1dnl  Intel P6 mpn_mod_34lsub1 -- remainder modulo 2^24-1.
2
3dnl  Copyright 2000-2002, 2004 Free Software Foundation, Inc.
4
5dnl  This file is part of the GNU MP Library.
6dnl
7dnl  The GNU MP Library is free software; you can redistribute it and/or modify
8dnl  it under the terms of either:
9dnl
10dnl    * the GNU Lesser General Public License as published by the Free
11dnl      Software Foundation; either version 3 of the License, or (at your
12dnl      option) any later version.
13dnl
14dnl  or
15dnl
16dnl    * the GNU General Public License as published by the Free Software
17dnl      Foundation; either version 2 of the License, or (at your option) any
18dnl      later version.
19dnl
20dnl  or both in parallel, as here.
21dnl
22dnl  The GNU MP Library is distributed in the hope that it will be useful, but
23dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
24dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
25dnl  for more details.
26dnl
27dnl  You should have received copies of the GNU General Public License and the
28dnl  GNU Lesser General Public License along with the GNU MP Library.  If not,
29dnl  see https://www.gnu.org/licenses/.
30
31include(`../config.m4')
32
33
34C P6: 2.0 cycles/limb
35
36C TODO
37C  Experiments with more unrolling indicate that 1.5 c/l is possible on P6-13
38C  with the current carry handling scheme.
39
40C mp_limb_t mpn_mod_34lsub1 (mp_srcptr src, mp_size_t size)
41C
42C Groups of three limbs are handled, with carry bits from 0mod3 into 1mod3
43C into 2mod3, but at that point going into a separate carries total so we
44C don't keep the carry flag live across the loop control.  Avoiding decl
45C lets us get to 2.0 c/l, as compared to the generic x86 code at 3.66.
46C
47
48defframe(PARAM_SIZE, 8)
49defframe(PARAM_SRC,  4)
50
51dnl  re-use parameter space
52define(SAVE_EBX, `PARAM_SIZE')
53define(SAVE_ESI, `PARAM_SRC')
54
55          TEXT
56          ALIGN(16)
57PROLOGUE(mpn_mod_34lsub1)
58deflit(`FRAME',0)
59
60          movl      PARAM_SIZE, %ecx
61          movl      PARAM_SRC, %edx
62
63          subl      $2, %ecx            C size-2
64          movl      (%edx), %eax                  C src[0]
65          ja        L(three_or_more)
66          jb        L(one)
67
68          C size==2
69
70          movl      4(%edx), %ecx                 C src[1]
71
72          movl      %eax, %edx                    C src[0]
73          shrl      $24, %eax           C src[0] high
74
75          andl      $0xFFFFFF, %edx               C src[0] low
76
77          addl      %edx, %eax
78          movl      %ecx, %edx                    C src[1]
79          shrl      $16, %ecx           C src[1] high
80
81          andl      $0xFFFF, %edx
82          addl      %ecx, %eax
83
84          shll      $8, %edx            C src[1] low
85
86          addl      %edx, %eax
87L(one):
88          ret
89
90
91L(three_or_more):
92          C eax     src[0], initial acc 0mod3
93          C ebx
94          C ecx     size-2
95          C edx     src
96          C esi
97          C edi
98          C ebp
99
100          movl      %ebx, SAVE_EBX
101          movl      4(%edx), %ebx                 C src[1], initial 1mod3
102          subl      $3, %ecx            C size-5
103
104          movl      %esi, SAVE_ESI
105          movl      8(%edx), %esi                 C src[2], initial 2mod3
106
107          pushl     %edi      FRAME_pushl()
108          movl      $0, %edi            C initial carries 0mod3
109          jng       L(done)                       C if size < 6
110
111
112L(top):
113          C eax     acc 0mod3
114          C ebx     acc 1mod3
115          C ecx     counter, limbs
116          C edx     src
117          C esi     acc 2mod3
118          C edi     carrys into 0mod3
119          C ebp
120
121          addl      12(%edx), %eax
122          adcl      16(%edx), %ebx
123          adcl      20(%edx), %esi
124          leal      12(%edx), %edx
125          adcl      $0, %edi
126
127          subl      $3, %ecx
128          jg        L(top)                        C at least 3 more to process
129
130
131L(done):
132          C ecx is -2, -1 or 0 representing 0, 1 or 2 more limbs respectively
133          cmpl      $-1, %ecx
134          jl        L(done_0)           C if -2, meaning 0 more limbs
135
136          C 1 or 2 more limbs
137          movl      $0, %ecx
138          je        L(done_1)           C if -1, meaning 1 more limb only
139          movl      16(%edx), %ecx
140L(done_1):
141          addl      12(%edx), %eax                C 0mod3
142          adcl      %ecx, %ebx                    C 1mod3
143          adcl      $0, %esi            C 2mod3
144          adcl      $0, %edi            C carries 0mod3
145
146L(done_0):
147          C eax     acc 0mod3
148          C ebx     acc 1mod3
149          C ecx
150          C edx
151          C esi     acc 2mod3
152          C edi     carries 0mod3
153          C ebp
154
155          movl      %eax, %ecx                    C 0mod3
156          shrl      $24, %eax           C 0mod3 high initial total
157
158          andl      $0xFFFFFF, %ecx               C 0mod3 low
159          movl      %edi, %edx                    C carries
160          shrl      $24, %edi           C carries high
161
162          addl      %ecx, %eax                    C add 0mod3 low
163          andl      $0xFFFFFF, %edx               C carries 0mod3 low
164          movl      %ebx, %ecx                    C 1mod3
165
166          shrl      $16, %ebx           C 1mod3 high
167          addl      %edi, %eax                    C add carries high
168          addl      %edx, %eax                    C add carries 0mod3 low
169
170          andl      $0xFFFF, %ecx                 C 1mod3 low mask
171          addl      %ebx, %eax                    C add 1mod3 high
172          movl      SAVE_EBX, %ebx
173
174          shll      $8, %ecx            C 1mod3 low
175          movl      %esi, %edx                    C 2mod3
176          popl      %edi      FRAME_popl()
177
178          shrl      $8, %esi            C 2mod3 high
179          andl      $0xFF, %edx                   C 2mod3 low mask
180          addl      %ecx, %eax                    C add 1mod3 low
181
182          shll      $16, %edx           C 2mod3 low
183          addl      %esi, %eax                    C add 2mod3 high
184          movl      SAVE_ESI, %esi
185
186          addl      %edx, %eax                    C add 2mod3 low
187
188          ret
189
190EPILOGUE()
191