![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to test-i386.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Qemu by Fabrice Bellard] / qemu / tests|
Return to test-i386.c CVS log | Up to [Qemu by Fabrice Bellard] / qemu / tests |
1.1 root 1: /*
2: * x86 CPU test
1.1.1.4 root 3: *
1.1 root 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
1.1.1.6 root 17: * along with this program; if not, see <http://www.gnu.org/licenses/>.
1.1 root 18: */
19: #define _GNU_SOURCE
20: #include <stdlib.h>
21: #include <stdio.h>
22: #include <string.h>
23: #include <inttypes.h>
24: #include <math.h>
25: #include <signal.h>
26: #include <setjmp.h>
27: #include <errno.h>
28: #include <sys/ucontext.h>
29: #include <sys/mman.h>
30:
31: #if !defined(__x86_64__)
1.1.1.5 root 32: //#define TEST_VM86
1.1 root 33: #define TEST_SEGS
34: #endif
35: //#define LINUX_VM86_IOPL_FIX
36: //#define TEST_P4_FLAGS
1.1.1.5 root 37: #ifdef __SSE__
1.1 root 38: #define TEST_SSE
39: #define TEST_CMOV 1
40: #define TEST_FCOMI 1
41: #else
1.1.1.5 root 42: #undef TEST_SSE
43: #define TEST_CMOV 1
44: #define TEST_FCOMI 1
1.1 root 45: #endif
46:
47: #if defined(__x86_64__)
48: #define FMT64X "%016lx"
49: #define FMTLX "%016lx"
50: #define X86_64_ONLY(x) x
51: #else
1.1.1.2 root 52: #define FMT64X "%016" PRIx64
1.1 root 53: #define FMTLX "%08lx"
54: #define X86_64_ONLY(x)
55: #endif
56:
57: #ifdef TEST_VM86
58: #include <asm/vm86.h>
59: #endif
60:
61: #define xglue(x, y) x ## y
62: #define glue(x, y) xglue(x, y)
63: #define stringify(s) tostring(s)
64: #define tostring(s) #s
65:
66: #define CC_C 0x0001
67: #define CC_P 0x0004
68: #define CC_A 0x0010
69: #define CC_Z 0x0040
70: #define CC_S 0x0080
71: #define CC_O 0x0800
72:
73: #define __init_call __attribute__ ((unused,__section__ ("initcall")))
74:
75: #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
76:
77: #if defined(__x86_64__)
78: static inline long i2l(long v)
79: {
80: return v | ((v ^ 0xabcd) << 32);
81: }
82: #else
83: static inline long i2l(long v)
84: {
85: return v;
86: }
87: #endif
88:
89: #define OP add
90: #include "test-i386.h"
91:
92: #define OP sub
93: #include "test-i386.h"
94:
95: #define OP xor
96: #include "test-i386.h"
97:
98: #define OP and
99: #include "test-i386.h"
100:
101: #define OP or
102: #include "test-i386.h"
103:
104: #define OP cmp
105: #include "test-i386.h"
106:
107: #define OP adc
108: #define OP_CC
109: #include "test-i386.h"
110:
111: #define OP sbb
112: #define OP_CC
113: #include "test-i386.h"
114:
115: #define OP inc
116: #define OP_CC
117: #define OP1
118: #include "test-i386.h"
119:
120: #define OP dec
121: #define OP_CC
122: #define OP1
123: #include "test-i386.h"
124:
125: #define OP neg
126: #define OP_CC
127: #define OP1
128: #include "test-i386.h"
129:
130: #define OP not
131: #define OP_CC
132: #define OP1
133: #include "test-i386.h"
134:
135: #undef CC_MASK
136: #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O)
137:
138: #define OP shl
139: #include "test-i386-shift.h"
140:
141: #define OP shr
142: #include "test-i386-shift.h"
143:
144: #define OP sar
145: #include "test-i386-shift.h"
146:
147: #define OP rol
148: #include "test-i386-shift.h"
149:
150: #define OP ror
151: #include "test-i386-shift.h"
152:
153: #define OP rcr
154: #define OP_CC
155: #include "test-i386-shift.h"
156:
157: #define OP rcl
158: #define OP_CC
159: #include "test-i386-shift.h"
160:
161: #define OP shld
162: #define OP_SHIFTD
163: #define OP_NOBYTE
164: #include "test-i386-shift.h"
165:
166: #define OP shrd
167: #define OP_SHIFTD
168: #define OP_NOBYTE
169: #include "test-i386-shift.h"
170:
171: /* XXX: should be more precise ? */
172: #undef CC_MASK
173: #define CC_MASK (CC_C)
174:
175: #define OP bt
176: #define OP_NOBYTE
177: #include "test-i386-shift.h"
178:
179: #define OP bts
180: #define OP_NOBYTE
181: #include "test-i386-shift.h"
182:
183: #define OP btr
184: #define OP_NOBYTE
185: #include "test-i386-shift.h"
186:
187: #define OP btc
188: #define OP_NOBYTE
189: #include "test-i386-shift.h"
190:
191: /* lea test (modrm support) */
192: #define TEST_LEAQ(STR)\
193: {\
194: asm("lea " STR ", %0"\
195: : "=r" (res)\
196: : "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\
197: printf("lea %s = " FMTLX "\n", STR, res);\
198: }
199:
200: #define TEST_LEA(STR)\
201: {\
202: asm("lea " STR ", %0"\
203: : "=r" (res)\
204: : "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\
205: printf("lea %s = " FMTLX "\n", STR, res);\
206: }
207:
208: #define TEST_LEA16(STR)\
209: {\
210: asm(".code16 ; .byte 0x67 ; leal " STR ", %0 ; .code32"\
211: : "=wq" (res)\
212: : "a" (eax), "b" (ebx), "c" (ecx), "d" (edx), "S" (esi), "D" (edi));\
213: printf("lea %s = %08lx\n", STR, res);\
214: }
215:
216:
217: void test_lea(void)
218: {
219: long eax, ebx, ecx, edx, esi, edi, res;
220: eax = i2l(0x0001);
221: ebx = i2l(0x0002);
222: ecx = i2l(0x0004);
223: edx = i2l(0x0008);
224: esi = i2l(0x0010);
225: edi = i2l(0x0020);
226:
227: TEST_LEA("0x4000");
228:
229: TEST_LEA("(%%eax)");
230: TEST_LEA("(%%ebx)");
231: TEST_LEA("(%%ecx)");
232: TEST_LEA("(%%edx)");
233: TEST_LEA("(%%esi)");
234: TEST_LEA("(%%edi)");
235:
236: TEST_LEA("0x40(%%eax)");
237: TEST_LEA("0x40(%%ebx)");
238: TEST_LEA("0x40(%%ecx)");
239: TEST_LEA("0x40(%%edx)");
240: TEST_LEA("0x40(%%esi)");
241: TEST_LEA("0x40(%%edi)");
242:
243: TEST_LEA("0x4000(%%eax)");
244: TEST_LEA("0x4000(%%ebx)");
245: TEST_LEA("0x4000(%%ecx)");
246: TEST_LEA("0x4000(%%edx)");
247: TEST_LEA("0x4000(%%esi)");
248: TEST_LEA("0x4000(%%edi)");
249:
250: TEST_LEA("(%%eax, %%ecx)");
251: TEST_LEA("(%%ebx, %%edx)");
252: TEST_LEA("(%%ecx, %%ecx)");
253: TEST_LEA("(%%edx, %%ecx)");
254: TEST_LEA("(%%esi, %%ecx)");
255: TEST_LEA("(%%edi, %%ecx)");
256:
257: TEST_LEA("0x40(%%eax, %%ecx)");
258: TEST_LEA("0x4000(%%ebx, %%edx)");
259:
260: TEST_LEA("(%%ecx, %%ecx, 2)");
261: TEST_LEA("(%%edx, %%ecx, 4)");
262: TEST_LEA("(%%esi, %%ecx, 8)");
263:
264: TEST_LEA("(,%%eax, 2)");
265: TEST_LEA("(,%%ebx, 4)");
266: TEST_LEA("(,%%ecx, 8)");
267:
268: TEST_LEA("0x40(,%%eax, 2)");
269: TEST_LEA("0x40(,%%ebx, 4)");
270: TEST_LEA("0x40(,%%ecx, 8)");
271:
272:
273: TEST_LEA("-10(%%ecx, %%ecx, 2)");
274: TEST_LEA("-10(%%edx, %%ecx, 4)");
275: TEST_LEA("-10(%%esi, %%ecx, 8)");
276:
277: TEST_LEA("0x4000(%%ecx, %%ecx, 2)");
278: TEST_LEA("0x4000(%%edx, %%ecx, 4)");
279: TEST_LEA("0x4000(%%esi, %%ecx, 8)");
280:
281: #if defined(__x86_64__)
282: TEST_LEAQ("0x4000");
283: TEST_LEAQ("0x4000(%%rip)");
284:
285: TEST_LEAQ("(%%rax)");
286: TEST_LEAQ("(%%rbx)");
287: TEST_LEAQ("(%%rcx)");
288: TEST_LEAQ("(%%rdx)");
289: TEST_LEAQ("(%%rsi)");
290: TEST_LEAQ("(%%rdi)");
291:
292: TEST_LEAQ("0x40(%%rax)");
293: TEST_LEAQ("0x40(%%rbx)");
294: TEST_LEAQ("0x40(%%rcx)");
295: TEST_LEAQ("0x40(%%rdx)");
296: TEST_LEAQ("0x40(%%rsi)");
297: TEST_LEAQ("0x40(%%rdi)");
298:
299: TEST_LEAQ("0x4000(%%rax)");
300: TEST_LEAQ("0x4000(%%rbx)");
301: TEST_LEAQ("0x4000(%%rcx)");
302: TEST_LEAQ("0x4000(%%rdx)");
303: TEST_LEAQ("0x4000(%%rsi)");
304: TEST_LEAQ("0x4000(%%rdi)");
305:
306: TEST_LEAQ("(%%rax, %%rcx)");
307: TEST_LEAQ("(%%rbx, %%rdx)");
308: TEST_LEAQ("(%%rcx, %%rcx)");
309: TEST_LEAQ("(%%rdx, %%rcx)");
310: TEST_LEAQ("(%%rsi, %%rcx)");
311: TEST_LEAQ("(%%rdi, %%rcx)");
312:
313: TEST_LEAQ("0x40(%%rax, %%rcx)");
314: TEST_LEAQ("0x4000(%%rbx, %%rdx)");
315:
316: TEST_LEAQ("(%%rcx, %%rcx, 2)");
317: TEST_LEAQ("(%%rdx, %%rcx, 4)");
318: TEST_LEAQ("(%%rsi, %%rcx, 8)");
319:
320: TEST_LEAQ("(,%%rax, 2)");
321: TEST_LEAQ("(,%%rbx, 4)");
322: TEST_LEAQ("(,%%rcx, 8)");
323:
324: TEST_LEAQ("0x40(,%%rax, 2)");
325: TEST_LEAQ("0x40(,%%rbx, 4)");
326: TEST_LEAQ("0x40(,%%rcx, 8)");
327:
328:
329: TEST_LEAQ("-10(%%rcx, %%rcx, 2)");
330: TEST_LEAQ("-10(%%rdx, %%rcx, 4)");
331: TEST_LEAQ("-10(%%rsi, %%rcx, 8)");
332:
333: TEST_LEAQ("0x4000(%%rcx, %%rcx, 2)");
334: TEST_LEAQ("0x4000(%%rdx, %%rcx, 4)");
335: TEST_LEAQ("0x4000(%%rsi, %%rcx, 8)");
336: #else
337: /* limited 16 bit addressing test */
338: TEST_LEA16("0x4000");
339: TEST_LEA16("(%%bx)");
340: TEST_LEA16("(%%si)");
341: TEST_LEA16("(%%di)");
342: TEST_LEA16("0x40(%%bx)");
343: TEST_LEA16("0x40(%%si)");
344: TEST_LEA16("0x40(%%di)");
345: TEST_LEA16("0x4000(%%bx)");
346: TEST_LEA16("0x4000(%%si)");
347: TEST_LEA16("(%%bx,%%si)");
348: TEST_LEA16("(%%bx,%%di)");
349: TEST_LEA16("0x40(%%bx,%%si)");
350: TEST_LEA16("0x40(%%bx,%%di)");
351: TEST_LEA16("0x4000(%%bx,%%si)");
352: TEST_LEA16("0x4000(%%bx,%%di)");
353: #endif
354: }
355:
356: #define TEST_JCC(JCC, v1, v2)\
357: {\
358: int res;\
359: asm("movl $1, %0\n\t"\
360: "cmpl %2, %1\n\t"\
361: "j" JCC " 1f\n\t"\
362: "movl $0, %0\n\t"\
363: "1:\n\t"\
364: : "=r" (res)\
365: : "r" (v1), "r" (v2));\
366: printf("%-10s %d\n", "j" JCC, res);\
367: \
368: asm("movl $0, %0\n\t"\
369: "cmpl %2, %1\n\t"\
370: "set" JCC " %b0\n\t"\
371: : "=r" (res)\
372: : "r" (v1), "r" (v2));\
373: printf("%-10s %d\n", "set" JCC, res);\
374: if (TEST_CMOV) {\
375: long val = i2l(1);\
376: long res = i2l(0x12345678);\
377: X86_64_ONLY(\
378: asm("cmpl %2, %1\n\t"\
379: "cmov" JCC "q %3, %0\n\t"\
380: : "=r" (res)\
381: : "r" (v1), "r" (v2), "m" (val), "0" (res));\
382: printf("%-10s R=" FMTLX "\n", "cmov" JCC "q", res);)\
383: asm("cmpl %2, %1\n\t"\
384: "cmov" JCC "l %k3, %k0\n\t"\
385: : "=r" (res)\
386: : "r" (v1), "r" (v2), "m" (val), "0" (res));\
387: printf("%-10s R=" FMTLX "\n", "cmov" JCC "l", res);\
388: asm("cmpl %2, %1\n\t"\
389: "cmov" JCC "w %w3, %w0\n\t"\
390: : "=r" (res)\
391: : "r" (v1), "r" (v2), "r" (1), "0" (res));\
392: printf("%-10s R=" FMTLX "\n", "cmov" JCC "w", res);\
393: } \
