summaryrefslogtreecommitdiff
path: root/src/lib/libcrypto/sha/asm
diff options
context:
space:
mode:
Diffstat (limited to 'src/lib/libcrypto/sha/asm')
-rw-r--r--src/lib/libcrypto/sha/asm/README1
-rw-r--r--src/lib/libcrypto/sha/asm/sha1-586.pl425
-rw-r--r--src/lib/libcrypto/sha/asm/sha1-ia64.pl549
3 files changed, 975 insertions, 0 deletions
diff --git a/src/lib/libcrypto/sha/asm/README b/src/lib/libcrypto/sha/asm/README
new file mode 100644
index 0000000000..b7e755765f
--- /dev/null
+++ b/src/lib/libcrypto/sha/asm/README
@@ -0,0 +1 @@
C2.pl works
diff --git a/src/lib/libcrypto/sha/asm/sha1-586.pl b/src/lib/libcrypto/sha/asm/sha1-586.pl
new file mode 100644
index 0000000000..041acc0348
--- /dev/null
+++ b/src/lib/libcrypto/sha/asm/sha1-586.pl
@@ -0,0 +1,425 @@
1#!/usr/local/bin/perl
2
3# It was noted that Intel IA-32 C compiler generates code which
4# performs ~30% *faster* on P4 CPU than original *hand-coded*
5# SHA1 assembler implementation. To address this problem (and
6# prove that humans are still better than machines:-), the
7# original code was overhauled, which resulted in following
8# performance changes:
9#
10# compared with original compared with Intel cc
11# assembler impl. generated code
12# Pentium -25% +37%
13# PIII/AMD +8% +16%
14# P4 +85%(!) +45%
15#
16# As you can see Pentium came out as looser:-( Yet I reckoned that
17# improvement on P4 outweights the loss and incorporate this
18# re-tuned code to 0.9.7 and later.
19# ----------------------------------------------------------------
20# Those who for any particular reason absolutely must score on
21# Pentium can replace this module with one from 0.9.6 distribution.
22# This "offer" shall be revoked the moment programming interface to
23# this module is changed, in which case this paragraph should be
24# removed.
25# ----------------------------------------------------------------
26# <appro@fy.chalmers.se>
27
28$normal=0;
29
30push(@INC,"perlasm","../../perlasm");
31require "x86asm.pl";
32
33&asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386");
34
35$A="eax";
36$B="ecx";
37$C="ebx";
38$D="edx";
39$E="edi";
40$T="esi";
41$tmp1="ebp";
42
43$off=9*4;
44
45@K=(0x5a827999,0x6ed9eba1,0x8f1bbcdc,0xca62c1d6);
46
47&sha1_block_data("sha1_block_asm_data_order");
48
49&asm_finish();
50
51sub Nn
52 {
53 local($p)=@_;
54 local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E);
55 return($n{$p});
56 }
57
58sub Np
59 {
60 local($p)=@_;
61 local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E);
62 local(%n)=($A,$B,$B,$C,$C,$D,$D,$E,$E,$T,$T,$A);
63 return($n{$p});
64 }
65
66sub Na
67 {
68 local($n)=@_;
69 return( (($n )&0x0f),
70 (($n+ 2)&0x0f),
71 (($n+ 8)&0x0f),
72 (($n+13)&0x0f),
73 (($n+ 1)&0x0f));
74 }
75
76sub X_expand
77 {
78 local($in)=@_;
79
80 &comment("First, load the words onto the stack in network byte order");
81 for ($i=0; $i<16; $i+=2)
82 {
83 &mov($A,&DWP(($i+0)*4,$in,"",0));# unless $i == 0;
84 &mov($B,&DWP(($i+1)*4,$in,"",0));
85 &bswap($A);
86 &bswap($B);
87 &mov(&swtmp($i+0),$A);
88 &mov(&swtmp($i+1),$B);
89 }
90
91 &comment("We now have the X array on the stack");
92 &comment("starting at sp-4");
93 }
94
95# Rules of engagement
96# F is always trashable at the start, the running total.
97# E becomes the next F so it can be trashed after it has been 'accumulated'
98# F becomes A in the next round. We don't need to access it much.
99# During the X update part, the result ends up in $X[$n0].
