root / tests / test-i386.c @ 298e01b6
History | View | Annotate | Download (68.2 kB)
1 |
/*
|
---|---|
2 |
* x86 CPU test
|
3 |
*
|
4 |
* Copyright (c) 2003 Fabrice Bellard
|
5 |
*
|
6 |
* This program is free software; you can redistribute it and/or modify
|
7 |
* it under the terms of the GNU General Public License as published by
|
8 |
* the Free Software Foundation; either version 2 of the License, or
|
9 |
* (at your option) any later version.
|
10 |
*
|
11 |
* This program is distributed in the hope that it will be useful,
|
12 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
14 |
* GNU General Public License for more details.
|
15 |
*
|
16 |
* You should have received a copy of the GNU General Public License
|
17 |
* along with this program; if not, write to the Free Software
|
18 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
|
19 |
*/
|
20 |
#define _GNU_SOURCE
|
21 |
#include <stdlib.h> |
22 |
#include <stdio.h> |
23 |
#include <string.h> |
24 |
#include <inttypes.h> |
25 |
#include <math.h> |
26 |
#include <signal.h> |
27 |
#include <setjmp.h> |
28 |
#include <errno.h> |
29 |
#include <sys/ucontext.h> |
30 |
#include <sys/mman.h> |
31 |
|
32 |
#if !defined(__x86_64__)
|
33 |
#define TEST_VM86
|
34 |
#define TEST_SEGS
|
35 |
#endif
|
36 |
//#define LINUX_VM86_IOPL_FIX
|
37 |
//#define TEST_P4_FLAGS
|
38 |
#if defined(__x86_64__)
|
39 |
#define TEST_SSE
|
40 |
#define TEST_CMOV 1 |
41 |
#define TEST_FCOMI 1 |
42 |
#else
|
43 |
//#define TEST_SSE
|
44 |
#define TEST_CMOV 0 |
45 |
#define TEST_FCOMI 0 |
46 |
#endif
|
47 |
|
48 |
#if defined(__x86_64__)
|
49 |
#define FMT64X "%016lx" |
50 |
#define FMTLX "%016lx" |
51 |
#define X86_64_ONLY(x) x
|
52 |
#else
|
53 |
#define FMT64X "%016" PRIx64 |
54 |
#define FMTLX "%08lx" |
55 |
#define X86_64_ONLY(x)
|
56 |
#endif
|
57 |
|
58 |
#ifdef TEST_VM86
|
59 |
#include <asm/vm86.h> |
60 |
#endif
|
61 |
|
62 |
#define xglue(x, y) x ## y |
63 |
#define glue(x, y) xglue(x, y)
|
64 |
#define stringify(s) tostring(s)
|
65 |
#define tostring(s) #s |
66 |
|
67 |
#define CC_C 0x0001 |
68 |
#define CC_P 0x0004 |
69 |
#define CC_A 0x0010 |
70 |
#define CC_Z 0x0040 |
71 |
#define CC_S 0x0080 |
72 |
#define CC_O 0x0800 |
73 |
|
74 |
#define __init_call __attribute__ ((unused,__section__ ("initcall"))) |
75 |
|
76 |
#define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
|
77 |
|
78 |
#if defined(__x86_64__)
|
79 |
static inline long i2l(long v) |
80 |
{ |
81 |
return v | ((v ^ 0xabcd) << 32); |
82 |
} |
83 |
#else
|
84 |
static inline long i2l(long v) |
85 |
{ |
86 |
return v;
|
87 |
} |
88 |
#endif
|
89 |
|
90 |
#define OP add
|
91 |
#include "test-i386.h" |
92 |
|
93 |
#define OP sub
|
94 |
#include "test-i386.h" |
95 |
|
96 |
#define OP xor
|
97 |
#include "test-i386.h" |
98 |
|
99 |
#define OP and
|
100 |
#include "test-i386.h" |
101 |
|
102 |
#define OP or
|
103 |
#include "test-i386.h" |
104 |
|
105 |
#define OP cmp
|
106 |
#include "test-i386.h" |
107 |
|
108 |
#define OP adc
|
109 |
#define OP_CC
|
110 |
#include "test-i386.h" |
111 |
|
112 |
#define OP sbb
|
113 |
#define OP_CC
|
114 |
#include "test-i386.h" |
115 |
|
116 |
#define OP inc
|
117 |
#define OP_CC
|
118 |
#define OP1
|
119 |
#include "test-i386.h" |
120 |
|
121 |
#define OP dec
|
122 |
#define OP_CC
|
123 |
#define OP1
|
124 |
#include "test-i386.h" |
125 |
|
126 |
#define OP neg
|
127 |
#define OP_CC
|
128 |
#define OP1
|
129 |
#include "test-i386.h" |
130 |
|
131 |
#define OP not
|
132 |
#define OP_CC
|
133 |
#define OP1
|
134 |
#include "test-i386.h" |
135 |
|
136 |
#undef CC_MASK
|
137 |
#define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O)
|
138 |
|
139 |
#define OP shl
|
140 |
#include "test-i386-shift.h" |
141 |
|
142 |
#define OP shr
|
143 |
#include "test-i386-shift.h" |
144 |
|
145 |
#define OP sar
|
146 |
#include "test-i386-shift.h" |
147 |
|
148 |
#define OP rol
|
149 |
#include "test-i386-shift.h" |
150 |
|
151 |
#define OP ror
|
152 |
#include "test-i386-shift.h" |
153 |
|
154 |
#define OP rcr
|
155 |
#define OP_CC
|
156 |
#include "test-i386-shift.h" |
157 |
|
158 |
#define OP rcl
|
159 |
#define OP_CC
|
160 |
#include "test-i386-shift.h" |
161 |
|
162 |
#define OP shld
|
163 |
#define OP_SHIFTD
|
164 |
#define OP_NOBYTE
|
165 |
#include "test-i386-shift.h" |
166 |
|
167 |
#define OP shrd
|
168 |
#define OP_SHIFTD
|
169 |
#define OP_NOBYTE
|
170 |
#include "test-i386-shift.h" |
171 |
|
172 |
/* XXX: should be more precise ? */
|
173 |
#undef CC_MASK
|
174 |
#define CC_MASK (CC_C)
|
175 |
|
176 |
#define OP bt
|
177 |
#define OP_NOBYTE
|
178 |
#include "test-i386-shift.h" |
179 |
|
180 |
#define OP bts
|
181 |
#define OP_NOBYTE
|
182 |
#include "test-i386-shift.h" |
183 |
|
184 |
#define OP btr
|
185 |
#define OP_NOBYTE
|
186 |
#include "test-i386-shift.h" |
187 |
|
188 |
#define OP btc
|
189 |
#define OP_NOBYTE
|
190 |
#include "test-i386-shift.h" |
191 |
|
192 |
/* lea test (modrm support) */
|
193 |
#define TEST_LEAQ(STR)\
|
194 |
{\ |
195 |
asm("lea " STR ", %0"\ |
196 |
: "=r" (res)\
|
197 |
: "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\ |
198 |
printf("lea %s = " FMTLX "\n", STR, res);\ |
199 |
} |
200 |
|
201 |
#define TEST_LEA(STR)\
|
202 |
{\ |
203 |
asm("lea " STR ", %0"\ |
204 |
: "=r" (res)\
|
205 |
: "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\ |
206 |
printf("lea %s = " FMTLX "\n", STR, res);\ |
207 |
} |
208 |
|
209 |
#define TEST_LEA16(STR)\
|
210 |
{\ |
211 |
asm(".code16 ; .byte 0x67 ; leal " STR ", %0 ; .code32"\ |
212 |
: "=wq" (res)\
|
213 |
: "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\ |
214 |
printf("lea %s = %08lx\n", STR, res);\
|
215 |
} |
216 |
|
217 |
|
218 |
void test_lea(void) |
219 |
{ |
220 |
long eax, ebx, ecx, edx, esi, edi, res;
|
221 |
eax = i2l(0x0001);
|
222 |
ebx = i2l(0x0002);
|
223 |
ecx = i2l(0x0004);
|
224 |
edx = i2l(0x0008);
|
225 |
esi = i2l(0x0010);
|
226 |
edi = i2l(0x0020);
|
227 |
|
228 |
TEST_LEA("0x4000");
|
229 |
|
230 |
TEST_LEA("(%%eax)");
|
231 |
TEST_LEA("(%%ebx)");
|
232 |
TEST_LEA("(%%ecx)");
|
233 |
TEST_LEA("(%%edx)");
|
234 |
TEST_LEA("(%%esi)");
|
235 |
TEST_LEA("(%%edi)");
|
236 |
|
237 |
TEST_LEA("0x40(%%eax)");
|
238 |
TEST_LEA("0x40(%%ebx)");
|
239 |
TEST_LEA("0x40(%%ecx)");
|
240 |
TEST_LEA("0x40(%%edx)");
|
241 |
TEST_LEA("0x40(%%esi)");
|
242 |
TEST_LEA("0x40(%%edi)");
|
243 |
|
244 |
TEST_LEA("0x4000(%%eax)");
|
245 |
TEST_LEA("0x4000(%%ebx)");
|
246 |
TEST_LEA("0x4000(%%ecx)");
|
247 |
TEST_LEA("0x4000(%%edx)");
|
248 |
TEST_LEA("0x4000(%%esi)");
|
249 |
TEST_LEA("0x4000(%%edi)");
|
250 |
|
251 |
TEST_LEA("(%%eax, %%ecx)");
|
252 |
TEST_LEA("(%%ebx, %%edx)");
|
253 |
TEST_LEA("(%%ecx, %%ecx)");
|
254 |
TEST_LEA("(%%edx, %%ecx)");
|
255 |
TEST_LEA("(%%esi, %%ecx)");
|
256 |
TEST_LEA("(%%edi, %%ecx)");
|
257 |
|
258 |
TEST_LEA("0x40(%%eax, %%ecx)");
|
259 |
TEST_LEA("0x4000(%%ebx, %%edx)");
|
260 |
|
261 |
TEST_LEA("(%%ecx, %%ecx, 2)");
|
262 |
TEST_LEA("(%%edx, %%ecx, 4)");
|
263 |
TEST_LEA("(%%esi, %%ecx, 8)");
|
264 |
|
265 |
TEST_LEA("(,%%eax, 2)");
|
266 |
TEST_LEA("(,%%ebx, 4)");
|
267 |
TEST_LEA("(,%%ecx, 8)");
|
268 |
|
269 |
TEST_LEA("0x40(,%%eax, 2)");
|
270 |
TEST_LEA("0x40(,%%ebx, 4)");
|
271 |
TEST_LEA("0x40(,%%ecx, 8)");
|
272 |
|
273 |
|
274 |
TEST_LEA("-10(%%ecx, %%ecx, 2)");
|
275 |
TEST_LEA("-10(%%edx, %%ecx, 4)");
|
276 |
TEST_LEA("-10(%%esi, %%ecx, 8)");
|
277 |
|
278 |
TEST_LEA("0x4000(%%ecx, %%ecx, 2)");
|
279 |
TEST_LEA("0x4000(%%edx, %%ecx, 4)");
|
280 |
TEST_LEA("0x4000(%%esi, %%ecx, 8)");
|
281 |
|
282 |
#if defined(__x86_64__)
|
283 |
TEST_LEAQ("0x4000");
|
284 |
TEST_LEAQ("0x4000(%%rip)");
|
285 |
|
286 |
TEST_LEAQ("(%%rax)");
|
287 |
TEST_LEAQ("(%%rbx)");
|
288 |
TEST_LEAQ("(%%rcx)");
|
289 |
TEST_LEAQ("(%%rdx)");
|
290 |
TEST_LEAQ("(%%rsi)");
|
291 |
TEST_LEAQ("(%%rdi)");
|
292 |
|
293 |
TEST_LEAQ("0x40(%%rax)");
|
294 |
TEST_LEAQ("0x40(%%rbx)");
|
295 |
TEST_LEAQ("0x40(%%rcx)");
|
296 |
TEST_LEAQ("0x40(%%rdx)");
|
297 |
TEST_LEAQ("0x40(%%rsi)");
|
298 |
TEST_LEAQ("0x40(%%rdi)");
|
299 |
|
300 |
TEST_LEAQ("0x4000(%%rax)");
|
301 |
TEST_LEAQ("0x4000(%%rbx)");
|
302 |
TEST_LEAQ("0x4000(%%rcx)");
|
303 |
TEST_LEAQ("0x4000(%%rdx)");
|
304 |
TEST_LEAQ("0x4000(%%rsi)");
|
305 |
TEST_LEAQ("0x4000(%%rdi)");
|
306 |
|
307 |
TEST_LEAQ("(%%rax, %%rcx)");
|
308 |
TEST_LEAQ("(%%rbx, %%rdx)");
|
309 |
TEST_LEAQ("(%%rcx, %%rcx)");
|
310 |
TEST_LEAQ("(%%rdx, %%rcx)");
|
311 |
TEST_LEAQ("(%%rsi, %%rcx)");
|
312 |
TEST_LEAQ("(%%rdi, %%rcx)");
|
313 |
|
314 |
TEST_LEAQ("0x40(%%rax, %%rcx)");
|
315 |
TEST_LEAQ("0x4000(%%rbx, %%rdx)");
|
316 |
|
317 |
TEST_LEAQ("(%%rcx, %%rcx, 2)");
|
318 |
TEST_LEAQ("(%%rdx, %%rcx, 4)");
|
319 |
TEST_LEAQ("(%%rsi, %%rcx, 8)");
|
320 |
|
321 |
TEST_LEAQ("(,%%rax, 2)");
|
322 |
TEST_LEAQ("(,%%rbx, 4)");
|
323 |
TEST_LEAQ("(,%%rcx, 8)");
|
324 |
|
325 |
TEST_LEAQ("0x40(,%%rax, 2)");
|
326 |
TEST_LEAQ("0x40(,%%rbx, 4)");
|
327 |
TEST_LEAQ("0x40(,%%rcx, 8)");
|
328 |
|
329 |
|
330 |
TEST_LEAQ("-10(%%rcx, %%rcx, 2)");
|
331 |
TEST_LEAQ("-10(%%rdx, %%rcx, 4)");
|
332 |
TEST_LEAQ("-10(%%rsi, %%rcx, 8)");
|
333 |
|
334 |
TEST_LEAQ("0x4000(%%rcx, %%rcx, 2)");
|
335 |
TEST_LEAQ("0x4000(%%rdx, %%rcx, 4)");
|
336 |
TEST_LEAQ("0x4000(%%rsi, %%rcx, 8)");
|
337 |
#else
|
338 |
/* limited 16 bit addressing test */
|
339 |
TEST_LEA16("0x4000");
|
340 |
TEST_LEA16("(%%bx)");
|
341 |
TEST_LEA16("(%%si)");
|
342 |
TEST_LEA16("(%%di)");
|
343 |
TEST_LEA16("0x40(%%bx)");
|
344 |
TEST_LEA16("0x40(%%si)");
|
345 |
TEST_LEA16("0x40(%%di)");
|
346 |
TEST_LEA16("0x4000(%%bx)");
|
347 |
TEST_LEA16("0x4000(%%si)");
|
348 |
TEST_LEA16("(%%bx,%%si)");
|
349 |
TEST_LEA16("(%%bx,%%di)");
|
350 |
TEST_LEA16("0x40(%%bx,%%si)");
|
351 |
TEST_LEA16("0x40(%%bx,%%di)");
|
352 |
TEST_LEA16("0x4000(%%bx,%%si)");
|
353 |
TEST_LEA16("0x4000(%%bx,%%di)");
|
354 |
#endif
|
355 |
} |
356 |
|
357 |
#define TEST_JCC(JCC, v1, v2)\
|
358 |
{\ |
359 |
int res;\
|
360 |
asm("movl $1, %0\n\t"\ |
361 |
"cmpl %2, %1\n\t"\
|
362 |
"j" JCC " 1f\n\t"\ |
363 |
"movl $0, %0\n\t"\
|
364 |
"1:\n\t"\
|
365 |
: "=r" (res)\
|
366 |
: "r" (v1), "r" (v2));\ |
