[Gmp-commit] /var/hg/gmp: 2 new changesets

mercurial at gmplib.org mercurial at gmplib.org
Sun May 13 01:17:55 CEST 2012


details:   /var/hg/gmp/rev/b5f1ad1e1745
changeset: 14965:b5f1ad1e1745
user:      Torbjorn Granlund <tege at gmplib.org>
date:      Sun May 13 01:16:38 2012 +0200
description:
Add ARM v5 mod_1s_2p.

details:   /var/hg/gmp/rev/b124cc447937
changeset: 14966:b124cc447937
user:      Torbjorn Granlund <tege at gmplib.org>
date:      Sun May 13 01:17:52 2012 +0200
description:
Add ARM rsh1add_n and rsh1sub_n.

diffstat:

 ChangeLog              |    5 +
 mpn/arm/rsh1aors_n.asm |  112 ++++++++++++++++++++++++++++++++++++++
 mpn/arm/v5/mod_1_2.asm |  144 +++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 261 insertions(+), 0 deletions(-)

diffs (276 lines):

diff -r 96ad88f37fa2 -r b124cc447937 ChangeLog
--- a/ChangeLog	Fri May 11 19:51:40 2012 +0200
+++ b/ChangeLog	Sun May 13 01:17:52 2012 +0200
@@ -1,3 +1,8 @@
+2012-05-13  Torbjorn Granlund  <tege at gmplib.org>
+
+	* mpn/arm/rsh1aors_n.asm: New file.
+	* mpn/arm/v5/mod_1_2.asm: New file.
+
 2012-05-11  Marc Glisse  <marc.glisse at inria.fr>
 
