4 * Copyright (c) 2003 Fabrice Bellard
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.
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.
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.
29 #include <sys/ucontext.h>
36 //#define LINUX_VM86_IOPL_FIX
37 //#define TEST_P4_FLAGS
40 #define xglue(x, y) x ## y
41 #define glue(x, y) xglue(x, y)
42 #define stringify(s) tostring(s)
43 #define tostring(s) #s
52 #define __init_call __attribute__ ((unused,__section__ (".initcall.init")))
54 static void *call_start __init_call = NULL;
56 #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
59 #include "test-i386.h"
62 #include "test-i386.h"
65 #include "test-i386.h"
68 #include "test-i386.h"
71 #include "test-i386.h"
74 #include "test-i386.h"
78 #include "test-i386.h"
82 #include "test-i386.h"
87 #include "test-i386.h"
92 #include "test-i386.h"
97 #include "test-i386.h"
102 #include "test-i386.h"
105 #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O)
108 #include "test-i386-shift.h"
111 #include "test-i386-shift.h"
114 #include "test-i386-shift.h"
117 #include "test-i386-shift.h"
120 #include "test-i386-shift.h"
124 #include "test-i386-shift.h"
128 #include "test-i386-shift.h"
133 #include "test-i386-shift.h"
138 #include "test-i386-shift.h"
140 /* XXX: should be more precise ? */
142 #define CC_MASK (CC_C)
146 #include "test-i386-shift.h"
150 #include "test-i386-shift.h"
154 #include "test-i386-shift.h"
158 #include "test-i386-shift.h"
160 /* lea test (modrm support) */
161 #define TEST_LEA(STR)\
163 asm("leal " STR ", %0"\
165 : "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\
166 printf("lea %s = %08x\n", STR, res);\
169 #define TEST_LEA16(STR)\
171 asm(".code16 ; .byte 0x67 ; leal " STR ", %0 ; .code32"\
173 : "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\
174 printf("lea %s = %08x\n", STR, res);\
180 int eax, ebx, ecx, edx, esi, edi, res;
197 TEST_LEA("0x40(%%eax)");
198 TEST_LEA("0x40(%%ebx)");
199 TEST_LEA("0x40(%%ecx)");
200 TEST_LEA("0x40(%%edx)");
201 TEST_LEA("0x40(%%esi)");
202 TEST_LEA("0x40(%%edi)");
204 TEST_LEA("0x4000(%%eax)");
205 TEST_LEA("0x4000(%%ebx)");
206 TEST_LEA("0x4000(%%ecx)");
207 TEST_LEA("0x4000(%%edx)");
208 TEST_LEA("0x4000(%%esi)");
209 TEST_LEA("0x4000(%%edi)");
211 TEST_LEA("(%%eax, %%ecx)");
212 TEST_LEA("(%%ebx, %%edx)");
213 TEST_LEA("(%%ecx, %%ecx)");
214 TEST_LEA("(%%edx, %%ecx)");
215 TEST_LEA("(%%esi, %%ecx)");
216 TEST_LEA("(%%edi, %%ecx)");
218 TEST_LEA("0x40(%%eax, %%ecx)");
219 TEST_LEA("0x4000(%%ebx, %%edx)");
221 TEST_LEA("(%%ecx, %%ecx, 2)");
222 TEST_LEA("(%%edx, %%ecx, 4)");
223 TEST_LEA("(%%esi, %%ecx, 8)");
225 TEST_LEA("(,%%eax, 2)");
226 TEST_LEA("(,%%ebx, 4)");
227 TEST_LEA("(,%%ecx, 8)");
229 TEST_LEA("0x40(,%%eax, 2)");
230 TEST_LEA("0x40(,%%ebx, 4)");
231 TEST_LEA("0x40(,%%ecx, 8)");
234 TEST_LEA("-10(%%ecx, %%ecx, 2)");
235 TEST_LEA("-10(%%edx, %%ecx, 4)");
236 TEST_LEA("-10(%%esi, %%ecx, 8)");
238 TEST_LEA("0x4000(%%ecx, %%ecx, 2)");
239 TEST_LEA("0x4000(%%edx, %%ecx, 4)");
240 TEST_LEA("0x4000(%%esi, %%ecx, 8)");
242 /* limited 16 bit addressing test */
243 TEST_LEA16("0x4000");
244 TEST_LEA16("(%%bx)");
245 TEST_LEA16("(%%si)");
246 TEST_LEA16("(%%di)");
247 TEST_LEA16("0x40(%%bx)");
248 TEST_LEA16("0x40(%%si)");
249 TEST_LEA16("0x40(%%di)");
250 TEST_LEA16("0x4000(%%bx)");
251 TEST_LEA16("0x4000(%%si)");
252 TEST_LEA16("(%%bx,%%si)");
253 TEST_LEA16("(%%bx,%%di)");
254 TEST_LEA16("0x40(%%bx,%%si)");
255 TEST_LEA16("0x40(%%bx,%%di)");
256 TEST_LEA16("0x4000(%%bx,%%si)");
257 TEST_LEA16("0x4000(%%bx,%%di)");
260 #define TEST_JCC(JCC, v1, v2)\
263 asm("movl $1, %0\n\t"\
269 : "r" (v1), "r" (v2));\
270 printf("%-10s %d\n", "j" JCC, res);\
272 asm("movl $0, %0\n\t"\
274 "set" JCC " %b0\n\t"\
276 : "r" (v1), "r" (v2));\
277 printf("%-10s %d\n", "set" JCC, res);\
279 asm("movl $0x12345678, %0\n\t"\
281 "cmov" JCC "l %3, %0\n\t"\
283 : "r" (v1), "r" (v2), "m" (1));\
284 printf("%-10s R=0x%08x\n", "cmov" JCC "l", res);\
285 asm("movl $0x12345678, %0\n\t"\
287 "cmov" JCC "w %w3, %w0\n\t"\
289 : "r" (v1), "r" (v2), "r" (1));\
290 printf("%-10s R=0x%08x\n", "cmov" JCC "w", res);\
294 /* various jump tests */
297 TEST_JCC("ne", 1, 1);
298 TEST_JCC("ne", 1, 0);
305 TEST_JCC("l", 1, -1);
307 TEST_JCC("le", 1, 1);
308 TEST_JCC("le", 1, 0);
309 TEST_JCC("le", 1, -1);
311 TEST_JCC("ge", 1, 1);
312 TEST_JCC("ge", 1, 0);
313 TEST_JCC("ge", -1, 1);
317 TEST_JCC("g", 1, -1);
321 TEST_JCC("b", 1, -1);
323 TEST_JCC("be", 1, 1);
324 TEST_JCC("be", 1, 0);
325 TEST_JCC("be", 1, -1);
327 TEST_JCC("ae", 1, 1);
328 TEST_JCC("ae", 1, 0);
329 TEST_JCC("ae", 1, -1);
333 TEST_JCC("a", 1, -1);
339 TEST_JCC("np", 1, 1);
340 TEST_JCC("np", 1, 0);
342 TEST_JCC("o", 0x7fffffff, 0);
343 TEST_JCC("o", 0x7fffffff, -1);
345 TEST_JCC("no", 0x7fffffff, 0);
346 TEST_JCC("no", 0x7fffffff, -1);
349 TEST_JCC("s", 0, -1);