367 |
printf("%-10s %d\n", "j" JCC, res);\ |
368 |
\ |
369 |
asm("movl $0, %0\n\t"\ |
370 |
"cmpl %2, %1\n\t"\
|
371 |
"set" JCC " %b0\n\t"\ |
372 |
: "=r" (res)\
|
373 |
: "r" (v1), "r" (v2));\ |
374 |
printf("%-10s %d\n", "set" JCC, res);\ |
375 |
if (TEST_CMOV) {\
|
376 |
long val = i2l(1);\ |
377 |
long res = i2l(0x12345678);\ |
378 |
X86_64_ONLY(\ |
379 |
asm("cmpl %2, %1\n\t"\ |
380 |
"cmov" JCC "q %3, %0\n\t"\ |
381 |
: "=r" (res)\
|
382 |
: "r" (v1), "r" (v2), "m" (val), "0" (res));\ |
383 |
printf("%-10s R=" FMTLX "\n", "cmov" JCC "q", res);)\ |
384 |
asm("cmpl %2, %1\n\t"\ |
385 |
"cmov" JCC "l %k3, %k0\n\t"\ |
386 |
: "=r" (res)\
|
387 |
: "r" (v1), "r" (v2), "m" (val), "0" (res));\ |
388 |
printf("%-10s R=" FMTLX "\n", "cmov" JCC "l", res);\ |
389 |
asm("cmpl %2, %1\n\t"\ |
390 |
"cmov" JCC "w %w3, %w0\n\t"\ |
391 |
: "=r" (res)\
|
392 |
: "r" (v1), "r" (v2), "r" (1), "0" (res));\ |
393 |
printf("%-10s R=" FMTLX "\n", "cmov" JCC "w", res);\ |
394 |
} \ |
395 |
} |
396 |
|
397 |
/* various jump tests */
|
398 |
void test_jcc(void) |
399 |
{ |
400 |
TEST_JCC("ne", 1, 1); |
401 |
TEST_JCC("ne", 1, 0); |
402 |
|
403 |
TEST_JCC("e", 1, 1); |
404 |
TEST_JCC("e", 1, 0); |
405 |
|
406 |
TEST_JCC("l", 1, 1); |
407 |
TEST_JCC("l", 1, 0); |
408 |
TEST_JCC("l", 1, -1); |
409 |
|
410 |
TEST_JCC("le", 1, 1); |
411 |
TEST_JCC("le", 1, 0); |
412 |
TEST_JCC("le", 1, -1); |
413 |
|
414 |
TEST_JCC("ge", 1, 1); |
415 |
TEST_JCC("ge", 1, 0); |
416 |
TEST_JCC("ge", -1, 1); |
417 |
|
418 |
TEST_JCC("g", 1, 1); |
419 |
TEST_JCC("g", 1, 0); |
420 |
TEST_JCC("g", 1, -1); |
421 |
|
422 |
TEST_JCC("b", 1, 1); |
423 |
TEST_JCC("b", 1, 0); |
424 |
TEST_JCC("b", 1, -1); |
425 |
|
426 |
TEST_JCC("be", 1, 1); |
427 |
TEST_JCC("be", 1, 0); |
428 |
TEST_JCC("be", 1, -1); |
429 |
|
430 |
TEST_JCC("ae", 1, 1); |
431 |
TEST_JCC("ae", 1, 0); |
432 |
TEST_JCC("ae", 1, -1); |
433 |
|
434 |
TEST_JCC("a", 1, 1); |
435 |
TEST_JCC("a", 1, 0); |
436 |
TEST_JCC("a", 1, -1); |
437 |
|
438 |
|
439 |
TEST_JCC("p", 1, 1); |
440 |
TEST_JCC("p", 1, 0); |
441 |
|
442 |
TEST_JCC("np", 1, 1); |
443 |
TEST_JCC("np", 1, 0); |
444 |
|
445 |
TEST_JCC("o", 0x7fffffff, 0); |
446 |
TEST_JCC("o", 0x7fffffff, -1); |
447 |
|
448 |
TEST_JCC("no", 0x7fffffff, 0); |
449 |
TEST_JCC("no", 0x7fffffff, -1); |
450 |
|
451 |
TEST_JCC("s", 0, 1); |
452 |
TEST_JCC("s", 0, -1); |
453 |
TEST_JCC("s", 0, 0); |
454 |
|
455 |
TEST_JCC("ns", 0, 1); |
456 |
TEST_JCC("ns", 0, -1); |
457 |
TEST_JCC("ns", 0, 0); |
458 |
} |
459 |
|
460 |
#undef CC_MASK
|
461 |
#ifdef TEST_P4_FLAGS
|
462 |
#define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
|
463 |
#else
|
464 |
#define CC_MASK (CC_O | CC_C)
|
465 |
#endif
|
466 |
|
467 |
#define OP mul
|
468 |
#include "test-i386-muldiv.h" |
469 |
|
470 |
#define OP imul
|
471 |
#include "test-i386-muldiv.h" |
472 |
|
473 |
void test_imulw2(long op0, long op1) |
474 |
{ |
475 |
long res, s1, s0, flags;
|
476 |
s0 = op0; |
477 |
s1 = op1; |
478 |
res = s0; |
479 |
flags = 0;
|
480 |
asm volatile ("push %4\n\t" |
481 |
"popf\n\t"
|
482 |
"imulw %w2, %w0\n\t"
|
483 |
"pushf\n\t"
|
484 |
"pop %1\n\t"
|
485 |
: "=q" (res), "=g" (flags) |
486 |
: "q" (s1), "0" (res), "1" (flags)); |
487 |
printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n", |
488 |
"imulw", s0, s1, res, flags & CC_MASK);
|
489 |
} |
490 |
|
491 |
void test_imull2(long op0, long op1) |
492 |
{ |
493 |
long res, s1, s0, flags;
|
494 |
s0 = op0; |
495 |
s1 = op1; |
496 |
res = s0; |
497 |
flags = 0;
|
498 |
asm volatile ("push %4\n\t" |
499 |
"popf\n\t"
|
500 |
"imull %k2, %k0\n\t"
|
501 |
"pushf\n\t"
|
502 |
"pop %1\n\t"
|
503 |
: "=q" (res), "=g" (flags) |
504 |
: "q" (s1), "0" (res), "1" (flags)); |
505 |
printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n", |
506 |
"imull", s0, s1, res, flags & CC_MASK);
|
507 |
} |
508 |
|
509 |
#if defined(__x86_64__)
|
510 |
void test_imulq2(long op0, long op1) |
511 |
{ |
512 |
long res, s1, s0, flags;
|
513 |
s0 = op0; |
514 |
s1 = op1; |
515 |
res = s0; |
516 |
flags = 0;
|
517 |
asm volatile ("push %4\n\t" |
518 |
"popf\n\t"
|
519 |
"imulq %2, %0\n\t"
|
520 |
"pushf\n\t"
|
521 |
"pop %1\n\t"
|
522 |
: "=q" (res), "=g" (flags) |
523 |
: "q" (s1), "0" (res), "1" (flags)); |
524 |
printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n", |
525 |
"imulq", s0, s1, res, flags & CC_MASK);
|
526 |
} |
527 |
#endif
|
528 |
|
529 |
#define TEST_IMUL_IM(size, rsize, op0, op1)\
|
530 |
{\ |
531 |
long res, flags, s1;\
|
532 |
flags = 0;\
|
533 |
res = 0;\
|
534 |
s1 = op1;\ |
535 |
asm volatile ("push %3\n\t"\ |
536 |
"popf\n\t"\
|
537 |
"imul" size " $" #op0 ", %" rsize "2, %" rsize "0\n\t" \ |
538 |
"pushf\n\t"\
|
539 |
"pop %1\n\t"\
|
540 |
: "=r" (res), "=g" (flags)\ |
541 |
: "r" (s1), "1" (flags), "0" (res));\ |
542 |
printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n",\ |
543 |
"imul" size " im", (long)op0, (long)op1, res, flags & CC_MASK);\ |
544 |
} |
545 |
|
546 |
|
547 |
#undef CC_MASK
|
548 |
#define CC_MASK (0) |
549 |
|
550 |
#define OP div
|
551 |
#include "test-i386-muldiv.h" |
552 |
|
553 |
#define OP idiv
|
554 |
#include "test-i386-muldiv.h" |
555 |
|
556 |
void test_mul(void) |
557 |
{ |
558 |
test_imulb(0x1234561d, 4); |
559 |
test_imulb(3, -4); |
560 |
test_imulb(0x80, 0x80); |
561 |
test_imulb(0x10, 0x10); |
562 |
|
563 |
test_imulw(0, 0x1234001d, 45); |
564 |
test_imulw(0, 23, -45); |
565 |
test_imulw(0, 0x8000, 0x8000); |
566 |
test_imulw(0, 0x100, 0x100); |
567 |
|
568 |
test_imull(0, 0x1234001d, 45); |
569 |
test_imull(0, 23, -45); |
570 |
test_imull(0, 0x80000000, 0x80000000); |
571 |
test_imull(0, 0x10000, 0x10000); |
572 |
|
573 |
test_mulb(0x1234561d, 4); |
574 |
test_mulb(3, -4); |
575 |
test_mulb(0x80, 0x80); |
576 |
test_mulb(0x10, 0x10); |
577 |
|
578 |
test_mulw(0, 0x1234001d, 45); |
579 |
test_mulw(0, 23, -45); |
580 |
test_mulw(0, 0x8000, 0x8000); |
581 |
test_mulw(0, 0x100, 0x100); |
582 |
|
583 |
test_mull(0, 0x1234001d, 45); |
584 |
test_mull(0, 23, -45); |
585 |
test_mull(0, 0x80000000, 0x80000000); |
586 |
test_mull(0, 0x10000, 0x10000); |
587 |
|
588 |
test_imulw2(0x1234001d, 45); |
589 |
test_imulw2(23, -45); |
590 |
test_imulw2(0x8000, 0x8000); |
591 |
test_imulw2(0x100, 0x100); |
592 |
|
593 |
test_imull2(0x1234001d, 45); |
594 |
test_imull2(23, -45); |
595 |
test_imull2(0x80000000, 0x80000000); |
596 |
test_imull2(0x10000, 0x10000); |
597 |
|
598 |
TEST_IMUL_IM("w", "w", 45, 0x1234); |
599 |
TEST_IMUL_IM("w", "w", -45, 23); |
600 |
TEST_IMUL_IM("w", "w", 0x8000, 0x80000000); |
601 |
TEST_IMUL_IM("w", "w", 0x7fff, 0x1000); |
602 |
|
603 |
TEST_IMUL_IM("l", "k", 45, 0x1234); |
604 |
TEST_IMUL_IM("l", "k", -45, 23); |
605 |
TEST_IMUL_IM("l", "k", 0x8000, 0x80000000); |
606 |
TEST_IMUL_IM("l", "k", 0x7fff, 0x1000); |
607 |
|
608 |
test_idivb(0x12341678, 0x127e); |
609 |
test_idivb(0x43210123, -5); |
610 |
test_idivb(0x12340004, -1); |
611 |
|
612 |
test_idivw(0, 0x12345678, 12347); |
613 |
test_idivw(0, -23223, -45); |
614 |
test_idivw(0, 0x12348000, -1); |
615 |
test_idivw(0x12343, 0x12345678, 0x81238567); |
616 |
|
617 |
test_idivl(0, 0x12345678, 12347); |
618 |
test_idivl(0, -233223, -45); |
619 |
test_idivl(0, 0x80000000, -1); |
620 |
test_idivl(0x12343, 0x12345678, 0x81234567); |
621 |
|
622 |
test_divb(0x12341678, 0x127e); |
623 |
test_divb(0x43210123, -5); |
624 |
test_divb(0x12340004, -1); |
625 |
|
626 |
test_divw(0, 0x12345678, 12347); |
627 |
test_divw(0, -23223, -45); |
628 |
test_divw(0, 0x12348000, -1); |
629 |
test_divw(0x12343, 0x12345678, 0x81238567); |
630 |
|
631 |
test_divl(0, 0x12345678, 12347); |
632 |
test_divl(0, -233223, -45); |
633 |
test_divl(0, 0x80000000, -1); |
634 |
test_divl(0x12343, 0x12345678, 0x81234567); |
635 |
|
636 |
#if defined(__x86_64__)
|
637 |
test_imulq(0, 0x1234001d1234001d, 45); |
638 |
test_imulq(0, 23, -45); |
639 |
test_imulq(0, 0x8000000000000000, 0x8000000000000000); |
640 |
test_imulq(0, 0x100000000, 0x100000000); |
641 |
|
642 |
test_mulq(0, 0x1234001d1234001d, 45); |
643 |
test_mulq(0, 23, -45); |
644 |
test_mulq(0, 0x8000000000000000, 0x8000000000000000); |
645 |
test_mulq(0, 0x100000000, 0x100000000); |
646 |
|
647 |
test_imulq2(0x1234001d1234001d, 45); |
648 |
test_imulq2(23, -45); |
649 |
test_imulq2(0x8000000000000000, 0x8000000000000000); |
650 |
test_imulq2(0x100000000, 0x100000000); |
651 |
|
652 |
TEST_IMUL_IM("q", "", 45, 0x12341234); |
653 |
TEST_IMUL_IM("q", "", -45, 23); |
654 |
TEST_IMUL_IM("q", "", 0x8000, 0x8000000000000000); |
655 |
TEST_IMUL_IM("q", "", 0x7fff, 0x10000000); |
656 |
|
657 |
test_idivq(0, 0x12345678abcdef, 12347); |
658 |
test_idivq(0, -233223, -45); |
659 |
test_idivq(0, 0x8000000000000000, -1); |
660 |
test_idivq(0x12343, 0x12345678, 0x81234567); |
661 |
|
662 |
test_divq(0, 0x12345678abcdef, 12347); |
663 |
test_divq(0, -233223, -45); |
664 |
test_divq(0, 0x8000000000000000, -1); |
665 |
test_divq(0x12343, 0x12345678, 0x81234567); |
666 |
#endif
|
667 |
} |
668 |
|
669 |
#define TEST_BSX(op, size, op0)\
|
670 |
{\ |
671 |
long res, val, resz;\
|
672 |
val = op0;\ |
673 |
asm("xor %1, %1\n"\ |
674 |
"mov $0x12345678, %0\n"\
|
675 |
#op " %" size "2, %" size "0 ; setz %b1" \ |
676 |
: "=r" (res), "=q" (resz)\ |
677 |
: "g" (val));\
|
678 |
printf("%-10s A=" FMTLX " R=" FMTLX " %ld\n", #op, val, res, resz);\ |
679 |
} |
680 |
|
681 |
void test_bsx(void) |
682 |
{ |
683 |
TEST_BSX(bsrw, "w", 0); |
684 |
TEST_BSX(bsrw, "w", 0x12340128); |
685 |
TEST_BSX(bsfw, "w", 0); |
686 |
TEST_BSX(bsfw, "w", 0x12340128); |
687 |
TEST_BSX(bsrl, "k", 0); |
688 |
TEST_BSX(bsrl, "k", 0x00340128); |
689 |
TEST_BSX(bsfl, "k", 0); |
690 |
TEST_BSX(bsfl, "k", 0x00340128); |
691 |
#if defined(__x86_64__)
|
692 |
TEST_BSX(bsrq, "", 0); |
693 |
TEST_BSX(bsrq, "", 0x003401281234); |
694 |
TEST_BSX(bsfq, "", 0); |
695 |
TEST_BSX(bsfq, "", 0x003401281234); |
696 |
#endif
|
697 |
} |
698 |
|
699 |
/**********************************************/
|
700 |
|
701 |
union float64u {
|
702 |
double d;
|
703 |
uint64_t l; |
704 |
}; |
705 |
|
706 |
union float64u q_nan = { .l = 0xFFF8000000000000LL }; |