394: }
395:
396: /* various jump tests */
397: void test_jcc(void)
398: {
399: TEST_JCC("ne", 1, 1);
400: TEST_JCC("ne", 1, 0);
401:
402: TEST_JCC("e", 1, 1);
403: TEST_JCC("e", 1, 0);
404:
405: TEST_JCC("l", 1, 1);
406: TEST_JCC("l", 1, 0);
407: TEST_JCC("l", 1, -1);
408:
409: TEST_JCC("le", 1, 1);
410: TEST_JCC("le", 1, 0);
411: TEST_JCC("le", 1, -1);
412:
413: TEST_JCC("ge", 1, 1);
414: TEST_JCC("ge", 1, 0);
415: TEST_JCC("ge", -1, 1);
416:
417: TEST_JCC("g", 1, 1);
418: TEST_JCC("g", 1, 0);
419: TEST_JCC("g", 1, -1);
420:
421: TEST_JCC("b", 1, 1);
422: TEST_JCC("b", 1, 0);
423: TEST_JCC("b", 1, -1);
424:
425: TEST_JCC("be", 1, 1);
426: TEST_JCC("be", 1, 0);
427: TEST_JCC("be", 1, -1);
428:
429: TEST_JCC("ae", 1, 1);
430: TEST_JCC("ae", 1, 0);
431: TEST_JCC("ae", 1, -1);
432:
433: TEST_JCC("a", 1, 1);
434: TEST_JCC("a", 1, 0);
435: TEST_JCC("a", 1, -1);
436:
437:
438: TEST_JCC("p", 1, 1);
439: TEST_JCC("p", 1, 0);
440:
441: TEST_JCC("np", 1, 1);
442: TEST_JCC("np", 1, 0);
443:
444: TEST_JCC("o", 0x7fffffff, 0);
445: TEST_JCC("o", 0x7fffffff, -1);
446:
447: TEST_JCC("no", 0x7fffffff, 0);
448: TEST_JCC("no", 0x7fffffff, -1);
449:
450: TEST_JCC("s", 0, 1);
451: TEST_JCC("s", 0, -1);
452: TEST_JCC("s", 0, 0);
453:
454: TEST_JCC("ns", 0, 1);
455: TEST_JCC("ns", 0, -1);
456: TEST_JCC("ns", 0, 0);
457: }
458:
1.1.1.5 root 459: #define TEST_LOOP(insn) \
460: {\
461: for(i = 0; i < sizeof(ecx_vals) / sizeof(long); i++) {\
462: ecx = ecx_vals[i];\
463: for(zf = 0; zf < 2; zf++) {\
464: asm("test %2, %2\n\t"\
465: "movl $1, %0\n\t"\
466: insn " 1f\n\t" \
467: "movl $0, %0\n\t"\
468: "1:\n\t"\
469: : "=a" (res)\
470: : "c" (ecx), "b" (!zf)); \
471: printf("%-10s ECX=" FMTLX " ZF=%ld r=%d\n", insn, ecx, zf, res); \
472: }\
473: }\
474: }
475:
476: void test_loop(void)
477: {
478: long ecx, zf;
479: const long ecx_vals[] = {
480: 0,
481: 1,
482: 0x10000,
483: 0x10001,
484: #if defined(__x86_64__)
485: 0x100000000L,
486: 0x100000001L,
487: #endif
488: };
489: int i, res;
490:
491: #if !defined(__x86_64__)
492: TEST_LOOP("jcxz");
493: TEST_LOOP("loopw");
494: TEST_LOOP("loopzw");
495: TEST_LOOP("loopnzw");
496: #endif
497:
498: TEST_LOOP("jecxz");
499: TEST_LOOP("loopl");
500: TEST_LOOP("loopzl");
501: TEST_LOOP("loopnzl");
502: }
503:
1.1 root 504: #undef CC_MASK
505: #ifdef TEST_P4_FLAGS
506: #define CC_MASK (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)
507: #else
508: #define CC_MASK (CC_O | CC_C)
509: #endif
510:
511: #define OP mul
512: #include "test-i386-muldiv.h"
513:
514: #define OP imul
515: #include "test-i386-muldiv.h"
516:
1.1.1.4 root 517: void test_imulw2(long op0, long op1)
1.1 root 518: {
519: long res, s1, s0, flags;
520: s0 = op0;
521: s1 = op1;
522: res = s0;
523: flags = 0;
524: asm volatile ("push %4\n\t"
525: "popf\n\t"
1.1.1.4 root 526: "imulw %w2, %w0\n\t"
1.1 root 527: "pushf\n\t"
528: "pop %1\n\t"
529: : "=q" (res), "=g" (flags)
530: : "q" (s1), "0" (res), "1" (flags));
531: printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n",
532: "imulw", s0, s1, res, flags & CC_MASK);
533: }
534:
1.1.1.4 root 535: void test_imull2(long op0, long op1)
1.1 root 536: {
537: long res, s1, s0, flags;
538: s0 = op0;
539: s1 = op1;
540: res = s0;
541: flags = 0;
542: asm volatile ("push %4\n\t"
543: "popf\n\t"
1.1.1.4 root 544: "imull %k2, %k0\n\t"
1.1 root 545: "pushf\n\t"
546: "pop %1\n\t"
547: : "=q" (res), "=g" (flags)
548: : "q" (s1), "0" (res), "1" (flags));
549: printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n",
550: "imull", s0, s1, res, flags & CC_MASK);
551: }
552:
553: #if defined(__x86_64__)
1.1.1.4 root 554: void test_imulq2(long op0, long op1)
1.1 root 555: {
556: long res, s1, s0, flags;
557: s0 = op0;
558: s1 = op1;
559: res = s0;
560: flags = 0;
561: asm volatile ("push %4\n\t"
562: "popf\n\t"
1.1.1.4 root 563: "imulq %2, %0\n\t"
1.1 root 564: "pushf\n\t"
565: "pop %1\n\t"
566: : "=q" (res), "=g" (flags)
567: : "q" (s1), "0" (res), "1" (flags));
568: printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n",
569: "imulq", s0, s1, res, flags & CC_MASK);
570: }
571: #endif
572:
573: #define TEST_IMUL_IM(size, rsize, op0, op1)\
574: {\
575: long res, flags, s1;\
576: flags = 0;\
577: res = 0;\
578: s1 = op1;\
579: asm volatile ("push %3\n\t"\
580: "popf\n\t"\
581: "imul" size " $" #op0 ", %" rsize "2, %" rsize "0\n\t" \
582: "pushf\n\t"\
583: "pop %1\n\t"\
584: : "=r" (res), "=g" (flags)\
585: : "r" (s1), "1" (flags), "0" (res));\
586: printf("%-10s A=" FMTLX " B=" FMTLX " R=" FMTLX " CC=%04lx\n",\
587: "imul" size " im", (long)op0, (long)op1, res, flags & CC_MASK);\
588: }
589:
590:
591: #undef CC_MASK
592: #define CC_MASK (0)
593:
594: #define OP div
595: #include "test-i386-muldiv.h"
596:
597: #define OP idiv
598: #include "test-i386-muldiv.h"
599:
600: void test_mul(void)
601: {
602: test_imulb(0x1234561d, 4);
603: test_imulb(3, -4);
604: test_imulb(0x80, 0x80);
605: test_imulb(0x10, 0x10);
606:
607: test_imulw(0, 0x1234001d, 45);
608: test_imulw(0, 23, -45);
609: test_imulw(0, 0x8000, 0x8000);
610: test_imulw(0, 0x100, 0x100);
611:
612: test_imull(0, 0x1234001d, 45);
613: test_imull(0, 23, -45);
614: test_imull(0, 0x80000000, 0x80000000);
615: test_imull(0, 0x10000, 0x10000);
616:
617: test_mulb(0x1234561d, 4);
618: test_mulb(3, -4);
619: test_mulb(0x80, 0x80);
620: test_mulb(0x10, 0x10);
621:
622: test_mulw(0, 0x1234001d, 45);
623: test_mulw(0, 23, -45);
624: test_mulw(0, 0x8000, 0x8000);
625: test_mulw(0, 0x100, 0x100);
626:
627: test_mull(0, 0x1234001d, 45);
628: test_mull(0, 23, -45);
629: test_mull(0, 0x80000000, 0x80000000);
630: test_mull(0, 0x10000, 0x10000);
631:
632: test_imulw2(0x1234001d, 45);
633: test_imulw2(23, -45);
634: test_imulw2(0x8000, 0x8000);
635: test_imulw2(0x100, 0x100);
636:
637: test_imull2(0x1234001d, 45);
638: test_imull2(23, -45);
639: test_imull2(0x80000000, 0x80000000);
640: test_imull2(0x10000, 0x10000);
641:
642: TEST_IMUL_IM("w", "w", 45, 0x1234);
643: TEST_IMUL_IM("w", "w", -45, 23);
644: TEST_IMUL_IM("w", "w", 0x8000, 0x80000000);
645: TEST_IMUL_IM("w", "w", 0x7fff, 0x1000);
646:
647: TEST_IMUL_IM("l", "k", 45, 0x1234);
648: TEST_IMUL_IM("l", "k", -45, 23);
649: TEST_IMUL_IM("l", "k", 0x8000, 0x80000000);
650: TEST_IMUL_IM("l", "k", 0x7fff, 0x1000);
651:
652: test_idivb(0x12341678, 0x127e);
653: test_idivb(0x43210123, -5);
654: test_idivb(0x12340004, -1);
655:
656: test_idivw(0, 0x12345678, 12347);
657: test_idivw(0, -23223, -45);
658: test_idivw(0, 0x12348000, -1);
659: test_idivw(0x12343, 0x12345678, 0x81238567);
660:
661: test_idivl(0, 0x12345678, 12347);
662: test_idivl(0, -233223, -45);
663: test_idivl(0, 0x80000000, -1);
664: test_idivl(0x12343, 0x12345678, 0x81234567);
665:
666: test_divb(0x12341678, 0x127e);
667: test_divb(0x43210123, -5);
668: test_divb(0x12340004, -1);
669:
670: test_divw(0, 0x12345678, 12347);
671: test_divw(0, -23223, -45);
672: test_divw(0, 0x12348000, -1);
673: test_divw(0x12343, 0x12345678, 0x81238567);
674:
675: test_divl(0, 0x12345678, 12347);
676: test_divl(0, -233223, -45);
677: test_divl(0, 0x80000000, -1);
678: test_divl(0x12343, 0x12345678, 0x81234567);
679:
680: #if defined(__x86_64__)
681: test_imulq(0, 0x1234001d1234001d, 45);
682: test_imulq(0, 23, -45);
683: test_imulq(0, 0x8000000000000000, 0x8000000000000000);
684: test_imulq(0, 0x100000000, 0x100000000);
685:
686: test_mulq(0, 0x1234001d1234001d, 45);
687: test_mulq(0, 23, -45);
688: test_mulq(0, 0x8000000000000000, 0x8000000000000000);
689: test_mulq(0, 0x100000000, 0x100000000);
690:
691: test_imulq2(0x1234001d1234001d, 45);
692: test_imulq2(23, -45);
693: test_imulq2(0x8000000000000000, 0x8000000000000000);
694: test_imulq2(0x100000000, 0x100000000);
695:
696: TEST_IMUL_IM("q", "", 45, 0x12341234);
697: TEST_IMUL_IM("q", "", -45, 23);
698: TEST_IMUL_IM("q", "", 0x8000, 0x8000000000000000);
699: TEST_IMUL_IM("q", "", 0x7fff, 0x10000000);
700:
701: test_idivq(0, 0x12345678abcdef, 12347);
702: test_idivq(0, -233223, -45);
703: test_idivq(0, 0x8000000000000000, -1);
704: test_idivq(0x12343, 0x12345678, 0x81234567);
705:
706: test_divq(0, 0x12345678abcdef, 12347);
707: test_divq(0, -233223, -45);
708: test_divq(0, 0x8000000000000000, -1);
709: test_divq(0x12343, 0x12345678, 0x81234567);
710: #endif
711: }
712:
713: #define TEST_BSX(op, size, op0)\
714: {\
715: long res, val, resz;\
716: val = op0;\
717: asm("xor %1, %1\n"\
718: "mov $0x12345678, %0\n"\
719: #op " %" size "2, %" size "0 ; setz %b1" \
1.1.1.5 root 720: : "=&r" (res), "=&q" (resz)\
721: : "r" (val));\
1.1 root 722: printf("%-10s A=" FMTLX " R=" FMTLX " %ld\n", #op, val, res, resz);\
723: }
724:
725: void test_bsx(void)
726: {
727: TEST_BSX(bsrw, "w", 0);
728: TEST_BSX(bsrw, "w", 0x12340128);
729: TEST_BSX(bsfw, "w", 0);
730: TEST_BSX(bsfw, "w", 0x12340128);
731: TEST_BSX(bsrl, "k", 0);
732: TEST_BSX(bsrl, "k", 0x00340128);
733: TEST_BSX(bsfl, "k", 0);
734: TEST_BSX(bsfl, "k", 0x00340128);
735: #if defined(__x86_64__)
736: TEST_BSX(bsrq, "", 0);
737: TEST_BSX(bsrq, "", 0x003401281234);
738: TEST_BSX(bsfq, "", 0);
739: TEST_BSX(bsfq, "", 0x003401281234);
740: #endif
741: }
742:
743: /**********************************************/
744:
745: union float64u {
746: double d;
747: uint64_t l;
748: };
749:
1.1.1.4 root 750: union float64u q_nan = { .l = 0xFFF8000000000000LL };
751: union float64u s_nan = { .l = 0xFFF0000000000000LL };
1.1 root 752:
753: void test_fops(double a, double b)
754: {
755: printf("a=%f b=%f a+b=%f\n", a, b, a + b);
756: printf("a=%f b=%f a-b=%f\n", a, b, a - b);
757: printf("a=%f b=%f a*b=%f\n", a, b, a * b);
758: printf("a=%f b=%f a/b=%f\n", a, b, a / b);