100
101sub BODY_00_15
102 {
103 local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
104
105 &comment("00_15 $n");
106
107 &mov($tmp1,$a);
108 &mov($f,$c); # f to hold F_00_19(b,c,d)
109 &rotl($tmp1,5); # tmp1=ROTATE(a,5)
110 &xor($f,$d);
111 &and($f,$b);
112 &rotr($b,2); # b=ROTATE(b,30)
113 &add($tmp1,$e); # tmp1+=e;
114 &mov($e,&swtmp($n)); # e becomes volatile and
115 # is loaded with xi
116 &xor($f,$d); # f holds F_00_19(b,c,d)
117 &lea($tmp1,&DWP($K,$tmp1,$e,1));# tmp1+=K_00_19+xi
118
119 &add($f,$tmp1); # f+=tmp1
120 }
121
122sub BODY_16_19
123 {
124 local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
125 local($n0,$n1,$n2,$n3,$np)=&Na($n);
126
127 &comment("16_19 $n");
128
129 &mov($f,&swtmp($n1)); # f to hold Xupdate(xi,xa,xb,xc,xd)
130 &mov($tmp1,$c); # tmp1 to hold F_00_19(b,c,d)
131 &xor($f,&swtmp($n0));
132 &xor($tmp1,$d);
133 &xor($f,&swtmp($n2));
134 &and($tmp1,$b); # tmp1 holds F_00_19(b,c,d)
135 &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd
136 &rotr($b,2); # b=ROTATE(b,30)
137 &xor($tmp1,$d); # tmp1=F_00_19(b,c,d)
138 &rotl($f,1); # f=ROATE(f,1)
139 &mov(&swtmp($n0),$f); # xi=f
140 &lea($f,&DWP($K,$f,$e,1)); # f+=K_00_19+e
141 &mov($e,$a); # e becomes volatile
142 &add($f,$tmp1); # f+=F_00_19(b,c,d)
143 &rotl($e,5); # e=ROTATE(a,5)
144 &add($f,$e); # f+=ROTATE(a,5)
145 }
146
147sub BODY_20_39
148 {
149 local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
150
151 &comment("20_39 $n");
152 local($n0,$n1,$n2,$n3,$np)=&Na($n);
153
154 &mov($f,&swtmp($n0)); # f to hold Xupdate(xi,xa,xb,xc,xd)
155 &mov($tmp1,$b); # tmp1 to hold F_20_39(b,c,d)
156 &xor($f,&swtmp($n1));
157 &rotr($b,2); # b=ROTATE(b,30)
158 &xor($f,&swtmp($n2));
159 &xor($tmp1,$c);
160 &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd
161 &xor($tmp1,$d); # tmp1 holds F_20_39(b,c,d)
162 &rotl($f,1); # f=ROTATE(f,1)
163 &mov(&swtmp($n0),$f); # xi=f
164 &lea($f,&DWP($K,$f,$e,1)); # f+=K_20_39+e
165 &mov($e,$a); # e becomes volatile
166 &rotl($e,5); # e=ROTATE(a,5)
167 &add($f,$tmp1); # f+=F_20_39(b,c,d)
168 &add($f,$e); # f+=ROTATE(a,5)
169 }
170
171sub BODY_40_59
172 {
173 local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_;
174
175 &comment("40_59 $n");
176 local($n0,$n1,$n2,$n3,$np)=&Na($n);
177
178 &mov($f,&swtmp($n0)); # f to hold Xupdate(xi,xa,xb,xc,xd)
179 &mov($tmp1,$b); # tmp1 to hold F_40_59(b,c,d)
180 &xor($f,&swtmp($n1));
181 &or($tmp1,$c);
182 &xor($f,&swtmp($n2));
183 &and($tmp1,$d);
184 &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd
185 &rotl($f,1); # f=ROTATE(f,1)
186 &mov(&swtmp($n0),$f); # xi=f
187 &lea($f,&DWP($K,$f,$e,1)); # f+=K_40_59+e
188 &mov($e,$b); # e becomes volatile and is used
189 # to calculate F_40_59(b,c,d)
190 &rotr($b,2); # b=ROTATE(b,30)
191 &and($e,$c);
192 &or($tmp1,$e); # tmp1 holds F_40_59(b,c,d)
193 &mov($e,$a);
194 &rotl($e,5); # e=ROTATE(a,5)
195 &add($tmp1,$e); # tmp1+=ROTATE(a,5)
196 &add($f,$tmp1); # f+=tmp1;
197 }
198
199sub BODY_60_79
200 {
201 &BODY_20_39(@_);
202 }
203
204sub sha1_block_host
205 {
206 local($name, $sclabel)=@_;
207
208 &function_begin_B($name,"");
209
210 # parameter 1 is the MD5_CTX structure.
211 # A 0
212 # B 4
213 # C 8
214 # D 12
215 # E 16
216
217 &mov("ecx", &wparam(2));
218 &push("esi");
219 &shl("ecx",6);
220 &mov("esi", &wparam(1));
221 &push("ebp");
222 &add("ecx","esi"); # offset to leave on
223 &push("ebx");
224 &mov("ebp", &wparam(0));
225 &push("edi");
226 &mov($D, &DWP(12,"ebp","",0));
227 &stack_push(18+9);
228 &mov($E, &DWP(16,"ebp","",0));
229 &mov($C, &DWP( 8,"ebp","",0));
230 &mov(&swtmp(17),"ecx");
231
232 &comment("First we need to setup the X array");
233
234 for ($i=0; $i<16; $i+=2)
235 {
236 &mov($A,&DWP(($i+0)*4,"esi","",0));# unless $i == 0;
237 &mov($B,&DWP(($i+1)*4,"esi","",0));
238 &mov(&swtmp($i+0),$A);
239 &mov(&swtmp($i+1),$B);
240 }
241 &jmp($sclabel);
242 &function_end_B($name);
243 }
244
245
246sub sha1_block_data
247 {
248 local($name)=@_;
249
250 &function_begin_B($name,"");
251
252 # parameter 1 is the MD5_CTX structure.
