From 0c1f3f869cd136e43479fc934eb4cfa3e4b1e3f5 Mon Sep 17 00:00:00 2001
From: roland <roland>
Date: Fri, 1 Mar 1996 18:43:01 +0000
Subject: [PATCH] Updated from /src/gmp-1.937

---
 sysdeps/alpha/alphaev5/add_n.s  | 175 +++++++++++++++++++++++-----------------
 sysdeps/alpha/alphaev5/lshift.s |  30 ++++---
 sysdeps/alpha/alphaev5/rshift.s |  28 +++----
 sysdeps/alpha/alphaev5/sub_n.s  | 148 +++++++++++++++++++++++++++++++++
 sysdeps/generic/add_n.c         |   8 +-
 sysdeps/generic/addmul_1.c      |   6 +-
 sysdeps/generic/cmp.c           |   8 +-
 sysdeps/generic/divmod_1.c      |   8 +-
 sysdeps/generic/lshift.c        |   8 +-
 sysdeps/generic/mod_1.c         |   8 +-
 sysdeps/generic/mul.c           |  36 +++++----
 sysdeps/generic/mul_1.c         |   6 +-
 sysdeps/generic/mul_n.c         | 110 ++++++++++++-------------
 sysdeps/generic/rshift.c        |   8 +-
 sysdeps/mips/addmul_1.s         |   4 +-
 sysdeps/mips/mips3/addmul_1.s   |   2 +-
 sysdeps/mips/mips3/mul_1.s      |   2 +-
 sysdeps/mips/mips3/submul_1.s   |   2 +-
 sysdeps/mips/mul_1.s            |   4 +-
 sysdeps/mips/submul_1.s         |   4 +-
 sysdeps/rs6000/add_n.s          |   2 +-
 sysdeps/rs6000/sub_n.s          |   2 +-
 22 files changed, 391 insertions(+), 218 deletions(-)
 create mode 100644 sysdeps/alpha/alphaev5/sub_n.s

diff --git a/sysdeps/alpha/alphaev5/add_n.s b/sysdeps/alpha/alphaev5/add_n.s
index 2aaf041774..66cf82b3c3 100644
--- a/sysdeps/alpha/alphaev5/add_n.s
+++ b/sysdeps/alpha/alphaev5/add_n.s
@@ -35,84 +35,113 @@
 __mpn_add_n:
 	.frame	$30,0,$26,0
 
-	ldq	$3,0($17)
-	ldq	$4,0($18)
-
-	subq	$19,1,$19
-	and	$19,4-1,$2	# number of limbs in first loop
-	bis	$31,$31,$0
-	beq	$2,.L0		# if multiple of 4 limbs, skip first loop
-
-	subq	$19,$2,$19
-
-.Loop0:	subq	$2,1,$2
+	or	$31,$31,$25		# clear cy
+	subq	$19,4,$19		# decr loop cnt
+	blt	$19,.Lend2		# if less than 4 limbs, goto 2nd loop
+ # Start software pipeline for 1st loop
+	ldq	$0,0($18)
+	ldq	$1,8($18)
+	ldq	$4,0($17)
 	ldq	$5,8($17)
-	addq	$4,$0,$4
-	ldq	$6,8($18)
-	cmpult	$4,$0,$1
-	addq	$3,$4,$4
-	cmpult	$4,$3,$0
-	stq	$4,0($16)
-	or	$0,$1,$0
-
-	addq	$17,8,$17
-	addq	$18,8,$18
-	bis	$5,$5,$3
-	bis	$6,$6,$4
-	addq	$16,8,$16
-	bne	$2,.Loop0
-
-.L0:	beq	$19,.Lend
-
+	addq	$17,32,$17		# update s1_ptr
+	ldq	$2,16($18)
+	addq	$0,$4,$20		# 1st main add
+	ldq	$3,24($18)
+	subq	$19,4,$19		# decr loop cnt
+	ldq	$6,-16($17)
+	cmpult	$20,$0,$25		# compute cy from last add
+	ldq	$7,-8($17)
+	addq	$1,$25,$28		# cy add
+	addq	$18,32,$18		# update s2_ptr
+	addq	$5,$28,$21		# 2nd main add
+	cmpult	$28,$25,$8		# compute cy from last add
+	blt	$19,.Lend1		# if less than 4 limbs remain, jump
+ # 1st loop handles groups of 4 limbs in a software pipeline
 	.align	4
-.Loop:	subq	$19,4,$19
-	unop
-
-	ldq	$6,8($18)
-	addq	$4,$0,$0
+.Loop:	cmpult	$21,$28,$25		# compute cy from last add
+	ldq	$0,0($18)
+	or	$8,$25,$25		# combine cy from the two adds
+	ldq	$1,8($18)
+	addq	$2,$25,$28		# cy add
+	ldq	$4,0($17)
+	addq	$28,$6,$22		# 3rd main add
 	ldq	$5,8($17)
-	cmpult	$0,$4,$1
-	ldq	$4,16($18)
-	addq	$3,$0,$20
-	cmpult	$20,$3,$0
-	ldq	$3,16($17)
-	or	$0,$1,$0
-	addq	$6,$0,$0
-	cmpult	$0,$6,$1
-	ldq	$6,24($18)
-	addq	$5,$0,$21
-	cmpult	$21,$5,$0
-	ldq	$5,24($17)
-	or	$0,$1,$0
-	addq	$4,$0,$0
-	cmpult	$0,$4,$1
-	ldq	$4,32($18)
-	addq	$3,$0,$22
-	cmpult	$22,$3,$0
-	ldq	$3,32($17)
-	or	$0,$1,$0
-	addq	$6,$0,$0
-	cmpult	$0,$6,$1
-	addq	$5,$0,$23
-	cmpult	$23,$5,$0
-	or	$0,$1,$0
-
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$22,$28,$25		# compute cy from last add
 	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
 	stq	$21,8($16)
-	stq	$22,16($16)
-	stq	$23,24($16)
-
-	addq	$17,32,$17
-	addq	$18,32,$18
-	addq	$16,32,$16
-	bne	$19,.Loop
+	addq	$3,$25,$28		# cy add
+	addq	$28,$7,$23		# 4th main add
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$23,$28,$25		# compute cy from last add
+	addq	$17,32,$17		# update s1_ptr
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,32,$16		# update res_ptr
+	addq	$0,$25,$28		# cy add
+	ldq	$2,16($18)
+	addq	$4,$28,$20		# 1st main add
+	ldq	$3,24($18)
+	cmpult	$28,$25,$8		# compute cy from last add
+	ldq	$6,-16($17)
+	cmpult	$20,$28,$25		# compute cy from last add
+	ldq	$7,-8($17)
+	or	$8,$25,$25		# combine cy from the two adds
+	subq	$19,4,$19		# decr loop cnt
+	stq	$22,-16($16)
+	addq	$1,$25,$28		# cy add
+	stq	$23,-8($16)
+	addq	$5,$28,$21		# 2nd main add
+	addq	$18,32,$18		# update s2_ptr
+	cmpult	$28,$25,$8		# compute cy from last add
+	bge	$19,.Loop
+ # Finish software pipeline for 1st loop
+.Lend1:	cmpult	$21,$28,$25		# compute cy from last add
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$2,$25,$28		# cy add
+	addq	$28,$6,$22		# 3rd main add
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$22,$28,$25		# compute cy from last add
+	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
+	stq	$21,8($16)
+	addq	$3,$25,$28		# cy add
+	addq	$28,$7,$23		# 4th main add
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$23,$28,$25		# compute cy from last add
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,32,$16		# update res_ptr
+	stq	$22,-16($16)
+	stq	$23,-8($16)
+.Lend2:	addq	$19,4,$19		# restore loop cnt
+	beq	$19,.Lret
+ # Start software pipeline for 2nd loop
+	ldq	$0,0($18)
+	ldq	$4,0($17)
+	subq	$19,1,$19
+	beq	$19,.Lend0
+ # 2nd loop handles remaining 1-3 limbs
+	.align	4
+.Loop0:	addq	$0,$25,$28		# cy add
+	ldq	$0,8($18)
+	addq	$4,$28,$20		# main add
+	ldq	$4,8($17)
+	addq	$18,8,$18
+	cmpult	$28,$25,$8		# compute cy from last add
+	addq	$17,8,$17
+	stq	$20,0($16)
+	cmpult	$20,$28,$25		# compute cy from last add
+	subq	$19,1,$19		# decr loop cnt
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,8,$16
+	bne	$19,.Loop0
+.Lend0:	addq	$0,$25,$28		# cy add
+	addq	$4,$28,$20		# main add
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$20,$28,$25		# compute cy from last add
+	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
 