759: printf("a=%f b=%f fmod(a, b)=%f\n", a, b, fmod(a, b));
760: printf("a=%f sqrt(a)=%f\n", a, sqrt(a));
761: printf("a=%f sin(a)=%f\n", a, sin(a));
762: printf("a=%f cos(a)=%f\n", a, cos(a));
763: printf("a=%f tan(a)=%f\n", a, tan(a));
764: printf("a=%f log(a)=%f\n", a, log(a));
765: printf("a=%f exp(a)=%f\n", a, exp(a));
766: printf("a=%f b=%f atan2(a, b)=%f\n", a, b, atan2(a, b));
767: /* just to test some op combining */
768: printf("a=%f asin(sin(a))=%f\n", a, asin(sin(a)));
769: printf("a=%f acos(cos(a))=%f\n", a, acos(cos(a)));
770: printf("a=%f atan(tan(a))=%f\n", a, atan(tan(a)));
771:
772: }
773:
774: void fpu_clear_exceptions(void)
775: {
776: struct __attribute__((packed)) {
777: uint16_t fpuc;
778: uint16_t dummy1;
779: uint16_t fpus;
780: uint16_t dummy2;
781: uint16_t fptag;
782: uint16_t dummy3;
783: uint32_t ignored[4];
784: long double fpregs[8];
785: } float_env32;
1.1.1.4 root 786:
1.1 root 787: asm volatile ("fnstenv %0\n" : : "m" (float_env32));
788: float_env32.fpus &= ~0x7f;
789: asm volatile ("fldenv %0\n" : : "m" (float_env32));
790: }
791:
792: /* XXX: display exception bits when supported */
793: #define FPUS_EMASK 0x0000
794: //#define FPUS_EMASK 0x007f
795:
796: void test_fcmp(double a, double b)
797: {
798: long eflags, fpus;
799:
800: fpu_clear_exceptions();
801: asm("fcom %2\n"
802: "fstsw %%ax\n"
803: : "=a" (fpus)
804: : "t" (a), "u" (b));
1.1.1.4 root 805: printf("fcom(%f %f)=%04lx \n",
1.1 root 806: a, b, fpus & (0x4500 | FPUS_EMASK));
807: fpu_clear_exceptions();
808: asm("fucom %2\n"
809: "fstsw %%ax\n"
810: : "=a" (fpus)
811: : "t" (a), "u" (b));
1.1.1.4 root 812: printf("fucom(%f %f)=%04lx\n",
1.1 root 813: a, b, fpus & (0x4500 | FPUS_EMASK));
814: if (TEST_FCOMI) {
815: /* test f(u)comi instruction */
816: fpu_clear_exceptions();
817: asm("fcomi %3, %2\n"
818: "fstsw %%ax\n"
819: "pushf\n"
820: "pop %0\n"
821: : "=r" (eflags), "=a" (fpus)
822: : "t" (a), "u" (b));
1.1.1.4 root 823: printf("fcomi(%f %f)=%04lx %02lx\n",
1.1 root 824: a, b, fpus & FPUS_EMASK, eflags & (CC_Z | CC_P | CC_C));
825: fpu_clear_exceptions();
826: asm("fucomi %3, %2\n"
827: "fstsw %%ax\n"
828: "pushf\n"
829: "pop %0\n"
830: : "=r" (eflags), "=a" (fpus)
831: : "t" (a), "u" (b));
1.1.1.4 root 832: printf("fucomi(%f %f)=%04lx %02lx\n",
1.1 root 833: a, b, fpus & FPUS_EMASK, eflags & (CC_Z | CC_P | CC_C));
834: }
835: fpu_clear_exceptions();
1.1.1.2 root 836: asm volatile("fxam\n"
837: "fstsw %%ax\n"
838: : "=a" (fpus)
839: : "t" (a));
840: printf("fxam(%f)=%04lx\n", a, fpus & 0x4700);
841: fpu_clear_exceptions();
1.1 root 842: }
843:
844: void test_fcvt(double a)
845: {
846: float fa;
847: long double la;
848: int16_t fpuc;
849: int i;
850: int64_t lla;
851: int ia;
852: int16_t wa;
853: double ra;
854:
855: fa = a;
856: la = a;
857: printf("(float)%f = %f\n", a, fa);
858: printf("(long double)%f = %Lf\n", a, la);
859: printf("a=" FMT64X "\n", *(uint64_t *)&a);
1.1.1.4 root 860: printf("la=" FMT64X " %04x\n", *(uint64_t *)&la,
1.1 root 861: *(unsigned short *)((char *)(&la) + 8));
862:
863: /* test all roundings */
864: asm volatile ("fstcw %0" : "=m" (fpuc));
865: for(i=0;i<4;i++) {
1.1.1.4 root 866: uint16_t val16;
867: val16 = (fpuc & ~0x0c00) | (i << 10);
868: asm volatile ("fldcw %0" : : "m" (val16));
1.1 root 869: asm volatile ("fist %0" : "=m" (wa) : "t" (a));
870: asm volatile ("fistl %0" : "=m" (ia) : "t" (a));
871: asm volatile ("fistpll %0" : "=m" (lla) : "t" (a) : "st");
872: asm volatile ("frndint ; fstl %0" : "=m" (ra) : "t" (a));
873: asm volatile ("fldcw %0" : : "m" (fpuc));
874: printf("(short)a = %d\n", wa);
875: printf("(int)a = %d\n", ia);
876: printf("(int64_t)a = " FMT64X "\n", lla);
877: printf("rint(a) = %f\n", ra);
878: }
879: }
880:
881: #define TEST(N) \
882: asm("fld" #N : "=t" (a)); \
883: printf("fld" #N "= %f\n", a);
884:
885: void test_fconst(void)
886: {
887: double a;
888: TEST(1);
889: TEST(l2t);
890: TEST(l2e);
891: TEST(pi);
892: TEST(lg2);
893: TEST(ln2);
894: TEST(z);
895: }
896:
897: void test_fbcd(double a)
898: {
899: unsigned short bcd[5];
900: double b;
901:
902: asm("fbstp %0" : "=m" (bcd[0]) : "t" (a) : "st");
903: asm("fbld %1" : "=t" (b) : "m" (bcd[0]));
1.1.1.4 root 904: printf("a=%f bcd=%04x%04x%04x%04x%04x b=%f\n",
1.1 root 905: a, bcd[4], bcd[3], bcd[2], bcd[1], bcd[0], b);
906: }
907:
908: #define TEST_ENV(env, save, restore)\
909: {\
910: memset((env), 0xaa, sizeof(*(env)));\
911: for(i=0;i<5;i++)\
912: asm volatile ("fldl %0" : : "m" (dtab[i]));\
913: asm volatile (save " %0\n" : : "m" (*(env)));\
914: asm volatile (restore " %0\n": : "m" (*(env)));\
915: for(i=0;i<5;i++)\
916: asm volatile ("fstpl %0" : "=m" (rtab[i]));\
917: for(i=0;i<5;i++)\
918: printf("res[%d]=%f\n", i, rtab[i]);\
919: printf("fpuc=%04x fpus=%04x fptag=%04x\n",\
920: (env)->fpuc,\
921: (env)->fpus & 0xff00,\
922: (env)->fptag);\
923: }
924:
925: void test_fenv(void)
926: {
927: struct __attribute__((packed)) {
928: uint16_t fpuc;
929: uint16_t dummy1;
930: uint16_t fpus;
931: uint16_t dummy2;
932: uint16_t fptag;
933: uint16_t dummy3;
934: uint32_t ignored[4];
935: long double fpregs[8];
936: } float_env32;
937: struct __attribute__((packed)) {
938: uint16_t fpuc;
939: uint16_t fpus;
940: uint16_t fptag;
941: uint16_t ignored[4];
942: long double fpregs[8];
943: } float_env16;
944: double dtab[8];
945: double rtab[8];
946: int i;
947:
948: for(i=0;i<8;i++)
949: dtab[i] = i + 1;
950:
951: TEST_ENV(&float_env16, "data16 fnstenv", "data16 fldenv");
952: TEST_ENV(&float_env16, "data16 fnsave", "data16 frstor");
953: TEST_ENV(&float_env32, "fnstenv", "fldenv");
954: TEST_ENV(&float_env32, "fnsave", "frstor");
955:
956: /* test for ffree */
957: for(i=0;i<5;i++)
958: asm volatile ("fldl %0" : : "m" (dtab[i]));
959: asm volatile("ffree %st(2)");
960: asm volatile ("fnstenv %0\n" : : "m" (float_env32));
961: asm volatile ("fninit");
962: printf("fptag=%04x\n", float_env32.fptag);
963: }
964:
965:
966: #define TEST_FCMOV(a, b, eflags, CC)\
967: {\
968: double res;\
969: asm("push %3\n"\
970: "popf\n"\
971: "fcmov" CC " %2, %0\n"\
972: : "=t" (res)\
973: : "0" (a), "u" (b), "g" (eflags));\
974: printf("fcmov%s eflags=0x%04lx-> %f\n", \
975: CC, (long)eflags, res);\
976: }
977:
978: void test_fcmov(void)
979: {
980: double a, b;
981: long eflags, i;
982:
983: a = 1.0;
984: b = 2.0;
985: for(i = 0; i < 4; i++) {
986: eflags = 0;
987: if (i & 1)
988: eflags |= CC_C;
989: if (i & 2)
990: eflags |= CC_Z;
991: TEST_FCMOV(a, b, eflags, "b");
992: TEST_FCMOV(a, b, eflags, "e");
993: TEST_FCMOV(a, b, eflags, "be");
994: TEST_FCMOV(a, b, eflags, "nb");
995: TEST_FCMOV(a, b, eflags, "ne");
996: TEST_FCMOV(a, b, eflags, "nbe");
997: }
998: TEST_FCMOV(a, b, 0, "u");
999: TEST_FCMOV(a, b, CC_P, "u");
1000: TEST_FCMOV(a, b, 0, "nu");
1001: TEST_FCMOV(a, b, CC_P, "nu");
1002: }
1003:
1004: void test_floats(void)
1005: {
1006: test_fops(2, 3);
1007: test_fops(1.4, -5);
1008: test_fcmp(2, -1);
1009: test_fcmp(2, 2);
1010: test_fcmp(2, 3);
1011: test_fcmp(2, q_nan.d);
1012: test_fcmp(q_nan.d, -1);
1.1.1.2 root 1013: test_fcmp(-1.0/0.0, -1);
1014: test_fcmp(1.0/0.0, -1);
1.1 root 1015: test_fcvt(0.5);
1016: test_fcvt(-0.5);
1017: test_fcvt(1.0/7.0);
1018: test_fcvt(-1.0/9.0);
1019: test_fcvt(32768);
1020: test_fcvt(-1e20);
1.1.1.2 root 1021: test_fcvt(-1.0/0.0);
1022: test_fcvt(1.0/0.0);
1023: test_fcvt(q_nan.d);
1.1 root 1024: test_fconst();
1.1.1.4 root 1025: test_fbcd(1234567890123456.0);
1026: test_fbcd(-123451234567890.0);
1.1 root 1027: test_fenv();
1028: if (TEST_CMOV) {
1029: test_fcmov();
1030: }
1031: }
1032:
1033: /**********************************************/
1034: #if !defined(__x86_64__)
1035:
1036: #define TEST_BCD(op, op0, cc_in, cc_mask)\
1037: {\
1038: int res, flags;\
1039: res = op0;\
1040: flags = cc_in;\
1041: asm ("push %3\n\t"\
1042: "popf\n\t"\
1043: #op "\n\t"\
1044: "pushf\n\t"\
1045: "pop %1\n\t"\
1046: : "=a" (res), "=g" (flags)\
1047: : "0" (res), "1" (flags));\
1048: printf("%-10s A=%08x R=%08x CCIN=%04x CC=%04x\n",\
1049: #op, op0, res, cc_in, flags & cc_mask);\
1050: }
1051:
1052: void test_bcd(void)
1053: {
1054: TEST_BCD(daa, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1055: TEST_BCD(daa, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1056: TEST_BCD(daa, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1057: TEST_BCD(daa, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1058: TEST_BCD(daa, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1059: TEST_BCD(daa, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1060: TEST_BCD(daa, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1061: TEST_BCD(daa, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1062: TEST_BCD(daa, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1063: TEST_BCD(daa, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1064: TEST_BCD(daa, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1065: TEST_BCD(daa, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1066: TEST_BCD(daa, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1067:
1068: TEST_BCD(das, 0x12340503, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1069: TEST_BCD(das, 0x12340506, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1070: TEST_BCD(das, 0x12340507, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1071: TEST_BCD(das, 0x12340559, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1072: TEST_BCD(das, 0x12340560, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1073: TEST_BCD(das, 0x1234059f, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1074: TEST_BCD(das, 0x123405a0, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1075: TEST_BCD(das, 0x12340503, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1076: TEST_BCD(das, 0x12340506, 0, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1077: TEST_BCD(das, 0x12340503, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1078: TEST_BCD(das, 0x12340506, CC_C, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1079: TEST_BCD(das, 0x12340503, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1080: TEST_BCD(das, 0x12340506, CC_C | CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_A));