253 # A 0
254 # B 4
255 # C 8
256 # D 12
257 # E 16
258
259 &mov("ecx", &wparam(2));
260 &push("esi");
261 &shl("ecx",6);
262 &mov("esi", &wparam(1));
263 &push("ebp");
264 &add("ecx","esi"); # offset to leave on
265 &push("ebx");
266 &mov("ebp", &wparam(0));
267 &push("edi");
268 &mov($D, &DWP(12,"ebp","",0));
269 &stack_push(18+9);
270 &mov($E, &DWP(16,"ebp","",0));
271 &mov($C, &DWP( 8,"ebp","",0));
272 &mov(&swtmp(17),"ecx");
273
274 &comment("First we need to setup the X array");
275
276 &set_label("start") unless $normal;
277
278 &X_expand("esi");
279 &mov(&wparam(1),"esi");
280
281 &set_label("shortcut", 0, 1);
282 &comment("");
283 &comment("Start processing");
284
285 # odd start
286 &mov($A, &DWP( 0,"ebp","",0));
287 &mov($B, &DWP( 4,"ebp","",0));
288 $X="esp";
289 &BODY_00_15(-2,$K[0],$X, 0,$A,$B,$C,$D,$E,$T);
290 &BODY_00_15( 0,$K[0],$X, 1,$T,$A,$B,$C,$D,$E);
291 &BODY_00_15( 0,$K[0],$X, 2,$E,$T,$A,$B,$C,$D);
292 &BODY_00_15( 0,$K[0],$X, 3,$D,$E,$T,$A,$B,$C);
293 &BODY_00_15( 0,$K[0],$X, 4,$C,$D,$E,$T,$A,$B);
294 &BODY_00_15( 0,$K[0],$X, 5,$B,$C,$D,$E,$T,$A);
295 &BODY_00_15( 0,$K[0],$X, 6,$A,$B,$C,$D,$E,$T);
296 &BODY_00_15( 0,$K[0],$X, 7,$T,$A,$B,$C,$D,$E);
297 &BODY_00_15( 0,$K[0],$X, 8,$E,$T,$A,$B,$C,$D);
298 &BODY_00_15( 0,$K[0],$X, 9,$D,$E,$T,$A,$B,$C);
299 &BODY_00_15( 0,$K[0],$X,10,$C,$D,$E,$T,$A,$B);
300 &BODY_00_15( 0,$K[0],$X,11,$B,$C,$D,$E,$T,$A);
301 &BODY_00_15( 0,$K[0],$X,12,$A,$B,$C,$D,$E,$T);
302 &BODY_00_15( 0,$K[0],$X,13,$T,$A,$B,$C,$D,$E);
303 &BODY_00_15( 0,$K[0],$X,14,$E,$T,$A,$B,$C,$D);
304 &BODY_00_15( 1,$K[0],$X,15,$D,$E,$T,$A,$B,$C);
305 &BODY_16_19(-1,$K[0],$X,16,$C,$D,$E,$T,$A,$B);
306 &BODY_16_19( 0,$K[0],$X,17,$B,$C,$D,$E,$T,$A);
307 &BODY_16_19( 0,$K[0],$X,18,$A,$B,$C,$D,$E,$T);
308 &BODY_16_19( 1,$K[0],$X,19,$T,$A,$B,$C,$D,$E);
309
310 &BODY_20_39(-1,$K[1],$X,20,$E,$T,$A,$B,$C,$D);
311 &BODY_20_39( 0,$K[1],$X,21,$D,$E,$T,$A,$B,$C);
312 &BODY_20_39( 0,$K[1],$X,22,$C,$D,$E,$T,$A,$B);
313 &BODY_20_39( 0,$K[1],$X,23,$B,$C,$D,$E,$T,$A);
314 &BODY_20_39( 0,$K[1],$X,24,$A,$B,$C,$D,$E,$T);
315 &BODY_20_39( 0,$K[1],$X,25,$T,$A,$B,$C,$D,$E);
316 &BODY_20_39( 0,$K[1],$X,26,$E,$T,$A,$B,$C,$D);
317 &BODY_20_39( 0,$K[1],$X,27,$D,$E,$T,$A,$B,$C);
318 &BODY_20_39( 0,$K[1],$X,28,$C,$D,$E,$T,$A,$B);
319 &BODY_20_39( 0,$K[1],$X,29,$B,$C,$D,$E,$T,$A);
320 &BODY_20_39( 0,$K[1],$X,30,$A,$B,$C,$D,$E,$T);
321 &BODY_20_39( 0,$K[1],$X,31,$T,$A,$B,$C,$D,$E);
322 &BODY_20_39( 0,$K[1],$X,32,$E,$T,$A,$B,$C,$D);
323 &BODY_20_39( 0,$K[1],$X,33,$D,$E,$T,$A,$B,$C);
324 &BODY_20_39( 0,$K[1],$X,34,$C,$D,$E,$T,$A,$B);
325 &BODY_20_39( 0,$K[1],$X,35,$B,$C,$D,$E,$T,$A);
326 &BODY_20_39( 0,$K[1],$X,36,$A,$B,$C,$D,$E,$T);
327 &BODY_20_39( 0,$K[1],$X,37,$T,$A,$B,$C,$D,$E);
328 &BODY_20_39( 0,$K[1],$X,38,$E,$T,$A,$B,$C,$D);
329 &BODY_20_39( 1,$K[1],$X,39,$D,$E,$T,$A,$B,$C);
330
331 &BODY_40_59(-1,$K[2],$X,40,$C,$D,$E,$T,$A,$B);
332 &BODY_40_59( 0,$K[2],$X,41,$B,$C,$D,$E,$T,$A);