707 |
union float64u s_nan = { .l = 0xFFF0000000000000LL }; |
708 |
|
709 |
void test_fops(double a, double b) |
710 |
{ |
711 |
printf("a=%f b=%f a+b=%f\n", a, b, a + b);
|
712 |
printf("a=%f b=%f a-b=%f\n", a, b, a - b);
|
713 |
printf("a=%f b=%f a*b=%f\n", a, b, a * b);
|
714 |
printf("a=%f b=%f a/b=%f\n", a, b, a / b);
|
715 |
printf("a=%f b=%f fmod(a, b)=%f\n", a, b, fmod(a, b));
|
716 |
printf("a=%f sqrt(a)=%f\n", a, sqrt(a));
|
717 |
printf("a=%f sin(a)=%f\n", a, sin(a));
|
718 |
printf("a=%f cos(a)=%f\n", a, cos(a));
|
719 |
printf("a=%f tan(a)=%f\n", a, tan(a));
|
720 |
printf("a=%f log(a)=%f\n", a, log(a));
|
721 |
printf("a=%f exp(a)=%f\n", a, exp(a));
|
722 |
printf("a=%f b=%f atan2(a, b)=%f\n", a, b, atan2(a, b));
|
723 |
/* just to test some op combining */
|
724 |
printf("a=%f asin(sin(a))=%f\n", a, asin(sin(a)));
|
725 |
printf("a=%f acos(cos(a))=%f\n", a, acos(cos(a)));
|
726 |
printf("a=%f atan(tan(a))=%f\n", a, atan(tan(a)));
|
727 |
|
728 |
} |
729 |
|
730 |
void fpu_clear_exceptions(void) |
731 |
{ |
732 |
struct __attribute__((packed)) {
|
733 |
uint16_t fpuc; |
734 |
uint16_t dummy1; |
735 |
uint16_t fpus; |
736 |
uint16_t dummy2; |
737 |
uint16_t fptag; |
738 |
uint16_t dummy3; |
739 |
uint32_t ignored[4];
|
740 |
long double fpregs[8]; |
741 |
} float_env32; |
742 |
|
743 |
asm volatile ("fnstenv %0\n" : : "m" (float_env32)); |
744 |
float_env32.fpus &= ~0x7f;
|
745 |
asm volatile ("fldenv %0\n" : : "m" (float_env32)); |
746 |
} |
747 |
|
748 |
/* XXX: display exception bits when supported */
|
749 |
#define FPUS_EMASK 0x0000 |
750 |
//#define FPUS_EMASK 0x007f
|
751 |
|
752 |
void test_fcmp(double a, double b) |
753 |
{ |
754 |
long eflags, fpus;
|
755 |
|
756 |
fpu_clear_exceptions(); |
757 |
asm("fcom %2\n" |
758 |
"fstsw %%ax\n"
|
759 |
: "=a" (fpus)
|
760 |
: "t" (a), "u" (b)); |
761 |
printf("fcom(%f %f)=%04lx \n",
|
762 |
a, b, fpus & (0x4500 | FPUS_EMASK));
|
763 |
fpu_clear_exceptions(); |
764 |
asm("fucom %2\n" |
765 |
"fstsw %%ax\n"
|
766 |
: "=a" (fpus)
|
767 |
: "t" (a), "u" (b)); |
768 |
printf("fucom(%f %f)=%04lx\n",
|
769 |
a, b, fpus & (0x4500 | FPUS_EMASK));
|
770 |
if (TEST_FCOMI) {
|
771 |
/* test f(u)comi instruction */
|
772 |
fpu_clear_exceptions(); |
773 |
asm("fcomi %3, %2\n" |
774 |
"fstsw %%ax\n"
|
775 |
"pushf\n"
|
776 |
"pop %0\n"
|
777 |
: "=r" (eflags), "=a" (fpus) |
778 |
: "t" (a), "u" (b)); |
779 |
printf("fcomi(%f %f)=%04lx %02lx\n",
|
780 |
a, b, fpus & FPUS_EMASK, eflags & (CC_Z | CC_P | CC_C)); |
781 |
fpu_clear_exceptions(); |
782 |
asm("fucomi %3, %2\n" |
783 |
"fstsw %%ax\n"
|
784 |
"pushf\n"
|
785 |
"pop %0\n"
|
786 |
: "=r" (eflags), "=a" (fpus) |
787 |
: "t" (a), "u" (b)); |
788 |
printf("fucomi(%f %f)=%04lx %02lx\n",
|
789 |
a, b, fpus & FPUS_EMASK, eflags & (CC_Z | CC_P | CC_C)); |
790 |
} |
791 |
fpu_clear_exceptions(); |
792 |
asm volatile("fxam\n" |
793 |
"fstsw %%ax\n"
|
794 |
: "=a" (fpus)
|
795 |
: "t" (a));
|
796 |
printf("fxam(%f)=%04lx\n", a, fpus & 0x4700); |
797 |
fpu_clear_exceptions(); |
798 |
} |
799 |
|
800 |
void test_fcvt(double a) |
801 |
{ |
802 |
float fa;
|
803 |
long double la; |
804 |
int16_t fpuc; |
805 |
int i;
|
806 |
int64_t lla; |
807 |
int ia;
|
808 |
int16_t wa; |
809 |
double ra;
|
810 |
|
811 |
fa = a; |
812 |
la = a; |
813 |
printf("(float)%f = %f\n", a, fa);
|
814 |
printf("(long double)%f = %Lf\n", a, la);
|
815 |
printf("a=" FMT64X "\n", *(uint64_t *)&a); |
816 |
printf("la=" FMT64X " %04x\n", *(uint64_t *)&la, |
817 |
*(unsigned short *)((char *)(&la) + 8)); |
818 |
|
819 |
/* test all roundings */
|
820 |
asm volatile ("fstcw %0" : "=m" (fpuc)); |
821 |
for(i=0;i<4;i++) { |
822 |
uint16_t val16; |
823 |
val16 = (fpuc & ~0x0c00) | (i << 10); |
824 |
asm volatile ("fldcw %0" : : "m" (val16)); |
825 |
asm volatile ("fist %0" : "=m" (wa) : "t" (a)); |
826 |
asm volatile ("fistl %0" : "=m" (ia) : "t" (a)); |
827 |
asm volatile ("fistpll %0" : "=m" (lla) : "t" (a) : "st"); |
828 |
asm volatile ("frndint ; fstl %0" : "=m" (ra) : "t" (a)); |
829 |
asm volatile ("fldcw %0" : : "m" (fpuc)); |
830 |
printf("(short)a = %d\n", wa);
|
831 |
printf("(int)a = %d\n", ia);
|
832 |
printf("(int64_t)a = " FMT64X "\n", lla); |
833 |
printf("rint(a) = %f\n", ra);
|
834 |
} |
835 |
} |
836 |
|
837 |
#define TEST(N) \
|
838 |
asm("fld" #N : "=t" (a)); \ |
839 |
printf("fld" #N "= %f\n", a); |
840 |
|
841 |
void test_fconst(void) |
842 |
{ |
843 |
double a;
|
844 |
TEST(1);
|
845 |
TEST(l2t); |
846 |
TEST(l2e); |
847 |
TEST(pi); |
848 |
TEST(lg2); |
849 |
TEST(ln2); |
850 |
TEST(z); |
851 |
} |
852 |
|
853 |
void test_fbcd(double a) |
854 |
{ |
855 |
unsigned short bcd[5]; |
856 |
double b;
|
857 |
|
858 |
asm("fbstp %0" : "=m" (bcd[0]) : "t" (a) : "st"); |
859 |
asm("fbld %1" : "=t" (b) : "m" (bcd[0])); |
860 |
printf("a=%f bcd=%04x%04x%04x%04x%04x b=%f\n",
|
861 |
a, bcd[4], bcd[3], bcd[2], bcd[1], bcd[0], b); |
862 |
} |
863 |
|
864 |
#define TEST_ENV(env, save, restore)\
|
865 |
{\ |
866 |
memset((env), 0xaa, sizeof(*(env)));\ |
867 |
for(i=0;i<5;i++)\ |
868 |
asm volatile ("fldl %0" : : "m" (dtab[i]));\ |
869 |
asm volatile (save " %0\n" : : "m" (*(env)));\ |
870 |
asm volatile (restore " %0\n": : "m" (*(env)));\ |
871 |
for(i=0;i<5;i++)\ |
872 |
asm volatile ("fstpl %0" : "=m" (rtab[i]));\ |
873 |
for(i=0;i<5;i++)\ |
874 |
printf("res[%d]=%f\n", i, rtab[i]);\
|
875 |
printf("fpuc=%04x fpus=%04x fptag=%04x\n",\
|
876 |
(env)->fpuc,\ |
877 |
(env)->fpus & 0xff00,\
|
878 |
(env)->fptag);\ |
879 |
} |
880 |
|
881 |
void test_fenv(void) |
882 |
{ |
883 |
struct __attribute__((packed)) {
|
884 |
uint16_t fpuc; |
885 |
uint16_t dummy1; |
886 |
uint16_t fpus; |
887 |
uint16_t dummy2; |
888 |
uint16_t fptag; |
889 |
uint16_t dummy3; |
890 |
uint32_t ignored[4];
|
891 |
long double fpregs[8]; |
892 |
} float_env32; |
893 |
struct __attribute__((packed)) {
|
894 |
uint16_t fpuc; |
895 |
uint16_t fpus; |
896 |
uint16_t fptag; |
897 |
uint16_t ignored[4];
|
898 |
long double fpregs[8]; |
899 |
} float_env16; |
900 |
double dtab[8]; |
901 |
double rtab[8]; |
902 |
int i;
|
903 |
|
904 |
for(i=0;i<8;i++) |
905 |
dtab[i] = i + 1;
|
906 |
|
907 |
TEST_ENV(&float_env16, "data16 fnstenv", "data16 fldenv"); |
908 |
TEST_ENV(&float_env16, "data16 fnsave", "data16 frstor"); |
909 |
TEST_ENV(&float_env32, "fnstenv", "fldenv"); |
910 |
TEST_ENV(&float_env32, "fnsave", "frstor"); |
911 |
|
912 |
/* test for ffree */
|
913 |
for(i=0;i<5;i++) |
914 |
asm volatile ("fldl %0" : : "m" (dtab[i])); |
915 |
asm volatile("ffree %st(2)"); |
916 |
asm volatile ("fnstenv %0\n" : : "m" (float_env32)); |
917 |
asm volatile ("fninit"); |
918 |
printf("fptag=%04x\n", float_env32.fptag);
|
919 |
} |
920 |
|
921 |
|
922 |
#define TEST_FCMOV(a, b, eflags, CC)\
|
923 |
{\ |
924 |
double res;\
|
925 |
asm("push %3\n"\ |
926 |
"popf\n"\
|
927 |
"fcmov" CC " %2, %0\n"\ |
928 |
: "=t" (res)\
|
929 |
: "0" (a), "u" (b), "g" (eflags));\ |
930 |
printf("fcmov%s eflags=0x%04lx-> %f\n", \
|
931 |
CC, (long)eflags, res);\
|
932 |
} |
933 |
|
934 |
void test_fcmov(void) |
935 |
{ |
936 |
double a, b;
|
937 |
long eflags, i;
|
938 |
|
939 |
a = 1.0; |
940 |
b = 2.0; |
941 |
for(i = 0; i < 4; i++) { |
942 |
eflags = 0;
|
943 |
if (i & 1) |
944 |
eflags |= CC_C; |
945 |
if (i & 2) |
946 |
eflags |= CC_Z; |
947 |
TEST_FCMOV(a, b, eflags, "b");
|
948 |
TEST_FCMOV(a, b, eflags, "e");
|
949 |
TEST_FCMOV(a, b, eflags, "be");
|
950 |
TEST_FCMOV(a, b, eflags, "nb");
|
951 |
TEST_FCMOV(a, b, eflags, "ne");
|
952 |
TEST_FCMOV(a, b, eflags, "nbe");
|
953 |
} |
954 |
TEST_FCMOV(a, b, 0, "u"); |
955 |
TEST_FCMOV(a, b, CC_P, "u");
|
956 |
TEST_FCMOV(a, b, 0, "nu"); |
957 |
TEST_FCMOV(a, b, CC_P, "nu");
|
958 |
} |
959 |
|
960 |
void test_floats(void) |
961 |
{ |
962 |
test_fops(2, 3); |
963 |
test_fops(1.4, -5); |
964 |
test_fcmp(2, -1); |
965 |
test_fcmp(2, 2); |
966 |
test_fcmp(2, 3); |
967 |
test_fcmp(2, q_nan.d);
|
968 |
test_fcmp(q_nan.d, -1);
|
969 |
test_fcmp(-1.0/0.0, -1); |
970 |
test_fcmp(1.0/0.0, -1); |
971 |
test_fcvt(0.5); |
972 |
test_fcvt(-0.5); |
973 |
test_fcvt(1.0/7.0); |
974 |
test_fcvt(-1.0/9.0); |
975 |
test_fcvt(32768);
|
976 |
test_fcvt(-1e20);
|
977 |
test_fcvt(-1.0/0.0); |
978 |
test_fcvt(1.0/0.0); |
979 |
test_fcvt(q_nan.d); |
980 |
test_fconst(); |
981 |
test_fbcd(1234567890123456.0); |
982 |
test_fbcd(-123451234567890.0); |
983 |
test_fenv(); |
984 |
if (TEST_CMOV) {
|
985 |
test_fcmov(); |
986 |
} |
987 |
} |
988 |
|
989 |
/**********************************************/
|
990 |
#if !defined(__x86_64__)
|
991 |
|
992 |
#define TEST_BCD(op, op0, cc_in, cc_mask)\
|
993 |
{\ |
994 |
int res, flags;\
|
995 |
res = op0;\ |
996 |
flags = cc_in;\ |
997 |
asm ("push %3\n\t"\ |
998 |
"popf\n\t"\
|
999 |
#op "\n\t"\ |
1000 |
"pushf\n\t"\
|
1001 |
"pop %1\n\t"\
|
1002 |
: "=a" (res), "=g" (flags)\ |
1003 |
: "0" (res), "1" (flags));\ |
1004 |
printf("%-10s A=%08x R=%08x CCIN=%04x CC=%04x\n",\
|
1005 |
#op, op0, res, cc_in, flags & cc_mask);\
|
1006 |
} |
1007 |
|
1008 |
void test_bcd(void) |
1009 |
{ |
1010 |
TEST_BCD(daa, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1011 |
TEST_BCD(daa, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1012 |
TEST_BCD(daa, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1013 |
TEST_BCD(daa, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1014 |
TEST_BCD(daa, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1015 |
TEST_BCD(daa, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1016 |
TEST_BCD(daa, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1017 |
TEST_BCD(daa, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A)); |
1018 |
TEST_BCD(daa, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A)); |
1019 |
TEST_BCD(daa, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1020 |
TEST_BCD(daa, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1021 |
TEST_BCD(daa, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1022 |
TEST_BCD(daa, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1023 |
|
1024 |
TEST_BCD(das, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1025 |