352 TEST_JCC("ns", 0, 1);
353 TEST_JCC("ns", 0, -1);
354 TEST_JCC("ns", 0, 0);
359 #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
361 #define CC_MASK (CC_O | CC_C)
365 #include "test-i386-muldiv.h"
368 #include "test-i386-muldiv.h"
370 void test_imulw2(int op0, int op1)
372 int res, s1, s0, flags;
377 asm volatile ("push %4\n\t"
382 : "=q" (res), "=g" (flags)
383 : "q" (s1), "0" (res), "1" (flags));
384 printf("%-10s A=%08x B=%08x R=%08x CC=%04x\n",
385 "imulw", s0, s1, res, flags & CC_MASK);
388 void test_imull2(int op0, int op1)
390 int res, s1, s0, flags;
395 asm volatile ("push %4\n\t"
400 : "=q" (res), "=g" (flags)
401 : "q" (s1), "0" (res), "1" (flags));
402 printf("%-10s A=%08x B=%08x R=%08x CC=%04x\n",
403 "imull", s0, s1, res, flags & CC_MASK);
406 #define TEST_IMUL_IM(size, size1, op0, op1)\
411 asm volatile ("push %3\n\t"\
413 "imul" size " $" #op0 ", %" size1 "2, %" size1 "0\n\t" \
416 : "=r" (res), "=g" (flags)\
417 : "r" (op1), "1" (flags), "0" (res));\
418 printf("%-10s A=%08x B=%08x R=%08x CC=%04x\n",\
419 "imul" size " im", op0, op1, res, flags & CC_MASK);\
427 #include "test-i386-muldiv.h"
430 #include "test-i386-muldiv.h"
434 test_imulb(0x1234561d, 4);
436 test_imulb(0x80, 0x80);
437 test_imulb(0x10, 0x10);
439 test_imulw(0, 0x1234001d, 45);
440 test_imulw(0, 23, -45);
441 test_imulw(0, 0x8000, 0x8000);
442 test_imulw(0, 0x100, 0x100);
444 test_imull(0, 0x1234001d, 45);
445 test_imull(0, 23, -45);
446 test_imull(0, 0x80000000, 0x80000000);
447 test_imull(0, 0x10000, 0x10000);
449 test_mulb(0x1234561d, 4);
451 test_mulb(0x80, 0x80);
452 test_mulb(0x10, 0x10);
454 test_mulw(0, 0x1234001d, 45);
455 test_mulw(0, 23, -45);
456 test_mulw(0, 0x8000, 0x8000);
457 test_mulw(0, 0x100, 0x100);
459 test_mull(0, 0x1234001d, 45);
460 test_mull(0, 23, -45);
461 test_mull(0, 0x80000000, 0x80000000);
462 test_mull(0, 0x10000, 0x10000);
464 test_imulw2(0x1234001d, 45);
465 test_imulw2(23, -45);
466 test_imulw2(0x8000, 0x8000);
467 test_imulw2(0x100, 0x100);
469 test_imull2(0x1234001d, 45);
470 test_imull2(23, -45);
471 test_imull2(0x80000000, 0x80000000);
472 test_imull2(0x10000, 0x10000);
474 TEST_IMUL_IM("w", "w", 45, 0x1234);
475 TEST_IMUL_IM("w", "w", -45, 23);
476 TEST_IMUL_IM("w", "w", 0x8000, 0x80000000);
477 TEST_IMUL_IM("w", "w", 0x7fff, 0x1000);
479 TEST_IMUL_IM("l", "", 45, 0x1234);
480 TEST_IMUL_IM("l", "", -45, 23);
481 TEST_IMUL_IM("l", "", 0x8000, 0x80000000);
482 TEST_IMUL_IM("l", "", 0x7fff, 0x1000);
484 test_idivb(0x12341678, 0x127e);
485 test_idivb(0x43210123, -5);
486 test_idivb(0x12340004, -1);
488 test_idivw(0, 0x12345678, 12347);
489 test_idivw(0, -23223, -45);
490 test_idivw(0, 0x12348000, -1);
491 test_idivw(0x12343, 0x12345678, 0x81238567);
493 test_idivl(0, 0x12345678, 12347);
494 test_idivl(0, -233223, -45);
495 test_idivl(0, 0x80000000, -1);
496 test_idivl(0x12343, 0x12345678, 0x81234567);
498 test_divb(0x12341678, 0x127e);
499 test_divb(0x43210123, -5);
500 test_divb(0x12340004, -1);
502 test_divw(0, 0x12345678, 12347);
503 test_divw(0, -23223, -45);
504 test_divw(0, 0x12348000, -1);
505 test_divw(0x12343, 0x12345678, 0x81238567);
507 test_divl(0, 0x12345678, 12347);
508 test_divl(0, -233223, -45);
509 test_divl(0, 0x80000000, -1);
510 test_divl(0x12343, 0x12345678, 0x81234567);
513 #define TEST_BSX(op, size, op0)\
518 "movl $0x12345678, %0\n"\
519 #op " %" size "2, %" size "0 ; setz %b1" \
520 : "=r" (res), "=q" (resz)\
522 printf("%-10s A=%08x R=%08x %d\n", #op, val, res, resz);\
527 TEST_BSX(bsrw, "w", 0);
528 TEST_BSX(bsrw, "w", 0x12340128);
529 TEST_BSX(bsrl, "", 0);
530 TEST_BSX(bsrl, "", 0x00340128);
531 TEST_BSX(bsfw, "w", 0);
532 TEST_BSX(bsfw, "w", 0x12340128);
533 TEST_BSX(bsfl, "", 0);
534 TEST_BSX(bsfl, "", 0x00340128);
537 /**********************************************/
539 void test_fops(double a, double b)
541 printf("a=%f b=%f a+b=%f\n", a, b, a + b);
542 printf("a=%f b=%f a-b=%f\n", a, b, a - b);
543 printf("a=%f b=%f a*b=%f\n", a, b, a * b);
544 printf("a=%f b=%f a/b=%f\n", a, b, a / b);
545 printf("a=%f b=%f fmod(a, b)=%f\n", a, b, fmod(a, b));
546 printf("a=%f sqrt(a)=%f\n", a, sqrt(a));
547 printf("a=%f sin(a)=%f\n", a, sin(a));
548 printf("a=%f cos(a)=%f\n", a, cos(a));
549 printf("a=%f tan(a)=%f\n", a, tan(a));
550 printf("a=%f log(a)=%f\n", a, log(a));
551 printf("a=%f exp(a)=%f\n", a, exp(a));
552 printf("a=%f b=%f atan2(a, b)=%f\n", a, b, atan2(a, b));
553 /* just to test some op combining */
554 printf("a=%f asin(sin(a))=%f\n", a, asin(sin(a)));
555 printf("a=%f acos(cos(a))=%f\n", a, acos(cos(a)));
556 printf("a=%f atan(tan(a))=%f\n", a, atan(tan(a)));
560 void test_fcmp(double a, double b)
562 printf("(%f<%f)=%d\n",
564 printf("(%f<=%f)=%d\n",
566 printf("(%f==%f)=%d\n",
568 printf("(%f>%f)=%d\n",
570 printf("(%f<=%f)=%d\n",
574 /* test f(u)comi instruction */
580 printf("fcomi(%f %f)=%08x\n", a, b, eflags & (CC_Z | CC_P | CC_C));
