cutils.c@ 68389

Last change on this file since 68389 was 37689, checked in by vboxsync, 13 years ago
recompiler: Merged in changes from 0.13.0.
Property svn:eol-style set to `native`
File size: 19.5 KB

Line
1	/*
2	* Simple C functions to supplement the C library
3	*
4	* Copyright (c) 2006 Fabrice Bellard
5	*
6	* Permission is hereby granted, free of charge, to any person obtaining a copy
7	* of this software and associated documentation files (the "Software"), to deal
8	* in the Software without restriction, including without limitation the rights
9	* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10	* copies of the Software, and to permit persons to whom the Software is
11	* furnished to do so, subject to the following conditions:
12	*
13	* The above copyright notice and this permission notice shall be included in
14	* all copies or substantial portions of the Software.
15	*
16	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19	* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20	* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21	* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22	* THE SOFTWARE.
23	*/
24	#include "qemu-common.h"
25	#include "host-utils.h"
26
27	#ifdef VBOX
28	# include "osdep.h"
29
30
31	static inline int toupper(int ch) {
32	if ( (unsigned int)(ch - 'a') < 26u )
33	ch += 'A' - 'a';
34	return ch;
35	}
36
37	/* Quick sort from OpenSolaris:
38	http://src.opensolaris.org/source/raw/onnv/onnv-gate/usr/src/common/util/qsort.c */
39	/*
40	* choose a median of 3 values
41	*
42	* note: cstyle specifically prohibits nested conditional operators
43	* but this is the only way to do the median of 3 function in-line
44	*/
45	#define med3(a, b, c) (cmp((a), (b)) < 0) \
46	? ((cmp((b), (c)) < 0) ? (b) : (cmp((a), (c)) < 0) ? (c) : (a)) \
47	: ((cmp((b), (c)) > 0) ? (b) : (cmp((a), (c)) > 0) ? (c) : (a))
48
49	#define THRESH_L 5 /* threshold for insertion sort */
50	#define THRESH_M3 20 /* threshold for median of 3 */
51	#define THRESH_M9 50 /* threshold for median of 9 */
52
53	typedef struct {
54	char *b_lim;
55	size_t nrec;
56	} stk_t;
57
58	/*
59	* The following swap functions should not create a stack frame
60	* the SPARC call / return instruction will be executed
61	* but the a save / restore will not be executed
62	* which means we won't do a window turn with the spill / fill overhead
63	* verify this by examining the assembly code
64	*/
65
66	/* ARGSUSED */
67	static void
68	swapp32(uint32_t r1, uint32_t r2, size_t cnt)
69	{
70	uint32_t temp;
71
72	temp = *r1;
73	r1++ = r2;
74	*r2++ = temp;
75	}
76
77	/* ARGSUSED */
78	static void
79	swapp64(uint64_t r1, uint64_t r2, size_t cnt)
80	{
81	uint64_t temp;
82
83	temp = *r1;
84	r1++ = r2;
85	*r2++ = temp;
86	}
87
88	static void
89	swapi(uint32_t r1, uint32_t r2, size_t cnt)
90	{
91	uint32_t temp;
92
93	/* character by character */
94	while (cnt--) {
95	temp = *r1;
96	r1++ = r2;
97	*r2++ = temp;
98	}
99	}
100
101	static void
102	swapb(char r1, char r2, size_t cnt)
103	{
104	char temp;
105
106	/* character by character */
107	while (cnt--) {
108	temp = *r1;
109	r1++ = r2;
110	*r2++ = temp;
111	}
112	}
113
114	/*
115	* qsort() is a general purpose, in-place sorting routine using a
116	* user provided call back function for comparisons. This implementation
117	* utilizes a ternary quicksort algorithm, and cuts over to an
118	* insertion sort for partitions involving fewer than THRESH_L records.
119	*
120	* Potential User Errors
121	* There is no return value from qsort, this function has no method
122	* of alerting the user that a sort did not work or could not work.
123	* We do not print an error message or exit the process or thread,
124	* Even if we can detect an error, We CANNOT silently return without
125	* sorting the data, if we did so the user could never be sure the
126	* sort completed successfully.
127	* It is possible we could change the return value of sort from void
128	* to int and return success or some error codes, but this gets into
129	* standards and compatibility issues.
130	*
131	* Examples of qsort parameter errors might be
132	* 1) record size (rsiz) equal to 0
133	* qsort will loop and never return.