TEST_BCD(das, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1026 |
TEST_BCD(das, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1027 |
TEST_BCD(das, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1028 |
TEST_BCD(das, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1029 |
TEST_BCD(das, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1030 |
TEST_BCD(das, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1031 |
TEST_BCD(das, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A)); |
1032 |
TEST_BCD(das, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A)); |
1033 |
TEST_BCD(das, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1034 |
TEST_BCD(das, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1035 |
TEST_BCD(das, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1036 |
TEST_BCD(das, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
|
1037 |
|
1038 |
TEST_BCD(aaa, 0x12340205, CC_A, (CC_C | CC_A));
|
1039 |
TEST_BCD(aaa, 0x12340306, CC_A, (CC_C | CC_A));
|
1040 |
TEST_BCD(aaa, 0x1234040a, CC_A, (CC_C | CC_A));
|
1041 |
TEST_BCD(aaa, 0x123405fa, CC_A, (CC_C | CC_A));
|
1042 |
TEST_BCD(aaa, 0x12340205, 0, (CC_C | CC_A)); |
1043 |
TEST_BCD(aaa, 0x12340306, 0, (CC_C | CC_A)); |
1044 |
TEST_BCD(aaa, 0x1234040a, 0, (CC_C | CC_A)); |
1045 |
TEST_BCD(aaa, 0x123405fa, 0, (CC_C | CC_A)); |
1046 |
|
1047 |
TEST_BCD(aas, 0x12340205, CC_A, (CC_C | CC_A));
|
1048 |
TEST_BCD(aas, 0x12340306, CC_A, (CC_C | CC_A));
|
1049 |
TEST_BCD(aas, 0x1234040a, CC_A, (CC_C | CC_A));
|
1050 |
TEST_BCD(aas, 0x123405fa, CC_A, (CC_C | CC_A));
|
1051 |
TEST_BCD(aas, 0x12340205, 0, (CC_C | CC_A)); |
1052 |
TEST_BCD(aas, 0x12340306, 0, (CC_C | CC_A)); |
1053 |
TEST_BCD(aas, 0x1234040a, 0, (CC_C | CC_A)); |
1054 |
TEST_BCD(aas, 0x123405fa, 0, (CC_C | CC_A)); |
1055 |
|
1056 |
TEST_BCD(aam, 0x12340547, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
|
1057 |
TEST_BCD(aad, 0x12340407, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
|
1058 |
} |
1059 |
#endif
|
1060 |
|
1061 |
#define TEST_XCHG(op, size, opconst)\
|
1062 |
{\ |
1063 |
long op0, op1;\
|
1064 |
op0 = i2l(0x12345678);\
|
1065 |
op1 = i2l(0xfbca7654);\
|
1066 |
asm(#op " %" size "0, %" size "1" \ |
1067 |
: "=q" (op0), opconst (op1) \
|
1068 |
: "0" (op0));\
|
1069 |
printf("%-10s A=" FMTLX " B=" FMTLX "\n",\ |
1070 |
#op, op0, op1);\
|
1071 |
} |
1072 |
|
1073 |
#define TEST_CMPXCHG(op, size, opconst, eax)\
|
1074 |
{\ |
1075 |
long op0, op1, op2;\
|
1076 |
op0 = i2l(0x12345678);\
|
1077 |
op1 = i2l(0xfbca7654);\
|
1078 |
op2 = i2l(eax);\ |
1079 |
asm(#op " %" size "0, %" size "1" \ |
1080 |
: "=q" (op0), opconst (op1) \
|
1081 |
: "0" (op0), "a" (op2));\ |
1082 |
printf("%-10s EAX=" FMTLX " A=" FMTLX " C=" FMTLX "\n",\ |
1083 |
#op, op2, op0, op1);\
|
1084 |
} |
1085 |
|
1086 |
void test_xchg(void) |
1087 |
{ |
1088 |
#if defined(__x86_64__)
|
1089 |
TEST_XCHG(xchgq, "", "+q"); |
1090 |
#endif
|
1091 |
TEST_XCHG(xchgl, "k", "+q"); |
1092 |
TEST_XCHG(xchgw, "w", "+q"); |
1093 |
TEST_XCHG(xchgb, "b", "+q"); |
1094 |
|
1095 |
#if defined(__x86_64__)
|
1096 |
TEST_XCHG(xchgq, "", "=m"); |
1097 |
#endif
|
1098 |
TEST_XCHG(xchgl, "k", "+m"); |
1099 |
TEST_XCHG(xchgw, "w", "+m"); |
1100 |
TEST_XCHG(xchgb, "b", "+m"); |
1101 |
|
1102 |
#if defined(__x86_64__)
|
1103 |
TEST_XCHG(xaddq, "", "+q"); |
1104 |
#endif
|
1105 |
TEST_XCHG(xaddl, "k", "+q"); |
1106 |
TEST_XCHG(xaddw, "w", "+q"); |
1107 |
TEST_XCHG(xaddb, "b", "+q"); |
1108 |
|
1109 |
{ |
1110 |
int res;
|
1111 |
res = 0x12345678;
|
1112 |
asm("xaddl %1, %0" : "=r" (res) : "0" (res)); |
1113 |
printf("xaddl same res=%08x\n", res);
|
1114 |
} |
1115 |
|
1116 |
#if defined(__x86_64__)
|
1117 |
TEST_XCHG(xaddq, "", "+m"); |
1118 |
#endif
|
1119 |
TEST_XCHG(xaddl, "k", "+m"); |
1120 |
TEST_XCHG(xaddw, "w", "+m"); |
1121 |
TEST_XCHG(xaddb, "b", "+m"); |
1122 |
|
1123 |
#if defined(__x86_64__)
|
1124 |
TEST_CMPXCHG(cmpxchgq, "", "+q", 0xfbca7654); |
1125 |
#endif
|
1126 |
TEST_CMPXCHG(cmpxchgl, "k", "+q", 0xfbca7654); |
1127 |
TEST_CMPXCHG(cmpxchgw, "w", "+q", 0xfbca7654); |
1128 |
TEST_CMPXCHG(cmpxchgb, "b", "+q", 0xfbca7654); |
1129 |
|
1130 |
#if defined(__x86_64__)
|
1131 |
TEST_CMPXCHG(cmpxchgq, "", "+q", 0xfffefdfc); |
1132 |
#endif
|
1133 |
TEST_CMPXCHG(cmpxchgl, "k", "+q", 0xfffefdfc); |
1134 |
TEST_CMPXCHG(cmpxchgw, "w", "+q", 0xfffefdfc); |
1135 |
TEST_CMPXCHG(cmpxchgb, "b", "+q", 0xfffefdfc); |
1136 |
|
1137 |
#if defined(__x86_64__)
|
1138 |
TEST_CMPXCHG(cmpxchgq, "", "+m", 0xfbca7654); |
1139 |
#endif
|
1140 |
TEST_CMPXCHG(cmpxchgl, "k", "+m", 0xfbca7654); |
1141 |
TEST_CMPXCHG(cmpxchgw, "w", "+m", 0xfbca7654); |
1142 |
TEST_CMPXCHG(cmpxchgb, "b", "+m", 0xfbca7654); |
1143 |
|
1144 |
#if defined(__x86_64__)
|
1145 |
TEST_CMPXCHG(cmpxchgq, "", "+m", 0xfffefdfc); |
1146 |
#endif
|
1147 |
TEST_CMPXCHG(cmpxchgl, "k", "+m", 0xfffefdfc); |
1148 |
TEST_CMPXCHG(cmpxchgw, "w", "+m", 0xfffefdfc); |
1149 |
TEST_CMPXCHG(cmpxchgb, "b", "+m", 0xfffefdfc); |
1150 |
|
1151 |
{ |
1152 |
uint64_t op0, op1, op2; |
1153 |
long i, eflags;
|
1154 |
|
1155 |
for(i = 0; i < 2; i++) { |
1156 |
op0 = 0x123456789abcdLL;
|
1157 |
if (i == 0) |
1158 |
op1 = 0xfbca765423456LL;
|
1159 |
else
|
1160 |
op1 = op0; |
1161 |
op2 = 0x6532432432434LL;
|
1162 |
asm("cmpxchg8b %1\n" |
1163 |
"pushf\n"
|
1164 |
"pop %2\n"
|
1165 |
: "=A" (op0), "=m" (op1), "=g" (eflags) |
1166 |
: "0" (op0), "m" (op1), "b" ((int)op2), "c" ((int)(op2 >> 32))); |
1167 |
printf("cmpxchg8b: op0=" FMT64X " op1=" FMT64X " CC=%02lx\n", |
1168 |
op0, op1, eflags & CC_Z); |
1169 |
} |
1170 |
} |
1171 |
} |
1172 |
|
1173 |
#ifdef TEST_SEGS
|
1174 |
/**********************************************/
|
1175 |
/* segmentation tests */
|
1176 |
|
1177 |
#include <sys/syscall.h> |
1178 |
#include <unistd.h> |
1179 |
#include <asm/ldt.h> |
1180 |
#include <linux/version.h> |
1181 |
|
1182 |
static inline int modify_ldt(int func, void * ptr, unsigned long bytecount) |
1183 |
{ |
1184 |
return syscall(__NR_modify_ldt, func, ptr, bytecount);
|
1185 |
} |
1186 |
|
1187 |
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 5, 66) |
1188 |
#define modify_ldt_ldt_s user_desc
|
1189 |
#endif
|
1190 |
|
1191 |
#define MK_SEL(n) (((n) << 3) | 7) |
1192 |
|
1193 |
uint8_t seg_data1[4096];
|
1194 |
uint8_t seg_data2[4096];
|
1195 |
|
1196 |
#define TEST_LR(op, size, seg, mask)\
|
1197 |
{\ |
1198 |
int res, res2;\
|
1199 |
res = 0x12345678;\
|
1200 |
asm (op " %" size "2, %" size "0\n" \ |
1201 |
"movl $0, %1\n"\
|
1202 |
"jnz 1f\n"\
|
1203 |
"movl $1, %1\n"\
|
1204 |
"1:\n"\
|
1205 |
: "=r" (res), "=r" (res2) : "m" (seg), "0" (res));\ |
1206 |
printf(op ": Z=%d %08x\n", res2, res & ~(mask));\
|
1207 |
} |
1208 |
|
1209 |
/* NOTE: we use Linux modify_ldt syscall */
|
1210 |
void test_segs(void) |
1211 |
{ |
1212 |
struct modify_ldt_ldt_s ldt;
|
1213 |
long long ldt_table[3]; |
1214 |
int res, res2;
|
1215 |
char tmp;
|
1216 |
struct {
|
1217 |
uint32_t offset; |
1218 |
uint16_t seg; |
1219 |
} __attribute__((packed)) segoff; |
1220 |
|
1221 |
ldt.entry_number = 1;
|
1222 |
ldt.base_addr = (unsigned long)&seg_data1; |
1223 |
ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12; |
1224 |
ldt.seg_32bit = 1;
|
1225 |
ldt.contents = MODIFY_LDT_CONTENTS_DATA; |
1226 |
ldt.read_exec_only = 0;
|
1227 |
ldt.limit_in_pages = 1;
|
1228 |
ldt.seg_not_present = 0;
|
1229 |
ldt.useable = 1;
|
1230 |
modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */ |
1231 |
|
1232 |
ldt.entry_number = 2;
|
1233 |
ldt.base_addr = (unsigned long)&seg_data2; |
1234 |
ldt.limit = (sizeof(seg_data2) + 0xfff) >> 12; |
1235 |
ldt.seg_32bit = 1;
|
1236 |
ldt.contents = MODIFY_LDT_CONTENTS_DATA; |
1237 |
ldt.read_exec_only = 0;
|
1238 |
ldt.limit_in_pages = 1;
|
1239 |
ldt.seg_not_present = 0;
|
1240 |
ldt.useable = 1;
|
1241 |
modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */ |
1242 |
|
1243 |
modify_ldt(0, &ldt_table, sizeof(ldt_table)); /* read ldt entries */ |
1244 |
#if 0
|
1245 |
{
|
1246 |
int i;
|
1247 |
for(i=0;i<3;i++)
|
1248 |
printf("%d: %016Lx\n", i, ldt_table[i]);
|
1249 |
}
|
1250 |
#endif
|
1251 |
/* do some tests with fs or gs */
|
1252 |
asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1))); |
1253 |
|
1254 |
seg_data1[1] = 0xaa; |
1255 |
seg_data2[1] = 0x55; |
1256 |
|
1257 |
asm volatile ("fs movzbl 0x1, %0" : "=r" (res)); |
1258 |
printf("FS[1] = %02x\n", res);
|
1259 |
|
1260 |
asm volatile ("pushl %%gs\n" |
1261 |
"movl %1, %%gs\n"
|
1262 |
"gs movzbl 0x1, %0\n"
|
1263 |
"popl %%gs\n"
|
1264 |
: "=r" (res)
|
1265 |
: "r" (MK_SEL(2))); |
1266 |
printf("GS[1] = %02x\n", res);
|
1267 |
|
1268 |
/* tests with ds/ss (implicit segment case) */
|
1269 |
tmp = 0xa5;
|
1270 |
asm volatile ("pushl %%ebp\n\t" |
1271 |
"pushl %%ds\n\t"
|
1272 |
"movl %2, %%ds\n\t"
|
1273 |
"movl %3, %%ebp\n\t"
|
1274 |
"movzbl 0x1, %0\n\t"
|
1275 |
"movzbl (%%ebp), %1\n\t"
|
1276 |
"popl %%ds\n\t"
|
1277 |
"popl %%ebp\n\t"
|
1278 |
: "=r" (res), "=r" (res2) |
1279 |
: "r" (MK_SEL(1)), "r" (&tmp)); |
1280 |
printf("DS[1] = %02x\n", res);
|
1281 |
printf("SS[tmp] = %02x\n", res2);
|
1282 |
|
1283 |
segoff.seg = MK_SEL(2);
|
1284 |
segoff.offset = 0xabcdef12;
|
1285 |
asm volatile("lfs %2, %0\n\t" |
1286 |
"movl %%fs, %1\n\t"
|
1287 |
: "=r" (res), "=g" (res2) |
1288 |
: "m" (segoff));
|
1289 |
printf("FS:reg = %04x:%08x\n", res2, res);
|
1290 |
|
1291 |
TEST_LR("larw", "w", MK_SEL(2), 0x0100); |
1292 |
TEST_LR("larl", "", MK_SEL(2), 0x0100); |
1293 |
TEST_LR("lslw", "w", MK_SEL(2), 0); |
1294 |
TEST_LR("lsll", "", MK_SEL(2), 0); |
1295 |
|
1296 |
TEST_LR("larw", "w", 0xfff8, 0); |
1297 |
TEST_LR("larl", "", 0xfff8, 0); |
1298 |
TEST_LR("lslw", "w", 0xfff8, 0); |
1299 |
TEST_LR("lsll", "", 0xfff8, 0); |
1300 |
} |
1301 |
|
1302 |
/* 16 bit code test */
|
1303 |
extern char code16_start, code16_end; |
1304 |
extern char code16_func1; |
1305 |
extern char code16_func2; |
1306 |
extern char code16_func3; |
1307 |
|
1308 |
void test_code16(void) |
1309 |
{ |
1310 |
struct modify_ldt_ldt_s ldt;
|
1311 |
int res, res2;
|
1312 |
|
1313 |
/* build a code segment */
