq	n0,1,n0
+	subq	n1,d,n1
+.LLp7:
+	stq	n1,0(rem_ptr)
+	bis	$31,n0,$0
+	ret	$31,($26),1
+
+	.end	__udiv_qrnnd
diff --git a/mpi/config.links b/mpi/config.links
index e48cf7a08..0560b7de5 100644
--- a/mpi/config.links
+++ b/mpi/config.links
@@ -2,9 +2,22 @@
 # this should set $mpi_ln_src and mpi_ln_dst.
 # Note: this is called from the above directory.
 
-echo '# created by config.links - do not edit' >./mpi/asm-syntax.h
+
+mpi_extra_modules=
+
+echo '/* created by config.links - do not edit */' >./mpi/asm-syntax.h
 
 case "${target}" in
+    i[345]86*-*-linuxaout* | i[345]86*-*-linuxoldld* | i[345]86*-*-*bsd*)
+	echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h
+	echo '#include "./i386/syntax.h"' >>./mpi/asm-syntax.h
+	path="i386"
+	;;
+    i[56]86*-*-linuxaout* | i[56]86*-*-linuxoldld* | i[56]86*-*-*bsd*)
+	echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h
+	echo '#include "./i386/syntax.h"' >>./mpi/asm-syntax.h
+	path="i586 i386"
+	;;
     i[3456]86*-*-*)
 	echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h
 	echo '#include "./i386/syntax.h"' >>./mpi/asm-syntax.h
@@ -13,7 +26,27 @@ case "${target}" in
     i[56]86*-*-* | pentium-*-* | pentiumpro-*-*)
 	echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h
 	echo '#include "./i586/syntax.h"' >>./mpi/asm-syntax.h
-	path="i586"
+	path="i586 i386"
+	;;
+    alpha*-*-*)
+	echo '/* configured for alpha */' >>./mpi/asm-syntax.h
+	path="alpha"
+	mpi_extra_modules="udiv-qrnnd"
+	;;
+    hppa7000*-*-*)
+	echo '/* configured for HPPA (pa7000) */' >>./mpi/asm-syntax.h
+	path="hppa1_1 hppa"
+	mpi_extra_modules="udiv-qrnnd"
+	;;
+    hppa1.0*-*-*)
+	echo '/* configured for HPPA 1.0 */' >>./mpi/asm-syntax.h
+	path="hppa"
+	mpi_extra_modules="udiv-qrnnd"
+	;;
+    hppa*-*-*)	# assume pa7100
+	echo '/* configured for HPPA (pa7100) */' >>./mpi/asm-syntax.h
+	path="pa7100 hppa1_1 hppa"
+	mpi_extra_modules="udiv-qrnnd"
 	;;
     *)
 	echo '/* No assembler modules configured */' >>./mpi/asm-syntax.h
@@ -21,9 +54,38 @@ case "${target}" in
 	;;
 esac
 
+case "${target}" in
+    *-*-linuxaout* | *-*-linuxoldld*)
+	needs_underscore="y"
+	;;
+    *-*-linux* | *-sysv* | *-solaris*)
+	needs_underscore="n"
+	;;
+    *)
+	needs_underscore="y"
+	;;
+esac
+
+
+# Make sysdep.h
+echo '/* created by config.links - do not edit */' >./mpi/sysdep.h
+if test "$needs_underscore" = "y" ; then
+    cat <<EOF >>./mpi/sysdep.h
+#if __STDC__
+#define C_SYMBOL_NAME(name) _##name
+#else
+#define C_SYMBOL_NAME(name) _/**/name
+#endif
+EOF
+else
+    cat <<EOF >>./mpi/sysdep.h
+#define C_SYMBOL_NAME(name) name
+EOF
+fi
+
 
 # fixme: grep these modules from Makefile.in
-mpi_ln_modules="mpih-add1 mpih-mul1 mpih-mul2 mpih-mul3 \
+mpi_ln_modules="${mpi_extra_modules} mpih-add1 mpih-mul1 mpih-mul2 mpih-mul3 \
 		 mpih-shift mpih-sub1"
 