-.Lend:	addq	$4,$0,$4
-	cmpult	$4,$0,$1
-	addq	$3,$4,$4
-	cmpult	$4,$3,$0
-	stq	$4,0($16)
-	or	$0,$1,$0
+.Lret:	or	$25,$31,$0		# return cy
 	ret	$31,($26),1
-
 	.end	__mpn_add_n
diff --git a/sysdeps/alpha/alphaev5/lshift.s b/sysdeps/alpha/alphaev5/lshift.s
index fdb089550f..392b4249b8 100644
--- a/sysdeps/alpha/alphaev5/lshift.s
+++ b/sysdeps/alpha/alphaev5/lshift.s
@@ -25,7 +25,7 @@
  # size		r18
  # cnt		r19
 
- # This code runs at 4.25 cycles/limb on the EV5.
+ # This code runs at 3.25 cycles/limb on the EV5.
 
 	.set	noreorder
 	.set	noat
@@ -44,11 +44,11 @@ __mpn_lshift:
 	and	$18,4-1,$28	# number of limbs in first loop
 	srl	$4,$20,$0	# compute function result
 
-	beq	$28,L0
+	beq	$28,.L0
 	subq	$18,$28,$18
 
 	.align	3
-Loop0:	ldq	$3,-16($17)
+.Loop0:	ldq	$3,-16($17)
 	subq	$16,8,$16
 	sll	$4,$19,$5
 	subq	$17,8,$17
@@ -57,17 +57,17 @@ Loop0:	ldq	$3,-16($17)
 	or	$3,$3,$4
 	or	$5,$6,$8
 	stq	$8,0($16)
-	bne	$28,Loop0
+	bne	$28,.Loop0
 
-L0:	sll	$4,$19,$24
-	beq	$18,Lend
+.L0:	sll	$4,$19,$24
+	beq	$18,.Lend
  # warm up phase 1
 	ldq	$1,-16($17)
 	subq	$18,4,$18
 	ldq	$2,-24($17)
 	ldq	$3,-32($17)
 	ldq	$4,-40($17)
-	beq	$18,Lcool1
+	beq	$18,.Lend1
  # warm up phase 2
 	srl	$1,$20,$7
 	sll	$1,$19,$21
@@ -84,10 +84,10 @@ L0:	sll	$4,$19,$24
 	sll	$4,$19,$24
 	ldq	$4,-72($17)
 	subq	$18,4,$18
-	beq	$18,Lcool1
+	beq	$18,.Lend2
 	.align  4
  # main loop
-Loop:	stq	$7,-8($16)
+.Loop:	stq	$7,-8($16)
 	or	$5,$22,$5
 	stq	$8,-16($16)
 	or	$6,$23,$6
@@ -113,16 +113,14 @@ Loop:	stq	$7,-8($16)
 	subq	$16,32,$16
 
 	srl	$4,$20,$6
-	ldq	$3,-96($17
+	ldq	$3,-96($17)
 	sll	$4,$19,$24
 	ldq	$4,-104($17)
 
 	subq	$17,32,$17
-	bne	$18,Loop
-	unop
-	unop
+	bne	$18,.Loop
  # cool down phase 2/1
-Lcool1:	stq	$7,-8($16)
+.Lend2:	stq	$7,-8($16)
 	or	$5,$22,$5
 	stq	$8,-16($16)
 	or	$6,$23,$6
@@ -150,7 +148,7 @@ Lcool1:	stq	$7,-8($16)
 	ret	$31,($26),1
 
  # cool down phase 1/1
-Lcool1:	srl	$1,$20,$7
+.Lend1:	srl	$1,$20,$7
 	sll	$1,$19,$21
 	srl	$2,$20,$8
 	sll	$2,$19,$22
@@ -170,6 +168,6 @@ Lcool1:	srl	$1,$20,$7
 	stq	$24,-40($16)
 	ret	$31,($26),1
 
-Lend	stq	$24,-8($16)
+.Lend:	stq	$24,-8($16)
 	ret	$31,($26),1
 	.end	__mpn_lshift
diff --git a/sysdeps/alpha/alphaev5/rshift.s b/sysdeps/alpha/alphaev5/rshift.s
index 1da9960b46..d20dde35b7 100644
--- a/sysdeps/alpha/alphaev5/rshift.s
+++ b/sysdeps/alpha/alphaev5/rshift.s
@@ -25,7 +25,7 @@
  # size		r18
  # cnt		r19
 
- # This code runs at 4.25 cycles/limb on the EV5.
+ # This code runs at 3.25 cycles/limb on the EV5.
 