333 &BODY_40_59( 0,$K[2],$X,42,$A,$B,$C,$D,$E,$T);
334 &BODY_40_59( 0,$K[2],$X,43,$T,$A,$B,$C,$D,$E);
335 &BODY_40_59( 0,$K[2],$X,44,$E,$T,$A,$B,$C,$D);
336 &BODY_40_59( 0,$K[2],$X,45,$D,$E,$T,$A,$B,$C);
337 &BODY_40_59( 0,$K[2],$X,46,$C,$D,$E,$T,$A,$B);
338 &BODY_40_59( 0,$K[2],$X,47,$B,$C,$D,$E,$T,$A);
339 &BODY_40_59( 0,$K[2],$X,48,$A,$B,$C,$D,$E,$T);
340 &BODY_40_59( 0,$K[2],$X,49,$T,$A,$B,$C,$D,$E);
341 &BODY_40_59( 0,$K[2],$X,50,$E,$T,$A,$B,$C,$D);
342 &BODY_40_59( 0,$K[2],$X,51,$D,$E,$T,$A,$B,$C);
343 &BODY_40_59( 0,$K[2],$X,52,$C,$D,$E,$T,$A,$B);
344 &BODY_40_59( 0,$K[2],$X,53,$B,$C,$D,$E,$T,$A);
345 &BODY_40_59( 0,$K[2],$X,54,$A,$B,$C,$D,$E,$T);
346 &BODY_40_59( 0,$K[2],$X,55,$T,$A,$B,$C,$D,$E);
347 &BODY_40_59( 0,$K[2],$X,56,$E,$T,$A,$B,$C,$D);
348 &BODY_40_59( 0,$K[2],$X,57,$D,$E,$T,$A,$B,$C);
349 &BODY_40_59( 0,$K[2],$X,58,$C,$D,$E,$T,$A,$B);
350 &BODY_40_59( 1,$K[2],$X,59,$B,$C,$D,$E,$T,$A);
351
352 &BODY_60_79(-1,$K[3],$X,60,$A,$B,$C,$D,$E,$T);
353 &BODY_60_79( 0,$K[3],$X,61,$T,$A,$B,$C,$D,$E);
354 &BODY_60_79( 0,$K[3],$X,62,$E,$T,$A,$B,$C,$D);
355 &BODY_60_79( 0,$K[3],$X,63,$D,$E,$T,$A,$B,$C);
356 &BODY_60_79( 0,$K[3],$X,64,$C,$D,$E,$T,$A,$B);
357 &BODY_60_79( 0,$K[3],$X,65,$B,$C,$D,$E,$T,$A);
358 &BODY_60_79( 0,$K[3],$X,66,$A,$B,$C,$D,$E,$T);
359 &BODY_60_79( 0,$K[3],$X,67,$T,$A,$B,$C,$D,$E);
360 &BODY_60_79( 0,$K[3],$X,68,$E,$T,$A,$B,$C,$D);
361 &BODY_60_79( 0,$K[3],$X,69,$D,$E,$T,$A,$B,$C);
362 &BODY_60_79( 0,$K[3],$X,70,$C,$D,$E,$T,$A,$B);
363 &BODY_60_79( 0,$K[3],$X,71,$B,$C,$D,$E,$T,$A);
364 &BODY_60_79( 0,$K[3],$X,72,$A,$B,$C,$D,$E,$T);
365 &BODY_60_79( 0,$K[3],$X,73,$T,$A,$B,$C,$D,$E);
366 &BODY_60_79( 0,$K[3],$X,74,$E,$T,$A,$B,$C,$D);
367 &BODY_60_79( 0,$K[3],$X,75,$D,$E,$T,$A,$B,$C);
368 &BODY_60_79( 0,$K[3],$X,76,$C,$D,$E,$T,$A,$B);
369 &BODY_60_79( 0,$K[3],$X,77,$B,$C,$D,$E,$T,$A);
370 &BODY_60_79( 0,$K[3],$X,78,$A,$B,$C,$D,$E,$T);
371 &BODY_60_79( 2,$K[3],$X,79,$T,$A,$B,$C,$D,$E);
372
373 &comment("End processing");
374 &comment("");
375 # D is the tmp value
376
377 # E -> A
378 # T -> B
379 # A -> C
380 # B -> D
381 # C -> E
382 # D -> T
383
384 &mov($tmp1,&wparam(0));
385
386 &mov($D, &DWP(12,$tmp1,"",0));
387 &add($D,$B);
388 &mov($B, &DWP( 4,$tmp1,"",0));
389 &add($B,$T);
390 &mov($T, $A);
391 &mov($A, &DWP( 0,$tmp1,"",0));
392 &mov(&DWP(12,$tmp1,"",0),$D);
393
394 &add($A,$E);
395 &mov($E, &DWP(16,$tmp1,"",0));
396 &add($E,$C);
397 &mov($C, &DWP( 8,$tmp1,"",0));
398 &add($C,$T);
399
400 &mov(&DWP( 0,$tmp1,"",0),$A);
401 &mov("esi",&wparam(1));
402 &mov(&DWP( 8,$tmp1,"",0),$C);
403 &add("esi",64);
404 &mov("eax",&swtmp(17));
405 &mov(&DWP(16,$tmp1,"",0),$E);
406 &cmp("esi","eax");
407 &mov(&DWP( 4,$tmp1,"",0),$B);
408 &jb(&label("start"));
409
410 &stack_pop(18+9);
411 &pop("edi");
412 &pop("ebx");
413 &pop("ebp");
414 &pop("esi");
415 &ret();
416
417 # keep a note of shortcut label so it can be used outside
418 # block.