1081:
1082: TEST_BCD(aaa, 0x12340205, CC_A, (CC_C | CC_A));
1083: TEST_BCD(aaa, 0x12340306, CC_A, (CC_C | CC_A));
1084: TEST_BCD(aaa, 0x1234040a, CC_A, (CC_C | CC_A));
1085: TEST_BCD(aaa, 0x123405fa, CC_A, (CC_C | CC_A));
1086: TEST_BCD(aaa, 0x12340205, 0, (CC_C | CC_A));
1087: TEST_BCD(aaa, 0x12340306, 0, (CC_C | CC_A));
1088: TEST_BCD(aaa, 0x1234040a, 0, (CC_C | CC_A));
1089: TEST_BCD(aaa, 0x123405fa, 0, (CC_C | CC_A));
1.1.1.4 root 1090:
1.1 root 1091: TEST_BCD(aas, 0x12340205, CC_A, (CC_C | CC_A));
1092: TEST_BCD(aas, 0x12340306, CC_A, (CC_C | CC_A));
1093: TEST_BCD(aas, 0x1234040a, CC_A, (CC_C | CC_A));
1094: TEST_BCD(aas, 0x123405fa, CC_A, (CC_C | CC_A));
1095: TEST_BCD(aas, 0x12340205, 0, (CC_C | CC_A));
1096: TEST_BCD(aas, 0x12340306, 0, (CC_C | CC_A));
1097: TEST_BCD(aas, 0x1234040a, 0, (CC_C | CC_A));
1098: TEST_BCD(aas, 0x123405fa, 0, (CC_C | CC_A));
1099:
1100: TEST_BCD(aam, 0x12340547, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
1101: TEST_BCD(aad, 0x12340407, CC_A, (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));
1102: }
1103: #endif
1104:
1105: #define TEST_XCHG(op, size, opconst)\
1106: {\
1107: long op0, op1;\
1108: op0 = i2l(0x12345678);\
1109: op1 = i2l(0xfbca7654);\
1110: asm(#op " %" size "0, %" size "1" \
1111: : "=q" (op0), opconst (op1) \
1.1.1.5 root 1112: : "0" (op0));\
1.1 root 1113: printf("%-10s A=" FMTLX " B=" FMTLX "\n",\
1114: #op, op0, op1);\
1115: }
1116:
1117: #define TEST_CMPXCHG(op, size, opconst, eax)\
1118: {\
1119: long op0, op1, op2;\
1120: op0 = i2l(0x12345678);\
1121: op1 = i2l(0xfbca7654);\
1122: op2 = i2l(eax);\
1123: asm(#op " %" size "0, %" size "1" \
1124: : "=q" (op0), opconst (op1) \
1.1.1.5 root 1125: : "0" (op0), "a" (op2));\
1.1 root 1126: printf("%-10s EAX=" FMTLX " A=" FMTLX " C=" FMTLX "\n",\
1127: #op, op2, op0, op1);\
1128: }
1129:
1130: void test_xchg(void)
1131: {
1132: #if defined(__x86_64__)
1.1.1.5 root 1133: TEST_XCHG(xchgq, "", "+q");
1.1 root 1134: #endif
1.1.1.5 root 1135: TEST_XCHG(xchgl, "k", "+q");
1136: TEST_XCHG(xchgw, "w", "+q");
1137: TEST_XCHG(xchgb, "b", "+q");
1.1 root 1138:
1139: #if defined(__x86_64__)
1140: TEST_XCHG(xchgq, "", "=m");
1141: #endif
1.1.1.5 root 1142: TEST_XCHG(xchgl, "k", "+m");
1143: TEST_XCHG(xchgw, "w", "+m");
1144: TEST_XCHG(xchgb, "b", "+m");
1.1 root 1145:
1146: #if defined(__x86_64__)
1.1.1.5 root 1147: TEST_XCHG(xaddq, "", "+q");
1.1 root 1148: #endif
1.1.1.5 root 1149: TEST_XCHG(xaddl, "k", "+q");
1150: TEST_XCHG(xaddw, "w", "+q");
1151: TEST_XCHG(xaddb, "b", "+q");
1.1 root 1152:
1153: {
1154: int res;
1155: res = 0x12345678;
1156: asm("xaddl %1, %0" : "=r" (res) : "0" (res));
1157: printf("xaddl same res=%08x\n", res);
1158: }
1159:
1160: #if defined(__x86_64__)
1.1.1.5 root 1161: TEST_XCHG(xaddq, "", "+m");
1.1 root 1162: #endif
1.1.1.5 root 1163: TEST_XCHG(xaddl, "k", "+m");
1164: TEST_XCHG(xaddw, "w", "+m");
1165: TEST_XCHG(xaddb, "b", "+m");
1.1 root 1166:
1167: #if defined(__x86_64__)
1.1.1.5 root 1168: TEST_CMPXCHG(cmpxchgq, "", "+q", 0xfbca7654);
1.1 root 1169: #endif
1.1.1.5 root 1170: TEST_CMPXCHG(cmpxchgl, "k", "+q", 0xfbca7654);
1171: TEST_CMPXCHG(cmpxchgw, "w", "+q", 0xfbca7654);
1172: TEST_CMPXCHG(cmpxchgb, "b", "+q", 0xfbca7654);
1.1 root 1173:
1174: #if defined(__x86_64__)
1.1.1.5 root 1175: TEST_CMPXCHG(cmpxchgq, "", "+q", 0xfffefdfc);
1.1 root 1176: #endif
1.1.1.5 root 1177: TEST_CMPXCHG(cmpxchgl, "k", "+q", 0xfffefdfc);
1178: TEST_CMPXCHG(cmpxchgw, "w", "+q", 0xfffefdfc);
1179: TEST_CMPXCHG(cmpxchgb, "b", "+q", 0xfffefdfc);
1.1 root 1180:
1181: #if defined(__x86_64__)
1.1.1.5 root 1182: TEST_CMPXCHG(cmpxchgq, "", "+m", 0xfbca7654);
1.1 root 1183: #endif
1.1.1.5 root 1184: TEST_CMPXCHG(cmpxchgl, "k", "+m", 0xfbca7654);
1185: TEST_CMPXCHG(cmpxchgw, "w", "+m", 0xfbca7654);
1186: TEST_CMPXCHG(cmpxchgb, "b", "+m", 0xfbca7654);
1.1 root 1187:
1188: #if defined(__x86_64__)
1.1.1.5 root 1189: TEST_CMPXCHG(cmpxchgq, "", "+m", 0xfffefdfc);
1.1 root 1190: #endif
1.1.1.5 root 1191: TEST_CMPXCHG(cmpxchgl, "k", "+m", 0xfffefdfc);
1192: TEST_CMPXCHG(cmpxchgw, "w", "+m", 0xfffefdfc);
1193: TEST_CMPXCHG(cmpxchgb, "b", "+m", 0xfffefdfc);
1.1 root 1194:
1195: {
1196: uint64_t op0, op1, op2;
1.1.1.5 root 1197: long eax, edx;
1.1 root 1198: long i, eflags;
1199:
1200: for(i = 0; i < 2; i++) {
1.1.1.4 root 1201: op0 = 0x123456789abcdLL;
1.1.1.5 root 1202: eax = i2l(op0 & 0xffffffff);
1203: edx = i2l(op0 >> 32);
1.1 root 1204: if (i == 0)
1.1.1.4 root 1205: op1 = 0xfbca765423456LL;
1.1 root 1206: else
1207: op1 = op0;
1.1.1.4 root 1208: op2 = 0x6532432432434LL;
1.1.1.5 root 1209: asm("cmpxchg8b %2\n"
1.1 root 1210: "pushf\n"
1.1.1.5 root 1211: "pop %3\n"
1212: : "=a" (eax), "=d" (edx), "=m" (op1), "=g" (eflags)
1213: : "0" (eax), "1" (edx), "m" (op1), "b" ((int)op2), "c" ((int)(op2 >> 32)));
1214: printf("cmpxchg8b: eax=" FMTLX " edx=" FMTLX " op1=" FMT64X " CC=%02lx\n",
1215: eax, edx, op1, eflags & CC_Z);
1.1 root 1216: }
1217: }
1218: }
1219:
1220: #ifdef TEST_SEGS
1221: /**********************************************/
1222: /* segmentation tests */
1223:
1.1.1.4 root 1224: #include <sys/syscall.h>
1225: #include <unistd.h>
1.1 root 1226: #include <asm/ldt.h>
1227: #include <linux/version.h>
1228:
1.1.1.4 root 1229: static inline int modify_ldt(int func, void * ptr, unsigned long bytecount)
1230: {
1231: return syscall(__NR_modify_ldt, func, ptr, bytecount);
1232: }
1.1 root 1233:
1234: #if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 5, 66)
1235: #define modify_ldt_ldt_s user_desc
1236: #endif
1237:
1238: #define MK_SEL(n) (((n) << 3) | 7)
1239:
1240: uint8_t seg_data1[4096];
1241: uint8_t seg_data2[4096];
1242:
1243: #define TEST_LR(op, size, seg, mask)\
1244: {\
1245: int res, res2;\
1.1.1.5 root 1246: uint16_t mseg = seg;\
1.1 root 1247: res = 0x12345678;\
1248: asm (op " %" size "2, %" size "0\n" \
1249: "movl $0, %1\n"\
1250: "jnz 1f\n"\
1251: "movl $1, %1\n"\
1252: "1:\n"\
1.1.1.5 root 1253: : "=r" (res), "=r" (res2) : "m" (mseg), "0" (res));\
1.1 root 1254: printf(op ": Z=%d %08x\n", res2, res & ~(mask));\
1255: }
1256:
1.1.1.5 root 1257: #define TEST_ARPL(op, size, op1, op2)\
1258: {\
1259: long a, b, c; \
1260: a = (op1); \
1261: b = (op2); \
1262: asm volatile(op " %" size "3, %" size "0\n"\
1263: "movl $0,%1\n"\
1264: "jnz 1f\n"\
1265: "movl $1,%1\n"\
1266: "1:\n"\
1267: : "=r" (a), "=r" (c) : "0" (a), "r" (b)); \
1268: printf(op size " A=" FMTLX " B=" FMTLX " R=" FMTLX " z=%ld\n",\
1269: (long)(op1), (long)(op2), a, c);\
1270: }
1271:
1.1 root 1272: /* NOTE: we use Linux modify_ldt syscall */
1273: void test_segs(void)
1274: {
1275: struct modify_ldt_ldt_s ldt;
1276: long long ldt_table[3];
1277: int res, res2;
1278: char tmp;
1279: struct {
1280: uint32_t offset;
1281: uint16_t seg;
1282: } __attribute__((packed)) segoff;
1283:
1284: ldt.entry_number = 1;
1285: ldt.base_addr = (unsigned long)&seg_data1;
1286: ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12;
1287: ldt.seg_32bit = 1;
1288: ldt.contents = MODIFY_LDT_CONTENTS_DATA;
1289: ldt.read_exec_only = 0;
1290: ldt.limit_in_pages = 1;
1291: ldt.seg_not_present = 0;
1292: ldt.useable = 1;
1293: modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1294:
1295: ldt.entry_number = 2;
1296: ldt.base_addr = (unsigned long)&seg_data2;
1297: ldt.limit = (sizeof(seg_data2) + 0xfff) >> 12;
1298: ldt.seg_32bit = 1;
1299: ldt.contents = MODIFY_LDT_CONTENTS_DATA;
1300: ldt.read_exec_only = 0;
1301: ldt.limit_in_pages = 1;
1302: ldt.seg_not_present = 0;
1303: ldt.useable = 1;
1304: modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1305:
1306: modify_ldt(0, &ldt_table, sizeof(ldt_table)); /* read ldt entries */
1307: #if 0
1308: {
1309: int i;
1310: for(i=0;i<3;i++)
1311: printf("%d: %016Lx\n", i, ldt_table[i]);
1312: }
1313: #endif
1314: /* do some tests with fs or gs */
1315: asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1)));
1316:
1317: seg_data1[1] = 0xaa;
1318: seg_data2[1] = 0x55;
1319:
1320: asm volatile ("fs movzbl 0x1, %0" : "=r" (res));
1321: printf("FS[1] = %02x\n", res);
1322:
1323: asm volatile ("pushl %%gs\n"
1324: "movl %1, %%gs\n"
1325: "gs movzbl 0x1, %0\n"
1326: "popl %%gs\n"
1327: : "=r" (res)
1328: : "r" (MK_SEL(2)));
1329: printf("GS[1] = %02x\n", res);
1330:
1331: /* tests with ds/ss (implicit segment case) */
1332: tmp = 0xa5;
1333: asm volatile ("pushl %%ebp\n\t"
1334: "pushl %%ds\n\t"
1335: "movl %2, %%ds\n\t"
1336: "movl %3, %%ebp\n\t"
1337: "movzbl 0x1, %0\n\t"
1338: "movzbl (%%ebp), %1\n\t"
1339: "popl %%ds\n\t"
1340: "popl %%ebp\n\t"
1341: : "=r" (res), "=r" (res2)
1342: : "r" (MK_SEL(1)), "r" (&tmp));
1343: printf("DS[1] = %02x\n", res);
1344: printf("SS[tmp] = %02x\n", res2);
1345:
1346: segoff.seg = MK_SEL(2);
1347: segoff.offset = 0xabcdef12;
1.1.1.4 root 1348: asm volatile("lfs %2, %0\n\t"
1.1 root 1349: "movl %%fs, %1\n\t"
1.1.1.4 root 1350: : "=r" (res), "=g" (res2)
1.1 root 1351: : "m" (segoff));
1352: printf("FS:reg = %04x:%08x\n", res2, res);
1353:
1354: TEST_LR("larw", "w", MK_SEL(2), 0x0100);
1355: TEST_LR("larl", "", MK_SEL(2), 0x0100);