 	.set	noreorder
 	.set	noat
@@ -42,11 +42,11 @@ __mpn_rshift:
 	and	$18,4-1,$28	# number of limbs in first loop
 	sll	$4,$20,$0	# compute function result
 
-	beq	$28,L0
+	beq	$28,.L0
 	subq	$18,$28,$18
 
 	.align	3
-Loop0:	ldq	$3,8($17)
+.Loop0:	ldq	$3,8($17)
 	addq	$16,8,$16
 	srl	$4,$19,$5
 	addq	$17,8,$17
@@ -55,17 +55,17 @@ Loop0:	ldq	$3,8($17)
 	or	$3,$3,$4
 	or	$5,$6,$8
 	stq	$8,-8($16)
-	bne	$28,Loop0
+	bne	$28,.Loop0
 
-L0:	srl	$4,$19,$24
-	beq	$18,Lend
+.L0:	srl	$4,$19,$24
+	beq	$18,.Lend
  # warm up phase 1
 	ldq	$1,8($17)
 	subq	$18,4,$18
 	ldq	$2,16($17)
 	ldq	$3,24($17)
 	ldq	$4,32($17)
-	beq	$18,Lcool1
+	beq	$18,.Lend1
  # warm up phase 2
 	sll	$1,$20,$7
 	srl	$1,$19,$21
@@ -82,10 +82,10 @@ L0:	srl	$4,$19,$24
 	srl	$4,$19,$24
 	ldq	$4,64($17)
 	subq	$18,4,$18
-	beq	$18,Lcool2
+	beq	$18,.Lend2
 	.align  4
  # main loop
-Loop:	stq	$7,0($16)
+.Loop:	stq	$7,0($16)
 	or	$5,$22,$5
 	stq	$8,8($16)
 	or	$6,$23,$6
@@ -116,11 +116,9 @@ Loop:	stq	$7,0($16)
 	ldq	$4,96($17)
 
 	addq	$17,32,$17
-	bne	$18,Loop
-	unop
-	unop
+	bne	$18,.Loop
  # cool down phase 2/1
-Lcool2:	stq	$7,0($16)
+.Lend2:	stq	$7,0($16)
 	or	$5,$22,$5
 	stq	$8,8($16)
 	or	$6,$23,$6
@@ -148,7 +146,7 @@ Lcool2:	stq	$7,0($16)
 	ret	$31,($26),1
 
  # cool down phase 1/1
-Lcool1:	sll	$1,$20,$7
+.Lend1:	sll	$1,$20,$7
 	srl	$1,$19,$21
 	sll	$2,$20,$8
 	srl	$2,$19,$22
@@ -168,6 +166,6 @@ Lcool1:	sll	$1,$20,$7
 	stq	$24,32($16)
 	ret	$31,($26),1
 
-Lend:	stq	$24,0($16)
+.Lend:	stq	$24,0($16)
 	ret	$31,($26),1
 	.end	__mpn_rshift
diff --git a/sysdeps/alpha/alphaev5/sub_n.s b/sysdeps/alpha/alphaev5/sub_n.s
new file mode 100644
index 0000000000..c9f3a4ec3a
--- /dev/null
+++ b/sysdeps/alpha/alphaev5/sub_n.s
@@ -0,0 +1,148 @@
+ # Alpha __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
+ # store difference in a third limb vector.
+
+ # Copyright (C) 1995 Free Software Foundation, Inc.
+
+ # This file is part of the GNU MP Library.
+
+ # The GNU MP Library is free software; you can redistribute it and/or modify
+ # it under the terms of the GNU Library General Public License as published by
+ # the Free Software Foundation; either version 2 of the License, or (at your
+ # option) any later version.
+
+ # The GNU MP 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 Library General Public
+ # License for more details.
+
+ # You should have received a copy of the GNU Library General Public License
+ # along with the GNU MP Library; see the file COPYING.LIB.  If not, write to
+ # the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+
+
+ # INPUT PARAMETERS
+ # res_ptr	$16
+ # s1_ptr	$17
+ # s2_ptr	$18
+ # size		$19
+
+	.set	noreorder
+	.set	noat
+.text
+	.align	3
+	.globl	__mpn_sub_n
+	.ent	__mpn_sub_n
+__mpn_sub_n:
+	.frame	$30,0,$26,0
+
+	or	$31,$31,$25		# clear cy
+	subq	$19,4,$19		# decr loop cnt
+	blt	$19,.Lend2		# if less than 4 limbs, goto 2nd loop
+ # Start software pipeline for 1st loop
+	ldq	$0,0($18)
+	ldq	$1,8($18)
+	ldq	$4,0($17)
+	ldq	$5,8($17)
+	addq	$17,32,$17		# update s1_ptr
+	ldq	$2,16($18)
+	subq	$4,$0,$20		# 1st main sub
+	ldq	$3,24($18)
+	subq	$19,4,$19		# decr loop cnt
+	ldq	$6,-16($17)
+	cmpult	$4,$20,$25		# compute cy from last sub
+	ldq	$7,-8($17)
+	addq	$1,$25,$28		# cy add
+	addq	$18,32,$18		# update s2_ptr
+	subq	$5,$28,$21		# 2nd main sub
+	cmpult	$28,$25,$8		# compute cy from last add
+	blt	$19,.Lend1		# if less than 4 limbs remain, jump
+ # 1st loop handles groups of 4 limbs in a software pipeline
+	.align	4
+.Loop:	cmpult	$5,$21,$25		# compute cy from last add
+	ldq	$0,0($18)
+	or	$8,$25,$25		# combine cy from the two adds
+	ldq	$1,8($18)
+	addq	$2,$25,$28		# cy add
+	ldq	$4,0($17)
+	subq	$6,$28,$22		# 3rd main sub
+	ldq	$5,8($17)
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$6,$22,$25		# compute cy from last add
+	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
+	stq	$21,8($16)
+	addq	$3,$25,$28		# cy add
+	subq	$7,$28,$23		# 4th main sub
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$7,$23,$25		# compute cy from last add
+	addq	$17,32,$17		# update s1_ptr
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,32,$16		# update res_ptr
+	addq	$0,$25,$28		# cy add
+	ldq	$2,16($18)
+	subq	$4,$28,$20		# 1st main sub
+	ldq	$3,24($18)
+	cmpult	$28,$25,$8		# compute cy from last add
+	ldq	$6,-16($17)
+	cmpult	$4,$20,$25		# compute cy from last add
+	ldq	$7,-8($17)
+	or	$8,$25,$25		# combine cy from the two adds
+	subq	$19,4,$19		# decr loop cnt
+	stq	$22,-16($16)
+	addq	$1,$25,$28		# cy add
+	stq	$23,-8($16)
+	subq	$5,$28,$21		# 2nd main sub
+	addq	$18,32,$18		# update s2_ptr
+	cmpult	$28,$25,$8		# compute cy from last add
+	bge	$19,.Loop
+ # Finish software pipeline for 1st loop
+.Lend1:	cmpult	$5,$21,$25		# compute cy from last add
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$2,$25,$28		# cy add
+	subq	$6,$28,$22		# 3rd main sub
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$6,$22,$25		# compute cy from last add
+	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
+	stq	$21,8($16)
+	addq	$3,$25,$28		# cy add
+	subq	$7,$28,$23		# 4th main sub
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$7,$23,$25		# compute cy from last add
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,32,$16		# update res_ptr
+	stq	$22,-16($16)
+	stq	$23,-8($16)
+.Lend2:	addq	$19,4,$19		# restore loop cnt
+	beq	$19,.Lret
+ # Start software pipeline for 2nd loop
+	ldq	$0,0($18)
+	ldq	$4,0($17)
+	subq	$19,1,$19
+	beq	$19,.Lend0
+ # 2nd loop handles remaining 1-3 limbs
+	.align	4
+.Loop0:	addq	$0,$25,$28		# cy add
+	ldq	$0,8($18)
+	subq	$4,$28,$20		# main sub
+	ldq	$1,8($17)
+	addq	$18,8,$18
+	cmpult	$28,$25,$8		# compute cy from last add
+	addq	$17,8,$17
+	stq	$20,0($16)
+	cmpult	$4,$20,$25		# compute cy from last add
+	subq	$19,1,$19		# decr loop cnt
+	or	$8,$25,$25		# combine cy from the two adds
+	addq	$16,8,$16
+	or	$1,$31,$4
+	bne	$19,.Loop0
+.Lend0:	addq	$0,$25,$28		# cy add
+	subq	$4,$28,$20		# main sub
+	cmpult	$28,$25,$8		# compute cy from last add
+	cmpult	$4,$20,$25		# compute cy from last add
+	stq	$20,0($16)
+	or	$8,$25,$25		# combine cy from the two adds
+
+.Lret:	or	$25,$31,$0		# return cy
+	ret	$31,($26),1
+	.end	__mpn_sub_n
diff --git a/sysdeps/generic/add_n.c b/sysdeps/generic/add_n.c
index 6989ab0628..647548d4c1 100644
--- a/sysdeps/generic/add_n.c
+++ b/sysdeps/generic/add_n.c
@@ -1,6 +1,6 @@
-/* __mpn_add_n -- Add two limb vectors of equal, non-zero length.
+/* mpn_add_n -- Add two limb vectors of equal, non-zero length.
 
-Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1992, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -23,9 +23,9 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_add_n (mp_ptr res_ptr, mp_srcptr s1_ptr, mp_srcptr s2_ptr, mp_size_t size)
+mpn_add_n (mp_ptr res_ptr, mp_srcptr s1_ptr, mp_srcptr s2_ptr, mp_size_t size)
 #else
-__mpn_add_n (res_ptr, s1_ptr, s2_ptr, size)
+mpn_add_n (res_ptr, s1_ptr, s2_ptr, size)
      register mp_ptr res_ptr;
      register mp_srcptr s1_ptr;
      register mp_srcptr s2_ptr;
diff --git a/sysdeps/generic/addmul_1.c b/sysdeps/generic/addmul_1.c
index fdf3541561..6156cab82c 100644
--- a/sysdeps/generic/addmul_1.c
+++ b/sysdeps/generic/addmul_1.c
@@ -1,9 +1,9 @@
-/* __mpn_addmul_1 -- multiply the S1_SIZE long limb vector pointed to by S1_PTR
+/* mpn_addmul_1 -- multiply the S1_SIZE long limb vector pointed to by S1_PTR
    by S2_LIMB, add the S1_SIZE least significant limbs of the product to the
    limb vector pointed to by RES_PTR.  Return the most significant limb of
    the product, adjusted for carry-out from the addition.
 
-Copyright (C) 1992, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1992, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -26,7 +26,7 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 #include "longlong.h"
 
 mp_limb
-__mpn_addmul_1 (res_ptr, s1_ptr, s1_size, s2_limb)
+mpn_addmul_1 (res_ptr, s1_ptr, s1_size, s2_limb)
      register mp_ptr res_ptr;
      register mp_srcptr s1_ptr;
      mp_size_t s1_size;
diff --git a/sysdeps/generic/cmp.c b/sysdeps/generic/cmp.c
index 144c88588f..e499b1ec44 100644
--- a/sysdeps/generic/cmp.c
+++ b/sysdeps/generic/cmp.c
@@ -1,6 +1,6 @@
-/* __mpn_cmp -- Compare two low-level natural-number integers.
+/* mpn_cmp -- Compare two low-level natural-number integers.
 
-Copyright (C) 1991, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -28,9 +28,9 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 int
 #if __STDC__
-__mpn_cmp (mp_srcptr op1_ptr, mp_srcptr op2_ptr, mp_size_t size)
+mpn_cmp (mp_srcptr op1_ptr, mp_srcptr op2_ptr, mp_size_t size)
 #else
-__mpn_cmp (op1_ptr, op2_ptr, size)
+mpn_cmp (op1_ptr, op2_ptr, size)
      mp_srcptr op1_ptr;
      mp_srcptr op2_ptr;
      mp_size_t size;
diff --git a/sysdeps/generic/divmod_1.c b/sysdeps/generic/divmod_1.c
index 2989d36708..c04032750c 100644
--- a/sysdeps/generic/divmod_1.c
+++ b/sysdeps/generic/divmod_1.c
@@ -1,4 +1,4 @@
-/* __mpn_divmod_1(quot_ptr, dividend_ptr, dividend_size, divisor_limb) --
+/* mpn_divmod_1(quot_ptr, dividend_ptr, dividend_size, divisor_limb) --
    Divide (DIVIDEND_PTR,,DIVIDEND_SIZE) by DIVISOR_LIMB.
    Write DIVIDEND_SIZE limbs of quotient at QUOT_PTR.
    Return the single-limb remainder.
@@ -6,7 +6,7 @@
 
    QUOT_PTR and DIVIDEND_PTR might point to the same limb.
 