|
1314 |
ldt.entry_number = 1;
|
1315 |
ldt.base_addr = (unsigned long)&code16_start; |
1316 |
ldt.limit = &code16_end - &code16_start; |
1317 |
ldt.seg_32bit = 0;
|
1318 |
ldt.contents = MODIFY_LDT_CONTENTS_CODE; |
1319 |
ldt.read_exec_only = 0;
|
1320 |
ldt.limit_in_pages = 0;
|
1321 |
ldt.seg_not_present = 0;
|
1322 |
ldt.useable = 1;
|
1323 |
modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */ |
1324 |
|
1325 |
/* call the first function */
|
1326 |
asm volatile ("lcall %1, %2" |
1327 |
: "=a" (res)
|
1328 |
: "i" (MK_SEL(1)), "i" (&code16_func1): "memory", "cc"); |
1329 |
printf("func1() = 0x%08x\n", res);
|
1330 |
asm volatile ("lcall %2, %3" |
1331 |
: "=a" (res), "=c" (res2) |
1332 |
: "i" (MK_SEL(1)), "i" (&code16_func2): "memory", "cc"); |
1333 |
printf("func2() = 0x%08x spdec=%d\n", res, res2);
|
1334 |
asm volatile ("lcall %1, %2" |
1335 |
: "=a" (res)
|
1336 |
: "i" (MK_SEL(1)), "i" (&code16_func3): "memory", "cc"); |
1337 |
printf("func3() = 0x%08x\n", res);
|
1338 |
} |
1339 |
#endif
|
1340 |
|
1341 |
#if defined(__x86_64__)
|
1342 |
asm(".globl func_lret\n" |
1343 |
"func_lret:\n"
|
1344 |
"movl $0x87654641, %eax\n"
|
1345 |
"lretq\n");
|
1346 |
#else
|
1347 |
asm(".globl func_lret\n" |
1348 |
"func_lret:\n"
|
1349 |
"movl $0x87654321, %eax\n"
|
1350 |
"lret\n"
|
1351 |
|
1352 |
".globl func_iret\n"
|
1353 |
"func_iret:\n"
|
1354 |
"movl $0xabcd4321, %eax\n"
|
1355 |
"iret\n");
|
1356 |
#endif
|
1357 |
|
1358 |
extern char func_lret; |
1359 |
extern char func_iret; |
1360 |
|
1361 |
void test_misc(void) |
1362 |
{ |
1363 |
char table[256]; |
1364 |
long res, i;
|
1365 |
|
1366 |
for(i=0;i<256;i++) table[i] = 256 - i; |
1367 |
res = 0x12345678;
|
1368 |
asm ("xlat" : "=a" (res) : "b" (table), "0" (res)); |
1369 |
printf("xlat: EAX=" FMTLX "\n", res); |
1370 |
|
1371 |
#if defined(__x86_64__)
|
1372 |
#if 0
|
1373 |
{
|
1374 |
/* XXX: see if Intel Core2 and AMD64 behavior really
|
1375 |
differ. Here we implemented the Intel way which is not
|
1376 |
compatible yet with QEMU. */
|
1377 |
static struct __attribute__((packed)) {
|
1378 |
uint64_t offset;
|
1379 |
uint16_t seg;
|
1380 |
} desc;
|
1381 |
long cs_sel;
|
1382 |
|
1383 |
asm volatile ("mov %%cs, %0" : "=r" (cs_sel));
|
1384 |
|
1385 |
asm volatile ("push %1\n"
|
1386 |
"call func_lret\n"
|
1387 |
: "=a" (res)
|
1388 |
: "r" (cs_sel) : "memory", "cc");
|
1389 |
printf("func_lret=" FMTLX "\n", res);
|
1390 |
|
1391 |
desc.offset = (long)&func_lret;
|
1392 |
desc.seg = cs_sel;
|
1393 |
|
1394 |
asm volatile ("xor %%rax, %%rax\n"
|
1395 |
"rex64 lcall *(%%rcx)\n"
|
1396 |
: "=a" (res)
|
1397 |
: "c" (&desc)
|
1398 |
: "memory", "cc");
|
1399 |
printf("func_lret2=" FMTLX "\n", res);
|
1400 |
|
1401 |
asm volatile ("push %2\n"
|
1402 |
"mov $ 1f, %%rax\n"
|
1403 |
"push %%rax\n"
|
1404 |
"rex64 ljmp *(%%rcx)\n"
|
1405 |
"1:\n"
|
1406 |
: "=a" (res)
|
1407 |
: "c" (&desc), "b" (cs_sel)
|
1408 |
: "memory", "cc");
|
1409 |
printf("func_lret3=" FMTLX "\n", res);
|
1410 |
}
|
1411 |
#endif
|
1412 |
#else
|
1413 |
asm volatile ("push %%cs ; call %1" |
1414 |
: "=a" (res)
|
1415 |
: "m" (func_lret): "memory", "cc"); |
1416 |
printf("func_lret=" FMTLX "\n", res); |
1417 |
|
1418 |
asm volatile ("pushf ; push %%cs ; call %1" |
1419 |
: "=a" (res)
|
1420 |
: "m" (func_iret): "memory", "cc"); |
1421 |
printf("func_iret=" FMTLX "\n", res); |
1422 |
#endif
|
1423 |
|
1424 |
#if defined(__x86_64__)
|
1425 |
/* specific popl test */
|
1426 |
asm volatile ("push $12345432 ; push $0x9abcdef ; pop (%%rsp) ; pop %0" |
1427 |
: "=g" (res));
|
1428 |
printf("popl esp=" FMTLX "\n", res); |
1429 |
#else
|
1430 |
/* specific popl test */
|
1431 |
asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popl (%%esp) ; popl %0" |
1432 |
: "=g" (res));
|
1433 |
printf("popl esp=" FMTLX "\n", res); |
1434 |
|
1435 |
/* specific popw test */
|
1436 |
asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popw (%%esp) ; addl $2, %%esp ; popl %0" |
1437 |
: "=g" (res));
|
1438 |
printf("popw esp=" FMTLX "\n", res); |
1439 |
#endif
|
1440 |
} |
1441 |
|
1442 |
uint8_t str_buffer[4096];
|
1443 |
|
1444 |
#define TEST_STRING1(OP, size, DF, REP)\
|
1445 |
{\ |
1446 |
long esi, edi, eax, ecx, eflags;\
|
1447 |
\ |
1448 |
esi = (long)(str_buffer + sizeof(str_buffer) / 2);\ |
1449 |
edi = (long)(str_buffer + sizeof(str_buffer) / 2) + 16;\ |
1450 |
eax = i2l(0x12345678);\
|
1451 |
ecx = 17;\
|
1452 |
\ |
1453 |
asm volatile ("push $0\n\t"\ |
1454 |
"popf\n\t"\
|
1455 |
DF "\n\t"\
|
1456 |
REP #OP size "\n\t"\ |
1457 |
"cld\n\t"\
|
1458 |
"pushf\n\t"\
|
1459 |
"pop %4\n\t"\
|
1460 |
: "=S" (esi), "=D" (edi), "=a" (eax), "=c" (ecx), "=g" (eflags)\ |
1461 |
: "0" (esi), "1" (edi), "2" (eax), "3" (ecx));\ |
1462 |
printf("%-10s ESI=" FMTLX " EDI=" FMTLX " EAX=" FMTLX " ECX=" FMTLX " EFL=%04x\n",\ |
1463 |
REP #OP size, esi, edi, eax, ecx,\
|
1464 |
(int)(eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)));\
|
1465 |
} |
1466 |
|
1467 |
#define TEST_STRING(OP, REP)\
|
1468 |
TEST_STRING1(OP, "b", "", REP);\ |
1469 |
TEST_STRING1(OP, "w", "", REP);\ |
1470 |
TEST_STRING1(OP, "l", "", REP);\ |
1471 |
X86_64_ONLY(TEST_STRING1(OP, "q", "", REP));\ |
1472 |
TEST_STRING1(OP, "b", "std", REP);\ |
1473 |
TEST_STRING1(OP, "w", "std", REP);\ |
1474 |
TEST_STRING1(OP, "l", "std", REP);\ |
1475 |
X86_64_ONLY(TEST_STRING1(OP, "q", "std", REP)) |
1476 |
|
1477 |
void test_string(void) |
1478 |
{ |
1479 |
int i;
|
1480 |
for(i = 0;i < sizeof(str_buffer); i++) |
1481 |
str_buffer[i] = i + 0x56;
|
1482 |
TEST_STRING(stos, "");
|
1483 |
TEST_STRING(stos, "rep ");
|
1484 |
TEST_STRING(lods, ""); /* to verify stos */ |
1485 |
TEST_STRING(lods, "rep ");
|
1486 |
TEST_STRING(movs, "");
|
1487 |
TEST_STRING(movs, "rep ");
|
1488 |
TEST_STRING(lods, ""); /* to verify stos */ |
1489 |
|
1490 |
/* XXX: better tests */
|
1491 |
TEST_STRING(scas, "");
|
1492 |
TEST_STRING(scas, "repz ");
|
1493 |
TEST_STRING(scas, "repnz ");
|
1494 |
TEST_STRING(cmps, "");
|
1495 |
TEST_STRING(cmps, "repz ");
|
1496 |
TEST_STRING(cmps, "repnz ");
|
1497 |
} |
1498 |
|
1499 |
#ifdef TEST_VM86
|
1500 |
/* VM86 test */
|
1501 |
|
1502 |
static inline void set_bit(uint8_t *a, unsigned int bit) |
1503 |
{ |
1504 |
a[bit / 8] |= (1 << (bit % 8)); |
1505 |
} |
1506 |
|
1507 |
static inline uint8_t *seg_to_linear(unsigned int seg, unsigned int reg) |
1508 |
{ |
1509 |
return (uint8_t *)((seg << 4) + (reg & 0xffff)); |
1510 |
} |
1511 |
|
1512 |
static inline void pushw(struct vm86_regs *r, int val) |
1513 |
{ |
1514 |
r->esp = (r->esp & ~0xffff) | ((r->esp - 2) & 0xffff); |
1515 |
*(uint16_t *)seg_to_linear(r->ss, r->esp) = val; |
1516 |
} |
1517 |
|
1518 |
static inline int vm86(int func, struct vm86plus_struct *v86) |
1519 |
{ |
1520 |
return syscall(__NR_vm86, func, v86);
|
1521 |
} |
1522 |
|
1523 |
extern char vm86_code_start; |
1524 |
extern char vm86_code_end; |
1525 |
|
1526 |
#define VM86_CODE_CS 0x100 |
1527 |
#define VM86_CODE_IP 0x100 |
1528 |
|
1529 |
void test_vm86(void) |
1530 |
{ |
1531 |
struct vm86plus_struct ctx;
|
1532 |
struct vm86_regs *r;
|
1533 |
uint8_t *vm86_mem; |
1534 |
int seg, ret;
|
1535 |
|
1536 |
vm86_mem = mmap((void *)0x00000000, 0x110000, |
1537 |
PROT_WRITE | PROT_READ | PROT_EXEC, |
1538 |
MAP_FIXED | MAP_ANON | MAP_PRIVATE, -1, 0); |
1539 |
if (vm86_mem == MAP_FAILED) {
|
1540 |
printf("ERROR: could not map vm86 memory");
|
1541 |
return;
|
1542 |
} |
1543 |
memset(&ctx, 0, sizeof(ctx)); |
1544 |
|
1545 |
/* init basic registers */
|
1546 |
r = &ctx.regs; |
1547 |
r->eip = VM86_CODE_IP; |
1548 |
r->esp = 0xfffe;
|
1549 |
seg = VM86_CODE_CS; |
1550 |
r->cs = seg; |
1551 |
r->ss = seg; |
1552 |
r->ds = seg; |
1553 |
r->es = seg; |
1554 |
r->fs = seg; |
1555 |
r->gs = seg; |
1556 |
r->eflags = VIF_MASK; |
1557 |
|
1558 |
/* move code to proper address. We use the same layout as a .com
|
1559 |
dos program. */
|
1560 |
memcpy(vm86_mem + (VM86_CODE_CS << 4) + VM86_CODE_IP,
|
1561 |
&vm86_code_start, &vm86_code_end - &vm86_code_start); |
1562 |
|
1563 |
/* mark int 0x21 as being emulated */
|
1564 |
set_bit((uint8_t *)&ctx.int_revectored, 0x21);
|
1565 |
|
1566 |
for(;;) {
|
1567 |
ret = vm86(VM86_ENTER, &ctx); |
1568 |
switch(VM86_TYPE(ret)) {
|
1569 |
case VM86_INTx:
|
1570 |
{ |
1571 |
int int_num, ah, v;
|
1572 |
|
1573 |
int_num = VM86_ARG(ret); |
1574 |
if (int_num != 0x21) |
1575 |
goto unknown_int;
|
1576 |
ah = (r->eax >> 8) & 0xff; |
1577 |
switch(ah) {
|
1578 |
case 0x00: /* exit */ |
1579 |
goto the_end;
|
1580 |
case 0x02: /* write char */ |
1581 |
{ |
1582 |
uint8_t c = r->edx; |
1583 |
putchar(c); |
1584 |
} |
1585 |
break;
|
1586 |
case 0x09: /* write string */ |
1587 |
{ |
1588 |
uint8_t c, *ptr; |
1589 |
ptr = seg_to_linear(r->ds, r->edx); |
1590 |
for(;;) {
|
1591 |
c = *ptr++; |
1592 |
if (c == '$') |
1593 |
break;
|
1594 |
putchar(c); |
1595 |
} |
1596 |
r->eax = (r->eax & ~0xff) | '$'; |
1597 |
} |
1598 |
break;
|
1599 |
case 0xff: /* extension: write eflags number in edx */ |
1600 |
v = (int)r->edx;
|
1601 |
#ifndef LINUX_VM86_IOPL_FIX
|
1602 |
v &= ~0x3000;
|
1603 |
#endif
|
1604 |
printf("%08x\n", v);
|
1605 |
break;
|
1606 |
default:
|
1607 |
unknown_int:
|
1608 |
printf("unsupported int 0x%02x\n", int_num);
|
1609 |
goto the_end;
|
1610 |
} |
1611 |
} |
1612 |
break;
|
1613 |
case VM86_SIGNAL:
|
1614 |
/* a signal came, we just ignore that */
|
1615 |
break;
|
1616 |
case VM86_STI:
|
1617 |
break;
|
1618 |
default:
|
1619 |
printf("ERROR: unhandled vm86 return code (0x%x)\n", ret);
|
1620 |
goto the_end;
|
1621 |
} |
1622 |
} |
1623 |
the_end:
|
1624 |
printf("VM86 end\n");
|
1625 |
munmap(vm86_mem, 0x110000);
|
1626 |
} |
1627 |
#endif
|
1628 |
|
1629 |
/* exception tests */
|
1630 |
#if defined(__i386__) && !defined(REG_EAX)
|
1631 |
#define REG_EAX EAX
|
1632 |
#define REG_EBX EBX
|
1633 |
#define REG_ECX ECX
|
1634 |
#define REG_EDX EDX
|
1635 |
#define REG_ESI ESI
|
1636 |
#define REG_EDI EDI
|
1637 |
#define REG_EBP EBP
|
1638 |
#define REG_ESP ESP
|
1639 |
#define REG_EIP EIP
|
1640 |
#define REG_EFL EFL
|
1641 |
#define REG_TRAPNO TRAPNO