1356: TEST_LR("lslw", "w", MK_SEL(2), 0);
1357: TEST_LR("lsll", "", MK_SEL(2), 0);
1358:
1359: TEST_LR("larw", "w", 0xfff8, 0);
1360: TEST_LR("larl", "", 0xfff8, 0);
1361: TEST_LR("lslw", "w", 0xfff8, 0);
1362: TEST_LR("lsll", "", 0xfff8, 0);
1.1.1.5 root 1363:
1364: TEST_ARPL("arpl", "w", 0x12345678 | 3, 0x762123c | 1);
1365: TEST_ARPL("arpl", "w", 0x12345678 | 1, 0x762123c | 3);
1366: TEST_ARPL("arpl", "w", 0x12345678 | 1, 0x762123c | 1);
1.1 root 1367: }
1368:
1369: /* 16 bit code test */
1370: extern char code16_start, code16_end;
1371: extern char code16_func1;
1372: extern char code16_func2;
1373: extern char code16_func3;
1374:
1375: void test_code16(void)
1376: {
1377: struct modify_ldt_ldt_s ldt;
1378: int res, res2;
1379:
1380: /* build a code segment */
1381: ldt.entry_number = 1;
1382: ldt.base_addr = (unsigned long)&code16_start;
1383: ldt.limit = &code16_end - &code16_start;
1384: ldt.seg_32bit = 0;
1385: ldt.contents = MODIFY_LDT_CONTENTS_CODE;
1386: ldt.read_exec_only = 0;
1387: ldt.limit_in_pages = 0;
1388: ldt.seg_not_present = 0;
1389: ldt.useable = 1;
1390: modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1391:
1392: /* call the first function */
1.1.1.4 root 1393: asm volatile ("lcall %1, %2"
1.1 root 1394: : "=a" (res)
1395: : "i" (MK_SEL(1)), "i" (&code16_func1): "memory", "cc");
1396: printf("func1() = 0x%08x\n", res);
1.1.1.4 root 1397: asm volatile ("lcall %2, %3"
1.1 root 1398: : "=a" (res), "=c" (res2)
1399: : "i" (MK_SEL(1)), "i" (&code16_func2): "memory", "cc");
1400: printf("func2() = 0x%08x spdec=%d\n", res, res2);
1.1.1.4 root 1401: asm volatile ("lcall %1, %2"
1.1 root 1402: : "=a" (res)
1403: : "i" (MK_SEL(1)), "i" (&code16_func3): "memory", "cc");
1404: printf("func3() = 0x%08x\n", res);
1405: }
1406: #endif
1407:
1408: #if defined(__x86_64__)
1409: asm(".globl func_lret\n"
1410: "func_lret:\n"
1411: "movl $0x87654641, %eax\n"
1412: "lretq\n");
1413: #else
1414: asm(".globl func_lret\n"
1415: "func_lret:\n"
1416: "movl $0x87654321, %eax\n"
1417: "lret\n"
1418:
1419: ".globl func_iret\n"
1420: "func_iret:\n"
1421: "movl $0xabcd4321, %eax\n"
1422: "iret\n");
1423: #endif
1424:
1425: extern char func_lret;
1426: extern char func_iret;
1427:
1428: void test_misc(void)
1429: {
1430: char table[256];
1431: long res, i;
1432:
1433: for(i=0;i<256;i++) table[i] = 256 - i;
1434: res = 0x12345678;
1435: asm ("xlat" : "=a" (res) : "b" (table), "0" (res));
1436: printf("xlat: EAX=" FMTLX "\n", res);
1437:
1438: #if defined(__x86_64__)
1.1.1.5 root 1439: #if 0
1.1 root 1440: {
1.1.1.5 root 1441: /* XXX: see if Intel Core2 and AMD64 behavior really
1442: differ. Here we implemented the Intel way which is not
1443: compatible yet with QEMU. */
1.1 root 1444: static struct __attribute__((packed)) {
1.1.1.5 root 1445: uint64_t offset;
1.1 root 1446: uint16_t seg;
1447: } desc;
1448: long cs_sel;
1449:
1450: asm volatile ("mov %%cs, %0" : "=r" (cs_sel));
1451:
1452: asm volatile ("push %1\n"
1.1.1.4 root 1453: "call func_lret\n"
1.1 root 1454: : "=a" (res)
1455: : "r" (cs_sel) : "memory", "cc");
1456: printf("func_lret=" FMTLX "\n", res);
1457:
1458: desc.offset = (long)&func_lret;
1459: desc.seg = cs_sel;
1.1.1.4 root 1460:
1.1 root 1461: asm volatile ("xor %%rax, %%rax\n"
1.1.1.5 root 1462: "rex64 lcall *(%%rcx)\n"
1.1 root 1463: : "=a" (res)
1.1.1.5 root 1464: : "c" (&desc)
1.1 root 1465: : "memory", "cc");
1466: printf("func_lret2=" FMTLX "\n", res);
1467:
1468: asm volatile ("push %2\n"
1469: "mov $ 1f, %%rax\n"
1470: "push %%rax\n"
1.1.1.5 root 1471: "rex64 ljmp *(%%rcx)\n"
1.1 root 1472: "1:\n"
1473: : "=a" (res)
1.1.1.5 root 1474: : "c" (&desc), "b" (cs_sel)
1.1 root 1475: : "memory", "cc");
1476: printf("func_lret3=" FMTLX "\n", res);
1477: }
1.1.1.5 root 1478: #endif
1.1 root 1479: #else
1.1.1.4 root 1480: asm volatile ("push %%cs ; call %1"
1.1 root 1481: : "=a" (res)
1482: : "m" (func_lret): "memory", "cc");
1483: printf("func_lret=" FMTLX "\n", res);
1484:
1.1.1.4 root 1485: asm volatile ("pushf ; push %%cs ; call %1"
1.1 root 1486: : "=a" (res)
1487: : "m" (func_iret): "memory", "cc");
1488: printf("func_iret=" FMTLX "\n", res);
1489: #endif
1490:
1491: #if defined(__x86_64__)
1492: /* specific popl test */
1493: asm volatile ("push $12345432 ; push $0x9abcdef ; pop (%%rsp) ; pop %0"
1494: : "=g" (res));
1495: printf("popl esp=" FMTLX "\n", res);
1496: #else
1497: /* specific popl test */
1498: asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popl (%%esp) ; popl %0"
1499: : "=g" (res));
1500: printf("popl esp=" FMTLX "\n", res);
1501:
1502: /* specific popw test */
1503: asm volatile ("pushl $12345432 ; pushl $0x9abcdef ; popw (%%esp) ; addl $2, %%esp ; popl %0"
1504: : "=g" (res));
1505: printf("popw esp=" FMTLX "\n", res);
1506: #endif
1507: }
1508:
1509: uint8_t str_buffer[4096];
1510:
1511: #define TEST_STRING1(OP, size, DF, REP)\
1512: {\
1513: long esi, edi, eax, ecx, eflags;\
1514: \
1515: esi = (long)(str_buffer + sizeof(str_buffer) / 2);\
1516: edi = (long)(str_buffer + sizeof(str_buffer) / 2) + 16;\
1517: eax = i2l(0x12345678);\
1518: ecx = 17;\
1519: \
1520: asm volatile ("push $0\n\t"\
1521: "popf\n\t"\
1522: DF "\n\t"\
1523: REP #OP size "\n\t"\
1524: "cld\n\t"\
1525: "pushf\n\t"\
1526: "pop %4\n\t"\
1527: : "=S" (esi), "=D" (edi), "=a" (eax), "=c" (ecx), "=g" (eflags)\
1528: : "0" (esi), "1" (edi), "2" (eax), "3" (ecx));\
1529: printf("%-10s ESI=" FMTLX " EDI=" FMTLX " EAX=" FMTLX " ECX=" FMTLX " EFL=%04x\n",\
1530: REP #OP size, esi, edi, eax, ecx,\
1531: (int)(eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A)));\
1532: }
1533:
1534: #define TEST_STRING(OP, REP)\
1535: TEST_STRING1(OP, "b", "", REP);\
1536: TEST_STRING1(OP, "w", "", REP);\
1537: TEST_STRING1(OP, "l", "", REP);\
1538: X86_64_ONLY(TEST_STRING1(OP, "q", "", REP));\
1539: TEST_STRING1(OP, "b", "std", REP);\
1540: TEST_STRING1(OP, "w", "std", REP);\
1541: TEST_STRING1(OP, "l", "std", REP);\
1542: X86_64_ONLY(TEST_STRING1(OP, "q", "std", REP))
1543:
1544: void test_string(void)
1545: {
1546: int i;
1547: for(i = 0;i < sizeof(str_buffer); i++)
1548: str_buffer[i] = i + 0x56;
1549: TEST_STRING(stos, "");
1550: TEST_STRING(stos, "rep ");
1551: TEST_STRING(lods, ""); /* to verify stos */
1.1.1.4 root 1552: TEST_STRING(lods, "rep ");
1.1 root 1553: TEST_STRING(movs, "");
1554: TEST_STRING(movs, "rep ");
1555: TEST_STRING(lods, ""); /* to verify stos */
1556:
1557: /* XXX: better tests */
1558: TEST_STRING(scas, "");
1559: TEST_STRING(scas, "repz ");
1560: TEST_STRING(scas, "repnz ");
1561: TEST_STRING(cmps, "");
1562: TEST_STRING(cmps, "repz ");
1563: TEST_STRING(cmps, "repnz ");
1564: }
1565:
1566: #ifdef TEST_VM86
1567: /* VM86 test */
1568:
1569: static inline void set_bit(uint8_t *a, unsigned int bit)
1570: {
1571: a[bit / 8] |= (1 << (bit % 8));
1572: }
1573:
1574: static inline uint8_t *seg_to_linear(unsigned int seg, unsigned int reg)
1575: {
1576: return (uint8_t *)((seg << 4) + (reg & 0xffff));
1577: }
1578:
1579: static inline void pushw(struct vm86_regs *r, int val)
1580: {
1581: r->esp = (r->esp & ~0xffff) | ((r->esp - 2) & 0xffff);
1582: *(uint16_t *)seg_to_linear(r->ss, r->esp) = val;
1583: }
1584:
1.1.1.4 root 1585: static inline int vm86(int func, struct vm86plus_struct *v86)
1586: {
1587: return syscall(__NR_vm86, func, v86);
1588: }
1.1 root 1589:
1590: extern char vm86_code_start;
1591: extern char vm86_code_end;
1592:
1593: #define VM86_CODE_CS 0x100
1594: #define VM86_CODE_IP 0x100
1595:
1596: void test_vm86(void)
1597: {
1598: struct vm86plus_struct ctx;
1599: struct vm86_regs *r;
1600: uint8_t *vm86_mem;
1601: int seg, ret;
1602:
1.1.1.4 root 1603: vm86_mem = mmap((void *)0x00000000, 0x110000,
1604: PROT_WRITE | PROT_READ | PROT_EXEC,
1.1 root 1605: MAP_FIXED | MAP_ANON | MAP_PRIVATE, -1, 0);
1606: if (vm86_mem == MAP_FAILED) {
1607: printf("ERROR: could not map vm86 memory");
1608: return;
1609: }
1610: memset(&ctx, 0, sizeof(ctx));
1611:
1612: /* init basic registers */
1613: r = &ctx.regs;
1614: r->eip = VM86_CODE_IP;
1615: r->esp = 0xfffe;
1616: seg = VM86_CODE_CS;
1617: r->cs = seg;
1618: r->ss = seg;
1619: r->ds = seg;
1620: r->es = seg;
1621: r->fs = seg;
1622: r->gs = seg;
1623: r->eflags = VIF_MASK;
1624:
1625: /* move code to proper address. We use the same layout as a .com
1626: dos program. */
1.1.1.4 root 1627: memcpy(vm86_mem + (VM86_CODE_CS << 4) + VM86_CODE_IP,
1.1 root 1628: &vm86_code_start, &vm86_code_end - &vm86_code_start);
1629:
1630: /* mark int 0x21 as being emulated */
1631: set_bit((uint8_t *)&ctx.int_revectored, 0x21);
1632:
1633: for(;;) {
1634: ret = vm86(VM86_ENTER, &ctx);
1635: switch(VM86_TYPE(ret)) {
1636: case VM86_INTx:
1637: {
1638: int int_num, ah, v;
1.1.1.4 root 1639:
1.1 root 1640: int_num = VM86_ARG(ret);
1641: if (int_num != 0x21)
1642: goto unknown_int;
1643: ah = (r->eax >> 8) & 0xff;
1644: switch(ah) {
1645: case 0x00: /* exit */
1646: goto the_end;
1647: case 0x02: /* write char */
1648: {
1649: uint8_t c = r->edx;
1650: putchar(c);
1651: }
1652: break;
1653: case 0x09: /* write string */
1654: {
1655: uint8_t c, *ptr;
1656: ptr = seg_to_linear(r->ds, r->edx);
1657: for(;;) {
1658: c = *ptr++;
1659: if (c == '$')
1660: break;
1661: putchar(c);
1662: }
1663: r->eax = (r->eax & ~0xff) | '$';
1664: }
1665: break;
1666: case 0xff: /* extension: write eflags number in edx */
1667: v = (int)r->edx;
1668: #ifndef LINUX_VM86_IOPL_FIX
1669: v &= ~0x3000;
1670: #endif
1671: printf("%08x\n", v);
1672: break;
1673: default:
1674: unknown_int:
1675: printf("unsupported int 0x%02x\n", int_num);
1676: goto the_end;
1677: }
1678: }