 mpi_ln_objects=
diff --git a/mpi/generic/distfiles b/mpi/generic/distfiles
new file mode 100644
index 000000000..1febb49dd
--- /dev/null
+++ b/mpi/generic/distfiles
@@ -0,0 +1,7 @@
+mpih-add1.c
+mpih-mul1.c
+mpih-mul2.c
+mpih-mul3.c
+mpih-shift.c
+mpih-sub1.c
+
diff --git a/mpi/hppa/distfiles b/mpi/hppa/distfiles
new file mode 100644
index 000000000..7ca77f801
--- /dev/null
+++ b/mpi/hppa/distfiles
@@ -0,0 +1,4 @@
+
+mpih-add1.S
+udiv-qrnnd.S
+
diff --git a/mpi/hppa/mpih-add1.S b/mpi/hppa/mpih-add1.S
new file mode 100644
index 000000000..a30e80b83
--- /dev/null
+++ b/mpi/hppa/mpih-add1.S
@@ -0,0 +1,70 @@
+/* hppa add_n -- Add two limb vectors of the same length > 0 and store
+ *		 sum in a third limb vector.
+ *
+ *	Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+ *	Copyright (c) 1997 by Werner Koch (dd9jn)
+ *
+ * This file is part of G10.
+ *
+ * G10 is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * G10 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
+ *
+ * Note: This code is heavily based on the GNU MP Library.
+ *	 Actually it's the same code with only minor changes in the
+ *	 way the data is stored; this is to support the abstraction
+ *	 of an optional secure memory allocation which may be used
+ *	 to avoid revealing of sensitive data due to paging etc.
+ *	 The GNU MP Library itself is published under the LGPL;
+ *	 however I decided to publish this code under the plain GPL.
+ */
+
+
+
+/*******************
+ *  mpi_limb_t
+ *  mpihelp_add_n( mpi_ptr_t res_ptr,	(gr26)
+ *		   mpi_ptr_t s1_ptr,	(gr25)
+ *		   mpi_ptr_t s2_ptr,	(gr24)
+ *		   mpi_size_t size)	(gr23)
+ *
+ * One might want to unroll this as for other processors, but it turns
+ * out that the data cache contention after a store makes such
+ * unrolling useless.  We can't come under 5 cycles/limb anyway.
+ */
+
+	.code
+	.export 	__mpihelp_add_n
+__mpihelp_add_n
+	.proc
+	.callinfo	frame=0,no_calls
+	.entry
+
+	ldws,ma 	4(0,%r25),%r20
+	ldws,ma 	4(0,%r24),%r19
+
+	addib,= 	-1,%r23,L$end	; check for (SIZE == 1)
+	 add		%r20,%r19,%r28	; add first limbs ignoring cy
+
+L$loop	ldws,ma 	4(0,%r25),%r20
+	ldws,ma 	4(0,%r24),%r19
+	stws,ma 	%r28,4(0,%r26)
+	addib,<>	-1,%r23,L$loop
+	 addc		%r20,%r19,%r28
+
+L$end	stws		%r28,0(0,%r26)
+	bv		0(%r2)
+	 addc		%r0,%r0,%r28
+
+	.exit
+	.procend
diff --git a/mpi/hppa/udiv-qrnnd.S b/mpi/hppa/udiv-qrnnd.S
new file mode 100644
index 000000000..849238349
--- /dev/null
+++ b/mpi/hppa/udiv-qrnnd.S
@@ -0,0 +1,297 @@
+/* HP-PA  __udiv_qrnnd division support, used from longlong.h.
+ *	  This version runs fast on pre-PA7000 CPUs.
+ *
+ *	Copyright (C) 1993, 1994 Free Software Foundation, Inc.
+ *	Copyright (c) 1997 by Werner Koch (dd9jn)
+ *
+ * This file is part of G10.
+ *
+ * G10 is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * G10 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
+ *
+ * Note: This code is heavily based on the GNU MP Library.
+ *	 Actually it's the same code with only minor changes in the
+ *	 way the data is stored; this is to support the abstraction
+ *	 of an optional secure memory allocation which may be used
+ *	 to avoid revealing of sensitive data due to paging etc.
+ *	 The GNU MP Library itself is published under the LGPL;
+ *	 however I decided to publish this code under the plain GPL.
+ */
+
+
+
+/* INPUT PARAMETERS
+ *   rem_ptr	   gr26
+ *   n1 	   gr25
+ *   n0 	   gr24
+ *   d		   gr23
+ *
+ *   The code size is a bit excessive.	We could merge the last two ds;addc
+ *   sequences by simply moving the "bb,< Odd" instruction down.  The only
+ *   trouble is the FFFFFFFF code that would need some hacking.
+ */
+
+	.code
+	.export 	__udiv_qrnnd
+__udiv_qrnnd
+	.proc
+	.callinfo	frame=0,no_calls
+	.entry
+
+	comb,<		%r23,0,L$largedivisor
+	 sub		%r0,%r23,%r1		; clear cy as side-effect
+	ds		%r0,%r1,%r0
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r23,%r25
+	addc		%r24,%r24,%r28
+	ds		%r25,%r23,%r25
+	comclr,>=	%r25,%r0,%r0
+	addl		%r25,%r23,%r25
+	stws		%r25,0(0,%r26)
+	bv		0(%r2)
+	 addc		%r28,%r28,%r28
+
+L$largedivisor
+	extru		%r24,31,1,%r19		; r19 = n0 & 1
+	bb,<		%r23,31,L$odd
+	 extru		%r23,30,31,%r22 	; r22 = d >> 1
+	shd		%r25,%r24,1,%r24	; r24 = new n0
+	extru		%r25,30,31,%r25 	; r25 = new n1
+	sub		%r0,%r22,%r21
+	ds		%r0,%r21,%r0
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	comclr,>=	%r25,%r0,%r0
+	addl		%r25,%r22,%r25
+	sh1addl 	%r25,%r19,%r25
+	stws		%r25,0(0,%r26)
+	bv		0(%r2)
+	 addc		%r24,%r24,%r28
+
+L$odd	addib,sv,n	1,%r22,L$FF..		; r22 = (d / 2 + 1)
+	shd		%r25,%r24,1,%r24	; r24 = new n0
+	extru		%r25,30,31,%r25 	; r25 = new n1
+	sub		%r0,%r22,%r21
+	ds		%r0,%r21,%r0
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r24
+	ds		%r25,%r22,%r25
+	addc		%r24,%r24,%r28
+	comclr,>=	%r25,%r0,%r0
+	addl		%r25,%r22,%r25
+	sh1addl 	%r25,%r19,%r25
+; We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25
+	add,nuv 	%r28,%r25,%r25
+	addl		%r25,%r1,%r25
+	addc		%r0,%r28,%r28
+	sub,<<		%r25,%r23,%r0
+	addl		%r25,%r1,%r25
+	stws		%r25,0(0,%r26)
+	bv		0(%r2)
+	 addc		%r0,%r28,%r28
+
+; This is just a special case of the code above.
+; We come here when d == 0xFFFFFFFF
+L$FF..	add,uv		%r25,%r24,%r24
+	sub,<<		%r24,%r23,%r0
+	ldo		1(%r24),%r24
+	stws		%r24,0(0,%r26)
+	bv		0(%r2)
+	 addc		%r0,%r25,%r28
+
+	.exit
+	.procend
diff --git a/mpi/i386/distfiles b/mpi/i386/distfiles
new file mode 100644
index 000000000..34de91574
--- /dev/null
+++ b/mpi/i386/distfiles
@@ -0,0 +1,8 @@
+mpih-add1.S
+mpih-mul1.S
+mpih-mul2.S
+mpih-mul3.S
+mpih-shift.S
+mpih-sub1.S
+syntax.h
+
diff --git a/mpi/mpi-inline.h b/mpi/mpi-inline.h
index 4d19942ad..03b5fbc86 100644
--- a/mpi/mpi-inline.h
+++ b/mpi/mpi-inline.h
@@ -123,5 +123,4 @@ mpihelp_sub( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size,
 }
 