134	* 2) record size (rsiz) less than 0
135	* rsiz is unsigned, so a negative value is insanely large
136	* 3) number of records (nrec) is 0
137	* This is legal - qsort will return without examining any records
138	* 4) number of records (nrec) is less than 0
139	* nrec is unsigned, so a negative value is insanely large.
140	* 5) nrec * rsiz > memory allocation for sort array
141	* a segment violation may occur
142	* corruption of other memory may occur
143	* 6) The base address of the sort array is invalid
144	* a segment violation may occur
145	* corruption of other memory may occur
146	* 7) The user call back function is invalid
147	* we may get alignment errors or segment violations
148	* we may jump into never-never land
149	*
150	* Some less obvious errors might be
151	* 8) The user compare function is not comparing correctly
152	* 9) The user compare function modifies the data records
153	*/
154
155	void
156	qemu_qsort(
157	void *basep,
158	size_t nrec,
159	size_t rsiz,
160	int (cmp)(const void , const void *))
161	{
162	size_t i; /* temporary variable */
163
164	/* variables used by swap */
165	void (swapf)(char , char *, size_t);
166	size_t loops;
167
168	/* variables used by sort */
169	stk_t stack[8 * sizeof (nrec) + 1];
170	stk_t *sp;
171	char b_lim; / bottom limit */
172	char b_dup; / bottom duplicate */
173	char b_par; / bottom partition */
174	char t_lim; / top limit */
175	char t_dup; / top duplicate */
176	char t_par; / top partition */
177	char m1, m2, m3; / median pointers */
178	uintptr_t d_bytelength; /* byte length of duplicate records */
179	int b_nrec;
180	int t_nrec;
181	int cv; /* results of compare (bottom / top) */
182
183	/*
184	* choose a swap function based on alignment and size
185	*
186	* The qsort function sorts an array of fixed length records.
187	* We have very limited knowledge about the data record itself.
188	* It may be that the data record is in the array we are sorting
189	* or it may be that the array contains pointers or indexes to
190	* the actual data record and all that we are sorting is the indexes.
191	*
192	* The following decision will choose an optimal swap function
193	* based on the size and alignment of the data records
194	* swapp64 will swap 64 bit pointers
195	* swapp32 will swap 32 bit pointers
196	* swapi will swap an array of 32 bit integers
197	* swapb will swap an array of 8 bit characters
198	*
199	* swapi and swapb will also require the variable loops to be set
200	* to control the length of the array being swapped
201	*/
202	if ((((uintptr_t)basep & (sizeof (uint64_t) - 1)) == 0) &&
203	(rsiz == sizeof (uint64_t))) {
204	loops = 1;
205	swapf = (void ()(char , char *, size_t))swapp64;
206	} else if ((((uintptr_t)basep & (sizeof (uint32_t) - 1)) == 0) &&
207	(rsiz == sizeof (uint32_t))) {
208	loops = 1;
209	swapf = (void ()(char , char *, size_t))swapp32;
210	} else if ((((uintptr_t)basep & (sizeof (uint32_t) - 1)) == 0) &&
211	((rsiz & (sizeof (uint32_t) - 1)) == 0)) {
212	loops = rsiz / sizeof (int);
213	swapf = (void ()(char , char *, size_t))swapi;
214	} else {
215	loops = rsiz;
216	swapf = swapb;
217	}
218
219	/*
220	* qsort is a partitioning sort
221	*
222	* the stack is the bookkeeping mechanism to keep track of all
223	* the partitions.
224	*
225	* each sort pass takes one partition and sorts it into two partitions.
226	* at the top of the loop we simply take the partition on the top
227	* of the stack and sort it. See the comments at the bottom
228	* of the loop regarding which partitions to add in what order.
229	*
230	* initially put the whole partition on the stack
231	*/
232	sp = stack;
233	sp->b_lim = (char *)basep;
234	sp->nrec = nrec;
235	sp++;
236	while (sp > stack) {
237	sp--;
238	b_lim = sp->b_lim;
239	nrec = sp->nrec;
240
241	/*
242	* a linear insertion sort i faster than a qsort for
243	* very small number of records (THRESH_L)
244	*
245	* if number records < threshold use linear insertion sort
246	*
247	* this also handles the special case where the partition
248	* 0 or 1 records length.