1679: break;
1680: case VM86_SIGNAL:
1681: /* a signal came, we just ignore that */
1682: break;
1683: case VM86_STI:
1684: break;
1685: default:
1686: printf("ERROR: unhandled vm86 return code (0x%x)\n", ret);
1687: goto the_end;
1688: }
1689: }
1690: the_end:
1691: printf("VM86 end\n");
1692: munmap(vm86_mem, 0x110000);
1693: }
1694: #endif
1695:
1696: /* exception tests */
1697: #if defined(__i386__) && !defined(REG_EAX)
1698: #define REG_EAX EAX
1699: #define REG_EBX EBX
1700: #define REG_ECX ECX
1701: #define REG_EDX EDX
1702: #define REG_ESI ESI
1703: #define REG_EDI EDI
1704: #define REG_EBP EBP
1705: #define REG_ESP ESP
1706: #define REG_EIP EIP
1707: #define REG_EFL EFL
1708: #define REG_TRAPNO TRAPNO
1709: #define REG_ERR ERR
1710: #endif
1711:
1712: #if defined(__x86_64__)
1713: #define REG_EIP REG_RIP
1714: #endif
1715:
1716: jmp_buf jmp_env;
1717: int v1;
1718: int tab[2];
1719:
1720: void sig_handler(int sig, siginfo_t *info, void *puc)
1721: {
1722: struct ucontext *uc = puc;
1723:
1724: printf("si_signo=%d si_errno=%d si_code=%d",
1725: info->si_signo, info->si_errno, info->si_code);
1726: printf(" si_addr=0x%08lx",
1727: (unsigned long)info->si_addr);
1728: printf("\n");
1729:
1730: printf("trapno=" FMTLX " err=" FMTLX,
1731: (long)uc->uc_mcontext.gregs[REG_TRAPNO],
1732: (long)uc->uc_mcontext.gregs[REG_ERR]);
1733: printf(" EIP=" FMTLX, (long)uc->uc_mcontext.gregs[REG_EIP]);
1734: printf("\n");
1735: longjmp(jmp_env, 1);
1736: }
1737:
1738: void test_exceptions(void)
1739: {
1740: struct sigaction act;
1741: volatile int val;
1.1.1.4 root 1742:
1.1 root 1743: act.sa_sigaction = sig_handler;
1744: sigemptyset(&act.sa_mask);
1745: act.sa_flags = SA_SIGINFO | SA_NODEFER;
1746: sigaction(SIGFPE, &act, NULL);
1747: sigaction(SIGILL, &act, NULL);
1748: sigaction(SIGSEGV, &act, NULL);
1749: sigaction(SIGBUS, &act, NULL);
1750: sigaction(SIGTRAP, &act, NULL);
1751:
1752: /* test division by zero reporting */
1753: printf("DIVZ exception:\n");
1754: if (setjmp(jmp_env) == 0) {
1755: /* now divide by zero */
1756: v1 = 0;
1757: v1 = 2 / v1;
1758: }
1759:
1760: #if !defined(__x86_64__)
1761: printf("BOUND exception:\n");
1762: if (setjmp(jmp_env) == 0) {
1763: /* bound exception */
1764: tab[0] = 1;
1765: tab[1] = 10;
1766: asm volatile ("bound %0, %1" : : "r" (11), "m" (tab[0]));
1767: }
1768: #endif
1769:
1770: #ifdef TEST_SEGS
1771: printf("segment exceptions:\n");
1772: if (setjmp(jmp_env) == 0) {
1773: /* load an invalid segment */
1774: asm volatile ("movl %0, %%fs" : : "r" ((0x1234 << 3) | 1));
1775: }
1776: if (setjmp(jmp_env) == 0) {
1777: /* null data segment is valid */
1778: asm volatile ("movl %0, %%fs" : : "r" (3));
1779: /* null stack segment */
1780: asm volatile ("movl %0, %%ss" : : "r" (3));
1781: }
1782:
1783: {
1784: struct modify_ldt_ldt_s ldt;
1785: ldt.entry_number = 1;
1786: ldt.base_addr = (unsigned long)&seg_data1;
1787: ldt.limit = (sizeof(seg_data1) + 0xfff) >> 12;
1788: ldt.seg_32bit = 1;
1789: ldt.contents = MODIFY_LDT_CONTENTS_DATA;
1790: ldt.read_exec_only = 0;
1791: ldt.limit_in_pages = 1;
1792: ldt.seg_not_present = 1;
1793: ldt.useable = 1;
1794: modify_ldt(1, &ldt, sizeof(ldt)); /* write ldt entry */
1.1.1.4 root 1795:
1.1 root 1796: if (setjmp(jmp_env) == 0) {
1797: /* segment not present */
1798: asm volatile ("movl %0, %%fs" : : "r" (MK_SEL(1)));
1799: }
1800: }
1801: #endif
1802:
1803: /* test SEGV reporting */
1804: printf("PF exception:\n");
1805: if (setjmp(jmp_env) == 0) {
1806: val = 1;
1807: /* we add a nop to test a weird PC retrieval case */
1808: asm volatile ("nop");
1809: /* now store in an invalid address */
1810: *(char *)0x1234 = 1;
1811: }
1812:
1813: /* test SEGV reporting */
1814: printf("PF exception:\n");
1815: if (setjmp(jmp_env) == 0) {
1816: val = 1;
1817: /* read from an invalid address */
1818: v1 = *(char *)0x1234;
1819: }
1.1.1.4 root 1820:
1.1 root 1821: /* test illegal instruction reporting */
1822: printf("UD2 exception:\n");
1823: if (setjmp(jmp_env) == 0) {
1824: /* now execute an invalid instruction */
1825: asm volatile("ud2");
1826: }
1827: printf("lock nop exception:\n");
1828: if (setjmp(jmp_env) == 0) {
1829: /* now execute an invalid instruction */
1830: asm volatile("lock nop");
1831: }
1.1.1.4 root 1832:
1.1 root 1833: printf("INT exception:\n");
1834: if (setjmp(jmp_env) == 0) {
1835: asm volatile ("int $0xfd");
1836: }
1837: if (setjmp(jmp_env) == 0) {
1838: asm volatile ("int $0x01");
1839: }
1840: if (setjmp(jmp_env) == 0) {
1841: asm volatile (".byte 0xcd, 0x03");
1842: }
1843: if (setjmp(jmp_env) == 0) {
1844: asm volatile ("int $0x04");
1845: }
1846: if (setjmp(jmp_env) == 0) {
1847: asm volatile ("int $0x05");
1848: }
1849:
1850: printf("INT3 exception:\n");
1851: if (setjmp(jmp_env) == 0) {
1852: asm volatile ("int3");
1853: }
1854:
1855: printf("CLI exception:\n");
1856: if (setjmp(jmp_env) == 0) {
1857: asm volatile ("cli");
1858: }
1859:
1860: printf("STI exception:\n");
1861: if (setjmp(jmp_env) == 0) {
1862: asm volatile ("cli");
1863: }
1864:
1865: #if !defined(__x86_64__)
1866: printf("INTO exception:\n");
1867: if (setjmp(jmp_env) == 0) {
1868: /* overflow exception */
1869: asm volatile ("addl $1, %0 ; into" : : "r" (0x7fffffff));
1870: }
1871: #endif
1872:
1873: printf("OUTB exception:\n");
1874: if (setjmp(jmp_env) == 0) {
1875: asm volatile ("outb %%al, %%dx" : : "d" (0x4321), "a" (0));
1876: }
1877:
1878: printf("INB exception:\n");
1879: if (setjmp(jmp_env) == 0) {
1880: asm volatile ("inb %%dx, %%al" : "=a" (val) : "d" (0x4321));
1881: }
1882:
1883: printf("REP OUTSB exception:\n");
1884: if (setjmp(jmp_env) == 0) {
1885: asm volatile ("rep outsb" : : "d" (0x4321), "S" (tab), "c" (1));
1886: }
1887:
1888: printf("REP INSB exception:\n");
1889: if (setjmp(jmp_env) == 0) {
1890: asm volatile ("rep insb" : : "d" (0x4321), "D" (tab), "c" (1));
1891: }
1892:
1893: printf("HLT exception:\n");
1894: if (setjmp(jmp_env) == 0) {
1895: asm volatile ("hlt");
1896: }
1897:
1898: printf("single step exception:\n");
1899: val = 0;
1900: if (setjmp(jmp_env) == 0) {
1901: asm volatile ("pushf\n"
1902: "orl $0x00100, (%%esp)\n"
1903: "popf\n"
1.1.1.4 root 1904: "movl $0xabcd, %0\n"
1.1 root 1905: "movl $0x0, %0\n" : "=m" (val) : : "cc", "memory");
1906: }
1907: printf("val=0x%x\n", val);
1908: }
1909:
1910: #if !defined(__x86_64__)
1911: /* specific precise single step test */
1912: void sig_trap_handler(int sig, siginfo_t *info, void *puc)
1913: {
1914: struct ucontext *uc = puc;
1915: printf("EIP=" FMTLX "\n", (long)uc->uc_mcontext.gregs[REG_EIP]);
1916: }
1917:
1918: const uint8_t sstep_buf1[4] = { 1, 2, 3, 4};
1919: uint8_t sstep_buf2[4];
1920:
1921: void test_single_step(void)
1922: {
1923: struct sigaction act;
1924: volatile int val;
1925: int i;
1926:
1927: val = 0;
1928: act.sa_sigaction = sig_trap_handler;
1929: sigemptyset(&act.sa_mask);
1930: act.sa_flags = SA_SIGINFO;
1931: sigaction(SIGTRAP, &act, NULL);
1932: asm volatile ("pushf\n"
1933: "orl $0x00100, (%%esp)\n"
1934: "popf\n"
1.1.1.4 root 1935: "movl $0xabcd, %0\n"
1.1 root 1936:
1937: /* jmp test */
1938: "movl $3, %%ecx\n"
1939: "1:\n"
1940: "addl $1, %0\n"
1941: "decl %%ecx\n"
1942: "jnz 1b\n"
1943:
1944: /* movsb: the single step should stop at each movsb iteration */
1945: "movl $sstep_buf1, %%esi\n"
1946: "movl $sstep_buf2, %%edi\n"
1947: "movl $0, %%ecx\n"
1948: "rep movsb\n"
1949: "movl $3, %%ecx\n"
1950: "rep movsb\n"
1951: "movl $1, %%ecx\n"
1952: "rep movsb\n"
1953:
1954: /* cmpsb: the single step should stop at each cmpsb iteration */
1955: "movl $sstep_buf1, %%esi\n"
1956: "movl $sstep_buf2, %%edi\n"
1957: "movl $0, %%ecx\n"
1958: "rep cmpsb\n"
1959: "movl $4, %%ecx\n"
1960: "rep cmpsb\n"
1.1.1.4 root 1961:
1.1 root 1962: /* getpid() syscall: single step should skip one
1963: instruction */
1964: "movl $20, %%eax\n"
1965: "int $0x80\n"
1966: "movl $0, %%eax\n"
1.1.1.4 root 1967:
1.1 root 1968: /* when modifying SS, trace is not done on the next
1969: instruction */
1970: "movl %%ss, %%ecx\n"
1971: "movl %%ecx, %%ss\n"
1972: "addl $1, %0\n"
1973: "movl $1, %%eax\n"
1974: "movl %%ecx, %%ss\n"
1975: "jmp 1f\n"
1976: "addl $1, %0\n"
1977: "1:\n"
1978: "movl $1, %%eax\n"
1979: "pushl %%ecx\n"
1980: "popl %%ss\n"
1981: "addl $1, %0\n"
1982: "movl $1, %%eax\n"
1.1.1.4 root 1983:
1.1 root 1984: "pushf\n"
1985: "andl $~0x00100, (%%esp)\n"
1986: "popf\n"
1.1.1.4 root 1987: : "=m" (val)
1988: :
1.1 root 1989: : "cc", "memory", "eax", "ecx", "esi", "edi");
1990: printf("val=%d\n", val);
1991: for(i = 0; i < 4; i++)
1992: printf("sstep_buf2[%d] = %d\n", i, sstep_buf2[i]);
1993: }
1994:
1995: /* self modifying code test */
1996: uint8_t code[] = {
1997: 0xb8, 0x1, 0x00, 0x00, 0x00, /* movl $1, %eax */
1998: 0xc3, /* ret */
1999: };
2000:
1.1.1.5 root 2001: asm(".section \".data\"\n"
2002: "smc_code2:\n"
1.1 root 2003: "movl 4(%esp), %eax\n"
2004: "movl %eax, smc_patch_addr2 + 1\n"
2005: "nop\n"
2006: "nop\n"
2007: "nop\n"
2008: "nop\n"
2009: "nop\n"
2010: "nop\n"
2011: "nop\n"
2012: "nop\n"
2013: "smc_patch_addr2:\n"
2014: "movl $1, %eax\n"
1.1.1.5 root 2015: "ret\n"
2016: ".previous\n"
2017: );
1.1 root 2018:
2019: typedef int FuncType(void);
2020: extern int smc_code2(int);
2021: void test_self_modifying_code(void)
2022: {
2023: int i;
2024: printf("self modifying code:\n");
2025: printf("func1 = 0x%x\n", ((FuncType *)code)());
2026: for(i = 2; i <= 4; i++) {
2027: code[1] = i;
2028: printf("func%d = 0x%x\n", i, ((FuncType *)code)());
2029: }
2030:
2031: /* more difficult test : the modified code is just after the
2032: modifying instruction. It is forbidden in Intel specs, but it