 
-
 #endif /*G10_MPI_INLINE_H*/
diff --git a/mpi/mpi-internal.h b/mpi/mpi-internal.h
index 93ed688ae..f084c7e8a 100644
--- a/mpi/mpi-internal.h
+++ b/mpi/mpi-internal.h
@@ -32,6 +32,21 @@
 
 #include "mpi.h"
 
+/* If KARATSUBA_THRESHOLD is not already defined, define it to a
+ * value which is good on most machines.  */
+
+/* tested 4, 16, 32 and 64, where 16 gave the best performance when
+ * checking a 768 and a 1024 bit ElGamal signature.
+ * (wk 22.12.97) */
+#ifndef KARATSUBA_THRESHOLD
+    #define KARATSUBA_THRESHOLD 16
+#endif
+
+/* The code can't handle KARATSUBA_THRESHOLD smaller than 2.  */
+#if KARATSUBA_THRESHOLD < 2
+    #undef KARATSUBA_THRESHOLD
+    #define KARATSUBA_THRESHOLD 2
+#endif
 
 
 typedef mpi_limb_t *mpi_ptr_t; /* pointer to a limb */
@@ -174,6 +189,9 @@ void mpihelp_mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp,
 						   mpi_size_t size);
 mpi_limb_t mpihelp_mul( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t usize,
 					 mpi_ptr_t vp, mpi_size_t vsize);
+void mpih_sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size );
+void mpih_sqr_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size,
+						mpi_ptr_t tspace);
 