-Copyright (C) 1991, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -41,11 +41,11 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_divmod_1 (mp_ptr quot_ptr,
+mpn_divmod_1 (mp_ptr quot_ptr,
 	      mp_srcptr dividend_ptr, mp_size_t dividend_size,
 	      mp_limb divisor_limb)
 #else
-__mpn_divmod_1 (quot_ptr, dividend_ptr, dividend_size, divisor_limb)
+mpn_divmod_1 (quot_ptr, dividend_ptr, dividend_size, divisor_limb)
      mp_ptr quot_ptr;
      mp_srcptr dividend_ptr;
      mp_size_t dividend_size;
diff --git a/sysdeps/generic/lshift.c b/sysdeps/generic/lshift.c
index 1ba09038dd..35794e4ea8 100644
--- a/sysdeps/generic/lshift.c
+++ b/sysdeps/generic/lshift.c
@@ -1,6 +1,6 @@
-/* __mpn_lshift -- Shift left low level.
+/* mpn_lshift -- Shift left low level.
 
-Copyright (C) 1991, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -32,11 +32,11 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_lshift (register mp_ptr wp,
+mpn_lshift (register mp_ptr wp,
 	    register mp_srcptr up, mp_size_t usize,
 	    register unsigned int cnt)
 #else
-__mpn_lshift (wp, up, usize, cnt)
+mpn_lshift (wp, up, usize, cnt)
      register mp_ptr wp;
      register mp_srcptr up;
      mp_size_t usize;
diff --git a/sysdeps/generic/mod_1.c b/sysdeps/generic/mod_1.c
index 8a49fb4be0..0842f6b1ee 100644
--- a/sysdeps/generic/mod_1.c
+++ b/sysdeps/generic/mod_1.c
@@ -1,4 +1,4 @@
-/* __mpn_mod_1(dividend_ptr, dividend_size, divisor_limb) --
+/* mpn_mod_1(dividend_ptr, dividend_size, divisor_limb) --
    Divide (DIVIDEND_PTR,,DIVIDEND_SIZE) by DIVISOR_LIMB.
    Return the single-limb remainder.
    There are no constraints on the value of the divisor.
@@ -38,10 +38,10 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_mod_1 (mp_srcptr dividend_ptr, mp_size_t dividend_size,
-	     mp_limb divisor_limb)
+mpn_mod_1 (mp_srcptr dividend_ptr, mp_size_t dividend_size,
+	   mp_limb divisor_limb)
 #else
-__mpn_mod_1 (dividend_ptr, dividend_size, divisor_limb)
+mpn_mod_1 (dividend_ptr, dividend_size, divisor_limb)
      mp_srcptr dividend_ptr;
      mp_size_t dividend_size;
      mp_limb divisor_limb;
diff --git a/sysdeps/generic/mul.c b/sysdeps/generic/mul.c
index cd2acb5127..3f3f41e99f 100644
--- a/sysdeps/generic/mul.c
+++ b/sysdeps/generic/mul.c
@@ -1,6 +1,6 @@
-/* __mpn_mul -- Multiply two natural numbers.
+/* mpn_mul -- Multiply two natural numbers.
 
-Copyright (C) 1991, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -43,11 +43,11 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_mul (mp_ptr prodp,
-	  mp_srcptr up, mp_size_t usize,
-	  mp_srcptr vp, mp_size_t vsize)
+mpn_mul (mp_ptr prodp,
+	 mp_srcptr up, mp_size_t usize,
+	 mp_srcptr vp, mp_size_t vsize)
 #else
-__mpn_mul (prodp, up, usize, vp, vsize)
+mpn_mul (prodp, up, usize, vp, vsize)
      mp_ptr prodp;
      mp_srcptr up;
      mp_size_t usize;
@@ -58,6 +58,7 @@ __mpn_mul (prodp, up, usize, vp, vsize)
   mp_ptr prod_endp = prodp + usize + vsize - 1;
   mp_limb cy;
   mp_ptr tspace;
+  TMP_DECL (marker);
 
   if (vsize < KARATSUBA_THRESHOLD)
     {
@@ -86,7 +87,7 @@ __mpn_mul (prodp, up, usize, vp, vsize)
 	  cy_limb = 0;
 	}
       else
-	cy_limb = __mpn_mul_1 (prodp, up, usize, v_limb);
+	cy_limb = mpn_mul_1 (prodp, up, usize, v_limb);
 
       prodp[usize] = cy_limb;
       prodp++;
@@ -100,10 +101,10 @@ __mpn_mul (prodp, up, usize, vp, vsize)
 	    {
 	      cy_limb = 0;
 	      if (v_limb == 1)
-		cy_limb = __mpn_add_n (prodp, prodp, up, usize);
+		cy_limb = mpn_add_n (prodp, prodp, up, usize);
 	    }
 	  else
-	    cy_limb = __mpn_addmul_1 (prodp, up, usize, v_limb);
+	    cy_limb = mpn_addmul_1 (prodp, up, usize, v_limb);
 
 	  prodp[usize] = cy_limb;
 	  prodp++;
@@ -111,7 +112,9 @@ __mpn_mul (prodp, up, usize, vp, vsize)
       return cy_limb;
     }
 
-  tspace = (mp_ptr) alloca (2 * vsize * BYTES_PER_MP_LIMB);
+  TMP_MARK (marker);
+
+  tspace = (mp_ptr) TMP_ALLOC (2 * vsize * BYTES_PER_MP_LIMB);
   MPN_MUL_N_RECURSE (prodp, up, vp, vsize, tspace);
 
   prodp += vsize;
@@ -119,12 +122,12 @@ __mpn_mul (prodp, up, usize, vp, vsize)
   usize -= vsize;
   if (usize >= vsize)
     {
-      mp_ptr tp = (mp_ptr) alloca (2 * vsize * BYTES_PER_MP_LIMB);
+      mp_ptr tp = (mp_ptr) TMP_ALLOC (2 * vsize * BYTES_PER_MP_LIMB);
       do
 	{
 	  MPN_MUL_N_RECURSE (tp, up, vp, vsize, tspace);
-	  cy = __mpn_add_n (prodp, prodp, tp, vsize);
-	  __mpn_add_1 (prodp + vsize, tp + vsize, vsize, cy);
+	  cy = mpn_add_n (prodp, prodp, tp, vsize);
+	  mpn_add_1 (prodp + vsize, tp + vsize, vsize, cy);
 	  prodp += vsize;
 	  up += vsize;
 	  usize -= vsize;
@@ -138,10 +141,11 @@ __mpn_mul (prodp, up, usize, vp, vsize)
 
   if (usize != 0)
     {
-      __mpn_mul (tspace, vp, vsize, up, usize);
-      cy = __mpn_add_n (prodp, prodp, tspace, vsize);
-      __mpn_add_1 (prodp + vsize, tspace + vsize, usize, cy);
+      mpn_mul (tspace, vp, vsize, up, usize);
+      cy = mpn_add_n (prodp, prodp, tspace, vsize);
+      mpn_add_1 (prodp + vsize, tspace + vsize, usize, cy);
     }
 
+  TMP_FREE (marker);
   return *prod_endp;
 }
diff --git a/sysdeps/generic/mul_1.c b/sysdeps/generic/mul_1.c
index 37dbc33031..01fdbbbc9b 100644
--- a/sysdeps/generic/mul_1.c
+++ b/sysdeps/generic/mul_1.c
@@ -1,7 +1,7 @@
-/* __mpn_mul_1 -- Multiply a limb vector with a single limb and
+/* mpn_mul_1 -- Multiply a limb vector with a single limb and
    store the product in a second limb vector.
 