584 void test_fcvt(double a)
597 printf("(float)%f = %f\n", a, fa);
598 printf("(long double)%f = %Lf\n", a, la);
599 printf("a=%016Lx\n", *(long long *)&a);
600 printf("la=%016Lx %04x\n", *(long long *)&la,
601 *(unsigned short *)((char *)(&la) + 8));
603 /* test all roundings */
604 asm volatile ("fstcw %0" : "=m" (fpuc));
606 asm volatile ("fldcw %0" : : "m" ((fpuc & ~0x0c00) | (i << 10)));
607 asm volatile ("fist %0" : "=m" (wa) : "t" (a));
608 asm volatile ("fistl %0" : "=m" (ia) : "t" (a));
609 asm volatile ("fistpll %0" : "=m" (lla) : "t" (a) : "st");
610 asm volatile ("frndint ; fstl %0" : "=m" (ra) : "t" (a));
611 asm volatile ("fldcw %0" : : "m" (fpuc));
612 printf("(short)a = %d\n", wa);
613 printf("(int)a = %d\n", ia);
614 printf("(int64_t)a = %Ld\n", lla);
615 printf("rint(a) = %f\n", ra);
620 asm("fld" #N : "=t" (a)); \
621 printf("fld" #N "= %f\n", a);
623 void test_fconst(void)
635 void test_fbcd(double a)
637 unsigned short bcd[5];
640 asm("fbstp %0" : "=m" (bcd[0]) : "t" (a) : "st");
641 asm("fbld %1" : "=t" (b) : "m" (bcd[0]));
642 printf("a=%f bcd=%04x%04x%04x%04x%04x b=%f\n",
643 a, bcd[4], bcd[3], bcd[2], bcd[1], bcd[0], b);
646 #define TEST_ENV(env, save, restore)\
648 memset((env), 0xaa, sizeof(*(env)));\
650 asm volatile ("fldl %0" : : "m" (dtab[i]));\
651 asm volatile (save " %0\n" : : "m" (*(env)));\
652 asm volatile (restore " %0\n": : "m" (*(env)));\
654 asm volatile ("fstpl %0" : "=m" (rtab[i]));\
656 printf("res[%d]=%f\n", i, rtab[i]);\
657 printf("fpuc=%04x fpus=%04x fptag=%04x\n",\
659 (env)->fpus & 0xff00,\
665 struct __attribute__((packed)) {
673 long double fpregs[8];
675 struct __attribute__((packed)) {
680 long double fpregs[8];
689 TEST_ENV(&float_env16, "data16 fnstenv", "data16 fldenv");
690 TEST_ENV(&float_env16, "data16 fnsave", "data16 frstor");
691 TEST_ENV(&float_env32, "fnstenv", "fldenv");
692 TEST_ENV(&float_env32, "fnsave", "frstor");
696 asm volatile ("fldl %0" : : "m" (dtab[i]));
697 asm volatile("ffree %st(2)");
698 asm volatile ("fnstenv %0\n" : : "m" (float_env32));
699 asm volatile ("fninit");
700 printf("fptag=%04x\n", float_env32.fptag);
704 #define TEST_FCMOV(a, b, eflags, CC)\
709 "fcmov" CC " %2, %0\n"\
711 : "0" (a), "u" (b), "g" (eflags));\
712 printf("fcmov%s eflags=0x%04x-> %f\n", \
716 void test_fcmov(void)
723 for(i = 0; i < 4; i++) {
729 TEST_FCMOV(a, b, eflags, "b");
730 TEST_FCMOV(a, b, eflags, "e");
731 TEST_FCMOV(a, b, eflags, "be");
732 TEST_FCMOV(a, b, eflags, "nb");
733 TEST_FCMOV(a, b, eflags, "ne");
734 TEST_FCMOV(a, b, eflags, "nbe");
736 TEST_FCMOV(a, b, 0, "u");
737 TEST_FCMOV(a, b, CC_P, "u");
738 TEST_FCMOV(a, b, 0, "nu");
739 TEST_FCMOV(a, b, CC_P, "nu");
742 void test_floats(void)
756 test_fbcd(1234567890123456);
757 test_fbcd(-123451234567890);
764 /**********************************************/
766 #define TEST_BCD(op, op0, cc_in, cc_mask)\
776 : "=a" (res), "=g" (flags)\
777 : "0" (res), "1" (flags));\
778 printf("%-10s A=%08x R=%08x CCIN=%04x CC=%04x\n",\
779 #op, op0, res, cc_in, flags & cc_mask);\
784 TEST_BCD(daa, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
785 TEST_BCD(daa, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
786 TEST_BCD(daa, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
787 TEST_BCD(daa, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
788 TEST_BCD(daa, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
789 TEST_BCD(daa, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
790 TEST_BCD(daa, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
791 TEST_BCD(daa, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
792 TEST_BCD(daa, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
793 TEST_BCD(daa, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
794 TEST_BCD(daa, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
795 TEST_BCD(daa, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
796 TEST_BCD(daa, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
798 TEST_BCD(das, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
799 TEST_BCD(das, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
800 TEST_BCD(das, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
801 TEST_BCD(das, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
802 TEST_BCD(das, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
803 TEST_BCD(das, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
804 TEST_BCD(das, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
805 TEST_BCD(das, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
806 TEST_BCD(das, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
807 TEST_BCD(das, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
808 TEST_BCD(das, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
809 TEST_BCD(das, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