249	*/
250	if (nrec < THRESH_L) {
251	/*
252	* Linear insertion sort
253	*/
254	t_par = b_lim;
255	for (i = 1; i < nrec; i++) {
256	t_par += rsiz;
257	b_par = t_par;
258	while (b_par > b_lim) {
259	b_par -= rsiz;
260	if ((*cmp)(b_par, b_par + rsiz) <= 0) {
261	break;
262	}
263	(*swapf)(b_par, b_par + rsiz, loops);
264	}
265	}
266
267	/*
268	* a linear insertion sort will put all records
269	* in their final position and will not create
270	* subpartitions.
271	*
272	* therefore when the insertion sort is complete
273	* just go to the top of the loop and get the
274	* next partition to sort.
275	*/
276	continue;
277	}
278
279	/* quicksort */
280
281	/*
282	* choose a pivot record
283	*
284	* Ideally the pivot record will divide the partition
285	* into two equal parts. however we have to balance the
286	* work involved in selecting the pivot record with the
287	* expected benefit.
288	*
289	* The choice of pivot record depends on the number of
290	* records in the partition
291	*
292	* for small partitions (nrec < THRESH_M3)
293	* we just select the record in the middle of the partition
294	*
295	* if (nrec >= THRESH_M3 && nrec < THRESH_M9)
296	* we select three values and choose the median of 3
297	*
298	* if (nrec >= THRESH_M9)
299	* then we use an approximate median of 9
300	* 9 records are selected and grouped in 3 groups of 3
301	* the median of each of these 3 groups is fed into another
302	* median of 3 decision.
303	*
304	* Each median of 3 decision is 2 or 3 compares,
305	* so median of 9 costs between 8 and 12 compares.
306	*
307	* i is byte distance between two consecutive samples
308	* m2 will point to the pivot record
309	*/
310	if (nrec < THRESH_M3) {
311	m2 = b_lim + (nrec / 2) * rsiz;
312	} else if (nrec < THRESH_M9) {
313	/* use median of 3 */
314	i = ((nrec - 1) / 2) * rsiz;
315	m2 = med3(b_lim, b_lim + i, b_lim + 2 * i);
316	} else {
317	/* approx median of 9 */
318	i = ((nrec - 1) / 8) * rsiz;
319	m1 = med3(b_lim, b_lim + i, b_lim + 2 * i);
320	m2 = med3(b_lim + 3 * i, b_lim + 4 * i, b_lim + 5 * i);
321	m3 = med3(b_lim + 6 * i, b_lim + 7 * i, b_lim + 8 * i);
322	m2 = med3(m1, m2, m3);
323	}
324
325	/*
326	* quick sort partitioning
327	*
328	* The partition limits are defined by bottom and top pointers
329	* b_lim and t_lim.
330	*
331	* qsort uses a fairly standard method of moving the
332	* partitioning pointers, b_par and t_par, to the middle of
333	* the partition and exchanging records that are in the
334	* wrong part of the partition.
335	*
336	* Two enhancements have been made to the basic algorithm.
337	* One for handling duplicate records and one to minimize
338	* the number of swaps.
339	*
340	* Two duplicate records pointers are (b_dup and t_dup) are
341	* initially set to b_lim and t_lim. Each time a record
342	* whose sort key value is equal to the pivot record is found
343	* it will be swapped with the record pointed to by
344	* b_dup or t_dup and the duplicate pointer will be
345	* incremented toward the center.
346	* When partitioning is complete, all the duplicate records
347	* will have been collected at the upper and lower limits of
348	* the partition and can easily be moved adjacent to the
349	* pivot record.
350	*
351	* The second optimization is to minimize the number of swaps.
352	* The pointer m2 points to the pivot record.
353	* During partitioning, if m2 is ever equal to the partitioning
354	* pointers, b_par or t_par, then b_par or t_par just moves
355	* onto the next record without doing a compare.
356	* If as a result of duplicate record detection,
357	* b_dup or t_dup is ever equal to m2,
358	* then m2 is changed to point to the duplicate record and
359	* b_dup or t_dup is incremented with out swapping records.
360	*
361	* When partitioning is done, we may not have the same pivot
362	* record that we started with, but we will have one with
363	* an equal sort key.