-Copyright (C) 1991, 1992, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1992, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -24,7 +24,7 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 #include "longlong.h"
 
 mp_limb
-__mpn_mul_1 (res_ptr, s1_ptr, s1_size, s2_limb)
+mpn_mul_1 (res_ptr, s1_ptr, s1_size, s2_limb)
      register mp_ptr res_ptr;
      register mp_srcptr s1_ptr;
      mp_size_t s1_size;
diff --git a/sysdeps/generic/mul_n.c b/sysdeps/generic/mul_n.c
index e37c5d8290..049f63dce8 100644
--- a/sysdeps/generic/mul_n.c
+++ b/sysdeps/generic/mul_n.c
@@ -1,6 +1,6 @@
-/* __mpn_mul_n -- Multiply two natural numbers of length n.
+/* mpn_mul_n -- Multiply two natural numbers of length n.
 
-Copyright (C) 1991, 1992, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1992, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -41,13 +41,6 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 #define KARATSUBA_THRESHOLD 2
 #endif
 
-void
-#if __STDC__
-____mpn_mul_n (mp_ptr, mp_srcptr, mp_srcptr, mp_size_t, mp_ptr);
-#else
-____mpn_mul_n ();
-#endif
-
 /* Handle simple cases with traditional multiplication.
 
    This is the most critical code of multiplication.  All multiplies rely
@@ -57,9 +50,9 @@ ____mpn_mul_n ();
 
 void
 #if __STDC__
-____mpn_mul_n_basecase (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
+impn_mul_n_basecase (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
 #else
-____mpn_mul_n_basecase (prodp, up, vp, size)
+impn_mul_n_basecase (prodp, up, vp, size)
      mp_ptr prodp;
      mp_srcptr up;
      mp_srcptr vp;
@@ -82,7 +75,7 @@ ____mpn_mul_n_basecase (prodp, up, vp, size)
       cy_limb = 0;
     }
   else
-    cy_limb = __mpn_mul_1 (prodp, up, size, v_limb);
+    cy_limb = mpn_mul_1 (prodp, up, size, v_limb);
 
   prodp[size] = cy_limb;
   prodp++;
@@ -96,10 +89,10 @@ ____mpn_mul_n_basecase (prodp, up, vp, size)
 	{
 	  cy_limb = 0;
 	  if (v_limb == 1)
-	    cy_limb = __mpn_add_n (prodp, prodp, up, size);
+	    cy_limb = mpn_add_n (prodp, prodp, up, size);
 	}
       else
-	cy_limb = __mpn_addmul_1 (prodp, up, size, v_limb);
+	cy_limb = mpn_addmul_1 (prodp, up, size, v_limb);
 
       prodp[size] = cy_limb;
       prodp++;
@@ -108,10 +101,10 @@ ____mpn_mul_n_basecase (prodp, up, vp, size)
 
 void
 #if __STDC__
-____mpn_mul_n (mp_ptr prodp,
+impn_mul_n (mp_ptr prodp,
 	     mp_srcptr up, mp_srcptr vp, mp_size_t size, mp_ptr tspace)
 #else
-____mpn_mul_n (prodp, up, vp, size, tspace)
+impn_mul_n (prodp, up, vp, size, tspace)
      mp_ptr prodp;
      mp_srcptr up;
      mp_srcptr vp;
@@ -135,9 +128,9 @@ ____mpn_mul_n (prodp, up, vp, size, tspace)
       mp_limb cy_limb;
 
       MPN_MUL_N_RECURSE (prodp, up, vp, esize, tspace);
-      cy_limb = __mpn_addmul_1 (prodp + esize, up, esize, vp[esize]);
+      cy_limb = mpn_addmul_1 (prodp + esize, up, esize, vp[esize]);
       prodp[esize + esize] = cy_limb;
-      cy_limb = __mpn_addmul_1 (prodp + esize, vp, size, up[esize]);
+      cy_limb = mpn_addmul_1 (prodp + esize, vp, size, up[esize]);
 
       prodp[esize + size] = cy_limb;
     }
@@ -170,24 +163,24 @@ ____mpn_mul_n (prodp, up, vp, size, tspace)
 
       /*** Product M.	 ________________
 			|_(U1-U0)(V0-V1)_|  */
-      if (__mpn_cmp (up + hsize, up, hsize) >= 0)
+      if (mpn_cmp (up + hsize, up, hsize) >= 0)
 	{
-	  __mpn_sub_n (prodp, up + hsize, up, hsize);
+	  mpn_sub_n (prodp, up + hsize, up, hsize);
 	  negflg = 0;
 	}
       else
 	{
-	  __mpn_sub_n (prodp, up, up + hsize, hsize);
+	  mpn_sub_n (prodp, up, up + hsize, hsize);
 	  negflg = 1;
 	}
-      if (__mpn_cmp (vp + hsize, vp, hsize) >= 0)
+      if (mpn_cmp (vp + hsize, vp, hsize) >= 0)
 	{
-	  __mpn_sub_n (prodp + hsize, vp + hsize, vp, hsize);
+	  mpn_sub_n (prodp + hsize, vp + hsize, vp, hsize);
 	  negflg ^= 1;
 	}
       else
 	{
-	  __mpn_sub_n (prodp + hsize, vp, vp + hsize, hsize);
+	  mpn_sub_n (prodp + hsize, vp, vp + hsize, hsize);
 	  /* No change of NEGFLG.  */
 	}
       /* Read temporary operands from low part of PROD.
@@ -197,13 +190,13 @@ ____mpn_mul_n (prodp, up, vp, size, tspace)
 
       /*** Add/copy product H.  */
       MPN_COPY (prodp + hsize, prodp + size, hsize);
-      cy = __mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);
+      cy = mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);
 
       /*** Add product M (if NEGFLG M is a negative number).  */
       if (negflg)
-	cy -= __mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);
+	cy -= mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);
       else
-	cy += __mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
+	cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
 
