aboutsummaryrefslogtreecommitdiff
path: root/vendor/gmp-6.3.0/mpn/x86_64/mod_34lsub1.asm
blob: 75421a6ec247a93439132edc3e71fadb26f029eb (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
dnl  AMD64 mpn_mod_34lsub1 -- remainder modulo 2^48-1.

dnl  Copyright 2000-2002, 2004, 2005, 2007, 2009-2012 Free Software Foundation,
dnl  Inc.

dnl  This file is part of the GNU MP Library.
dnl
dnl  The GNU MP Library is free software; you can redistribute it and/or modify
dnl  it under the terms of either:
dnl
dnl    * the GNU Lesser General Public License as published by the Free
dnl      Software Foundation; either version 3 of the License, or (at your
dnl      option) any later version.
dnl
dnl  or
dnl
dnl    * the GNU General Public License as published by the Free Software
dnl      Foundation; either version 2 of the License, or (at your option) any
dnl      later version.
dnl
dnl  or both in parallel, as here.
dnl
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 General Public License
dnl  for more details.
dnl
dnl  You should have received copies of the GNU General Public License and the
dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
dnl  see https://www.gnu.org/licenses/.

include(`../config.m4')


C	    cycles/limb
C AMD K8,K9	 0.67	   0.583 is possible with zero-reg instead of $0, 4-way
C AMD K10	 0.67	   this seems hard to beat
C AMD bd1	 1
C AMD bd2	 1
C AMD bd3	 ?
C AMD bd4	 ?
C AMD zen	 0.62
C AMD bobcat	 1.07
C AMD jaguar	 1
C Intel P4	 7.35	   terrible, use old code
C Intel core2	 1.25	   1+epsilon with huge unrolling
C Intel NHM	 1.15	   this seems hard to beat
C Intel SBR	 0.93
C Intel IBR	 0.93
C Intel HWL	 0.82
C Intel BWL	 0.64
C Intel SKY	 0.60
C Intel atom	 2.5
C Intel SLM      1.59
C VIA nano	 1.25	   this seems hard to beat

C INPUT PARAMETERS
define(`ap',	%rdi)
define(`n',	%rsi)

C mp_limb_t mpn_mod_34lsub1 (mp_srcptr up, mp_size_t n)

C TODO
C  * Review feed-in and wind-down code.

ABI_SUPPORT(DOS64)
ABI_SUPPORT(STD64)

ASM_START()
	TEXT
	ALIGN(32)
PROLOGUE(mpn_mod_34lsub1)
	FUNC_ENTRY(2)

	mov	$0x0000FFFFFFFFFFFF, %r11

	mov	(ap), %rax

	cmp	$2, %rsi
	ja	L(gt2)

	jb	L(one)

	mov	8(ap), %rsi
	mov	%rax, %rdx
	shr	$48, %rax		C src[0] low

	and	%r11, %rdx		C src[0] high
	add	%rdx, %rax
	mov	R32(%rsi), R32(%rdx)

	shr	$32, %rsi		C src[1] high
	add	%rsi, %rax

	shl	$16, %rdx		C src[1] low
	add	%rdx, %rax
L(one):	FUNC_EXIT()
	ret


C Don't change this, the wind-down code is not able to handle greater values
define(UNROLL,3)

L(gt2):	mov	8(ap), %rcx
	mov	16(ap), %rdx
	xor	%r9, %r9
	add	$24, ap
	sub	$eval(UNROLL*3+3), %rsi
	jc	L(end)
	ALIGN(16)
L(top):
	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
forloop(i,1,UNROLL-1,`dnl
	add	eval(i*24)(ap), %rax
	adc	eval(i*24+8)(ap), %rcx
	adc	eval(i*24+16)(ap), %rdx
	adc	$0, %r9
')dnl
	add	$eval(UNROLL*24), ap
	sub	$eval(UNROLL*3), %rsi
	jnc	L(top)

L(end):
	lea	L(tab)(%rip), %r8
ifdef(`PIC',
`	movslq	36(%r8,%rsi,4), %r10
	add	%r10, %r8
	jmp	*%r8
',`
	jmp	*72(%r8,%rsi,8)
')
	JUMPTABSECT
	ALIGN(8)
L(tab):	JMPENT(	L(0), L(tab))
	JMPENT(	L(1), L(tab))
	JMPENT(	L(2), L(tab))
	JMPENT(	L(3), L(tab))
	JMPENT(	L(4), L(tab))
	JMPENT(	L(5), L(tab))
	JMPENT(	L(6), L(tab))
	JMPENT(	L(7), L(tab))
	JMPENT(	L(8), L(tab))
	TEXT

L(6):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
	add	$24, ap
L(3):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	jmp	L(cj1)

L(7):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
	add	$24, ap
L(4):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
	add	$24, ap
L(1):	add	(ap), %rax
	adc	$0, %rcx
	jmp	L(cj2)

L(8):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
	add	$24, ap
L(5):	add	(ap), %rax
	adc	8(ap), %rcx
	adc	16(ap), %rdx
	adc	$0, %r9
	add	$24, ap
L(2):	add	(ap), %rax
	adc	8(ap), %rcx

L(cj2):	adc	$0, %rdx
L(cj1):	adc	$0, %r9
L(0):	add	%r9, %rax
	adc	$0, %rcx
	adc	$0, %rdx
	adc	$0, %rax

	mov	%rax, %rdi		C 0mod3
	shr	$48, %rax		C 0mod3 high

	and	%r11, %rdi		C 0mod3 low
	mov	R32(%rcx), R32(%r10)	C 1mod3

	shr	$32, %rcx		C 1mod3 high

	add	%rdi, %rax		C apply 0mod3 low
	movzwl	%dx, R32(%rdi)		C 2mod3
	shl	$16, %r10		C 1mod3 low

	add	%rcx, %rax		C apply 1mod3 high
	shr	$16, %rdx		C 2mod3 high

	add	%r10, %rax		C apply 1mod3 low
	shl	$32, %rdi		C 2mod3 low

	add	%rdx, %rax		C apply 2mod3 high
	add	%rdi, %rax		C apply 2mod3 low

	FUNC_EXIT()
	ret
EPILOGUE()