364	*/
365	b_dup = b_par = b_lim;
366	t_dup = t_par = t_lim = b_lim + rsiz * (nrec - 1);
367	for (;;) {
368
369	/* move bottom pointer up */
370	for (; b_par <= t_par; b_par += rsiz) {
371	if (b_par == m2) {
372	continue;
373	}
374	cv = cmp(b_par, m2);
375	if (cv > 0) {
376	break;
377	}
378	if (cv == 0) {
379	if (b_dup == m2) {
380	m2 = b_par;
381	} else if (b_dup != b_par) {
382	(*swapf)(b_dup, b_par, loops);
383	}
384	b_dup += rsiz;
385	}
386	}
387
388	/* move top pointer down */
389	for (; b_par < t_par; t_par -= rsiz) {
390	if (t_par == m2) {
391	continue;
392	}
393	cv = cmp(t_par, m2);
394	if (cv < 0) {
395	break;
396	}
397	if (cv == 0) {
398	if (t_dup == m2) {
399	m2 = t_par;
400	} else if (t_dup != t_par) {
401	(*swapf)(t_dup, t_par, loops);
402	}
403	t_dup -= rsiz;
404	}
405	}
406
407	/* break if we are done partitioning */
408	if (b_par >= t_par) {
409	break;
410	}
411
412	/* exchange records at upper and lower break points */
413	(*swapf)(b_par, t_par, loops);
414	b_par += rsiz;
415	t_par -= rsiz;
416	}
417
418	/*
419	* partitioning is now complete
420	*
421	* there are two termination conditions from the partitioning
422	* loop above. Either b_par or t_par have crossed or
423	* they are equal.
424	*
425	* we need to swap the pivot record to its final position
426	* m2 could be in either the upper or lower partitions
427	* or it could already be in its final position
428	*/
429	/*
430	* R[b_par] > R[m2]
431	* R[t_par] < R[m2]
432	*/
433	if (t_par < b_par) {
434	if (m2 < t_par) {
435	(*swapf)(m2, t_par, loops);
436	m2 = b_par = t_par;
437	} else if (m2 > b_par) {
438	(*swapf)(m2, b_par, loops);
439	m2 = t_par = b_par;
440	} else {
441	b_par = t_par = m2;
442	}
443	} else {
444	if (m2 < t_par) {
445	t_par = b_par = t_par - rsiz;
446	}
447	if (m2 != b_par) {
448	(*swapf)(m2, b_par, loops);
449	}
450	m2 = t_par;
451	}
452
453	/*
454	* move bottom duplicates next to pivot
455	* optimized to eliminate overlap
456	*/
457	d_bytelength = b_dup - b_lim;
458	if (b_par - b_dup < d_bytelength) {
459	b_dup = b_lim + (b_par - b_dup);
460	}
461	while (b_dup > b_lim) {
462	b_dup -= rsiz;
463	b_par -= rsiz;
464	(*swapf)(b_dup, b_par, loops);
465	}
466	b_par = m2 - d_bytelength;
467
468	/*
469	* move top duplicates next to pivot
470	*/
471	d_bytelength = t_lim - t_dup;
472	if (t_dup - t_par < d_bytelength) {
473	t_dup = t_lim - (t_dup - t_par);
474	}
475	while (t_dup < t_lim) {
476	t_dup += rsiz;
477	t_par += rsiz;
478	(*swapf)(t_dup, t_par, loops);
479	}
480	t_par = m2 + d_bytelength;
481
482	/*
483	* when a qsort pass completes there are three partitions
484	* 1) the lower contains all records less than pivot
485	* 2) the upper contains all records greater than pivot
486	* 3) the pivot partition contains all record equal to pivot
487	*
488	* all records in the pivot partition are in their final
489	* position and do not need to be accounted for by the stack
490	*
491	* when adding partitions to the stack
492	* it is important to add the largest partition first
493	* to prevent stack overflow.