|
1642 |
#define REG_ERR ERR
|
1643 |
#endif
|
1644 |
|
1645 |
#if defined(__x86_64__)
|
1646 |
#define REG_EIP REG_RIP
|
1647 |
#endif
|
1648 |
|
1649 |
jmp_buf jmp_env; |
1650 |
int v1;
|
1651 |
int tab[2]; |
1652 |
|
1653 |
void sig_handler(int sig, siginfo_t *info, void *puc) |
1654 |
{ |
1655 |
struct ucontext *uc = puc;
|
1656 |
|
1657 |
printf("si_signo=%d si_errno=%d si_code=%d",
|
1658 |
info->si_signo, info->si_errno, info->si_code); |
1659 |
printf(" si_addr=0x%08lx",
|
1660 |
(unsigned long)info->si_addr); |
1661 |
printf("\n");
|
1662 |
|
1663 |
printf("trapno=" FMTLX " err=" FMTLX, |
1664 |
(long)uc->uc_mcontext.gregs[REG_TRAPNO],
|
1665 |
(long)uc->uc_mcontext.gregs[REG_ERR]);
|
1666 |
printf(" EIP=" FMTLX, (long)uc->uc_mcontext.gregs[REG_EIP]); |
1667 |
printf("\n");
|
1668 |
longjmp(jmp_env, 1);
|
1669 |
} |
1670 |
|
1671 |
void test_exceptions(void) |
1672 |
{ |
1673 |
struct sigaction act;
|
1674 |
volatile int val; |
1675 |
|
1676 |
act.sa_sigaction = sig_handler; |
1677 |
sigemptyset(&act.sa_mask); |
1678 |
act.sa_flags = SA_SIGINFO | SA_NODEFER; |
1679 |
sigaction(SIGFPE, &act, NULL);
|
1680 |
sigaction(SIGILL, &act, NULL);
|
1681 |
sigaction(SIGSEGV, &act, NULL);
|
1682 |
sigaction(SIGBUS, &act, NULL);
|
1683 |
sigaction(SIGTRAP, &act, NULL);
|
1684 |
|
1685 |
/* test division by zero reporting */
|
1686 |
printf("DIVZ exception:\n");
|
1687 |
if (setjmp(jmp_env) == 0) { |
1688 |
/* now divide by zero */
|
1689 |
v1 = 0;
|
1690 |
v1 = 2 / v1;
|
1691 |
} |
1692 |
|
1693 |
#if !defined(__x86_64__)
|
1694 |
printf("BOUND exception:\n");
|
1695 |
if (setjmp(jmp_env) == 0) { |
1696 |
/* bound exception */
|
1697 |
tab[0] = 1; |
1698 |
tab[1] = 10; |
1699 |
asm volatile ("bound %0, %1" : : "r" (11), "m" (tab[0])); |
1700 |
} |
1701 |
#endif
|
1702 |
|
1703 |
#ifdef TEST_SEGS
|
1704 |
printf("segment exceptions:\n");
|
1705 |
if (setjmp(jmp_env) == 0) { |
1706 |
/* load an invalid segment */
|
1707 |
asm volatile ("movl %0, %%fs" : : "r" ((0x1234 << 3) | 1)); |
1708 |
} |
1709 |
if (setjmp(jmp_env) == 0) { |
1710 |
/* null data segment is valid */
|
1711 |
asm volatile ("movl %0, %%fs" : : "r" (3)); |
1712 |
/* null stack segment */
|
1713 |
asm volatile ("movl %0, %%ss" : : "r" (3)); |
1714 |
} |
1715 |
|
1716 |
{ |
1717 |
struct modify_ldt_ldt_s ldt;
|
1718 |
ldt.entry_number = 1;
|
1719 |
ldt.base_addr = (unsigned long)&seg_data1; |
1720 |
ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12; |
1721 |
ldt.seg_32bit = 1;
|
1722 |
ldt.contents = MODIFY_LDT_CONTENTS_DATA; |
1723 |
ldt.read_exec_only = 0;
|
1724 |
ldt.limit_in_pages = 1;
|
1725 |
ldt.seg_not_present = 1;
|
1726 |
ldt.useable = 1;
|
1727 |
modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */ |
1728 |
|
1729 |
if (setjmp(jmp_env) == 0) { |
1730 |
/* segment not present */
|
1731 |
asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1))); |
1732 |
} |
1733 |
} |
1734 |
#endif
|
1735 |
|
1736 |
/* test SEGV reporting */
|
1737 |
printf("PF exception:\n");
|
1738 |
if (setjmp(jmp_env) == 0) { |
1739 |
val = 1;
|
1740 |
/* we add a nop to test a weird PC retrieval case */
|
1741 |
asm volatile ("nop"); |
1742 |
/* now store in an invalid address */
|
1743 |
*(char *)0x1234 = 1; |
1744 |
} |
1745 |
|
1746 |
/* test SEGV reporting */
|
1747 |
printf("PF exception:\n");
|
1748 |
if (setjmp(jmp_env) == 0) { |
1749 |
val = 1;
|
1750 |
/* read from an invalid address */
|
1751 |
v1 = *(char *)0x1234; |
1752 |
} |
1753 |
|
1754 |
/* test illegal instruction reporting */
|
1755 |
printf("UD2 exception:\n");
|
1756 |
if (setjmp(jmp_env) == 0) { |
1757 |
/* now execute an invalid instruction */
|
1758 |
asm volatile("ud2"); |
1759 |
} |
1760 |
printf("lock nop exception:\n");
|
1761 |
if (setjmp(jmp_env) == 0) { |
1762 |
/* now execute an invalid instruction */
|
1763 |
asm volatile("lock nop"); |
1764 |
} |
1765 |
|
1766 |
printf("INT exception:\n");
|
1767 |
if (setjmp(jmp_env) == 0) { |
1768 |
asm volatile ("int $0xfd"); |
1769 |
} |
1770 |
if (setjmp(jmp_env) == 0) { |
1771 |
asm volatile ("int $0x01"); |
1772 |
} |
1773 |
if (setjmp(jmp_env) == 0) { |
1774 |
asm volatile (".byte 0xcd, 0x03"); |
1775 |
} |
1776 |
if (setjmp(jmp_env) == 0) { |
1777 |
asm volatile ("int $0x04"); |
1778 |
} |
1779 |
if (setjmp(jmp_env) == 0) { |
1780 |
asm volatile ("int $0x05"); |
1781 |
} |
1782 |
|
1783 |
printf("INT3 exception:\n");
|
1784 |
if (setjmp(jmp_env) == 0) { |
1785 |
asm volatile ("int3"); |
1786 |
} |
1787 |
|
1788 |
printf("CLI exception:\n");
|
1789 |
if (setjmp(jmp_env) == 0) { |
1790 |
asm volatile ("cli"); |
1791 |
} |
1792 |
|
1793 |
printf("STI exception:\n");
|
1794 |
if (setjmp(jmp_env) == 0) { |
1795 |
asm volatile ("cli"); |
1796 |
} |
1797 |
|
1798 |
#if !defined(__x86_64__)
|
1799 |
printf("INTO exception:\n");
|
1800 |
if (setjmp(jmp_env) == 0) { |
1801 |
/* overflow exception */
|
1802 |
asm volatile ("addl $1, %0 ; into" : : "r" (0x7fffffff)); |
1803 |
} |
1804 |
#endif
|
1805 |
|
1806 |
printf("OUTB exception:\n");
|
1807 |
if (setjmp(jmp_env) == 0) { |
1808 |
asm volatile ("outb %%al, %%dx" : : "d" (0x4321), "a" (0)); |
1809 |
} |
1810 |
|
1811 |
printf("INB exception:\n");
|
1812 |
if (setjmp(jmp_env) == 0) { |
1813 |
asm volatile ("inb %%dx, %%al" : "=a" (val) : "d" (0x4321)); |
1814 |
} |
1815 |
|
1816 |
printf("REP OUTSB exception:\n");
|
1817 |
if (setjmp(jmp_env) == 0) { |
1818 |
asm volatile ("rep outsb" : : "d" (0x4321), "S" (tab), "c" (1)); |
1819 |
} |
1820 |
|
1821 |
printf("REP INSB exception:\n");
|
1822 |
if (setjmp(jmp_env) == 0) { |
1823 |
asm volatile ("rep insb" : : "d" (0x4321), "D" (tab), "c" (1)); |
1824 |
} |
1825 |
|
1826 |
printf("HLT exception:\n");
|
1827 |
if (setjmp(jmp_env) == 0) { |
1828 |
asm volatile ("hlt"); |
1829 |
} |
1830 |
|
1831 |
printf("single step exception:\n");
|
1832 |
val = 0;
|
1833 |
if (setjmp(jmp_env) == 0) { |
1834 |
asm volatile ("pushf\n" |
1835 |
"orl $0x00100, (%%esp)\n"
|
1836 |
"popf\n"
|
1837 |
"movl $0xabcd, %0\n"
|
1838 |
"movl $0x0, %0\n" : "=m" (val) : : "cc", "memory"); |
1839 |
} |
1840 |
printf("val=0x%x\n", val);
|
1841 |
} |
1842 |
|
1843 |
#if !defined(__x86_64__)
|
1844 |
/* specific precise single step test */
|
1845 |
void sig_trap_handler(int sig, siginfo_t *info, void *puc) |
1846 |
{ |
1847 |
struct ucontext *uc = puc;
|
1848 |
printf("EIP=" FMTLX "\n", (long)uc->uc_mcontext.gregs[REG_EIP]); |
1849 |
} |
1850 |
|
1851 |
const uint8_t sstep_buf1[4] = { 1, 2, 3, 4}; |
1852 |
uint8_t sstep_buf2[4];
|
1853 |
|
1854 |
void test_single_step(void) |
1855 |
{ |
1856 |
struct sigaction act;
|
1857 |
volatile int val; |
1858 |
int i;
|
1859 |
|
1860 |
val = 0;
|
1861 |
act.sa_sigaction = sig_trap_handler; |
1862 |
sigemptyset(&act.sa_mask); |
1863 |
act.sa_flags = SA_SIGINFO; |
1864 |
sigaction(SIGTRAP, &act, NULL);
|
1865 |
asm volatile ("pushf\n" |
1866 |
"orl $0x00100, (%%esp)\n"
|
1867 |
"popf\n"
|
1868 |
"movl $0xabcd, %0\n"
|
1869 |
|
1870 |
/* jmp test */
|
1871 |
"movl $3, %%ecx\n"
|
1872 |
"1:\n"
|
1873 |
"addl $1, %0\n"
|
1874 |
"decl %%ecx\n"
|
1875 |
"jnz 1b\n"
|
1876 |
|
1877 |
/* movsb: the single step should stop at each movsb iteration */
|
1878 |
"movl $sstep_buf1, %%esi\n"
|
1879 |
"movl $sstep_buf2, %%edi\n"
|
1880 |
"movl $0, %%ecx\n"
|
1881 |
"rep movsb\n"
|
1882 |
"movl $3, %%ecx\n"
|
1883 |
"rep movsb\n"
|
1884 |
"movl $1, %%ecx\n"
|
1885 |
"rep movsb\n"
|
1886 |
|
1887 |
/* cmpsb: the single step should stop at each cmpsb iteration */
|
1888 |
"movl $sstep_buf1, %%esi\n"
|
1889 |
"movl $sstep_buf2, %%edi\n"
|
1890 |
"movl $0, %%ecx\n"
|
1891 |
"rep cmpsb\n"
|
1892 |
"movl $4, %%ecx\n"
|
1893 |
"rep cmpsb\n"
|
1894 |
|
1895 |
/* getpid() syscall: single step should skip one
|
1896 |
instruction */
|
1897 |
"movl $20, %%eax\n"
|
1898 |
"int $0x80\n"
|
1899 |
"movl $0, %%eax\n"
|
1900 |
|
1901 |
/* when modifying SS, trace is not done on the next
|
1902 |
instruction */
|
1903 |
"movl %%ss, %%ecx\n"
|
1904 |
"movl %%ecx, %%ss\n"
|
1905 |
"addl $1, %0\n"
|
1906 |
"movl $1, %%eax\n"
|
1907 |
"movl %%ecx, %%ss\n"
|
1908 |
"jmp 1f\n"
|
1909 |
"addl $1, %0\n"
|
1910 |
"1:\n"
|
1911 |
"movl $1, %%eax\n"
|
1912 |
"pushl %%ecx\n"
|
1913 |
"popl %%ss\n"
|
1914 |
"addl $1, %0\n"
|
1915 |
"movl $1, %%eax\n"
|
1916 |
|
1917 |
"pushf\n"
|
1918 |
"andl $~0x00100, (%%esp)\n"
|
1919 |
"popf\n"
|
1920 |
: "=m" (val)
|
1921 |
: |
1922 |
: "cc", "memory", "eax", "ecx", "esi", "edi"); |
1923 |
printf("val=%d\n", val);
|
1924 |
for(i = 0; i < 4; i++) |
1925 |
printf("sstep_buf2[%d] = %d\n", i, sstep_buf2[i]);
|
1926 |
} |
1927 |
|
1928 |
/* self modifying code test */
|
1929 |
uint8_t code[] = { |
1930 |
0xb8, 0x1, 0x00, 0x00, 0x00, /* movl $1, %eax */ |
1931 |
0xc3, /* ret */ |
1932 |
}; |
1933 |
|
1934 |
asm("smc_code2:\n" |
1935 |
"movl 4(%esp), %eax\n"
|
1936 |
"movl %eax, smc_patch_addr2 + 1\n"
|
1937 |
"nop\n"
|
1938 |
"nop\n"
|
1939 |
"nop\n"
|
1940 |
"nop\n"
|
1941 |
"nop\n"
|
1942 |
"nop\n"
|
1943 |
"nop\n"
|
1944 |
"nop\n"
|
1945 |
"smc_patch_addr2:\n"
|
1946 |
"movl $1, %eax\n"
|
1947 |
"ret\n");
|
1948 |
|
1949 |
typedef int FuncType(void); |
1950 |
extern int smc_code2(int); |
1951 |
void test_self_modifying_code(void) |
1952 |
{ |
1953 |
int i;
|
1954 |
|
1955 |
printf("self modifying code:\n");
|
1956 |
printf("func1 = 0x%x\n", ((FuncType *)code)());
|
1957 |
for(i = 2; i <= 4; i++) { |
1958 |
code[1] = i;
|
1959 |
printf("func%d = 0x%x\n", i, ((FuncType *)code)());
|
1960 |
} |
1961 |
|
1962 |
/* more difficult test : the modified code is just after the
|
1963 |
modifying instruction. It is forbidden in Intel specs, but it
|
1964 |
is used by old DOS programs */
|
1965 |
for(i = 2; i <= 4; i++) { |
1966 |
printf("smc_code2(%d) = %d\n", i, smc_code2(i));
|
1967 |
} |
1968 |
} |
1969 |
#endif
|
1970 |
|
1971 |
long enter_stack[4096]; |
1972 |
|
1973 |
#if defined(__x86_64__)
|
1974 |
#define RSP "%%rsp" |
1975 |
#define RBP "%%rbp" |
1976 |
#else
|
1977 |
#define RSP "%%esp" |
1978 |
#define RBP "%%ebp" |
1979 |
#endif
|
1980 |
|
1981 |
#define TEST_ENTER(size, stack_type, level)\
|
1982 |
{\ |