419 my $sclabel = &label("shortcut");
420
421 &function_end_B($name);
422 # Putting this here avoids problems with MASM in debugging mode
423 &sha1_block_host("sha1_block_asm_host_order", $sclabel);
424 }
425
diff --git a/src/lib/libcrypto/sha/asm/sha1-ia64.pl b/src/lib/libcrypto/sha/asm/sha1-ia64.pl
new file mode 100644
index 0000000000..cb9dfad124
--- /dev/null
+++ b/src/lib/libcrypto/sha/asm/sha1-ia64.pl
@@ -0,0 +1,549 @@
1#!/usr/bin/env perl
2#
3# ====================================================================
4# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
5# project. Rights for redistribution and usage in source and binary
6# forms are granted according to the OpenSSL license.
7# ====================================================================
8#
9# Eternal question is what's wrong with compiler generated code? The
10# trick is that it's possible to reduce the number of shifts required
11# to perform rotations by maintaining copy of 32-bit value in upper
12# bits of 64-bit register. Just follow mux2 and shrp instructions...
13# Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which
14# is >50% better than HP C and >2x better than gcc. As of this moment
15# performance under little-endian OS such as Linux and Windows will be
16# a bit lower, because data has to be picked in reverse byte-order.
17# It's possible to resolve this issue by implementing third function,
18# sha1_block_asm_data_order_aligned, which would temporarily flip
19# BE field in User Mask register...
20
21$code=<<___;
22.ident \"sha1-ia64.s, version 1.0\"
23.ident \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
24.explicit
25
26___
27
28
29if ($^O eq "hpux") {
30 $ADDP="addp4";
31 for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
32} else { $ADDP="add"; }
33for (@ARGV) { $big_endian=1 if (/\-DB_ENDIAN/);
34 $big_endian=0 if (/\-DL_ENDIAN/); }
35if (!defined($big_endian))
36 { $big_endian=(unpack('L',pack('N',1))==1); }
37
38#$human=1;
39if ($human) { # useful for visual code auditing...
40 ($A,$B,$C,$D,$E,$T) = ("A","B","C","D","E","T");
41 ($h0,$h1,$h2,$h3,$h4) = ("h0","h1","h2","h3","h4");
42 ($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
43 ( "K_00_19","K_20_39","K_40_59","K_60_79" );
44 @X= ( "X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7",
45 "X8", "X9","X10","X11","X12","X13","X14","X15" );
46}
47else {
48 ($A,$B,$C,$D,$E,$T) = ("loc0","loc1","loc2","loc3","loc4","loc5");
49 ($h0,$h1,$h2,$h3,$h4) = ("loc6","loc7","loc8","loc9","loc10");
50 ($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
51 ( "r14", "r15", "loc11", "loc12" );
52 @X= ( "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
53 "r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31" );
54}
55
56sub BODY_00_15 {
57local *code=shift;
58local ($i,$a,$b,$c,$d,$e,$f,$unaligned)=@_;
59
60if ($unaligned) {
61 $code.=<<___;
62{ .mmi; ld1 tmp0=[inp],2 // MSB
63 ld1 tmp1=[tmp3],2 };;
64{ .mmi; ld1 tmp2=[inp],2
65 ld1 $X[$i&0xf]=[tmp3],2 // LSB
66 dep tmp1=tmp0,tmp1,8,8 };;
67{ .mii; cmp.ne p16,p0=r0,r0 // no misaligned prefetch
68 dep $X[$i&0xf]=tmp2,$X[$i&0xf],8,8;;
69 dep $X[$i&0xf]=tmp1,$X[$i&0xf],16,16 };;
70{ .mmi; nop.m 0
71___
72 }
73elsif ($i<15) {
74 $code.=<<___;
75{ .mmi; ld4 $X[($i+1)&0xf]=[inp],4 // prefetch
76___
77 }
78else {
79 $code.=<<___;
80{ .mmi; nop.m 0
81___
82 }
83if ($i<15) {
84 $code.=<<___;
85 and tmp0=$c,$b
86 dep.z tmp5=$a,5,27 } // a<<5
87{ .mmi; andcm tmp1=$d,$b
88 add tmp4=$e,$K_00_19 };;
89{ .mmi; or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d)
90 add $f=tmp4,$X[$i&0xf] // f=xi+e+K_00_19
91 extr.u tmp1=$a,27,5 };; // a>>27
92{ .mib; add $f=$f,tmp0 // f+=F_00_19(b,c,d)
93 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
94{ .mib; or tmp1=tmp1,tmp5 // ROTATE(a,5)
95 mux2 tmp6=$a,0x44 };; // see b in next iteration
96{ .mii; add $f=$f,tmp1 // f+=ROTATE(a,5)
97 mux2 $X[$i&0xf]=$X[$i&0xf],0x44
98 nop.i 0 };;
99
100___
101 }
102else {
103 $code.=<<___;
104 and tmp0=$c,$b
105 dep.z tmp5=$a,5,27 } // a<<5 ;;?