810 TEST_BCD(das, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
812 TEST_BCD(aaa, 0x12340205, CC_A, (CC_C | CC_A));
813 TEST_BCD(aaa, 0x12340306, CC_A, (CC_C | CC_A));
814 TEST_BCD(aaa, 0x1234040a, CC_A, (CC_C | CC_A));
815 TEST_BCD(aaa, 0x123405fa, CC_A, (CC_C | CC_A));
816 TEST_BCD(aaa, 0x12340205, 0, (CC_C | CC_A));
817 TEST_BCD(aaa, 0x12340306, 0, (CC_C | CC_A));
818 TEST_BCD(aaa, 0x1234040a, 0, (CC_C | CC_A));
819 TEST_BCD(aaa, 0x123405fa, 0, (CC_C | CC_A));
821 TEST_BCD(aas, 0x12340205, CC_A, (CC_C | CC_A));
822 TEST_BCD(aas, 0x12340306, CC_A, (CC_C | CC_A));
823 TEST_BCD(aas, 0x1234040a, CC_A, (CC_C | CC_A));
824 TEST_BCD(aas, 0x123405fa, CC_A, (CC_C | CC_A));
825 TEST_BCD(aas, 0x12340205, 0, (CC_C | CC_A));
826 TEST_BCD(aas, 0x12340306, 0, (CC_C | CC_A));
827 TEST_BCD(aas, 0x1234040a, 0, (CC_C | CC_A));
828 TEST_BCD(aas, 0x123405fa, 0, (CC_C | CC_A));
830 TEST_BCD(aam, 0x12340547, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
831 TEST_BCD(aad, 0x12340407, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
834 #define TEST_XCHG(op, size, opconst)\
839 asm(#op " %" size "0, %" size "1" \
840 : "=q" (op0), opconst (op1) \
841 : "0" (op0), "1" (op1));\
842 printf("%-10s A=%08x B=%08x\n",\
846 #define TEST_CMPXCHG(op, size, opconst, eax)\
851 asm(#op " %" size "0, %" size "1" \
852 : "=q" (op0), opconst (op1) \
853 : "0" (op0), "1" (op1), "a" (eax));\
854 printf("%-10s EAX=%08x A=%08x C=%08x\n",\
855 #op, eax, op0, op1);\
860 TEST_XCHG(xchgl, "", "=q");
861 TEST_XCHG(xchgw, "w", "=q");
862 TEST_XCHG(xchgb, "b", "=q");
864 TEST_XCHG(xchgl, "", "=m");
865 TEST_XCHG(xchgw, "w", "=m");
866 TEST_XCHG(xchgb, "b", "=m");
868 TEST_XCHG(xaddl, "", "=q");
869 TEST_XCHG(xaddw, "w", "=q");
870 TEST_XCHG(xaddb, "b", "=q");
875 asm("xaddl %1, %0" : "=r" (res) : "0" (res));
876 printf("xaddl same res=%08x\n", res);
879 TEST_XCHG(xaddl, "", "=m");
880 TEST_XCHG(xaddw, "w", "=m");
881 TEST_XCHG(xaddb, "b", "=m");
883 TEST_CMPXCHG(cmpxchgl, "", "=q", 0xfbca7654);
884 TEST_CMPXCHG(cmpxchgw, "w", "=q", 0xfbca7654);
885 TEST_CMPXCHG(cmpxchgb, "b", "=q", 0xfbca7654);
887 TEST_CMPXCHG(cmpxchgl, "", "=q", 0xfffefdfc);
888 TEST_CMPXCHG(cmpxchgw, "w", "=q", 0xfffefdfc);
889 TEST_CMPXCHG(cmpxchgb, "b", "=q", 0xfffefdfc);
891 TEST_CMPXCHG(cmpxchgl, "", "=m", 0xfbca7654);
892 TEST_CMPXCHG(cmpxchgw, "w", "=m", 0xfbca7654);
893 TEST_CMPXCHG(cmpxchgb, "b", "=m", 0xfbca7654);
895 TEST_CMPXCHG(cmpxchgl, "", "=m", 0xfffefdfc);
896 TEST_CMPXCHG(cmpxchgw, "w", "=m", 0xfffefdfc);
897 TEST_CMPXCHG(cmpxchgb, "b", "=m", 0xfffefdfc);
900 uint64_t op0, op1, op2;
903 for(i = 0; i < 2; i++) {
904 op0 = 0x123456789abcd;
906 op1 = 0xfbca765423456;
909 op2 = 0x6532432432434;
913 : "=A" (op0), "=m" (op1), "=g" (eflags)
914 : "0" (op0), "m" (op1), "b" ((int)op2), "c" ((int)(op2 >> 32)));
915 printf("cmpxchg8b: op0=%016llx op1=%016llx CC=%02x\n",
916 op0, op1, eflags & CC_Z);
921 /**********************************************/
922 /* segmentation tests */
925 #include <linux/unistd.h>
926 #include <linux/version.h>
928 _syscall3(int, modify_ldt, int, func, void *, ptr, unsigned long, bytecount)
930 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 5, 66)
931 #define modify_ldt_ldt_s user_desc
934 uint8_t seg_data1[4096];
935 uint8_t seg_data2[4096];
937 #define MK_SEL(n) (((n) << 3) | 7)
939 #define TEST_LR(op, size, seg, mask)\
943 asm (op " %" size "2, %" size "0\n" \
948 : "=r" (res), "=r" (res2) : "m" (seg), "0" (res));\
949 printf(op ": Z=%d %08x\n", res2, res & ~(mask));\
952 /* NOTE: we use Linux modify_ldt syscall */
955 struct modify_ldt_ldt_s ldt;
956 long long ldt_table[3];
962 } __attribute__((packed)) segoff;
964 ldt.entry_number = 1;
965 ldt.base_addr = (unsigned long)&seg_data1;
966 ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12;
968 ldt.contents = MODIFY_LDT_CONTENTS_DATA;
969 ldt.read_exec_only = 0;
970 ldt.limit_in_pages = 1;
971 ldt.seg_not_present = 0;
973 modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
975 ldt.entry_number = 2;
976 ldt.base_addr = (unsigned long)&seg_data2;
977 ldt.limit = (sizeof(seg_data2) + 0xfff) >> 12;
979 ldt.contents = MODIFY_LDT_CONTENTS_DATA;
980 ldt.read_exec_only = 0;
981 ldt.limit_in_pages = 1;
982 ldt.seg_not_present = 0;
984 modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
986 modify_ldt(0, &ldt_table, sizeof(ldt_table)); /* read ldt entries */
991 printf("%d: %016Lx\n", i, ldt_table[i]);
994 /* do some tests with fs or gs */
995 asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1)));
1000 asm volatile ("fs movzbl 0x1, %0" : "=r" (res));
1001 printf("FS[1] = %02x\n", res);
1003 asm volatile ("pushl %%gs\n"
1005 "gs movzbl 0x1, %0\n"
1009 printf("GS[1] = %02x\n", res);
1011 /* tests with ds/ss (implicit segment case) */
1013 asm volatile ("pushl %%ebp\n\t"
1016 "movl %3, %%ebp\n\t"
1017 "movzbl 0x1, %0\n\t"
1018 "movzbl (%%ebp), %1\n\t"