1983 |
long esp_save, esp_val, ebp_val, ebp_save, i;\
|
1984 |
stack_type *ptr, *stack_end, *stack_ptr;\ |
1985 |
memset(enter_stack, 0, sizeof(enter_stack));\ |
1986 |
stack_end = stack_ptr = (stack_type *)(enter_stack + 4096);\
|
1987 |
ebp_val = (long)stack_ptr;\
|
1988 |
for(i=1;i<=32;i++)\ |
1989 |
*--stack_ptr = i;\ |
1990 |
esp_val = (long)stack_ptr;\
|
1991 |
asm("mov " RSP ", %[esp_save]\n"\ |
1992 |
"mov " RBP ", %[ebp_save]\n"\ |
1993 |
"mov %[esp_val], " RSP "\n"\ |
1994 |
"mov %[ebp_val], " RBP "\n"\ |
1995 |
"enter" size " $8, $" #level "\n"\ |
1996 |
"mov " RSP ", %[esp_val]\n"\ |
1997 |
"mov " RBP ", %[ebp_val]\n"\ |
1998 |
"mov %[esp_save], " RSP "\n"\ |
1999 |
"mov %[ebp_save], " RBP "\n"\ |
2000 |
: [esp_save] "=r" (esp_save),\
|
2001 |
[ebp_save] "=r" (ebp_save),\
|
2002 |
[esp_val] "=r" (esp_val),\
|
2003 |
[ebp_val] "=r" (ebp_val)\
|
2004 |
: "[esp_val]" (esp_val),\
|
2005 |
"[ebp_val]" (ebp_val));\
|
2006 |
printf("level=%d:\n", level);\
|
2007 |
printf("esp_val=" FMTLX "\n", esp_val - (long)stack_end);\ |
2008 |
printf("ebp_val=" FMTLX "\n", ebp_val - (long)stack_end);\ |
2009 |
for(ptr = (stack_type *)esp_val; ptr < stack_end; ptr++)\
|
2010 |
printf(FMTLX "\n", (long)ptr[0]);\ |
2011 |
} |
2012 |
|
2013 |
static void test_enter(void) |
2014 |
{ |
2015 |
#if defined(__x86_64__)
|
2016 |
TEST_ENTER("q", uint64_t, 0); |
2017 |
TEST_ENTER("q", uint64_t, 1); |
2018 |
TEST_ENTER("q", uint64_t, 2); |
2019 |
TEST_ENTER("q", uint64_t, 31); |
2020 |
#else
|
2021 |
TEST_ENTER("l", uint32_t, 0); |
2022 |
TEST_ENTER("l", uint32_t, 1); |
2023 |
TEST_ENTER("l", uint32_t, 2); |
2024 |
TEST_ENTER("l", uint32_t, 31); |
2025 |
#endif
|
2026 |
|
2027 |
TEST_ENTER("w", uint16_t, 0); |
2028 |
TEST_ENTER("w", uint16_t, 1); |
2029 |
TEST_ENTER("w", uint16_t, 2); |
2030 |
TEST_ENTER("w", uint16_t, 31); |
2031 |
} |
2032 |
|
2033 |
#ifdef TEST_SSE
|
2034 |
|
2035 |
typedef int __m64 __attribute__ ((__mode__ (__V2SI__))); |
2036 |
typedef float __m128 __attribute__ ((__mode__(__V4SF__))); |
2037 |
|
2038 |
typedef union { |
2039 |
double d[2]; |
2040 |
float s[4]; |
2041 |
uint32_t l[4];
|
2042 |
uint64_t q[2];
|
2043 |
__m128 dq; |
2044 |
} XMMReg; |
2045 |
|
2046 |
static uint64_t __attribute__((aligned(16))) test_values[4][2] = { |
2047 |
{ 0x456723c698694873, 0xdc515cff944a58ec }, |
2048 |
{ 0x1f297ccd58bad7ab, 0x41f21efba9e3e146 }, |
2049 |
{ 0x007c62c2085427f8, 0x231be9e8cde7438d }, |
2050 |
{ 0x0f76255a085427f8, 0xc233e9e8c4c9439a }, |
2051 |
}; |
2052 |
|
2053 |
#define SSE_OP(op)\
|
2054 |
{\ |
2055 |
asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\ |
2056 |
printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\ |
2057 |
#op,\
|
2058 |
a.q[1], a.q[0],\ |
2059 |
b.q[1], b.q[0],\ |
2060 |
r.q[1], r.q[0]);\ |
2061 |
} |
2062 |
|
2063 |
#define SSE_OP2(op)\
|
2064 |
{\ |
2065 |
int i;\
|
2066 |
for(i=0;i<2;i++) {\ |
2067 |
a.q[0] = test_values[2*i][0];\ |
2068 |
a.q[1] = test_values[2*i][1];\ |
2069 |
b.q[0] = test_values[2*i+1][0];\ |
2070 |
b.q[1] = test_values[2*i+1][1];\ |
2071 |
SSE_OP(op);\ |
2072 |
}\ |
2073 |
} |
2074 |
|
2075 |
#define MMX_OP2(op)\
|
2076 |
{\ |
2077 |
int i;\
|
2078 |
for(i=0;i<2;i++) {\ |
2079 |
a.q[0] = test_values[2*i][0];\ |
2080 |
b.q[0] = test_values[2*i+1][0];\ |
2081 |
asm volatile (#op " %2, %0" : "=y" (r.q[0]) : "0" (a.q[0]), "y" (b.q[0]));\ |
2082 |
printf("%-9s: a=" FMT64X " b=" FMT64X " r=" FMT64X "\n",\ |
2083 |
#op,\
|
2084 |
a.q[0],\
|
2085 |
b.q[0],\
|
2086 |
r.q[0]);\
|
2087 |
}\ |
2088 |
SSE_OP2(op);\ |
2089 |
} |
2090 |
|
2091 |
#define SHUF_OP(op, ib)\
|
2092 |
{\ |
2093 |
a.q[0] = test_values[0][0];\ |
2094 |
a.q[1] = test_values[0][1];\ |
2095 |
b.q[0] = test_values[1][0];\ |
2096 |
b.q[1] = test_values[1][1];\ |
2097 |
asm volatile (#op " $" #ib ", %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\ |
2098 |
printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\ |
2099 |
#op,\
|
2100 |
a.q[1], a.q[0],\ |
2101 |
b.q[1], b.q[0],\ |
2102 |
ib,\ |
2103 |
r.q[1], r.q[0]);\ |
2104 |
} |
2105 |
|
2106 |
#define PSHUF_OP(op, ib)\
|
2107 |
{\ |
2108 |
int i;\
|
2109 |
for(i=0;i<2;i++) {\ |
2110 |
a.q[0] = test_values[2*i][0];\ |
2111 |
a.q[1] = test_values[2*i][1];\ |
2112 |
asm volatile (#op " $" #ib ", %1, %0" : "=x" (r.dq) : "x" (a.dq));\ |
2113 |
printf("%-9s: a=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\ |
2114 |
#op,\
|
2115 |
a.q[1], a.q[0],\ |
2116 |
ib,\ |
2117 |
r.q[1], r.q[0]);\ |
2118 |
}\ |
2119 |
} |
2120 |
|
2121 |
#define SHIFT_IM(op, ib)\
|
2122 |
{\ |
2123 |
int i;\
|
2124 |
for(i=0;i<2;i++) {\ |
2125 |
a.q[0] = test_values[2*i][0];\ |
2126 |
a.q[1] = test_values[2*i][1];\ |
2127 |
asm volatile (#op " $" #ib ", %0" : "=x" (r.dq) : "0" (a.dq));\ |
2128 |
printf("%-9s: a=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\ |
2129 |
#op,\
|
2130 |
a.q[1], a.q[0],\ |
2131 |
ib,\ |
2132 |
r.q[1], r.q[0]);\ |
2133 |
}\ |
2134 |
} |
2135 |
|
2136 |
#define SHIFT_OP(op, ib)\
|
2137 |
{\ |
2138 |
int i;\
|
2139 |
SHIFT_IM(op, ib);\ |
2140 |
for(i=0;i<2;i++) {\ |
2141 |
a.q[0] = test_values[2*i][0];\ |
2142 |
a.q[1] = test_values[2*i][1];\ |
2143 |
b.q[0] = ib;\
|
2144 |
b.q[1] = 0;\ |
2145 |
asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\ |
2146 |
printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\ |
2147 |
#op,\
|
2148 |
a.q[1], a.q[0],\ |
2149 |
b.q[1], b.q[0],\ |
2150 |
r.q[1], r.q[0]);\ |
2151 |
}\ |
2152 |
} |
2153 |
|
2154 |
#define MOVMSK(op)\
|
2155 |
{\ |
2156 |
int i, reg;\
|
2157 |
for(i=0;i<2;i++) {\ |
2158 |
a.q[0] = test_values[2*i][0];\ |
2159 |
a.q[1] = test_values[2*i][1];\ |
2160 |
asm volatile (#op " %1, %0" : "=r" (reg) : "x" (a.dq));\ |
2161 |
printf("%-9s: a=" FMT64X "" FMT64X " r=%08x\n",\ |
2162 |
#op,\
|
2163 |
a.q[1], a.q[0],\ |
2164 |
reg);\ |
2165 |
}\ |
2166 |
} |
2167 |
|
2168 |
#define SSE_OPS(a) \
|
2169 |
SSE_OP(a ## ps);\ |
2170 |
SSE_OP(a ## ss); |
2171 |
|
2172 |
#define SSE_OPD(a) \
|
2173 |
SSE_OP(a ## pd);\ |
2174 |
SSE_OP(a ## sd); |
2175 |
|
2176 |
#define SSE_COMI(op, field)\
|
2177 |
{\ |
2178 |
unsigned int eflags;\ |
2179 |
XMMReg a, b;\ |
2180 |
a.field[0] = a1;\
|
2181 |
b.field[0] = b1;\
|
2182 |
asm volatile (#op " %2, %1\n"\ |
2183 |
"pushf\n"\
|
2184 |
"pop %0\n"\
|
2185 |
: "=m" (eflags)\
|
2186 |
: "x" (a.dq), "x" (b.dq));\ |
2187 |
printf("%-9s: a=%f b=%f cc=%04x\n",\
|
2188 |
#op, a1, b1,\
|
2189 |
eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));\ |
2190 |
} |
2191 |
|
2192 |
void test_sse_comi(double a1, double b1) |
2193 |
{ |
2194 |
SSE_COMI(ucomiss, s); |
2195 |
SSE_COMI(ucomisd, d); |
2196 |
SSE_COMI(comiss, s); |
2197 |
SSE_COMI(comisd, d); |
2198 |
} |
2199 |
|
2200 |
#define CVT_OP_XMM(op)\
|
2201 |
{\ |
2202 |
asm volatile (#op " %1, %0" : "=x" (r.dq) : "x" (a.dq));\ |
2203 |
printf("%-9s: a=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\ |
2204 |
#op,\
|
2205 |
a.q[1], a.q[0],\ |
2206 |
r.q[1], r.q[0]);\ |
2207 |
} |
2208 |
|
2209 |
/* Force %xmm0 usage to avoid the case where both register index are 0
|
2210 |
to test intruction decoding more extensively */
|
2211 |
#define CVT_OP_XMM2MMX(op)\
|
2212 |
{\ |
2213 |
asm volatile (#op " %1, %0" : "=y" (r.q[0]) : "x" (a.dq) \ |
2214 |
: "%xmm0");\
|
2215 |
printf("%-9s: a=" FMT64X "" FMT64X " r=" FMT64X "\n",\ |
2216 |
#op,\
|
2217 |
a.q[1], a.q[0],\ |
2218 |
r.q[0]);\
|
2219 |
} |
2220 |
|
2221 |
#define CVT_OP_MMX2XMM(op)\
|
2222 |
{\ |
2223 |
asm volatile (#op " %1, %0" : "=x" (r.dq) : "y" (a.q[0]));\ |
2224 |
printf("%-9s: a=" FMT64X " r=" FMT64X "" FMT64X "\n",\ |
2225 |
#op,\
|
2226 |
a.q[0],\
|
2227 |
r.q[1], r.q[0]);\ |
2228 |
} |
2229 |
|
2230 |
#define CVT_OP_REG2XMM(op)\
|
2231 |
{\ |
2232 |
asm volatile (#op " %1, %0" : "=x" (r.dq) : "r" (a.l[0]));\ |
2233 |
printf("%-9s: a=%08x r=" FMT64X "" FMT64X "\n",\ |
2234 |
#op,\
|
2235 |
a.l[0],\
|
2236 |
r.q[1], r.q[0]);\ |
2237 |
} |
2238 |
|
2239 |
#define CVT_OP_XMM2REG(op)\
|
2240 |
{\ |
2241 |
asm volatile (#op " %1, %0" : "=r" (r.l[0]) : "x" (a.dq));\ |
2242 |
printf("%-9s: a=" FMT64X "" FMT64X " r=%08x\n",\ |
2243 |
#op,\
|
2244 |
a.q[1], a.q[0],\ |
2245 |
r.l[0]);\
|
2246 |
} |
2247 |
|
2248 |
struct fpxstate {
|
2249 |
uint16_t fpuc; |
2250 |
uint16_t fpus; |
2251 |
uint16_t fptag; |
2252 |
uint16_t fop; |
2253 |
uint32_t fpuip; |
2254 |
uint16_t cs_sel; |
2255 |
uint16_t dummy0; |
2256 |
uint32_t fpudp; |
2257 |
uint16_t ds_sel; |
2258 |
uint16_t dummy1; |
2259 |
uint32_t mxcsr; |
2260 |
uint32_t mxcsr_mask; |
2261 |
uint8_t fpregs1[8 * 16]; |
2262 |
uint8_t xmm_regs[8 * 16]; |
2263 |
uint8_t dummy2[224];
|
2264 |
}; |
2265 |
|
2266 |
static struct fpxstate fpx_state __attribute__((aligned(16))); |
2267 |
static struct fpxstate fpx_state2 __attribute__((aligned(16))); |
2268 |
|
2269 |
void test_fxsave(void) |
2270 |
{ |
2271 |
struct fpxstate *fp = &fpx_state;
|
2272 |
struct fpxstate *fp2 = &fpx_state2;
|
2273 |
int i, nb_xmm;
|
2274 |
XMMReg a, b; |
2275 |
a.q[0] = test_values[0][0]; |
2276 |
a.q[1] = test_values[0][1]; |
2277 |
b.q[0] = test_values[1][0]; |
2278 |
b.q[1] = test_values[1][1]; |
2279 |
|
2280 |
asm("movdqa %2, %%xmm0\n" |
2281 |
"movdqa %3, %%xmm7\n"
|
2282 |
#if defined(__x86_64__)
|
2283 |
"movdqa %2, %%xmm15\n"
|
2284 |
#endif
|
2285 |
" fld1\n"
|
2286 |
" fldpi\n"
|
2287 |
" fldln2\n"
|
2288 |
" fxsave %0\n"
|
2289 |
" fxrstor %0\n"
|
2290 |
" fxsave %1\n"
|
2291 |
" fninit\n"
|
2292 |
: "=m" (*(uint32_t *)fp2), "=m" (*(uint32_t *)fp) |
2293 |
: "m" (a), "m" (b)); |
2294 |
printf("fpuc=%04x\n", fp->fpuc);
|
2295 |
printf("fpus=%04x\n", fp->fpus);
|
2296 |
printf("fptag=%04x\n", fp->fptag);
|
2297 |
for(i = 0; i < 3; i++) { |
2298 |
printf("ST%d: " FMT64X " %04x\n", |
2299 |
i, |
2300 |
*(uint64_t *)&fp->fpregs1[i * 16],
|
2301 |
*(uint16_t *)&fp->fpregs1[i * 16 + 8]); |
2302 |
} |
2303 |
printf("mxcsr=%08x\n", fp->mxcsr & 0x1f80); |
2304 |
#if defined(__x86_64__)
|
2305 |
nb_xmm = 16;
|
2306 |
#else
|
2307 |
nb_xmm = 8;
|
2308 |
#endif
|
2309 |
for(i = 0; i < nb_xmm; i++) { |
2310 |
printf("xmm%d: " FMT64X "" FMT64X "\n", |
2311 |
i, |
2312 |
*(uint64_t *)&fp->xmm_regs[i * 16],