       /*** Product L.	 ________________  ________________
 			|________________||____U0 x V0_____|  */
@@ -214,22 +207,22 @@ ____mpn_mul_n (prodp, up, vp, size, tspace)
 
       /*** Add/copy Product L (twice).  */
 
-      cy += __mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
+      cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
       if (cy)
-	__mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);
+	mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);
 
       MPN_COPY (prodp, tspace, hsize);
-      cy = __mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
+      cy = mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
       if (cy)
-	__mpn_add_1 (prodp + size, prodp + size, size, 1);
+	mpn_add_1 (prodp + size, prodp + size, size, 1);
     }
 }
 
 void
 #if __STDC__
-____mpn_sqr_n_basecase (mp_ptr prodp, mp_srcptr up, mp_size_t size)
+impn_sqr_n_basecase (mp_ptr prodp, mp_srcptr up, mp_size_t size)
 #else
-____mpn_sqr_n_basecase (prodp, up, size)
+impn_sqr_n_basecase (prodp, up, size)
      mp_ptr prodp;
      mp_srcptr up;
      mp_size_t size;
@@ -251,7 +244,7 @@ ____mpn_sqr_n_basecase (prodp, up, size)
       cy_limb = 0;
     }
   else
-    cy_limb = __mpn_mul_1 (prodp, up, size, v_limb);
+    cy_limb = mpn_mul_1 (prodp, up, size, v_limb);
 
   prodp[size] = cy_limb;
   prodp++;
@@ -265,10 +258,10 @@ ____mpn_sqr_n_basecase (prodp, up, size)
 	{
 	  cy_limb = 0;
 	  if (v_limb == 1)
-	    cy_limb = __mpn_add_n (prodp, prodp, up, size);
+	    cy_limb = mpn_add_n (prodp, prodp, up, size);
 	}
       else
-	cy_limb = __mpn_addmul_1 (prodp, up, size, v_limb);
+	cy_limb = mpn_addmul_1 (prodp, up, size, v_limb);
 
       prodp[size] = cy_limb;
       prodp++;
@@ -277,10 +270,10 @@ ____mpn_sqr_n_basecase (prodp, up, size)
 
 void
 #if __STDC__
-____mpn_sqr_n (mp_ptr prodp,
+impn_sqr_n (mp_ptr prodp,
 	     mp_srcptr up, mp_size_t size, mp_ptr tspace)
 #else
-____mpn_sqr_n (prodp, up, size, tspace)
+impn_sqr_n (prodp, up, size, tspace)
      mp_ptr prodp;
      mp_srcptr up;
      mp_size_t size;
@@ -303,9 +296,9 @@ ____mpn_sqr_n (prodp, up, size, tspace)
       mp_limb cy_limb;
 
       MPN_SQR_N_RECURSE (prodp, up, esize, tspace);
-      cy_limb = __mpn_addmul_1 (prodp + esize, up, esize, up[esize]);
+      cy_limb = mpn_addmul_1 (prodp + esize, up, esize, up[esize]);
       prodp[esize + esize] = cy_limb;
-      cy_limb = __mpn_addmul_1 (prodp + esize, up, size, up[esize]);
+      cy_limb = mpn_addmul_1 (prodp + esize, up, size, up[esize]);
 
       prodp[esize + size] = cy_limb;
     }
@@ -322,13 +315,13 @@ ____mpn_sqr_n (prodp, up, size, tspace)
 
       /*** Product M.	 ________________
 			|_(U1-U0)(U0-U1)_|  */
-      if (__mpn_cmp (up + hsize, up, hsize) >= 0)
+      if (mpn_cmp (up + hsize, up, hsize) >= 0)
 	{
-	  __mpn_sub_n (prodp, up + hsize, up, hsize);
+	  mpn_sub_n (prodp, up + hsize, up, hsize);
 	}
       else
 	{
-	  __mpn_sub_n (prodp, up, up + hsize, hsize);
+	  mpn_sub_n (prodp, up, up + hsize, hsize);
 	}
 
       /* Read temporary operands from low part of PROD.
@@ -338,10 +331,10 @@ ____mpn_sqr_n (prodp, up, size, tspace)
 
       /*** Add/copy product H.  */
       MPN_COPY (prodp + hsize, prodp + size, hsize);
-      cy = __mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);
+      cy = mpn_add_n (prodp + size, prodp + size, prodp + size + hsize, hsize);
 
       /*** Add product M (if NEGFLG M is a negative number).  */
-      cy -= __mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);
+      cy -= mpn_sub_n (prodp + hsize, prodp + hsize, tspace, size);
 
       /*** Product L.	 ________________  ________________
 			|________________||____U0 x U0_____|  */
@@ -352,53 +345,56 @@ ____mpn_sqr_n (prodp, up, size, tspace)
 
       /*** Add/copy Product L (twice).  */
 
-      cy += __mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
+      cy += mpn_add_n (prodp + hsize, prodp + hsize, tspace, size);
       if (cy)
-	__mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);
+	mpn_add_1 (prodp + hsize + size, prodp + hsize + size, hsize, cy);
 
       MPN_COPY (prodp, tspace, hsize);
-      cy = __mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
+      cy = mpn_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize);
       if (cy)
-	__mpn_add_1 (prodp + size, prodp + size, size, 1);
+	mpn_add_1 (prodp + size, prodp + size, size, 1);
     }
 }
 
 /* This should be made into an inline function in gmp.h.  */
 inline void
 #if __STDC__
-__mpn_mul_n (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
+mpn_mul_n (mp_ptr prodp, mp_srcptr up, mp_srcptr vp, mp_size_t size)
 #else
-__mpn_mul_n (prodp, up, vp, size)
+mpn_mul_n (prodp, up, vp, size)
      mp_ptr prodp;
      mp_srcptr up;
      mp_srcptr vp;
      mp_size_t size;
 #endif
 {
+  TMP_DECL (marker);
+  TMP_MARK (marker);
   if (up == vp)
     {
       if (size < KARATSUBA_THRESHOLD)
 	{
-	  ____mpn_sqr_n_basecase (prodp, up, size);
+	  impn_sqr_n_basecase (prodp, up, size);
 	}
       else
 	{
 	  mp_ptr tspace;
-	  tspace = (mp_ptr) alloca (2 * size * BYTES_PER_MP_LIMB);
-	  ____mpn_sqr_n (prodp, up, size, tspace);
+	  tspace = (mp_ptr) TMP_ALLOC (2 * size * BYTES_PER_MP_LIMB);
+	  impn_sqr_n (prodp, up, size, tspace);
 	}
     }
   else
     {
       if (size < KARATSUBA_THRESHOLD)
 	{
-	  ____mpn_mul_n_basecase (prodp, up, vp, size);
+	  impn_mul_n_basecase (prodp, up, vp, size);
 	}
       else
 	{
 	  mp_ptr tspace;
-	  tspace = (mp_ptr) alloca (2 * size * BYTES_PER_MP_LIMB);
-	  ____mpn_mul_n (prodp, up, vp, size, tspace);
+	  tspace = (mp_ptr) TMP_ALLOC (2 * size * BYTES_PER_MP_LIMB);
+	  impn_mul_n (prodp, up, vp, size, tspace);
 	}
     }
+  TMP_FREE (marker);
 }
diff --git a/sysdeps/generic/rshift.c b/sysdeps/generic/rshift.c
index 966cc7bcad..7ce02e07f7 100644
--- a/sysdeps/generic/rshift.c
+++ b/sysdeps/generic/rshift.c
@@ -1,6 +1,6 @@
-/* __mpn_rshift -- Shift right a low-level natural-number integer.
+/* mpn_rshift -- Shift right a low-level natural-number integer.
 