1021 : "=r" (res), "=r" (res2)
1022 : "r" (MK_SEL(1)), "r" (&tmp));
1023 printf("DS[1] = %02x\n", res);
1024 printf("SS[tmp] = %02x\n", res2);
1026 segoff.seg = MK_SEL(2);
1027 segoff.offset = 0xabcdef12;
1028 asm volatile("lfs %2, %0\n\t"
1030 : "=r" (res), "=g" (res2)
1032 printf("FS:reg = %04x:%08x\n", res2, res);
1034 TEST_LR("larw", "w", MK_SEL(2), 0x0100);
1035 TEST_LR("larl", "", MK_SEL(2), 0x0100);
1036 TEST_LR("lslw", "w", MK_SEL(2), 0);
1037 TEST_LR("lsll", "", MK_SEL(2), 0);
1039 TEST_LR("larw", "w", 0xfff8, 0);
1040 TEST_LR("larl", "", 0xfff8, 0);
1041 TEST_LR("lslw", "w", 0xfff8, 0);
1042 TEST_LR("lsll", "", 0xfff8, 0);
1045 /* 16 bit code test */
1046 extern char code16_start, code16_end;
1047 extern char code16_func1;
1048 extern char code16_func2;
1049 extern char code16_func3;
1051 void test_code16(void)
1053 struct modify_ldt_ldt_s ldt;
1056 /* build a code segment */
1057 ldt.entry_number = 1;
1058 ldt.base_addr = (unsigned long)&code16_start;
1059 ldt.limit = &code16_end - &code16_start;
1061 ldt.contents = MODIFY_LDT_CONTENTS_CODE;
1062 ldt.read_exec_only = 0;
1063 ldt.limit_in_pages = 0;
1064 ldt.seg_not_present = 0;
1066 modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1068 /* call the first function */
1069 asm volatile ("lcall %1, %2"
1071 : "i" (MK_SEL(1)), "i" (&code16_func1): "memory", "cc");
1072 printf("func1() = 0x%08x\n", res);
1073 asm volatile ("lcall %2, %3"
1074 : "=a" (res), "=c" (res2)
1075 : "i" (MK_SEL(1)), "i" (&code16_func2): "memory", "cc");
1076 printf("func2() = 0x%08x spdec=%d\n", res, res2);
1077 asm volatile ("lcall %1, %2"
1079 : "i" (MK_SEL(1)), "i" (&code16_func3): "memory", "cc");
1080 printf("func3() = 0x%08x\n", res);
1083 extern char func_lret32;
1084 extern char func_iret32;
1086 void test_misc(void)
1091 for(i=0;i<256;i++) table[i] = 256 - i;
1093 asm ("xlat" : "=a" (res) : "b" (table), "0" (res));
1094 printf("xlat: EAX=%08x\n", res);
1096 asm volatile ("pushl %%cs ; call %1"
1098 : "m" (func_lret32): "memory", "cc");
1099 printf("func_lret32=%x\n", res);
1101 asm volatile ("pushfl ; pushl %%cs ; call %1"
1103 : "m" (func_iret32): "memory", "cc");
1104 printf("func_iret32=%x\n", res);
1106 /* specific popl test */
1107 asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popl (%%esp) ; popl %0"
1109 printf("popl esp=%x\n", res);
1111 /* specific popw test */
1112 asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popw (%%esp) ; addl $2, %%esp ; popl %0"
1114 printf("popw esp=%x\n", res);
1117 uint8_t str_buffer[4096];
1119 #define TEST_STRING1(OP, size, DF, REP)\
1121 int esi, edi, eax, ecx, eflags;\
1123 esi = (long)(str_buffer + sizeof(str_buffer) / 2);\
1124 edi = (long)(str_buffer + sizeof(str_buffer) / 2) + 16;\
1128 asm volatile ("pushl $0\n\t"\
1131 REP #OP size "\n\t"\
1135 : "=S" (esi), "=D" (edi), "=a" (eax), "=c" (ecx), "=g" (eflags)\
1136 : "0" (esi), "1" (edi), "2" (eax), "3" (ecx));\
1137 printf("%-10s ESI=%08x EDI=%08x EAX=%08x ECX=%08x EFL=%04x\n",\
1138 REP #OP size, esi, edi, eax, ecx,\
1139 eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));\
1142 #define TEST_STRING(OP, REP)\
1143 TEST_STRING1(OP, "b", "", REP);\
1144 TEST_STRING1(OP, "w", "", REP);\
1145 TEST_STRING1(OP, "l", "", REP);\
1146 TEST_STRING1(OP, "b", "std", REP);\
1147 TEST_STRING1(OP, "w", "std", REP);\
1148 TEST_STRING1(OP, "l", "std", REP)
1150 void test_string(void)
1153 for(i = 0;i < sizeof(str_buffer); i++)
1154 str_buffer[i] = i + 0x56;
1155 TEST_STRING(stos, "");
1156 TEST_STRING(stos, "rep ");
1157 TEST_STRING(lods, ""); /* to verify stos */
1158 TEST_STRING(lods, "rep ");
1159 TEST_STRING(movs, "");
1160 TEST_STRING(movs, "rep ");
1161 TEST_STRING(lods, ""); /* to verify stos */
1163 /* XXX: better tests */
1164 TEST_STRING(scas, "");
1165 TEST_STRING(scas, "repz ");
1166 TEST_STRING(scas, "repnz ");
1167 TEST_STRING(cmps, "");
1168 TEST_STRING(cmps, "repz ");
1169 TEST_STRING(cmps, "repnz ");
1174 static inline void set_bit(uint8_t *a, unsigned int bit)
1176 a[bit / 8] |= (1 << (bit % 8));
1179 static inline uint8_t *seg_to_linear(unsigned int seg, unsigned int reg)
1181 return (uint8_t *)((seg << 4) + (reg & 0xffff));
1184 static inline void pushw(struct vm86_regs *r, int val)
1186 r->esp = (r->esp & ~0xffff) | ((r->esp - 2) & 0xffff);
1187 *(uint16_t *)seg_to_linear(r->ss, r->esp) = val;
1190 #undef __syscall_return
1191 #define __syscall_return(type, res) \
1193 return (type) (res); \
1196 _syscall2(int, vm86, int, func, struct vm86plus_struct *, v86)
1198 extern char vm86_code_start;
1199 extern char vm86_code_end;
1201 #define VM86_CODE_CS 0x100
1202 #define VM86_CODE_IP 0x100
1204 void test_vm86(void)
1206 struct vm86plus_struct ctx;
1207 struct vm86_regs *r;
1211 vm86_mem = mmap((void *)0x00000000, 0x110000,
1212 PROT_WRITE | PROT_READ | PROT_EXEC,
1213 MAP_FIXED | MAP_ANON | MAP_PRIVATE, -1, 0);