2033: is used by old DOS programs */
2034: for(i = 2; i <= 4; i++) {
2035: printf("smc_code2(%d) = %d\n", i, smc_code2(i));
2036: }
2037: }
2038: #endif
2039:
2040: long enter_stack[4096];
2041:
2042: #if defined(__x86_64__)
2043: #define RSP "%%rsp"
2044: #define RBP "%%rbp"
2045: #else
2046: #define RSP "%%esp"
2047: #define RBP "%%ebp"
2048: #endif
2049:
1.1.1.7 ! root 2050: #if !defined(__x86_64__)
! 2051: /* causes an infinite loop, disable it for now. */
! 2052: #define TEST_ENTER(size, stack_type, level)
! 2053: #else
1.1 root 2054: #define TEST_ENTER(size, stack_type, level)\
2055: {\
2056: long esp_save, esp_val, ebp_val, ebp_save, i;\
2057: stack_type *ptr, *stack_end, *stack_ptr;\
2058: memset(enter_stack, 0, sizeof(enter_stack));\
2059: stack_end = stack_ptr = (stack_type *)(enter_stack + 4096);\
2060: ebp_val = (long)stack_ptr;\
2061: for(i=1;i<=32;i++)\
2062: *--stack_ptr = i;\
2063: esp_val = (long)stack_ptr;\
2064: asm("mov " RSP ", %[esp_save]\n"\
2065: "mov " RBP ", %[ebp_save]\n"\
2066: "mov %[esp_val], " RSP "\n"\
2067: "mov %[ebp_val], " RBP "\n"\
2068: "enter" size " $8, $" #level "\n"\
2069: "mov " RSP ", %[esp_val]\n"\
2070: "mov " RBP ", %[ebp_val]\n"\
2071: "mov %[esp_save], " RSP "\n"\
2072: "mov %[ebp_save], " RBP "\n"\
2073: : [esp_save] "=r" (esp_save),\
2074: [ebp_save] "=r" (ebp_save),\
2075: [esp_val] "=r" (esp_val),\
2076: [ebp_val] "=r" (ebp_val)\
2077: : "[esp_val]" (esp_val),\
2078: "[ebp_val]" (ebp_val));\
2079: printf("level=%d:\n", level);\
2080: printf("esp_val=" FMTLX "\n", esp_val - (long)stack_end);\
2081: printf("ebp_val=" FMTLX "\n", ebp_val - (long)stack_end);\
2082: for(ptr = (stack_type *)esp_val; ptr < stack_end; ptr++)\
2083: printf(FMTLX "\n", (long)ptr[0]);\
2084: }
1.1.1.7 ! root 2085: #endif
1.1 root 2086:
2087: static void test_enter(void)
2088: {
2089: #if defined(__x86_64__)
2090: TEST_ENTER("q", uint64_t, 0);
2091: TEST_ENTER("q", uint64_t, 1);
2092: TEST_ENTER("q", uint64_t, 2);
2093: TEST_ENTER("q", uint64_t, 31);
2094: #else
2095: TEST_ENTER("l", uint32_t, 0);
2096: TEST_ENTER("l", uint32_t, 1);
2097: TEST_ENTER("l", uint32_t, 2);
2098: TEST_ENTER("l", uint32_t, 31);
2099: #endif
2100:
2101: TEST_ENTER("w", uint16_t, 0);
2102: TEST_ENTER("w", uint16_t, 1);
2103: TEST_ENTER("w", uint16_t, 2);
2104: TEST_ENTER("w", uint16_t, 31);
2105: }
2106:
2107: #ifdef TEST_SSE
2108:
2109: typedef int __m64 __attribute__ ((__mode__ (__V2SI__)));
1.1.1.5 root 2110: typedef float __m128 __attribute__ ((__mode__(__V4SF__)));
1.1 root 2111:
2112: typedef union {
2113: double d[2];
2114: float s[4];
2115: uint32_t l[4];
2116: uint64_t q[2];
2117: __m128 dq;
2118: } XMMReg;
2119:
2120: static uint64_t __attribute__((aligned(16))) test_values[4][2] = {
2121: { 0x456723c698694873, 0xdc515cff944a58ec },
2122: { 0x1f297ccd58bad7ab, 0x41f21efba9e3e146 },
2123: { 0x007c62c2085427f8, 0x231be9e8cde7438d },
2124: { 0x0f76255a085427f8, 0xc233e9e8c4c9439a },
2125: };
2126:
2127: #define SSE_OP(op)\
2128: {\
2129: asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\
2130: printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\
2131: #op,\
2132: a.q[1], a.q[0],\
2133: b.q[1], b.q[0],\
2134: r.q[1], r.q[0]);\
2135: }
2136:
2137: #define SSE_OP2(op)\
2138: {\
2139: int i;\
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] = test_values[2*i+1][0];\
2144: b.q[1] = test_values[2*i+1][1];\
2145: SSE_OP(op);\
2146: }\
2147: }
2148:
2149: #define MMX_OP2(op)\
2150: {\
2151: int i;\
2152: for(i=0;i<2;i++) {\
2153: a.q[0] = test_values[2*i][0];\
2154: b.q[0] = test_values[2*i+1][0];\
2155: asm volatile (#op " %2, %0" : "=y" (r.q[0]) : "0" (a.q[0]), "y" (b.q[0]));\
2156: printf("%-9s: a=" FMT64X " b=" FMT64X " r=" FMT64X "\n",\
2157: #op,\
2158: a.q[0],\
2159: b.q[0],\
2160: r.q[0]);\
2161: }\
2162: SSE_OP2(op);\
2163: }
2164:
2165: #define SHUF_OP(op, ib)\
2166: {\
2167: a.q[0] = test_values[0][0];\
2168: a.q[1] = test_values[0][1];\
2169: b.q[0] = test_values[1][0];\
2170: b.q[1] = test_values[1][1];\
2171: asm volatile (#op " $" #ib ", %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\
2172: printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\
2173: #op,\
2174: a.q[1], a.q[0],\
2175: b.q[1], b.q[0],\
2176: ib,\
2177: r.q[1], r.q[0]);\
2178: }
2179:
2180: #define PSHUF_OP(op, ib)\
2181: {\
2182: int i;\
2183: for(i=0;i<2;i++) {\
2184: a.q[0] = test_values[2*i][0];\
2185: a.q[1] = test_values[2*i][1];\
2186: asm volatile (#op " $" #ib ", %1, %0" : "=x" (r.dq) : "x" (a.dq));\
2187: printf("%-9s: a=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\
2188: #op,\
2189: a.q[1], a.q[0],\
2190: ib,\
2191: r.q[1], r.q[0]);\
2192: }\
2193: }
2194:
2195: #define SHIFT_IM(op, ib)\
2196: {\
2197: int i;\
2198: for(i=0;i<2;i++) {\
2199: a.q[0] = test_values[2*i][0];\
2200: a.q[1] = test_values[2*i][1];\
2201: asm volatile (#op " $" #ib ", %0" : "=x" (r.dq) : "0" (a.dq));\
2202: printf("%-9s: a=" FMT64X "" FMT64X " ib=%02x r=" FMT64X "" FMT64X "\n",\
2203: #op,\
2204: a.q[1], a.q[0],\
2205: ib,\
2206: r.q[1], r.q[0]);\
2207: }\
2208: }
2209:
2210: #define SHIFT_OP(op, ib)\
2211: {\
2212: int i;\
2213: SHIFT_IM(op, ib);\
2214: for(i=0;i<2;i++) {\
2215: a.q[0] = test_values[2*i][0];\
2216: a.q[1] = test_values[2*i][1];\
2217: b.q[0] = ib;\
2218: b.q[1] = 0;\
2219: asm volatile (#op " %2, %0" : "=x" (r.dq) : "0" (a.dq), "x" (b.dq));\
2220: printf("%-9s: a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\
2221: #op,\
2222: a.q[1], a.q[0],\
2223: b.q[1], b.q[0],\
2224: r.q[1], r.q[0]);\
2225: }\
2226: }
2227:
2228: #define MOVMSK(op)\
2229: {\
2230: int i, reg;\
2231: for(i=0;i<2;i++) {\
2232: a.q[0] = test_values[2*i][0];\
2233: a.q[1] = test_values[2*i][1];\
2234: asm volatile (#op " %1, %0" : "=r" (reg) : "x" (a.dq));\
2235: printf("%-9s: a=" FMT64X "" FMT64X " r=%08x\n",\
2236: #op,\
2237: a.q[1], a.q[0],\
2238: reg);\
2239: }\
2240: }
2241:
2242: #define SSE_OPS(a) \
2243: SSE_OP(a ## ps);\
2244: SSE_OP(a ## ss);
2245:
2246: #define SSE_OPD(a) \
2247: SSE_OP(a ## pd);\
2248: SSE_OP(a ## sd);
2249:
2250: #define SSE_COMI(op, field)\
2251: {\
2252: unsigned int eflags;\
2253: XMMReg a, b;\
2254: a.field[0] = a1;\
2255: b.field[0] = b1;\
2256: asm volatile (#op " %2, %1\n"\
2257: "pushf\n"\
2258: "pop %0\n"\
2259: : "=m" (eflags)\
2260: : "x" (a.dq), "x" (b.dq));\
2261: printf("%-9s: a=%f b=%f cc=%04x\n",\
2262: #op, a1, b1,\
2263: eflags & (CC_C | CC_P | CC_Z | CC_S | CC_O | CC_A));\
2264: }
2265:
2266: void test_sse_comi(double a1, double b1)
2267: {
2268: SSE_COMI(ucomiss, s);
2269: SSE_COMI(ucomisd, d);
2270: SSE_COMI(comiss, s);
2271: SSE_COMI(comisd, d);
2272: }
2273:
2274: #define CVT_OP_XMM(op)\
2275: {\
2276: asm volatile (#op " %1, %0" : "=x" (r.dq) : "x" (a.dq));\
2277: printf("%-9s: a=" FMT64X "" FMT64X " r=" FMT64X "" FMT64X "\n",\
2278: #op,\
2279: a.q[1], a.q[0],\
2280: r.q[1], r.q[0]);\
2281: }
2282:
1.1.1.2 root 2283: /* Force %xmm0 usage to avoid the case where both register index are 0
2284: to test intruction decoding more extensively */
1.1 root 2285: #define CVT_OP_XMM2MMX(op)\
2286: {\
1.1.1.2 root 2287: asm volatile (#op " %1, %0" : "=y" (r.q[0]) : "x" (a.dq) \
1.1.1.5 root 2288: : "%xmm0"); \
2289: asm volatile("emms\n"); \
1.1 root 2290: printf("%-9s: a=" FMT64X "" FMT64X " r=" FMT64X "\n",\
2291: #op,\
2292: a.q[1], a.q[0],\
2293: r.q[0]);\
2294: }
2295:
2296: #define CVT_OP_MMX2XMM(op)\
2297: {\
2298: asm volatile (#op " %1, %0" : "=x" (r.dq) : "y" (a.q[0]));\
1.1.1.5 root 2299: asm volatile("emms\n"); \
1.1 root 2300: printf("%-9s: a=" FMT64X " r=" FMT64X "" FMT64X "\n",\
2301: #op,\
2302: a.q[0],\
2303: r.q[1], r.q[0]);\
2304: }
2305:
2306: #define CVT_OP_REG2XMM(op)\
2307: {\
2308: asm volatile (#op " %1, %0" : "=x" (r.dq) : "r" (a.l[0]));\
2309: printf("%-9s: a=%08x r=" FMT64X "" FMT64X "\n",\
2310: #op,\
2311: a.l[0],\
2312: r.q[1], r.q[0]);\
2313: }
2314:
2315: #define CVT_OP_XMM2REG(op)\
2316: {\
2317: asm volatile (#op " %1, %0" : "=r" (r.l[0]) : "x" (a.dq));\
2318: printf("%-9s: a=" FMT64X "" FMT64X " r=%08x\n",\
2319: #op,\
2320: a.q[1], a.q[0],\
2321: r.l[0]);\
2322: }
2323:
2324: struct fpxstate {
2325: uint16_t fpuc;
2326: uint16_t fpus;
2327: uint16_t fptag;
2328: uint16_t fop;
2329: uint32_t fpuip;
2330: uint16_t cs_sel;
2331: uint16_t dummy0;
2332: uint32_t fpudp;
2333: uint16_t ds_sel;
2334: uint16_t dummy1;
2335: uint32_t mxcsr;
2336: uint32_t mxcsr_mask;
2337: uint8_t fpregs1[8 * 16];
2338: uint8_t xmm_regs[8 * 16];
2339: uint8_t dummy2[224];
2340: };
2341:
2342: static struct fpxstate fpx_state __attribute__((aligned(16)));
2343: static struct fpxstate fpx_state2 __attribute__((aligned(16)));
2344:
2345: void test_fxsave(void)
2346: {
2347: struct fpxstate *fp = &fpx_state;
2348: struct fpxstate *fp2 = &fpx_state2;
2349: int i, nb_xmm;
2350: XMMReg a, b;
2351: a.q[0] = test_values[0][0];
2352: a.q[1] = test_values[0][1];
2353: b.q[0] = test_values[1][0];
2354: b.q[1] = test_values[1][1];
2355:
2356: asm("movdqa %2, %%xmm0\n"
2357: "movdqa %3, %%xmm7\n"
2358: #if defined(__x86_64__)
2359: "movdqa %2, %%xmm15\n"
2360: #endif
2361: " fld1\n"
2362: " fldpi\n"
2363: " fldln2\n"
2364: " fxsave %0\n"
2365: " fxrstor %0\n"
2366: " fxsave %1\n"
2367: " fninit\n"
1.1.1.4 root 2368: : "=m" (*(uint32_t *)fp2), "=m" (*(uint32_t *)fp)
1.1 root 2369: : "m" (a), "m" (b));
2370: printf("fpuc=%04x\n", fp->fpuc);
2371: printf("fpus=%04x\n", fp->fpus);
2372: printf("fptag=%04x\n", fp->fptag);
2373: for(i = 0; i < 3; i++) {
2374: printf("ST%d: " FMT64X " %04x\n",
1.1.1.4 root 2375: i,
1.1 root 2376: *(uint64_t *)&fp->fpregs1[i * 16],
