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
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
|
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# March 2010
#
# The module implements "4-bit" GCM GHASH function and underlying
# single multiplication operation in GF(2^128). "4-bit" means that it
# uses 256 bytes per-key table [+128 bytes shared table]. Performance
# results are for streamed GHASH subroutine on UltraSPARC pre-Tx CPU
# and are expressed in cycles per processed byte, less is better:
#
# gcc 3.3.x cc 5.2 this assembler
#
# 32-bit build 81.4 43.3 12.6 (+546%/+244%)
# 64-bit build 20.2 21.2 12.6 (+60%/+68%)
#
# Here is data collected on UltraSPARC T1 system running Linux:
#
# gcc 4.4.1 this assembler
#
# 32-bit build 566 50 (+1000%)
# 64-bit build 56 50 (+12%)
#
# I don't quite understand why difference between 32-bit and 64-bit
# compiler-generated code is so big. Compilers *were* instructed to
# generate code for UltraSPARC and should have used 64-bit registers
# for Z vector (see C code) even in 32-bit build... Oh well, it only
# means more impressive improvement coefficients for this assembler
# module;-) Loops are aggressively modulo-scheduled in respect to
# references to input data and Z.hi updates to achieve 12 cycles
# timing. To anchor to something else, sha1-sparcv9.pl spends 11.6
# cycles to process one byte on UltraSPARC pre-Tx CPU and ~24 on T1.
$bits=32;
for (@ARGV) { $bits=64 if (/\-m64/ || /\-xarch\=v9/); }
if ($bits==64) { $bias=2047; $frame=192; }
else { $bias=0; $frame=112; }
$output=shift;
open STDOUT,">$output";
$Zhi="%o0"; # 64-bit values
$Zlo="%o1";
$Thi="%o2";
$Tlo="%o3";
$rem="%o4";
$tmp="%o5";
$nhi="%l0"; # small values and pointers
$nlo="%l1";
$xi0="%l2";
$xi1="%l3";
$rem_4bit="%l4";
$remi="%l5";
$Htblo="%l6";
$cnt="%l7";
$Xi="%i0"; # input argument block
$Htbl="%i1";
$inp="%i2";
$len="%i3";
$code.=<<___;
.section ".text",#alloc,#execinstr
.align 64
rem_4bit:
.long `0x0000<<16`,0,`0x1C20<<16`,0,`0x3840<<16`,0,`0x2460<<16`,0
.long `0x7080<<16`,0,`0x6CA0<<16`,0,`0x48C0<<16`,0,`0x54E0<<16`,0
.long `0xE100<<16`,0,`0xFD20<<16`,0,`0xD940<<16`,0,`0xC560<<16`,0
.long `0x9180<<16`,0,`0x8DA0<<16`,0,`0xA9C0<<16`,0,`0xB5E0<<16`,0
.type rem_4bit,#object
.size rem_4bit,(.-rem_4bit)
.globl gcm_ghash_4bit
.align 32
gcm_ghash_4bit:
save %sp,-$frame,%sp
ldub [$inp+15],$nlo
ldub [$Xi+15],$xi0
ldub [$Xi+14],$xi1
add $len,$inp,$len
add $Htbl,8,$Htblo
1: call .+8
add %o7,rem_4bit-1b,$rem_4bit
.Louter:
xor $xi0,$nlo,$nlo
and $nlo,0xf0,$nhi
and $nlo,0x0f,$nlo
sll $nlo,4,$nlo
ldx [$Htblo+$nlo],$Zlo
ldx [$Htbl+$nlo],$Zhi
ldub [$inp+14],$nlo
ldx [$Htblo+$nhi],$Tlo
and $Zlo,0xf,$remi
ldx [$Htbl+$nhi],$Thi
sll $remi,3,$remi
ldx [$rem_4bit+$remi],$rem
srlx $Zlo,4,$Zlo
mov 13,$cnt
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $xi1,$nlo,$nlo
and $Zlo,0xf,$remi
and $nlo,0xf0,$nhi
and $nlo,0x0f,$nlo
ba .Lghash_inner
sll $nlo,4,$nlo
.align 32
.Lghash_inner:
ldx [$Htblo+$nlo],$Tlo
sll $remi,3,$remi
xor $Thi,$Zhi,$Zhi
ldx [$Htbl+$nlo],$Thi
srlx $Zlo,4,$Zlo
xor $rem,$Zhi,$Zhi
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
ldub [$inp+$cnt],$nlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