-Copyright (C) 1991, 1993, 1994 Free Software Foundation, Inc.
+Copyright (C) 1991, 1993, 1994, 1996 Free Software Foundation, Inc.
 
 This file is part of the GNU MP Library.
 
@@ -32,11 +32,11 @@ the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
 
 mp_limb
 #if __STDC__
-__mpn_rshift (register mp_ptr wp,
+mpn_rshift (register mp_ptr wp,
 	    register mp_srcptr up, mp_size_t usize,
 	    register unsigned int cnt)
 #else
-__mpn_rshift (wp, up, usize, cnt)
+mpn_rshift (wp, up, usize, cnt)
      register mp_ptr wp;
      register mp_srcptr up;
      mp_size_t usize;
diff --git a/sysdeps/mips/addmul_1.s b/sysdeps/mips/addmul_1.s
index abc2fb8dcf..917af1bac3 100644
--- a/sysdeps/mips/addmul_1.s
+++ b/sysdeps/mips/addmul_1.s
@@ -1,7 +1,7 @@
  # MIPS __mpn_addmul_1 -- Multiply a limb vector with a single limb and
  # add the product to a second limb vector.
 
- # Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+ # Copyright (C) 1992, 1994, 1996 Free Software Foundation, Inc.
 
  # This file is part of the GNU MP Library.
 
@@ -63,7 +63,7 @@ Loop:	lw	$10,0($4)
 	addu	$2,$2,$10
 	sw	$3,0($4)
 	addiu	$4,$4,4
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 addu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/mips/mips3/addmul_1.s b/sysdeps/mips/mips3/addmul_1.s
index 7af0172614..7dbc9ad41d 100644
--- a/sysdeps/mips/mips3/addmul_1.s
+++ b/sysdeps/mips/mips3/addmul_1.s
@@ -63,7 +63,7 @@ Loop:	ld	$10,0($4)
 	daddu	$2,$2,$10
 	sd	$3,0($4)
 	daddiu	$4,$4,8
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 daddu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/mips/mips3/mul_1.s b/sysdeps/mips/mips3/mul_1.s
index 87954e5bc3..8376a02546 100644
--- a/sysdeps/mips/mips3/mul_1.s
+++ b/sysdeps/mips/mips3/mul_1.s
@@ -59,7 +59,7 @@ Loop:	mflo	$10
 	sltu	$2,$10,$2	# carry from previous addition -> $2
 	sd	$10,0($4)
 	daddiu	$4,$4,8
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 daddu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/mips/mips3/submul_1.s b/sysdeps/mips/mips3/submul_1.s
index f28c6a5167..f041f6c0b4 100644
--- a/sysdeps/mips/mips3/submul_1.s
+++ b/sysdeps/mips/mips3/submul_1.s
@@ -63,7 +63,7 @@ Loop:	ld	$10,0($4)
 	daddu	$2,$2,$10
 	sd	$3,0($4)
 	daddiu	$4,$4,8
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 daddu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/mips/mul_1.s b/sysdeps/mips/mul_1.s
index 01327e22d8..6f5324cef4 100644
--- a/sysdeps/mips/mul_1.s
+++ b/sysdeps/mips/mul_1.s
@@ -1,7 +1,7 @@
  # MIPS __mpn_mul_1 -- Multiply a limb vector with a single limb and
  # store the product in a second limb vector.
 
- # Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+ # Copyright (C) 1992, 1994, 1996 Free Software Foundation, Inc.
 
  # This file is part of the GNU MP Library.
 
@@ -59,7 +59,7 @@ Loop:	mflo	$10
 	sltu	$2,$10,$2	# carry from previous addition -> $2
 	sw	$10,0($4)
 	addiu	$4,$4,4
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 addu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/mips/submul_1.s b/sysdeps/mips/submul_1.s
index 616dd1b47c..a78072a1e9 100644
--- a/sysdeps/mips/submul_1.s
+++ b/sysdeps/mips/submul_1.s
@@ -1,7 +1,7 @@
  # MIPS __mpn_submul_1 -- Multiply a limb vector with a single limb and
  # subtract the product from a second limb vector.
 
- # Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+ # Copyright (C) 1992, 1994, 1996 Free Software Foundation, Inc.
 
  # This file is part of the GNU MP Library.
 
@@ -63,7 +63,7 @@ Loop:	lw	$10,0($4)
 	addu	$2,$2,$10
 	sw	$3,0($4)
 	addiu	$4,$4,4
-	bne	$6,$0,Loop	# should be "bnel"
+	bne	$6,$0,Loop
 	 addu	$2,$9,$2	# add high product limb and carry from addition
 
  # cool down phase 1
diff --git a/sysdeps/rs6000/add_n.s b/sysdeps/rs6000/add_n.s
index 7090cf1b00..e2536d5887 100644
--- a/sysdeps/rs6000/add_n.s
+++ b/sysdeps/rs6000/add_n.s
@@ -1,6 +1,6 @@
 # IBM POWER __mpn_add_n -- Add two limb vectors of equal, non-zero length.
 
-# Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+# Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
 
 # This file is part of the GNU MP Library.
 
diff --git a/sysdeps/rs6000/sub_n.s b/sysdeps/rs6000/sub_n.s
index 40fe7d68bd..c57675b106 100644
--- a/sysdeps/rs6000/sub_n.s
+++ b/sysdeps/rs6000/sub_n.s
@@ -1,7 +1,7 @@
 # IBM POWER __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and
 # store difference in a third limb vector.
 
-# Copyright (C) 1992, 1994 Free Software Foundation, Inc.
+# Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
 
 # This file is part of the GNU MP Library.
 
-- 
2.11.0