1214 if (vm86_mem == MAP_FAILED) {
1215 printf("ERROR: could not map vm86 memory");
1218 memset(&ctx, 0, sizeof(ctx));
1220 /* init basic registers */
1222 r->eip = VM86_CODE_IP;
1231 r->eflags = VIF_MASK;
1233 /* move code to proper address. We use the same layout as a .com
1235 memcpy(vm86_mem + (VM86_CODE_CS << 4) + VM86_CODE_IP,
1236 &vm86_code_start, &vm86_code_end - &vm86_code_start);
1238 /* mark int 0x21 as being emulated */
1239 set_bit((uint8_t *)&ctx.int_revectored, 0x21);
1242 ret = vm86(VM86_ENTER, &ctx);
1243 switch(VM86_TYPE(ret)) {
1248 int_num = VM86_ARG(ret);
1249 if (int_num != 0x21)
1251 ah = (r->eax >> 8) & 0xff;
1253 case 0x00: /* exit */
1255 case 0x02: /* write char */
1261 case 0x09: /* write string */
1264 ptr = seg_to_linear(r->ds, r->edx);
1271 r->eax = (r->eax & ~0xff) | '$';
1274 case 0xff: /* extension: write eflags number in edx */
1276 #ifndef LINUX_VM86_IOPL_FIX
1279 printf("%08x\n", v);
1283 printf("unsupported int 0x%02x\n", int_num);
1289 /* a signal came, we just ignore that */
1294 printf("ERROR: unhandled vm86 return code (0x%x)\n", ret);
1299 printf("VM86 end\n");
1300 munmap(vm86_mem, 0x110000);
1303 /* exception tests */
1315 #define REG_TRAPNO TRAPNO
1323 void sig_handler(int sig, siginfo_t *info, void *puc)
1325 struct ucontext *uc = puc;
1327 printf("si_signo=%d si_errno=%d si_code=%d",
1328 info->si_signo, info->si_errno, info->si_code);
1329 printf(" si_addr=0x%08lx",
1330 (unsigned long)info->si_addr);
1333 printf("trapno=0x%02x err=0x%08x",
1334 uc->uc_mcontext.gregs[REG_TRAPNO],
1335 uc->uc_mcontext.gregs[REG_ERR]);
1336 printf(" EIP=0x%08x", uc->uc_mcontext.gregs[REG_EIP]);
1338 longjmp(jmp_env, 1);
1341 void test_exceptions(void)
1343 struct modify_ldt_ldt_s ldt;
1344 struct sigaction act;
1347 act.sa_sigaction = sig_handler;
1348 sigemptyset(&act.sa_mask);
1349 act.sa_flags = SA_SIGINFO;
1350 sigaction(SIGFPE, &act, NULL);
1351 sigaction(SIGILL, &act, NULL);
1352 sigaction(SIGSEGV, &act, NULL);
1353 sigaction(SIGBUS, &act, NULL);
1354 sigaction(SIGTRAP, &act, NULL);
1356 /* test division by zero reporting */
1357 printf("DIVZ exception:\n");
1358 if (setjmp(jmp_env) == 0) {
1359 /* now divide by zero */
1364 printf("BOUND exception:\n");
1365 if (setjmp(jmp_env) == 0) {
1366 /* bound exception */
1369 asm volatile ("bound %0, %1" : : "r" (11), "m" (tab[0]));
1372 printf("segment exceptions:\n");
1373 if (setjmp(jmp_env) == 0) {
1374 /* load an invalid segment */
1375 asm volatile ("movl %0, %%fs" : : "r" ((0x1234 << 3) | 1));
1377 if (setjmp(jmp_env) == 0) {
1378 /* null data segment is valid */
1379 asm volatile ("movl %0, %%fs" : : "r" (3));
1380 /* null stack segment */
1381 asm volatile ("movl %0, %%ss" : : "r" (3));
1384 ldt.entry_number = 1;
1385 ldt.base_addr = (unsigned long)&seg_data1;
1386 ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12;
1388 ldt.contents = MODIFY_LDT_CONTENTS_DATA;
1389 ldt.read_exec_only = 0;
1390 ldt.limit_in_pages = 1;
1391 ldt.seg_not_present = 1;
1393 modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1395 if (setjmp(jmp_env) == 0) {
1396 /* segment not present */
1397 asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1)));
1400 /* test SEGV reporting */
1401 printf("PF exception:\n");
1402 if (setjmp(jmp_env) == 0) {
1404 /* we add a nop to test a weird PC retrieval case */
1405 asm volatile ("nop");
1406 /* now store in an invalid address */
1407 *(char *)0x1234 = 1;
1410 /* test SEGV reporting */
1411 printf("PF exception:\n");
1412 if (setjmp(jmp_env) == 0) {
1414 /* read from an invalid address */
1415 v1 = *(char *)0x1234;
1418 /* test illegal instruction reporting */
1419 printf("UD2 exception:\n");
1420 if (setjmp(jmp_env) == 0) {
1421 /* now execute an invalid instruction */
1422 asm volatile("ud2");
1424 printf("lock nop exception:\n");
1425 if (setjmp(jmp_env) == 0) {
1426 /* now execute an invalid instruction */
1427 asm volatile("lock nop");
1430 printf("INT exception:\n");
1431 if (setjmp(jmp_env) == 0) {
1432 asm volatile ("int $0xfd");
1434 if (setjmp(jmp_env) == 0) {
1435 asm volatile ("int $0x01");
1437 if (setjmp(jmp_env) == 0) {
1438 asm volatile (".byte 0xcd, 0x03");
1440 if (setjmp(jmp_env) == 0) {
1441 asm volatile ("int $0x04");
1443 if (setjmp(jmp_env) == 0) {
1444 asm volatile ("int $0x05");
1447 printf("INT3 exception:\n");
1448 if (setjmp(jmp_env) == 0) {
1449 asm volatile ("int3");
1452 printf("CLI exception:\n");
1453 if (setjmp(jmp_env) == 0) {
1454 asm volatile ("cli");
1457 printf("STI exception:\n");
1458 if (setjmp(jmp_env) == 0) {
1459 asm volatile ("cli");
1462 printf("INTO exception:\n");
1463 if (setjmp(jmp_env) == 0) {
1464 /* overflow exception */
1465 asm volatile ("addl $1, %0 ; into" : : "r" (0x7fffffff));
1468 printf("OUTB exception:\n");
1469 if (setjmp(jmp_env) == 0) {
1470 asm volatile ("outb %%al, %%dx" : : "d" (0x4321), "a" (0));