2377: *(uint16_t *)&fp->fpregs1[i * 16 + 8]);
2378: }
2379: printf("mxcsr=%08x\n", fp->mxcsr & 0x1f80);
2380: #if defined(__x86_64__)
2381: nb_xmm = 16;
2382: #else
2383: nb_xmm = 8;
2384: #endif
2385: for(i = 0; i < nb_xmm; i++) {
2386: printf("xmm%d: " FMT64X "" FMT64X "\n",
1.1.1.4 root 2387: i,
1.1 root 2388: *(uint64_t *)&fp->xmm_regs[i * 16],
2389: *(uint64_t *)&fp->xmm_regs[i * 16 + 8]);
2390: }
2391: }
2392:
2393: void test_sse(void)
2394: {
2395: XMMReg r, a, b;
2396: int i;
2397:
2398: MMX_OP2(punpcklbw);
2399: MMX_OP2(punpcklwd);
2400: MMX_OP2(punpckldq);
2401: MMX_OP2(packsswb);
2402: MMX_OP2(pcmpgtb);
2403: MMX_OP2(pcmpgtw);
2404: MMX_OP2(pcmpgtd);
2405: MMX_OP2(packuswb);
2406: MMX_OP2(punpckhbw);
2407: MMX_OP2(punpckhwd);
2408: MMX_OP2(punpckhdq);
2409: MMX_OP2(packssdw);
2410: MMX_OP2(pcmpeqb);
2411: MMX_OP2(pcmpeqw);
2412: MMX_OP2(pcmpeqd);
2413:
2414: MMX_OP2(paddq);
2415: MMX_OP2(pmullw);
2416: MMX_OP2(psubusb);
2417: MMX_OP2(psubusw);
2418: MMX_OP2(pminub);
2419: MMX_OP2(pand);
2420: MMX_OP2(paddusb);
2421: MMX_OP2(paddusw);
2422: MMX_OP2(pmaxub);
2423: MMX_OP2(pandn);
2424:
2425: MMX_OP2(pmulhuw);
2426: MMX_OP2(pmulhw);
1.1.1.4 root 2427:
1.1 root 2428: MMX_OP2(psubsb);
2429: MMX_OP2(psubsw);
2430: MMX_OP2(pminsw);
2431: MMX_OP2(por);
2432: MMX_OP2(paddsb);
2433: MMX_OP2(paddsw);
2434: MMX_OP2(pmaxsw);
2435: MMX_OP2(pxor);
2436: MMX_OP2(pmuludq);
2437: MMX_OP2(pmaddwd);
2438: MMX_OP2(psadbw);
2439: MMX_OP2(psubb);
2440: MMX_OP2(psubw);
2441: MMX_OP2(psubd);
2442: MMX_OP2(psubq);
2443: MMX_OP2(paddb);
2444: MMX_OP2(paddw);
2445: MMX_OP2(paddd);
2446:
2447: MMX_OP2(pavgb);
2448: MMX_OP2(pavgw);
2449:
2450: asm volatile ("pinsrw $1, %1, %0" : "=y" (r.q[0]) : "r" (0x12345678));
2451: printf("%-9s: r=" FMT64X "\n", "pinsrw", r.q[0]);
2452:
2453: asm volatile ("pinsrw $5, %1, %0" : "=x" (r.dq) : "r" (0x12345678));
2454: printf("%-9s: r=" FMT64X "" FMT64X "\n", "pinsrw", r.q[1], r.q[0]);
2455:
2456: a.q[0] = test_values[0][0];
2457: a.q[1] = test_values[0][1];
2458: asm volatile ("pextrw $1, %1, %0" : "=r" (r.l[0]) : "y" (a.q[0]));
2459: printf("%-9s: r=%08x\n", "pextrw", r.l[0]);
2460:
2461: asm volatile ("pextrw $5, %1, %0" : "=r" (r.l[0]) : "x" (a.dq));
2462: printf("%-9s: r=%08x\n", "pextrw", r.l[0]);
2463:
2464: asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "y" (a.q[0]));
2465: printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]);
1.1.1.4 root 2466:
1.1 root 2467: asm volatile ("pmovmskb %1, %0" : "=r" (r.l[0]) : "x" (a.dq));
2468: printf("%-9s: r=%08x\n", "pmovmskb", r.l[0]);
2469:
2470: {
2471: r.q[0] = -1;
2472: r.q[1] = -1;
2473:
2474: a.q[0] = test_values[0][0];
2475: a.q[1] = test_values[0][1];
2476: b.q[0] = test_values[1][0];
2477: b.q[1] = test_values[1][1];
1.1.1.4 root 2478: asm volatile("maskmovq %1, %0" :
1.1 root 2479: : "y" (a.q[0]), "y" (b.q[0]), "D" (&r)
1.1.1.4 root 2480: : "memory");
2481: printf("%-9s: r=" FMT64X " a=" FMT64X " b=" FMT64X "\n",
2482: "maskmov",
2483: r.q[0],
2484: a.q[0],
1.1 root 2485: b.q[0]);
1.1.1.4 root 2486: asm volatile("maskmovdqu %1, %0" :
1.1 root 2487: : "x" (a.dq), "x" (b.dq), "D" (&r)
1.1.1.4 root 2488: : "memory");
2489: printf("%-9s: r=" FMT64X "" FMT64X " a=" FMT64X "" FMT64X " b=" FMT64X "" FMT64X "\n",
2490: "maskmov",
2491: r.q[1], r.q[0],
2492: a.q[1], a.q[0],
1.1 root 2493: b.q[1], b.q[0]);
2494: }
2495:
2496: asm volatile ("emms");
2497:
2498: SSE_OP2(punpcklqdq);
2499: SSE_OP2(punpckhqdq);
2500: SSE_OP2(andps);
2501: SSE_OP2(andpd);
2502: SSE_OP2(andnps);
2503: SSE_OP2(andnpd);
2504: SSE_OP2(orps);
2505: SSE_OP2(orpd);
2506: SSE_OP2(xorps);
2507: SSE_OP2(xorpd);
2508:
2509: SSE_OP2(unpcklps);
2510: SSE_OP2(unpcklpd);
2511: SSE_OP2(unpckhps);
2512: SSE_OP2(unpckhpd);
2513:
2514: SHUF_OP(shufps, 0x78);
2515: SHUF_OP(shufpd, 0x02);
2516:
2517: PSHUF_OP(pshufd, 0x78);
2518: PSHUF_OP(pshuflw, 0x78);
2519: PSHUF_OP(pshufhw, 0x78);
2520:
2521: SHIFT_OP(psrlw, 7);
2522: SHIFT_OP(psrlw, 16);
2523: SHIFT_OP(psraw, 7);
2524: SHIFT_OP(psraw, 16);
2525: SHIFT_OP(psllw, 7);
2526: SHIFT_OP(psllw, 16);
2527:
2528: SHIFT_OP(psrld, 7);
2529: SHIFT_OP(psrld, 32);
2530: SHIFT_OP(psrad, 7);
2531: SHIFT_OP(psrad, 32);
2532: SHIFT_OP(pslld, 7);
2533: SHIFT_OP(pslld, 32);
2534:
2535: SHIFT_OP(psrlq, 7);
2536: SHIFT_OP(psrlq, 32);
2537: SHIFT_OP(psllq, 7);
2538: SHIFT_OP(psllq, 32);
2539:
2540: SHIFT_IM(psrldq, 16);
2541: SHIFT_IM(psrldq, 7);
2542: SHIFT_IM(pslldq, 16);
2543: SHIFT_IM(pslldq, 7);
2544:
2545: MOVMSK(movmskps);
2546: MOVMSK(movmskpd);
2547:
2548: /* FPU specific ops */
2549:
2550: {
2551: uint32_t mxcsr;
2552: asm volatile("stmxcsr %0" : "=m" (mxcsr));
2553: printf("mxcsr=%08x\n", mxcsr & 0x1f80);
2554: asm volatile("ldmxcsr %0" : : "m" (mxcsr));
2555: }
2556:
2557: test_sse_comi(2, -1);
2558: test_sse_comi(2, 2);
2559: test_sse_comi(2, 3);
2560: test_sse_comi(2, q_nan.d);
2561: test_sse_comi(q_nan.d, -1);
2562:
2563: for(i = 0; i < 2; i++) {
2564: a.s[0] = 2.7;
2565: a.s[1] = 3.4;
2566: a.s[2] = 4;
2567: a.s[3] = -6.3;
2568: b.s[0] = 45.7;
2569: b.s[1] = 353.4;
2570: b.s[2] = 4;
2571: b.s[3] = 56.3;
2572: if (i == 1) {
2573: a.s[0] = q_nan.d;
2574: b.s[3] = q_nan.d;
2575: }
2576:
2577: SSE_OPS(add);
2578: SSE_OPS(mul);
2579: SSE_OPS(sub);
2580: SSE_OPS(min);
2581: SSE_OPS(div);
2582: SSE_OPS(max);
2583: SSE_OPS(sqrt);
2584: SSE_OPS(cmpeq);
2585: SSE_OPS(cmplt);
2586: SSE_OPS(cmple);
2587: SSE_OPS(cmpunord);
2588: SSE_OPS(cmpneq);
2589: SSE_OPS(cmpnlt);
2590: SSE_OPS(cmpnle);
2591: SSE_OPS(cmpord);
1.1.1.4 root 2592:
2593:
1.1 root 2594: a.d[0] = 2.7;
2595: a.d[1] = -3.4;
2596: b.d[0] = 45.7;
2597: b.d[1] = -53.4;
2598: if (i == 1) {
2599: a.d[0] = q_nan.d;
2600: b.d[1] = q_nan.d;
2601: }
2602: SSE_OPD(add);
2603: SSE_OPD(mul);
2604: SSE_OPD(sub);
2605: SSE_OPD(min);
2606: SSE_OPD(div);
2607: SSE_OPD(max);
2608: SSE_OPD(sqrt);
2609: SSE_OPD(cmpeq);
2610: SSE_OPD(cmplt);
2611: SSE_OPD(cmple);
2612: SSE_OPD(cmpunord);
2613: SSE_OPD(cmpneq);
2614: SSE_OPD(cmpnlt);
2615: SSE_OPD(cmpnle);
2616: SSE_OPD(cmpord);
2617: }
2618:
2619: /* float to float/int */
2620: a.s[0] = 2.7;
2621: a.s[1] = 3.4;
2622: a.s[2] = 4;
2623: a.s[3] = -6.3;
2624: CVT_OP_XMM(cvtps2pd);
2625: CVT_OP_XMM(cvtss2sd);
2626: CVT_OP_XMM2MMX(cvtps2pi);
2627: CVT_OP_XMM2MMX(cvttps2pi);
2628: CVT_OP_XMM2REG(cvtss2si);
2629: CVT_OP_XMM2REG(cvttss2si);
2630: CVT_OP_XMM(cvtps2dq);
2631: CVT_OP_XMM(cvttps2dq);
2632:
2633: a.d[0] = 2.6;
2634: a.d[1] = -3.4;
2635: CVT_OP_XMM(cvtpd2ps);
2636: CVT_OP_XMM(cvtsd2ss);
2637: CVT_OP_XMM2MMX(cvtpd2pi);
2638: CVT_OP_XMM2MMX(cvttpd2pi);
2639: CVT_OP_XMM2REG(cvtsd2si);
2640: CVT_OP_XMM2REG(cvttsd2si);
2641: CVT_OP_XMM(cvtpd2dq);
2642: CVT_OP_XMM(cvttpd2dq);
2643:
1.1.1.2 root 2644: /* sse/mmx moves */
2645: CVT_OP_XMM2MMX(movdq2q);
2646: CVT_OP_MMX2XMM(movq2dq);
2647:
1.1 root 2648: /* int to float */
2649: a.l[0] = -6;
2650: a.l[1] = 2;
2651: a.l[2] = 100;
2652: a.l[3] = -60000;
2653: CVT_OP_MMX2XMM(cvtpi2ps);
2654: CVT_OP_MMX2XMM(cvtpi2pd);
2655: CVT_OP_REG2XMM(cvtsi2ss);
2656: CVT_OP_REG2XMM(cvtsi2sd);
2657: CVT_OP_XMM(cvtdq2ps);
2658: CVT_OP_XMM(cvtdq2pd);
2659:
2660: /* XXX: test PNI insns */
2661: #if 0
2662: SSE_OP2(movshdup);
2663: #endif
2664: asm volatile ("emms");
2665: }
2666:
2667: #endif
2668:
1.1.1.3 root 2669: #define TEST_CONV_RAX(op)\
2670: {\
2671: unsigned long a, r;\
2672: a = i2l(0x8234a6f8);\
2673: r = a;\
2674: asm volatile(#op : "=a" (r) : "0" (r));\
2675: printf("%-10s A=" FMTLX " R=" FMTLX "\n", #op, a, r);\
2676: }
2677:
2678: #define TEST_CONV_RAX_RDX(op)\
2679: {\
2680: unsigned long a, d, r, rh; \
2681: a = i2l(0x8234a6f8);\
2682: d = i2l(0x8345a1f2);\
2683: r = a;\
2684: rh = d;\
2685: asm volatile(#op : "=a" (r), "=d" (rh) : "0" (r), "1" (rh)); \
2686: printf("%-10s A=" FMTLX " R=" FMTLX ":" FMTLX "\n", #op, a, r, rh); \
2687: }
2688:
2689: void test_conv(void)
2690: {
2691: TEST_CONV_RAX(cbw);
2692: TEST_CONV_RAX(cwde);
2693: #if defined(__x86_64__)
2694: TEST_CONV_RAX(cdqe);
2695: #endif
2696:
2697: TEST_CONV_RAX_RDX(cwd);
2698: TEST_CONV_RAX_RDX(cdq);
2699: #if defined(__x86_64__)
2700: TEST_CONV_RAX_RDX(cqo);
2701: #endif
1.1.1.5 root 2702:
2703: {
2704: unsigned long a, r;
2705: a = i2l(0x12345678);
2706: asm volatile("bswapl %k0" : "=r" (r) : "0" (a));
2707: printf("%-10s: A=" FMTLX " R=" FMTLX "\n", "bswapl", a, r);
2708: }
2709: #if defined(__x86_64__)
2710: {
2711: unsigned long a, r;
2712: a = i2l(0x12345678);
2713: asm volatile("bswapq %0" : "=r" (r) : "0" (a));
2714: printf("%-10s: A=" FMTLX " R=" FMTLX "\n", "bswapq", a, r);
2715: }
2716: #endif
1.1.1.3 root 2717: }
2718:
1.1 root 2719: extern void *__start_initcall;
2720: extern void *__stop_initcall;
2721:
2722:
2723: int main(int argc, char **argv)
2724: {
2725: void **ptr;
2726: void (*func)(void);
2727:
2728: ptr = &__start_initcall;
2729: while (ptr != &__stop_initcall) {
2730: func = *ptr++;
2731: func();
2732: }
2733: test_bsx();
2734: test_mul();
2735: test_jcc();
1.1.1.5 root 2736: test_loop();
1.1 root 2737: test_floats();
2738: #if !defined(__x86_64__)
2739: test_bcd();
2740: #endif
2741: test_xchg();
2742: test_string();
2743: test_misc();
2744: test_lea();
2745: #ifdef TEST_SEGS
2746: test_segs();
2747: test_code16();
2748: #endif
2749: #ifdef TEST_VM86
2750: test_vm86();
2751: #endif
2752: #if !defined(__x86_64__)
1.1.1.5 root 2753: test_exceptions();
1.1 root 2754: test_self_modifying_code();
2755: test_single_step();
2756: #endif
2757: test_enter();
1.1.1.3 root 2758: test_conv();
1.1 root 2759: #ifdef TEST_SSE
2760: test_sse();
2761: test_fxsave();
2762: #endif
2763: return 0;
2764: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.