106{ .mmi; andcm tmp1=$d,$b
107 add tmp4=$e,$K_00_19 };;
108{ .mmi; or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d)
109 add $f=tmp4,$X[$i&0xf] // f=xi+e+K_00_19
110 extr.u tmp1=$a,27,5 } // a>>27
111{ .mmi; xor tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf] // +1
112 xor tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
113 nop.i 0 };;
114{ .mmi; add $f=$f,tmp0 // f+=F_00_19(b,c,d)
115 xor tmp2=tmp2,tmp3 // +1
116 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
117{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
118 mux2 tmp6=$a,0x44 };; // see b in next iteration
119{ .mii; add $f=$f,tmp1 // f+=ROTATE(a,5)
120 shrp $e=tmp2,tmp2,31 // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
121 mux2 $X[$i&0xf]=$X[$i&0xf],0x44 };;
122
123___
124 }
125}
126
127sub BODY_16_19 {
128local *code=shift;
129local ($i,$a,$b,$c,$d,$e,$f)=@_;
130
131$code.=<<___;
132{ .mmi; mov $X[$i&0xf]=$f // Xupdate
133 and tmp0=$c,$b
134 dep.z tmp5=$a,5,27 } // a<<5
135{ .mmi; andcm tmp1=$d,$b
136 add tmp4=$e,$K_00_19 };;
137{ .mmi; or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d)
138 add $f=$f,tmp4 // f+=e+K_00_19
139 extr.u tmp1=$a,27,5 } // a>>27
140{ .mmi; xor tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf] // +1
141 xor tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
142 nop.i 0 };;
143{ .mmi; add $f=$f,tmp0 // f+=F_00_19(b,c,d)
144 xor tmp2=tmp2,tmp3 // +1
145 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
146{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
147 mux2 tmp6=$a,0x44 };; // see b in next iteration
148{ .mii; add $f=$f,tmp1 // f+=ROTATE(a,5)
149 shrp $e=tmp2,tmp2,31 // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
150 nop.i 0 };;
151
152___
153}
154
155sub BODY_20_39 {
156local *code=shift;
157local ($i,$a,$b,$c,$d,$e,$f,$Konst)=@_;
158 $Konst = $K_20_39 if (!defined($Konst));
159
160if ($i<79) {
161$code.=<<___;
162{ .mib; mov $X[$i&0xf]=$f // Xupdate
163 dep.z tmp5=$a,5,27 } // a<<5
164{ .mib; xor tmp0=$c,$b
165 add tmp4=$e,$Konst };;
166{ .mmi; xor tmp0=tmp0,$d // F_20_39(b,c,d)=b^c^d
167 add $f=$f,tmp4 // f+=e+K_20_39
168 extr.u tmp1=$a,27,5 } // a>>27
169{ .mmi; xor tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf] // +1
170 xor tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
171 nop.i 0 };;
172{ .mmi; add $f=$f,tmp0 // f+=F_20_39(b,c,d)
173 xor tmp2=tmp2,tmp3 // +1
174 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
175{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
176 mux2 tmp6=$a,0x44 };; // see b in next iteration
177{ .mii; add $f=$f,tmp1 // f+=ROTATE(a,5)
178 shrp $e=tmp2,tmp2,31 // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
179 nop.i 0 };;
180
181___
182}
183else {
184$code.=<<___;
185{ .mib; mov $X[$i&0xf]=$f // Xupdate
186 dep.z tmp5=$a,5,27 } // a<<5
187{ .mib; xor tmp0=$c,$b
188 add tmp4=$e,$Konst };;
189{ .mib; xor tmp0=tmp0,$d // F_20_39(b,c,d)=b^c^d
190 extr.u tmp1=$a,27,5 } // a>>27
191{ .mib; add $f=$f,tmp4 // f+=e+K_20_39
192 add $h1=$h1,$a };; // wrap up
193{ .mmi;
194(p16) ld4.s $X[0]=[inp],4 // non-faulting prefetch
195 add $f=$f,tmp0 // f+=F_20_39(b,c,d)
196 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30) ;;?