|
2313 |
*(uint64_t *)&fp->xmm_regs[i * 16 + 8]); |
2314 |
} |
2315 |
} |
2316 |
|
2317 |
void test_sse(void) |
2318 |
{ |
2319 |
XMMReg r, a, b; |
2320 |
int i;
|
2321 |
|
2322 |
MMX_OP2(punpcklbw); |
2323 |
MMX_OP2(punpcklwd); |
2324 |
MMX_OP2(punpckldq); |
2325 |
MMX_OP2(packsswb); |
2326 |
MMX_OP2(pcmpgtb); |
2327 |
MMX_OP2(pcmpgtw); |
2328 |
MMX_OP2(pcmpgtd); |
2329 |
MMX_OP2(packuswb); |
2330 |
MMX_OP2(punpckhbw); |
2331 |
MMX_OP2(punpckhwd); |
2332 |
MMX_OP2(punpckhdq); |
2333 |
MMX_OP2(packssdw); |
2334 |
MMX_OP2(pcmpeqb); |
2335 |
MMX_OP2(pcmpeqw); |
2336 |
MMX_OP2(pcmpeqd); |
2337 |
|
2338 |
MMX_OP2(paddq); |
2339 |
MMX_OP2(pmullw); |
2340 |
MMX_OP2(psubusb); |
2341 |
MMX_OP2(psubusw); |
2342 |
MMX_OP2(pminub); |
2343 |
MMX_OP2(pand); |
2344 |
MMX_OP2(paddusb); |
2345 |
MMX_OP2(paddusw); |
2346 |
MMX_OP2(pmaxub); |
2347 |
MMX_OP2(pandn); |
2348 |
|
2349 |
MMX_OP2(pmulhuw); |
2350 |
MMX_OP2(pmulhw); |
2351 |
|
2352 |
MMX_OP2(psubsb); |
2353 |
MMX_OP2(psubsw); |
2354 |
MMX_OP2(pminsw); |
2355 |
MMX_OP2(por); |
2356 |
MMX_OP2(paddsb); |
2357 |
MMX_OP2(paddsw); |
2358 |
MMX_OP2(pmaxsw); |
2359 |
MMX_OP2(pxor); |
2360 |
MMX_OP2(pmuludq); |
2361 |
MMX_OP2(pmaddwd); |
2362 |
MMX_OP2(psadbw); |
2363 |
MMX_OP2(psubb); |
2364 |
MMX_OP2(psubw); |
2365 |
MMX_OP2(psubd); |
2366 |
MMX_OP2(psubq); |
2367 |
MMX_OP2(paddb); |
2368 |
MMX_OP2(paddw); |
2369 |
MMX_OP2(paddd); |
2370 |
|
2371 |
MMX_OP2(pavgb); |
2372 |
MMX_OP2(pavgw); |
2373 |
|
2374 |
asm volatile ("pinsrw $1, %1, %0" : "=y" (r.q[0]) : "r" (0x12345678)); |
2375 |
printf("%-9s: r=" FMT64X "\n", "pinsrw", r.q[0]); |
2376 |
|
2377 |
asm volatile ("pinsrw $5, %1, %0" : "=x" (r.dq) : "r" (0x12345678)); |
2378 |
printf("%-9s: r=" FMT64X "" FMT64X "\n", "pinsrw", r.q[1], r.q[0]); |
2379 |
|
2380 |
a.q[0] = test_values[0][0]; |
2381 |
a.q[1] = test_values[0][1]; |
2382 |
asm volatile ("pextrw $1, %1, %0" : "=r" (r.l[0]) : "y" (a.q[0])); |
2383 |
printf("%-9s: r=%08x\n", "pextrw", r.l[0]); |
2384 |
|
2385 |
asm volatile ("pextrw $5, %1, %0" : "=r" (r.l[0]) : "x" (a.dq)); |
2386 |
printf("%-9s: r=%08x\n", "pextrw", r.l[0]); |
2387 |
|
2388 |
asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "y" (a.q[0])); |
2389 |
printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]); |
2390 |
|
2391 |
asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "x" (a.dq)); |
2392 |
printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]); |
2393 |
|
2394 |
{ |
2395 |
r.q[0] = -1; |
2396 |
r.q[1] = -1; |
2397 |
|
2398 |
a.q[0] = test_values[0][0]; |
2399 |
a.q[1] = test_values[0][1]; |
2400 |
b.q[0] = test_values[1][0]; |
2401 |
b.q[1] = test_values[1][1]; |
2402 |
asm volatile("maskmovq %1, %0" : |
2403 |
: "y" (a.q[0]), "y" (b.q[0]), "D" (&r) |
2404 |
: "memory");
|
2405 |
printf("%-9s: r=" FMT64X " a=" FMT64X " b=" FMT64X "\n", |
2406 |
"maskmov",
|
2407 |
r.q[0],
|
2408 |
a.q[0],
|
2409 |
b.q[0]);
|
2410 |
asm volatile("maskmovdqu %1, %0" : |
2411 |
: "x" (a.dq), "x" (b.dq), "D" (&r) |
2412 |
: "memory");
|
2413 |
printf("%-9s: r=" FMT64X "" FMT64X " a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X "\n", |
2414 |
"maskmov",
|
2415 |
r.q[1], r.q[0], |
2416 |
a.q[1], a.q[0], |
2417 |
b.q[1], b.q[0]); |
2418 |
} |
2419 |
|
2420 |
asm volatile ("emms"); |
2421 |
|
2422 |
SSE_OP2(punpcklqdq); |
2423 |
SSE_OP2(punpckhqdq); |
2424 |
SSE_OP2(andps); |
2425 |
SSE_OP2(andpd); |
2426 |
SSE_OP2(andnps); |
2427 |
SSE_OP2(andnpd); |
2428 |
SSE_OP2(orps); |
2429 |
SSE_OP2(orpd); |
2430 |
SSE_OP2(xorps); |
2431 |
SSE_OP2(xorpd); |
2432 |
|
2433 |
SSE_OP2(unpcklps); |
2434 |
SSE_OP2(unpcklpd); |
2435 |
SSE_OP2(unpckhps); |
2436 |
SSE_OP2(unpckhpd); |
2437 |
|
2438 |
SHUF_OP(shufps, 0x78);
|
2439 |
SHUF_OP(shufpd, 0x02);
|
2440 |
|
2441 |
PSHUF_OP(pshufd, 0x78);
|
2442 |
PSHUF_OP(pshuflw, 0x78);
|
2443 |
PSHUF_OP(pshufhw, 0x78);
|
2444 |
|
2445 |
SHIFT_OP(psrlw, 7);
|
2446 |
SHIFT_OP(psrlw, 16);
|
2447 |
SHIFT_OP(psraw, 7);
|
2448 |
SHIFT_OP(psraw, 16);
|
2449 |
SHIFT_OP(psllw, 7);
|
2450 |
SHIFT_OP(psllw, 16);
|
2451 |
|
2452 |
SHIFT_OP(psrld, 7);
|
2453 |
SHIFT_OP(psrld, 32);
|
2454 |
SHIFT_OP(psrad, 7);
|
2455 |
SHIFT_OP(psrad, 32);
|
2456 |
SHIFT_OP(pslld, 7);
|
2457 |
SHIFT_OP(pslld, 32);
|
2458 |
|
2459 |
SHIFT_OP(psrlq, 7);
|
2460 |
SHIFT_OP(psrlq, 32);
|
2461 |
SHIFT_OP(psllq, 7);
|
2462 |
SHIFT_OP(psllq, 32);
|
2463 |
|
2464 |
SHIFT_IM(psrldq, 16);
|
2465 |
SHIFT_IM(psrldq, 7);
|
2466 |
SHIFT_IM(pslldq, 16);
|
2467 |
SHIFT_IM(pslldq, 7);
|
2468 |
|
2469 |
MOVMSK(movmskps); |
2470 |
MOVMSK(movmskpd); |
2471 |
|
2472 |
/* FPU specific ops */
|
2473 |
|
2474 |
{ |
2475 |
uint32_t mxcsr; |
2476 |
asm volatile("stmxcsr %0" : "=m" (mxcsr)); |
2477 |
printf("mxcsr=%08x\n", mxcsr & 0x1f80); |
2478 |
asm volatile("ldmxcsr %0" : : "m" (mxcsr)); |
2479 |
} |
2480 |
|
2481 |
test_sse_comi(2, -1); |
2482 |
test_sse_comi(2, 2); |
2483 |
test_sse_comi(2, 3); |
2484 |
test_sse_comi(2, q_nan.d);
|
2485 |
test_sse_comi(q_nan.d, -1);
|
2486 |
|
2487 |
for(i = 0; i < 2; i++) { |
2488 |
a.s[0] = 2.7; |
2489 |
a.s[1] = 3.4; |
2490 |
a.s[2] = 4; |
2491 |
a.s[3] = -6.3; |
2492 |
b.s[0] = 45.7; |
2493 |
b.s[1] = 353.4; |
2494 |
b.s[2] = 4; |
2495 |
b.s[3] = 56.3; |
2496 |
if (i == 1) { |
2497 |
a.s[0] = q_nan.d;
|
2498 |
b.s[3] = q_nan.d;
|
2499 |
} |
2500 |
|
2501 |
SSE_OPS(add); |
2502 |
SSE_OPS(mul); |
2503 |
SSE_OPS(sub); |
2504 |
SSE_OPS(min); |
2505 |
SSE_OPS(div); |
2506 |
SSE_OPS(max); |
2507 |
SSE_OPS(sqrt); |
2508 |
SSE_OPS(cmpeq); |
2509 |
SSE_OPS(cmplt); |
2510 |
SSE_OPS(cmple); |
2511 |
SSE_OPS(cmpunord); |
2512 |
SSE_OPS(cmpneq); |
2513 |
SSE_OPS(cmpnlt); |
2514 |
SSE_OPS(cmpnle); |
2515 |
SSE_OPS(cmpord); |
2516 |
|
2517 |
|
2518 |
a.d[0] = 2.7; |
2519 |
a.d[1] = -3.4; |
2520 |
b.d[0] = 45.7; |
2521 |
b.d[1] = -53.4; |
2522 |
if (i == 1) { |
2523 |
a.d[0] = q_nan.d;
|
2524 |
b.d[1] = q_nan.d;
|
2525 |
} |
2526 |
SSE_OPD(add); |
2527 |
SSE_OPD(mul); |
2528 |
SSE_OPD(sub); |
2529 |
SSE_OPD(min); |
2530 |
SSE_OPD(div); |
2531 |
SSE_OPD(max); |
2532 |
SSE_OPD(sqrt); |
2533 |
SSE_OPD(cmpeq); |
2534 |
SSE_OPD(cmplt); |
2535 |
SSE_OPD(cmple); |
2536 |
SSE_OPD(cmpunord); |
2537 |
SSE_OPD(cmpneq); |
2538 |
SSE_OPD(cmpnlt); |
2539 |
SSE_OPD(cmpnle); |
2540 |
SSE_OPD(cmpord); |
2541 |
} |
2542 |
|
2543 |
/* float to float/int */
|
2544 |
a.s[0] = 2.7; |
2545 |
a.s[1] = 3.4; |
2546 |
a.s[2] = 4; |
2547 |
a.s[3] = -6.3; |
2548 |
CVT_OP_XMM(cvtps2pd); |
2549 |
CVT_OP_XMM(cvtss2sd); |
2550 |
CVT_OP_XMM2MMX(cvtps2pi); |
2551 |
CVT_OP_XMM2MMX(cvttps2pi); |
2552 |
CVT_OP_XMM2REG(cvtss2si); |
2553 |
CVT_OP_XMM2REG(cvttss2si); |
2554 |
CVT_OP_XMM(cvtps2dq); |
2555 |
CVT_OP_XMM(cvttps2dq); |
2556 |
|
2557 |
a.d[0] = 2.6; |
2558 |
a.d[1] = -3.4; |
2559 |
CVT_OP_XMM(cvtpd2ps); |
2560 |
CVT_OP_XMM(cvtsd2ss); |
2561 |
CVT_OP_XMM2MMX(cvtpd2pi); |
2562 |
CVT_OP_XMM2MMX(cvttpd2pi); |
2563 |
CVT_OP_XMM2REG(cvtsd2si); |
2564 |
CVT_OP_XMM2REG(cvttsd2si); |
2565 |
CVT_OP_XMM(cvtpd2dq); |
2566 |
CVT_OP_XMM(cvttpd2dq); |
2567 |
|
2568 |
/* sse/mmx moves */
|
2569 |
CVT_OP_XMM2MMX(movdq2q); |
2570 |
CVT_OP_MMX2XMM(movq2dq); |
2571 |
|
2572 |
/* int to float */
|
2573 |
a.l[0] = -6; |
2574 |
a.l[1] = 2; |
2575 |
a.l[2] = 100; |
2576 |
a.l[3] = -60000; |
2577 |
CVT_OP_MMX2XMM(cvtpi2ps); |
2578 |
CVT_OP_MMX2XMM(cvtpi2pd); |
2579 |
CVT_OP_REG2XMM(cvtsi2ss); |
2580 |
CVT_OP_REG2XMM(cvtsi2sd); |
2581 |
CVT_OP_XMM(cvtdq2ps); |
2582 |
CVT_OP_XMM(cvtdq2pd); |
2583 |
|
2584 |
/* XXX: test PNI insns */
|
2585 |
#if 0
|
2586 |
SSE_OP2(movshdup);
|
2587 |
#endif
|
2588 |
asm volatile ("emms"); |
2589 |
} |
2590 |
|
2591 |
#endif
|
2592 |
|
2593 |
#define TEST_CONV_RAX(op)\
|
2594 |
{\ |
2595 |
unsigned long a, r;\ |
2596 |
a = i2l(0x8234a6f8);\
|
2597 |
r = a;\ |
2598 |
asm volatile(#op : "=a" (r) : "0" (r));\ |
2599 |
printf("%-10s A=" FMTLX " R=" FMTLX "\n", #op, a, r);\ |
2600 |
} |
2601 |
|
2602 |
#define TEST_CONV_RAX_RDX(op)\
|
2603 |
{\ |
2604 |
unsigned long a, d, r, rh; \ |
2605 |
a = i2l(0x8234a6f8);\
|
2606 |
d = i2l(0x8345a1f2);\
|
2607 |
r = a;\ |
2608 |
rh = d;\ |
2609 |
asm volatile(#op : "=a" (r), "=d" (rh) : "0" (r), "1" (rh)); \ |
2610 |
printf("%-10s A=" FMTLX " R=" FMTLX ":" FMTLX "\n", #op, a, r, rh); \ |
2611 |
} |
2612 |
|
2613 |
void test_conv(void) |
2614 |
{ |
2615 |
TEST_CONV_RAX(cbw); |
2616 |
TEST_CONV_RAX(cwde); |
2617 |
#if defined(__x86_64__)
|
2618 |
TEST_CONV_RAX(cdqe); |
2619 |
#endif
|
2620 |
|
2621 |
TEST_CONV_RAX_RDX(cwd); |
2622 |
TEST_CONV_RAX_RDX(cdq); |
2623 |
#if defined(__x86_64__)
|
2624 |
TEST_CONV_RAX_RDX(cqo); |
2625 |
#endif
|
2626 |
|
2627 |
{ |
2628 |
unsigned long a, r; |
2629 |
a = i2l(0x12345678);
|
2630 |
asm volatile("bswapl %k0" : "=r" (r) : "0" (a)); |
2631 |
printf("%-10s: A=" FMTLX " R=" FMTLX "\n", "bswapl", a, r); |
2632 |
} |
2633 |
#if defined(__x86_64__)
|
2634 |
{ |
2635 |
unsigned long a, r; |
2636 |
a = i2l(0x12345678);
|
2637 |
asm volatile("bswapq %0" : "=r" (r) : "0" (a)); |
2638 |
printf("%-10s: A=" FMTLX " R=" FMTLX "\n", "bswapq", a, r); |
2639 |
} |
2640 |
#endif
|
2641 |
} |
2642 |
|
2643 |
extern void *__start_initcall; |
2644 |
extern void *__stop_initcall; |
2645 |
|
2646 |
|
2647 |
int main(int argc, char **argv) |
2648 |
{ |
2649 |
void **ptr;
|
2650 |
void (*func)(void); |
2651 |
|
2652 |
ptr = &__start_initcall; |
2653 |
while (ptr != &__stop_initcall) {
|
2654 |
func = *ptr++; |
2655 |
func(); |
2656 |
} |
2657 |
test_bsx(); |
2658 |
test_mul(); |
2659 |
test_jcc(); |
2660 |
test_floats(); |
2661 |
#if !defined(__x86_64__)
|
2662 |
test_bcd(); |
2663 |
#endif
|
2664 |
test_xchg(); |
2665 |
test_string(); |
2666 |
test_misc(); |
2667 |
test_lea(); |
2668 |
#ifdef TEST_SEGS
|
2669 |
test_segs(); |
2670 |
test_code16(); |
2671 |
#endif
|
2672 |
#ifdef TEST_VM86
|
2673 |
test_vm86(); |
2674 |
#endif
|
2675 |
#if !defined(__x86_64__)
|
2676 |
test_exceptions(); |
2677 |
test_self_modifying_code(); |
2678 |
test_single_step(); |
2679 |
#endif
|
2680 |
test_enter(); |
2681 |
test_conv(); |
2682 |
#ifdef TEST_SSE
|
2683 |
test_sse(); |
2684 |
test_fxsave(); |
2685 |
#endif
|
2686 |
return 0; |
2687 |
} |