 /*-- mpihelp-mul_1.c (or xxx/cpu/*.S) --*/
 mpi_limb_t mpihelp_mul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr,
diff --git a/mpi/mpi-pow.c b/mpi/mpi-pow.c
index 43514567b..fcf500c36 100644
--- a/mpi/mpi-pow.c
+++ b/mpi/mpi-pow.c
@@ -51,6 +51,8 @@ mpi_powm( MPI res, MPI base, MPI exp, MPI mod)
     mpi_ptr_t mp_marker=NULL, bp_marker=NULL, ep_marker=NULL;
     mpi_ptr_t xp_marker=NULL;
     int assign_rp=0;
+    mpi_ptr_t tspace = NULL;
+    mpi_size_t tsize;
 
     esize = exp->nlimbs;
     msize = mod->nlimbs;
@@ -179,7 +181,23 @@ mpi_powm( MPI res, MPI base, MPI exp, MPI mod)
 		mpi_ptr_t tp;
 		mpi_size_t xsize;
 
-		mpihelp_mul_n(xp, rp, rp, rsize);
+		/*mpihelp_mul_n(xp, rp, rp, rsize);*/
+		if( rsize < KARATSUBA_THRESHOLD )
+		    mpih_sqr_n_basecase( xp, rp, rsize );
+		else {
+		    if( !tspace ) {
+			tsize = 2 * rsize;
+			tspace = mpi_alloc_limb_space( tsize, 0 );
+		    }
+		    else if( tsize < (2*rsize) ) {
+			mpi_free_limb_space( tspace );
+			tsize = 2 * rsize;
+			tspace = mpi_alloc_limb_space( tsize, 0 );
+
+		    }
+		    mpih_sqr_n( xp, rp, rsize, tspace );
+		}
+
 		xsize = 2 * rsize;
 		if( xsize > msize ) {
 		    mpihelp_divrem(xp + msize, 0, xp, xsize, mp, msize);
@@ -258,5 +276,6 @@ mpi_powm( MPI res, MPI base, MPI exp, MPI mod)
     if( bp_marker ) mpi_free_limb_space( bp_marker );
     if( ep_marker ) mpi_free_limb_space( ep_marker );
     if( xp_marker ) mpi_free_limb_space( xp_marker );
+    if( tspace )    mpi_free_limb_space( tspace );
 }
 
diff --git a/mpi/mpih-mul.c b/mpi/mpih-mul.c
index 0e52488ec..3b6b732b6 100644
--- a/mpi/mpih-mul.c
+++ b/mpi/mpih-mul.c
@@ -33,17 +33,6 @@
 #include "mpi-internal.h"
 #include "longlong.h"
 
-/* If KARATSUBA_THRESHOLD is not already defined, define it to a
- * value which is good on most machines.  */
-#ifndef KARATSUBA_THRESHOLD
-    #define KARATSUBA_THRESHOLD 32
-#endif
-
-/* The code can't handle KARATSUBA_THRESHOLD smaller than 2.  */
-#if KARATSUBA_THRESHOLD < 2
-    #undef KARATSUBA_THRESHOLD
-    #define KARATSUBA_THRESHOLD 2
-#endif
 
 
 #define MPN_MUL_N_RECURSE(prodp, up, vp, size, tspace) \
@@ -57,9 +46,9 @@
 #define MPN_SQR_N_RECURSE(prodp, up, size, tspace) \
     do {					    \
 	if ((size) < KARATSUBA_THRESHOLD)	    \
-	    sqr_n_basecase (prodp, up, size);	    \
+	    mpih_sqr_n_basecase (prodp, up, size);	 \
 	else					    \
-	    sqr_n (prodp, up, size, tspace);	    \
+	    mpih_sqr_n (prodp, up, size, tspace);	 \
     } while (0);
 
 
@@ -235,8 +224,8 @@ mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp,
 }
 
 
-static void
-sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size )
+void
+mpih_sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size )
 {
     mpi_size_t i;
     mpi_limb_t cy_limb;
@@ -276,8 +265,8 @@ sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size )
 }
 
 
-static void
-sqr_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size, mpi_ptr_t tspace)
+void
+mpih_sqr_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size, mpi_ptr_t tspace)
 {
     if( size & 1 ) {
 	/* The size is odd, the code code below doesn't handle that.
@@ -361,11 +350,11 @@ mpihelp_mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp, mpi_size_t size)
 
     if( up == vp ) {
 	if( size < KARATSUBA_THRESHOLD )
-	    sqr_n_basecase( prodp, up, size );
+	    mpih_sqr_n_basecase( prodp, up, size );
 	else {
 	    mpi_ptr_t tspace;
 	    tspace = mpi_alloc_limb_space( 2 * size, 0 );
-	    sqr_n( prodp, up, size, tspace );
+	    mpih_sqr_n( prodp, up, size, tspace );
 	    mpi_free_limb_space( tspace );
 	}
     }