197{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
198 add $h3=$h3,$c };; // wrap up
199{ .mib; add tmp3=1,inp // used in unaligned codepath
200 add $f=$f,tmp1 } // f+=ROTATE(a,5)
201{ .mib; add $h2=$h2,$b // wrap up
202 add $h4=$h4,$d };; // wrap up
203
204___
205}
206}
207
208sub BODY_40_59 {
209local *code=shift;
210local ($i,$a,$b,$c,$d,$e,$f)=@_;
211
212$code.=<<___;
213{ .mmi; mov $X[$i&0xf]=$f // Xupdate
214 and tmp0=$c,$b
215 dep.z tmp5=$a,5,27 } // a<<5
216{ .mmi; and tmp1=$d,$b
217 add tmp4=$e,$K_40_59 };;
218{ .mmi; or tmp0=tmp0,tmp1 // (b&c)|(b&d)
219 add $f=$f,tmp4 // f+=e+K_40_59
220 extr.u tmp1=$a,27,5 } // a>>27
221{ .mmi; and tmp4=$c,$d
222 xor tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf] // +1
223 xor tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
224 };;
225{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
226 xor tmp2=tmp2,tmp3 // +1
227 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
228{ .mmi; or tmp0=tmp0,tmp4 // F_40_59(b,c,d)=(b&c)|(b&d)|(c&d)
229 mux2 tmp6=$a,0x44 };; // see b in next iteration
230{ .mii; add $f=$f,tmp0 // f+=F_40_59(b,c,d)
231 shrp $e=tmp2,tmp2,31;; // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
232 add $f=$f,tmp1 };; // f+=ROTATE(a,5)
233
234___
235}
236sub BODY_60_79 { &BODY_20_39(@_,$K_60_79); }
237
238$code.=<<___;
239.text
240
241tmp0=r8;
242tmp1=r9;
243tmp2=r10;
244tmp3=r11;
245ctx=r32; // in0
246inp=r33; // in1
247
248// void sha1_block_asm_host_order(SHA_CTX *c,const void *p,size_t num);
249.global sha1_block_asm_host_order#
250.proc sha1_block_asm_host_order#
251.align 32
252sha1_block_asm_host_order:
253 .prologue
254 .fframe 0
255 .save ar.pfs,r0
256 .save ar.lc,r3
257{ .mmi; alloc tmp1=ar.pfs,3,15,0,0
258 $ADDP tmp0=4,ctx
259 mov r3=ar.lc }
260{ .mmi; $ADDP ctx=0,ctx
261 $ADDP inp=0,inp
262 mov r2=pr };;
263tmp4=in2;
264tmp5=loc13;
265tmp6=loc14;
266 .body
267{ .mlx; ld4 $h0=[ctx],8
268 movl $K_00_19=0x5a827999 }
269{ .mlx; ld4 $h1=[tmp0],8
270 movl $K_20_39=0x6ed9eba1 };;
271{ .mlx; ld4 $h2=[ctx],8
272 movl $K_40_59=0x8f1bbcdc }
273{ .mlx; ld4 $h3=[tmp0]
274 movl $K_60_79=0xca62c1d6 };;
275{ .mmi; ld4 $h4=[ctx],-16
276 add in2=-1,in2 // adjust num for ar.lc
277 mov ar.ec=1 };;
278{ .mmi; ld4 $X[0]=[inp],4 // prefetch
279 cmp.ne p16,p0=r0,in2 // prefecth at loop end
280 mov ar.lc=in2 };; // brp.loop.imp: too far
281
282.Lhtop:
283{ .mmi; mov $A=$h0
284 mov $B=$h1
285 mux2 tmp6=$h1,0x44 }
286{ .mmi; mov $C=$h2
287 mov $D=$h3
288 mov $E=$h4 };;
289
290___
291
292 &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T);
293 &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E);
294 &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D);
295 &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C);
296 &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B);
297 &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A);
298 &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T);
299 &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E);
300 &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D);
301 &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C);
302 &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B);
303 &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A);
304 &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T);
305 &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E);
306 &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D);
307 &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C);
308
309 &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
310 &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
311 &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
312 &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
313
314 &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
315 &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
316 &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
317 &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
318 &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
319 &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
320 &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
321 &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
322 &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
323 &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
324 &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
325 &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
326 &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
327 &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
328 &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
329 &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
330 &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
331 &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
332 &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
333 &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
334
335 &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
336 &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
337 &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
338 &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
339 &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
340 &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
341 &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
342 &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
343 &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
344 &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
345 &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
346 &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
347 &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
348 &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
349 &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
350 &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
351 &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
352 &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
353 &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
354 &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
355
356 &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
357 &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
358 &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
359 &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
360 &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
361 &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
362 &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
363 &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
364 &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
365 &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
366 &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
367 &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
368 &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
369 &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
370 &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
371 &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
372 &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
373 &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