ldub [$Xi+$cnt],$xi1
xor $Thi,$Zhi,$Zhi
and $Zlo,0xf,$remi
ldx [$Htblo+$nhi],$Tlo
sll $remi,3,$remi
xor $rem,$Zhi,$Zhi
ldx [$Htbl+$nhi],$Thi
srlx $Zlo,4,$Zlo
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $xi1,$nlo,$nlo
srlx $Zhi,4,$Zhi
and $nlo,0xf0,$nhi
addcc $cnt,-1,$cnt
xor $Zlo,$tmp,$Zlo
and $nlo,0x0f,$nlo
xor $Tlo,$Zlo,$Zlo
sll $nlo,4,$nlo
blu .Lghash_inner
and $Zlo,0xf,$remi
ldx [$Htblo+$nlo],$Tlo
sll $remi,3,$remi
xor $Thi,$Zhi,$Zhi
ldx [$Htbl+$nlo],$Thi
srlx $Zlo,4,$Zlo
xor $rem,$Zhi,$Zhi
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
add $inp,16,$inp
cmp $inp,$len
be,pn `$bits==64?"%xcc":"%icc"`,.Ldone
and $Zlo,0xf,$remi
ldx [$Htblo+$nhi],$Tlo
sll $remi,3,$remi
xor $rem,$Zhi,$Zhi
ldx [$Htbl+$nhi],$Thi
srlx $Zlo,4,$Zlo
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
ldub [$inp+15],$nlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
stx $Zlo,[$Xi+8]
xor $rem,$Zhi,$Zhi
stx $Zhi,[$Xi]
srl $Zlo,8,$xi1
and $Zlo,0xff,$xi0
ba .Louter
and $xi1,0xff,$xi1
.align 32
.Ldone:
ldx [$Htblo+$nhi],$Tlo
sll $remi,3,$remi
xor $rem,$Zhi,$Zhi
ldx [$Htbl+$nhi],$Thi
srlx $Zlo,4,$Zlo
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
stx $Zlo,[$Xi+8]
xor $rem,$Zhi,$Zhi
stx $Zhi,[$Xi]
ret
restore
.type gcm_ghash_4bit,#function
.size gcm_ghash_4bit,(.-gcm_ghash_4bit)
___
undef $inp;
undef $len;
$code.=<<___;
.globl gcm_gmult_4bit
.align 32
gcm_gmult_4bit:
save %sp,-$frame,%sp
ldub [$Xi+15],$nlo
add $Htbl,8,$Htblo
1: call .+8
add %o7,rem_4bit-1b,$rem_4bit
and $nlo,0xf0,$nhi
and $nlo,0x0f,$nlo
sll $nlo,4,$nlo
ldx [$Htblo+$nlo],$Zlo
ldx [$Htbl+$nlo],$Zhi
ldub [$Xi+14],$nlo
ldx [$Htblo+$nhi],$Tlo
and $Zlo,0xf,$remi
ldx [$Htbl+$nhi],$Thi
sll $remi,3,$remi
ldx [$rem_4bit+$remi],$rem
srlx $Zlo,4,$Zlo
mov 13,$cnt
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
and $Zlo,0xf,$remi
and $nlo,0xf0,$nhi
and $nlo,0x0f,$nlo
ba .Lgmult_inner
sll $nlo,4,$nlo
.align 32
.Lgmult_inner:
ldx [$Htblo+$nlo],$Tlo
sll $remi,3,$remi
xor $Thi,$Zhi,$Zhi
ldx [$Htbl+$nlo],$Thi
srlx $Zlo,4,$Zlo
xor $rem,$Zhi,$Zhi
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
ldub [$Xi+$cnt],$nlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
and $Zlo,0xf,$remi
ldx [$Htblo+$nhi],$Tlo
sll $remi,3,$remi
xor $rem,$Zhi,$Zhi
ldx [$Htbl+$nhi],$Thi
srlx $Zlo,4,$Zlo
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
srlx $Zhi,4,$Zhi
and $nlo,0xf0,$nhi
addcc $cnt,-1,$cnt
xor $Zlo,$tmp,$Zlo
and $nlo,0x0f,$nlo
xor $Tlo,$Zlo,$Zlo
sll $nlo,4,$nlo
blu .Lgmult_inner
and $Zlo,0xf,$remi
ldx [$Htblo+$nlo],$Tlo
sll $remi,3,$remi
xor $Thi,$Zhi,$Zhi
ldx [$Htbl+$nlo],$Thi
srlx $Zlo,4,$Zlo
xor $rem,$Zhi,$Zhi
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
and $Zlo,0xf,$remi
ldx [$Htblo+$nhi],$Tlo
sll $remi,3,$remi
xor $rem,$Zhi,$Zhi
ldx [$Htbl+$nhi],$Thi
srlx $Zlo,4,$Zlo
ldx [$rem_4bit+$remi],$rem
sllx $Zhi,60,$tmp
xor $Tlo,$Zlo,$Zlo
srlx $Zhi,4,$Zhi
xor $Zlo,$tmp,$Zlo
xor $Thi,$Zhi,$Zhi
stx $Zlo,[$Xi+8]
xor $rem,$Zhi,$Zhi
stx $Zhi,[$Xi]
ret
restore
.type gcm_gmult_4bit,#function
.size gcm_gmult_4bit,(.-gcm_gmult_4bit)
.asciz "GHASH for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
.align 4
___
$code =~ s/\`([^\`]*)\`/eval $1/gem;
print $code;
close STDOUT;
|