1473 printf("INB exception:\n");
1474 if (setjmp(jmp_env) == 0) {
1475 asm volatile ("inb %%dx, %%al" : "=a" (val) : "d" (0x4321));
1478 printf("REP OUTSB exception:\n");
1479 if (setjmp(jmp_env) == 0) {
1480 asm volatile ("rep outsb" : : "d" (0x4321), "S" (tab), "c" (1));
1483 printf("REP INSB exception:\n");
1484 if (setjmp(jmp_env) == 0) {
1485 asm volatile ("rep insb" : : "d" (0x4321), "D" (tab), "c" (1));
1488 printf("HLT exception:\n");
1489 if (setjmp(jmp_env) == 0) {
1490 asm volatile ("hlt");
1493 printf("single step exception:\n");
1495 if (setjmp(jmp_env) == 0) {
1496 asm volatile ("pushf\n"
1497 "orl $0x00100, (%%esp)\n"
1499 "movl $0xabcd, %0\n"
1500 "movl $0x0, %0\n" : "=m" (val) : : "cc", "memory");
1502 printf("val=0x%x\n", val);
1505 /* specific precise single step test */
1506 void sig_trap_handler(int sig, siginfo_t *info, void *puc)
1508 struct ucontext *uc = puc;
1509 printf("EIP=0x%08x\n", uc->uc_mcontext.gregs[REG_EIP]);
1512 const uint8_t sstep_buf1[4] = { 1, 2, 3, 4};
1513 uint8_t sstep_buf2[4];
1515 void test_single_step(void)
1517 struct sigaction act;
1522 act.sa_sigaction = sig_trap_handler;
1523 sigemptyset(&act.sa_mask);
1524 act.sa_flags = SA_SIGINFO;
1525 sigaction(SIGTRAP, &act, NULL);
1526 asm volatile ("pushf\n"
1527 "orl $0x00100, (%%esp)\n"
1529 "movl $0xabcd, %0\n"
1538 /* movsb: the single step should stop at each movsb iteration */
1539 "movl $sstep_buf1, %%esi\n"
1540 "movl $sstep_buf2, %%edi\n"
1548 /* cmpsb: the single step should stop at each cmpsb iteration */
1549 "movl $sstep_buf1, %%esi\n"
1550 "movl $sstep_buf2, %%edi\n"
1556 /* getpid() syscall: single step should skip one
1562 /* when modifying SS, trace is not done on the next
1564 "movl %%ss, %%ecx\n"
1565 "movl %%ecx, %%ss\n"
1568 "movl %%ecx, %%ss\n"
1579 "andl $~0x00100, (%%esp)\n"
1583 : "cc", "memory", "eax", "ecx", "esi", "edi");
1584 printf("val=%d\n", val);
1585 for(i = 0; i < 4; i++)
1586 printf("sstep_buf2[%d] = %d\n", i, sstep_buf2[i]);
1589 /* self modifying code test */
1591 0xb8, 0x1, 0x00, 0x00, 0x00, /* movl $1, %eax */
1596 "movl 4(%esp), %eax\n"
1597 "movl %eax, smc_patch_addr2 + 1\n"
1606 "smc_patch_addr2:\n"
1610 typedef int FuncType(void);
1611 extern int smc_code2(int);
1612 void test_self_modifying_code(void)
1616 printf("self modifying code:\n");
1617 printf("func1 = 0x%x\n", ((FuncType *)code)());
1618 for(i = 2; i <= 4; i++) {
1620 printf("func%d = 0x%x\n", i, ((FuncType *)code)());
1623 /* more difficult test : the modified code is just after the
1624 modifying instruction. It is forbidden in Intel specs, but it
1625 is used by old DOS programs */
1626 for(i = 2; i <= 4; i++) {
1627 printf("smc_code2(%d) = %d\n", i, smc_code2(i));
1631 int enter_stack[4096];
1633 #define TEST_ENTER(size, stack_type, level)\
1635 int esp_save, esp_val, ebp_val, ebp_save, i;\
1636 stack_type *ptr, *stack_end, *stack_ptr;\
1637 memset(enter_stack, 0, sizeof(enter_stack));\
1638 stack_end = stack_ptr = (stack_type *)(enter_stack + 4096);\
1639 ebp_val = (long)stack_ptr;\
1642 esp_val = (long)stack_ptr;\
1643 asm("movl %%esp, %[esp_save]\n"\
1644 "movl %%ebp, %[ebp_save]\n"\
1645 "movl %[esp_val], %%esp\n"\
1646 "movl %[ebp_val], %%ebp\n"\
1647 "enter" size " $12, $" #level "\n"\
1648 "movl %%esp, %[esp_val]\n"\
1649 "movl %%ebp, %[ebp_val]\n"\
1650 "movl %[esp_save], %%esp\n"\
1651 "movl %[ebp_save], %%ebp\n"\
1652 : [esp_save] "=r" (esp_save),\
1653 [ebp_save] "=r" (ebp_save),\
1654 [esp_val] "=r" (esp_val),\
1655 [ebp_val] "=r" (ebp_val)\
1656 : "[esp_val]" (esp_val),\
1657 "[ebp_val]" (ebp_val));\
1658 printf("level=%d:\n", level);\
1659 printf("esp_val=0x%08lx\n", esp_val - (long)stack_end);\
1660 printf("ebp_val=0x%08lx\n", ebp_val - (long)stack_end);\
1661 for(ptr = (stack_type *)esp_val; ptr < stack_end; ptr++)\
1662 printf("%08x\n", ptr[0]);\
1665 static void test_enter(void)
1667 TEST_ENTER("l", uint32_t, 0);
1668 TEST_ENTER("l", uint32_t, 1);
1669 TEST_ENTER("l", uint32_t, 2);
1670 TEST_ENTER("l", uint32_t, 31);
1672 TEST_ENTER("w", uint16_t, 0);
1673 TEST_ENTER("w", uint16_t, 1);
1674 TEST_ENTER("w", uint16_t, 2);
1675 TEST_ENTER("w", uint16_t, 31);
1680 typedef int __m64 __attribute__ ((__mode__ (__V2SI__)));
1681 typedef int __m128 __attribute__ ((__mode__(__V4SF__)));
1691 static uint64_t __attribute__((aligned(16))) test_values[4][2] = {
1692 { 0x456723c698694873, 0xdc515cff944a58ec },
1693 { 0x1f297ccd58bad7ab, 0x41f21efba9e3e146 },
1694 { 0x007c62c2085427f8, 0x231be9e8cde7438d },
1695 { 0x0f76255a085427f8, 0xc233e9e8c4c9439a },
1700 asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\
1701 printf("%-9s: a=%016llx%016llx b=%016llx%016llx r=%016llx%016llx\n",\
1708 #define SSE_OP2(op)\
1712 a.q[0] = test_values[2*i][0];\
1713 a.q[1] = test_values[2*i][1];\
1714 b.q[0] = test_values[2*i+1][0];\
1715 b.q[1] = test_values[2*i+1][1];\