 	* gmpxx.h (explicit operator bool): New functions.
diff -r 96ad88f37fa2 -r b124cc447937 mpn/arm/rsh1aors_n.asm
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/mpn/arm/rsh1aors_n.asm	Sun May 13 01:17:52 2012 +0200
@@ -0,0 +1,112 @@
+dnl  ARM mpn_rsh1add_n and mpn_rsh1sub_n.
+
+dnl  Contributed to the GNU project by Torbjorn Granlund.
+
+dnl  Copyright 2012 Free Software Foundation, Inc.
+
+dnl  This file is part of the GNU MP Library.
+
+dnl  The GNU MP Library is free software; you can redistribute it and/or modify
+dnl  it under the terms of the GNU Lesser General Public License as published
+dnl  by the Free Software Foundation; either version 3 of the License, or (at
+dnl  your option) any later version.
+
+dnl  The GNU MP Library is distributed in the hope that it will be useful, but
+dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
+dnl  License for more details.
+
+dnl  You should have received a copy of the GNU Lesser General Public License
+dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C	     cycles/limb
+C StrongARM	 ?
+C XScale	 ?
+C Cortex-A8	 ?
+C Cortex-A9	3.64-3.7
+C Cortex-A15	 ?
+
+C TODO
+C  * Not optimised.
+
+define(`rp', `r0')
+define(`up', `r1')
+define(`vp', `r2')
+define(`n',  `r3')
+
+ifdef(`OPERATION_rsh1add_n', `
+  define(`ADDSUB',	adds)
+  define(`ADDSUBC',	adcs)
+  define(`RSTCY',	`cmn	$1, $1')
+  define(`func',	mpn_rsh1add_n)
+  define(`func_nc',	mpn_rsh1add_nc)')
+ifdef(`OPERATION_rsh1sub_n', `
+  define(`ADDSUB',	subs)
+  define(`ADDSUBC',	sbcs)
+  define(`RSTCY',
+	`mvn	$2, #0x80000000
+	cmp	$2, $1')
+  define(`func',	mpn_rsh1sub_n)
+  define(`func_nc',	mpn_rsh1sub_nc)')
+
+MULFUNC_PROLOGUE(mpn_rsh1add_n mpn_rsh1sub_n)
+
+ASM_START()
+PROLOGUE(func)
+	push	{r4-r11}
+	ldr	r4, [up], #4
+	ldr	r8, [vp], #4
+	ADDSUB	r4, r4, r8
+	rrxs	r12, r7
+	and	r11, r4, #1	C return value
+	subs	n, n, #4
+	blo	L(end)
+
+L(top):	ldmia	up!, {r5,r6,r7}
+	ldmia	vp!, {r8,r9,r10}
+	cmn	r12, r12
+	ADDSUBC	r5, r5, r8
+	ADDSUBC	r6, r6, r9
+	ADDSUBC	r7, r7, r10
+	rrxs	r12, r7
+	rrxs	r6, r6
+	rrxs	r5, r5
+	rrxs	r4, r4
+	subs	n, n, #3
+	stmia	rp!, {r4,r5,r6}
+	mov	r4, r7
+	bhs	L(top)
+
+L(end):	cmn	n, #2
+	bls	L(e2)
+	ldm	up, {r5,r6}
+	ldm	vp, {r8,r9}
+	cmn	r12, r12
+	ADDSUBC	r5, r5, r8
+	ADDSUBC	r6, r6, r9
+	rrxs	r12, r6
+	rrxs	r5, r5
+	rrxs	r4, r4
+	stmia	rp!, {r4,r5}
+	mov	r4, r6
+	b	L(e1)
+
+L(e2):	bne	L(e1)
+	ldr	r5, [up, #0]
+	ldr	r8, [vp, #0]
+	cmn	r12, r12
+	ADDSUBC	r5, r5, r8
+	rrxs	r12, r5
+	rrxs	r4, r4
+	str	r4, [rp], #4
+	mov	r4, r5
+
+L(e1):	RSTCY(	r12, r1)
+	rrxs	r4, r4
+	str	r4, [rp, #0]
+	mov	r0, r11
+	pop	{r4-r11}
+	bx	r14
+EPILOGUE()
diff -r 96ad88f37fa2 -r b124cc447937 mpn/arm/v5/mod_1_2.asm
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/mpn/arm/v5/mod_1_2.asm	Sun May 13 01:17:52 2012 +0200
@@ -0,0 +1,144 @@
+dnl  ARM mpn_mod_1s_2p
+
+dnl  Contributed to the GNU project by Torbjorn Granlund.
+
+dnl  Copyright 2012 Free Software Foundation, Inc.
+
+dnl  This file is part of the GNU MP Library.
+
+dnl  The GNU MP Library is free software; you can redistribute it and/or modify
+dnl  it under the terms of the GNU Lesser General Public License as published
+dnl  by the Free Software Foundation; either version 3 of the License, or (at
+dnl  your option) any later version.
+
+dnl  The GNU MP Library is distributed in the hope that it will be useful, but
+dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
+dnl  License for more details.
+
+dnl  You should have received a copy of the GNU Lesser General Public License
+dnl  along with the GNU MP Library.  If not, see http://www.gnu.org/licenses/.
+
+include(`../config.m4')
+
+C	     cycles/limb
+C StrongARM	 ?
+C XScale	 ?
+C Cortex-A8	 ?
+C Cortex-A9	 4.25
+C Cortex-A15	 ?
+
+define(`ap', `r0')
+define(`n',  `r1')
+define(`d',  `r2')
+define(`cps',`r3')
+
+ASM_START()
+PROLOGUE(mpn_mod_1s_2p)
+	push	{r4-r10}
+	tst	n, #1
+	add	r7, r3, #8
+	ldmia	r7, {r7, r8, r12}	C load B1, B2, B3
+	add	ap, ap, n, lsl #2	C put ap at operand end
+	beq	L(evn)
+
+L(odd):	subs	n, n, #1
+	beq	L(1)
+	ldmdb	ap!, {r4,r6,r9}
+	mov	r10, #0
+	umlal	r4, r10, r6, r7
+	umlal	r4, r10, r9, r8
+	b	L(com)
+
+L(evn):	ldmdb	ap!, {r4,r10}
+L(com):	subs	n, n, #2
+	ble	L(end)
+	ldmdb	ap!, {r5,r6}
+	b	L(mid)
+
+L(top):	mov	r9, #0
+	umlal	r5, r9, r6, r7		C B1
+	umlal	r5, r9, r4, r8		C B2
+	ldmdb	ap!, {r4,r6}
+	umlal	r5, r9, r10, r12	C B3
+	ble	L(xit)
+	mov	r10, #0
+	umlal	r4, r10, r6, r7		C B1
+	umlal	r4, r10, r5, r8		C B2
+	ldmdb	ap!, {r5,r6}
+	umlal	r4, r10, r9, r12	C B3
+L(mid):	subs	n, n, #4
+	bge	L(top)
+
+	mov	r9, #0
+	umlal	r5, r9, r6, r7		C B1
+	umlal	r5, r9, r4, r8		C B2
+	umlal	r5, r9, r10, r12	C B3
+	mov	r4, r5
+
+L(end):	movge	   r9, r10		C executed iff coming via xit
+	ldr	r6, [r3, #4]		C cps[1] = cnt
+	mov	r5, #0
+	umlal	r4, r5, r9, r7
+	mov	r7, r5, lsl r6
+L(x):	rsb	r1, r6, #32
+	orr	r8, r7, r4, lsr r1
+	mov	r9, r4, lsl r6
+	ldr	r5, [r3, #0]
+	add	r0, r8, #1
+	umull	r12, r1, r8, r5
+	adds	r4, r12, r9
+	adc	r1, r1, r0
+	mul	r5, r2, r1
+	sub	r9, r9, r5
+	cmp	r9, r4
+	addhi	r9, r9, r2
+	cmp	r2, r9
+	subls	r9, r9, r2
+	mov	r0, r9, lsr r6
+	pop	{r4-r10}
+	bx	r14
+
+L(xit):	mov	r10, #0
+	umlal	r4, r10, r6, r7		C B1
+	umlal	r4, r10, r5, r8		C B2
+	umlal	r4, r10, r9, r12	C B3
+	b	L(end)
+
+L(1):	ldr	r6, [r3, #4]		C cps[1] = cnt
+	ldr	r4, [ap, #-4]		C ap[0]
+	mov	r7, #0
+	b	L(x)
+EPILOGUE()
+
+PROLOGUE(mpn_mod_1s_2p_cps)
+	push	{r4-r8, r14}
+	clz	r4, r1
+	mov	r5, r1, lsl r4		C b <<= cnt
+	mov	r6, r0			C r6 = cps
+	mov	r0, r5
+	bl	mpn_invert_limb
+	rsb	r3, r4, #32
+	mov	r3, r0, lsr r3
+	mov	r2, #1
+	orr	r3, r3, r2, lsl r4
+	rsb	r1, r5, #0
+	mul	r2, r1, r3
+	umull	r3, r12, r2, r0
+	add	r12, r2, r12
+	mvn	r12, r12
+	mul	r1, r5, r12
+	cmp	r1, r3
+	addhi	r1, r1, r5
+	umull	r12, r7, r1, r0
+	add	r7, r1, r7
+	mvn	r7, r7
+	mul	r3, r5, r7
+	cmp	r3, r12
+	addhi	r3, r3, r5
+	mov	r5, r2, lsr r4
+	mov	r7, r1, lsr r4
+	mov	r8, r3, lsr r4
+	stmia	r6, {r0,r4,r5,r7,r8}	C fill cps
+	pop	{r4-r8, pc}
+EPILOGUE()


More information about the gmp-commit mailing list