374 &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
375 &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
376
377$code.=<<___;
378{ .mmb; add $h0=$h0,$E
379 nop.m 0
380 br.ctop.dptk.many .Lhtop };;
381.Lhend:
382{ .mmi; add tmp0=4,ctx
383 mov ar.lc=r3 };;
384{ .mmi; st4 [ctx]=$h0,8
385 st4 [tmp0]=$h1,8 };;
386{ .mmi; st4 [ctx]=$h2,8
387 st4 [tmp0]=$h3 };;
388{ .mib; st4 [ctx]=$h4,-16
389 mov pr=r2,0x1ffff
390 br.ret.sptk.many b0 };;
391.endp sha1_block_asm_host_order#
392___
393
394
395$code.=<<___;
396// void sha1_block_asm_data_order(SHA_CTX *c,const void *p,size_t num);
397.global sha1_block_asm_data_order#
398.proc sha1_block_asm_data_order#
399.align 32
400sha1_block_asm_data_order:
401___
402$code.=<<___ if ($big_endian);
403{ .mmi; and r2=3,inp };;
404{ .mib; cmp.eq p6,p0=r0,r2
405(p6) br.dptk.many sha1_block_asm_host_order };;
406___
407$code.=<<___;
408 .prologue
409 .fframe 0
410 .save ar.pfs,r0
411 .save ar.lc,r3
412{ .mmi; alloc tmp1=ar.pfs,3,15,0,0
413 $ADDP tmp0=4,ctx
414 mov r3=ar.lc }
415{ .mmi; $ADDP ctx=0,ctx
416 $ADDP inp=0,inp
417 mov r2=pr };;
418tmp4=in2;
419tmp5=loc13;
420tmp6=loc14;
421 .body
422{ .mlx; ld4 $h0=[ctx],8
423 movl $K_00_19=0x5a827999 }
424{ .mlx; ld4 $h1=[tmp0],8
425 movl $K_20_39=0x6ed9eba1 };;
426{ .mlx; ld4 $h2=[ctx],8
427 movl $K_40_59=0x8f1bbcdc }
428{ .mlx; ld4 $h3=[tmp0]
429 movl $K_60_79=0xca62c1d6 };;
430{ .mmi; ld4 $h4=[ctx],-16
431 add in2=-1,in2 // adjust num for ar.lc
432 mov ar.ec=1 };;
433{ .mmi; nop.m 0
434 add tmp3=1,inp
435 mov ar.lc=in2 };; // brp.loop.imp: too far
436
437.Ldtop:
438{ .mmi; mov $A=$h0
439 mov $B=$h1
440 mux2 tmp6=$h1,0x44 }
441{ .mmi; mov $C=$h2
442 mov $D=$h3
443 mov $E=$h4 };;
444
445___
446
447 &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T,1);
448 &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E,1);
449 &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D,1);
450 &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C,1);
451 &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B,1);
452 &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A,1);
453 &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T,1);
454 &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E,1);
455 &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D,1);
456 &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C,1);
457 &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B,1);
458 &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A,1);
459 &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T,1);
460 &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E,1);
461 &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D,1);
462 &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C,1);
463
464 &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B);
465 &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A);
466 &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T);
467 &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E);
468
469 &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D);
470 &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C);
471 &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B);
472 &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A);
473 &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T);
474 &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E);
475 &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D);
476 &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C);
477 &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B);
478 &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A);
479 &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T);
480 &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E);
481 &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D);
482 &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C);
483 &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B);
484 &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A);
485 &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T);
486 &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E);
487 &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D);
488 &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C);
489
490 &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B);
491 &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A);
492 &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T);
493 &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E);
494 &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D);
495 &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C);
496 &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B);
497 &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A);
498 &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T);
499 &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E);
500 &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D);
501 &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C);
502 &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B);
503 &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A);
504 &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T);
505 &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E);
506 &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D);
507 &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C);
508 &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B);
509 &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A);
510
511 &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T);
512 &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E);
513 &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D);
514 &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C);
515 &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B);
516 &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A);
517 &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T);
518 &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E);
519 &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D);
520 &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C);
521 &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B);
522 &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A);
523 &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T);
524 &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E);
525 &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D);
526 &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C);
527 &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B);
528 &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A);
529 &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T);
530 &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E);
531
532$code.=<<___;
533{ .mmb; add $h0=$h0,$E
534 nop.m 0
535 br.ctop.dptk.many .Ldtop };;
536.Ldend:
537{ .mmi; add tmp0=4,ctx
538 mov ar.lc=r3 };;
539{ .mmi; st4 [ctx]=$h0,8
540 st4 [tmp0]=$h1,8 };;
541{ .mmi; st4 [ctx]=$h2,8
542 st4 [tmp0]=$h3 };;
543{ .mib; st4 [ctx]=$h4,-16
544 mov pr=r2,0x1ffff
545 br.ret.sptk.many b0 };;
546.endp sha1_block_asm_data_order#
547___
548
549print $code;
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-12-10 04:10:14 +0000