1720 #define MMX_OP2(op)\
1724 a.q[0] = test_values[2*i][0];\
1725 b.q[0] = test_values[2*i+1][0];\
1726 asm volatile (#op " %2, %0" : "=y" (r.q[0]) : "0" (a.q[0]), "y" (b.q[0]));\
1727 printf("%-9s: a=%016llx b=%016llx r=%016llx\n",\
1736 #define PSHUF_OP(op, ib)\
1740 a.q[0] = test_values[2*i][0];\
1741 a.q[1] = test_values[2*i][1];\
1742 asm volatile (#op " $" #ib ", %1, %0" : "=x" (r.dq) : "x" (a.dq));\
1743 printf("%-9s: a=%016llx%016llx ib=%02x r=%016llx%016llx\n",\
1751 #define SHIFT_IM(op, ib)\
1755 a.q[0] = test_values[2*i][0];\
1756 a.q[1] = test_values[2*i][1];\
1757 asm volatile (#op " $" #ib ", %0" : "=x" (r.dq) : "0" (a.dq));\
1758 printf("%-9s: a=%016llx%016llx ib=%02x r=%016llx%016llx\n",\
1766 #define SHIFT_OP(op, ib)\
1771 a.q[0] = test_values[2*i][0];\
1772 a.q[1] = test_values[2*i][1];\
1775 asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\
1776 printf("%-9s: a=%016llx%016llx b=%016llx%016llx r=%016llx%016llx\n",\
1788 a.q[0] = test_values[2*i][0];\
1789 a.q[1] = test_values[2*i][1];\
1790 asm volatile (#op " %1, %0" : "=r" (reg) : "x" (a.dq));\
1791 printf("%-9s: a=%016llx%016llx r=%08x\n",\
1798 #define SSE_OPS(a) \
1802 #define SSE_OPD(a) \
1806 #define SSE_COMI(op, field)\
1808 unsigned int eflags;\
1812 asm volatile (#op " %2, %1\n"\
1816 : "x" (a.dq), "x" (b.dq));\
1817 printf("%-9s: a=%f b=%f cc=%04x\n",\
1819 eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));\
1822 void test_sse_comi(double a1, double b1)
1824 SSE_COMI(ucomiss, s);
1825 SSE_COMI(ucomisd, d);
1826 SSE_COMI(comiss, s);
1827 SSE_COMI(comisd, d);
1830 #define CVT_OP_XMM(op)\
1832 asm volatile (#op " %1, %0" : "=x" (r.dq) : "x" (a.dq));\
1833 printf("%-9s: a=%016llx%016llx r=%016llx%016llx\n",\
1839 #define CVT_OP_XMM2MMX(op)\
1841 asm volatile (#op " %1, %0" : "=y" (r.q[0]) : "x" (a.dq));\
1842 printf("%-9s: a=%016llx%016llx r=%016llx\n",\
1848 #define CVT_OP_MMX2XMM(op)\
1850 asm volatile (#op " %1, %0" : "=x" (r.dq) : "y" (a.q[0]));\
1851 printf("%-9s: a=%016llx r=%016llx%016llx\n",\
1857 #define CVT_OP_REG2XMM(op)\
1859 asm volatile (#op " %1, %0" : "=x" (r.dq) : "r" (a.l[0]));\
1860 printf("%-9s: a=%08x r=%016llx%016llx\n",\
1866 #define CVT_OP_XMM2REG(op)\
1868 asm volatile (#op " %1, %0" : "=r" (r.l[0]) : "x" (a.dq));\
1869 printf("%-9s: a=%016llx%016llx r=%08x\n",\
1931 asm volatile ("pinsrw $1, %1, %0" : "=y" (r.q[0]) : "r" (0x12345678));
1932 printf("%-9s: r=%016llx\n", "pinsrw", r.q[0]);
1934 asm volatile ("pinsrw $5, %1, %0" : "=x" (r.dq) : "r" (0x12345678));
1935 printf("%-9s: r=%016llx%016llx\n", "pinsrw", r.q[1], r.q[0]);
1937 a.q[0] = test_values[0][0];
1938 a.q[1] = test_values[0][1];
1939 asm volatile ("pextrw $1, %1, %0" : "=r" (r.l[0]) : "y" (a.q[0]));
1940 printf("%-9s: r=%08x\n", "pextrw", r.l[0]);
1942 asm volatile ("pextrw $5, %1, %0" : "=r" (r.l[0]) : "x" (a.dq));
1943 printf("%-9s: r=%08x\n", "pextrw", r.l[0]);
1945 asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "y" (a.q[0]));
1946 printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]);
1948 asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "x" (a.dq));
1949 printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]);
1951 asm volatile ("emms");
1953 SSE_OP2(punpcklqdq);
1954 SSE_OP2(punpckhqdq);
1969 PSHUF_OP(shufps, 0x78);
1970 PSHUF_OP(shufpd, 0x02);
1972 PSHUF_OP(pshufd, 0x78);
1973 PSHUF_OP(pshuflw, 0x78);
1974 PSHUF_OP(pshufhw, 0x78);
1977 SHIFT_OP(psrlw, 16);
1979 SHIFT_OP(psraw, 16);
1981 SHIFT_OP(psllw, 16);
1984 SHIFT_OP(psrld, 32);
1986 SHIFT_OP(psrad, 32);
1988 SHIFT_OP(pslld, 32);
1991 SHIFT_OP(psrlq, 32);
1993 SHIFT_OP(psllq, 32);
1995 SHIFT_IM(psrldq, 16);
1996 SHIFT_IM(psrldq, 7);
1997 SHIFT_IM(pslldq, 16);
1998 SHIFT_IM(pslldq, 7);
2003 /* FPU specific ops */
2007 asm volatile("stmxcsr %0" : "=m" (mxcsr));
2008 printf("mxcsr=%08x\n", mxcsr & 0x1f80);
2009 asm volatile("ldmxcsr %0" : : "m" (mxcsr));
2012 test_sse_comi(2, -1);
2013 test_sse_comi(2, 2);
2014 test_sse_comi(2, 3);
2060 /* float to float/int */
2065 CVT_OP_XMM(cvtps2pd);
2066 CVT_OP_XMM(cvtss2sd);
2067 CVT_OP_XMM2MMX(cvtps2pi);
2068 CVT_OP_XMM2MMX(cvttps2pi);
2069 CVT_OP_XMM2REG(cvtss2si);
2070 CVT_OP_XMM2REG(cvttss2si);
2071 CVT_OP_XMM(cvtps2dq);
2072 CVT_OP_XMM(cvttps2dq);
2076 CVT_OP_XMM(cvtpd2ps);
2077 CVT_OP_XMM(cvtsd2ss);
2078 CVT_OP_XMM2MMX(cvtpd2pi);
2079 CVT_OP_XMM2MMX(cvttpd2pi);
2080 CVT_OP_XMM2REG(cvtsd2si);
2081 CVT_OP_XMM2REG(cvttsd2si);
2082 CVT_OP_XMM(cvtpd2dq);
2083 CVT_OP_XMM(cvttpd2dq);
2090 CVT_OP_MMX2XMM(cvtpi2ps);
2091 CVT_OP_MMX2XMM(cvtpi2pd);
2092 CVT_OP_REG2XMM(cvtsi2ss);
2093 CVT_OP_REG2XMM(cvtsi2sd);
2094 CVT_OP_XMM(cvtdq2ps);
2095 CVT_OP_XMM(cvtdq2pd);
2097 /* XXX: test PNI insns */
2105 static void *call_end __init_call = NULL;
2107 int main(int argc, char **argv)
2112 ptr = &call_start + 1;
2113 while (*ptr != NULL) {
2132 test_self_modifying_code();