494	*
495	* calculate number of unsorted records in top and bottom
496	* push resulting partitions on stack
497	*/
498	b_nrec = (b_par - b_lim) / rsiz;
499	t_nrec = (t_lim - t_par) / rsiz;
500	if (b_nrec < t_nrec) {
501	sp->b_lim = t_par + rsiz;
502	sp->nrec = t_nrec;
503	sp++;
504	sp->b_lim = b_lim;
505	sp->nrec = b_nrec;
506	sp++;
507	} else {
508	sp->b_lim = b_lim;
509	sp->nrec = b_nrec;
510	sp++;
511	sp->b_lim = t_par + rsiz;
512	sp->nrec = t_nrec;
513	sp++;
514	}
515	}
516	}
517
518	#endif /* VBOX */
519
520	void pstrcpy(char buf, int buf_size, const char str)
521	{
522	int c;
523	char *q = buf;
524
525	if (buf_size <= 0)
526	return;
527
528	for(;;) {
529	c = *str++;
530	if (c == 0 \|\| q >= buf + buf_size - 1)
531	break;
532	*q++ = c;
533	}
534	*q = '\0';
535	}
536
537	/* strcat and truncate. */
538	char pstrcat(char buf, int buf_size, const char *s)
539	{
540	int len;
541	len = strlen(buf);
542	if (len < buf_size)
543	pstrcpy(buf + len, buf_size - len, s);
544	return buf;
545	}
546
547	int strstart(const char str, const char val, const char **ptr)
548	{
549	const char p, q;
550	p = str;
551	q = val;
552	while (*q != '\0') {
553	if (p != q)
554	return 0;
555	p++;
556	q++;
557	}
558	if (ptr)
559	*ptr = p;
560	return 1;
561	}
562
563	int stristart(const char str, const char val, const char **ptr)
564	{
565	const char p, q;
566	p = str;
567	q = val;
568	while (*q != '\0') {
569	if (qemu_toupper(p) != qemu_toupper(q))
570	return 0;
571	p++;
572	q++;
573	}
574	if (ptr)
575	*ptr = p;
576	return 1;
577	}
578
579	/* XXX: use host strnlen if available ? */
580	int qemu_strnlen(const char *s, int max_len)
581	{
582	int i;
583
584	for(i = 0; i < max_len; i++) {
585	if (s[i] == '\0') {
586	break;
587	}
588	}
589	return i;
590	}
591
592	#ifndef VBOX
593	time_t mktimegm(struct tm *tm)
594	{
595	time_t t;
596	int y = tm->tm_year + 1900, m = tm->tm_mon + 1, d = tm->tm_mday;
597	if (m < 3) {
598	m += 12;
599	y--;
600	}
601	t = 86400 * (d + (153 * m - 457) / 5 + 365 * y + y / 4 - y / 100 +
602	y / 400 - 719469);
603	t += 3600 * tm->tm_hour + 60 * tm->tm_min + tm->tm_sec;
604	return t;
605	}
606	#endif /* !VBOX */
607
608	int qemu_fls(int i)
609	{
610	return 32 - clz32(i);
611	}
612
613	#ifndef VBOX
614
615	/*
616	* Make sure data goes on disk, but if possible do not bother to
617	* write out the inode just for timestamp updates.
618	*
619	* Unfortunately even in 2009 many operating systems do not support
620	* fdatasync and have to fall back to fsync.
621	*/
622	int qemu_fdatasync(int fd)
623	{
624	#ifdef CONFIG_FDATASYNC
625	return fdatasync(fd);
626	#else
627	return fsync(fd);
628	#endif
629	}
630
631	/* io vectors */
632
633	void qemu_iovec_init(QEMUIOVector *qiov, int alloc_hint)
634	{
635	qiov->iov = qemu_malloc(alloc_hint * sizeof(struct iovec));
636	qiov->niov = 0;
637	qiov->nalloc = alloc_hint;
638	qiov->size = 0;
639	}
640
641	void qemu_iovec_init_external(QEMUIOVector qiov, struct iovec iov, int niov)
642	{
643	int i;
644
645	qiov->iov = iov;
646	qiov->niov = niov;
647	qiov->nalloc = -1;
648	qiov->size = 0;
649	for (i = 0; i < niov; i++)
650	qiov->size += iov[i].iov_len;
651	}
652
653	void qemu_iovec_add(QEMUIOVector qiov, void base, size_t len)
654	{
655	assert(qiov->nalloc != -1);
656
657	if (qiov->niov == qiov->nalloc) {
658	qiov->nalloc = 2 * qiov->nalloc + 1;
659	qiov->iov = qemu_realloc(qiov->iov, qiov->nalloc * sizeof(struct iovec));
660	}
661	qiov->iov[qiov->niov].iov_base = base;
662	qiov->iov[qiov->niov].iov_len = len;
663	qiov->size += len;
664	++qiov->niov;
665	}
666
667	/*
668	* Copies iovecs from src to the end dst until src is completely copied or the
669	* total size of the copied iovec reaches size. The size of the last copied
670	* iovec is changed in order to fit the specified total size if it isn't a
671	* perfect fit already.
672	*/
673	void qemu_iovec_concat(QEMUIOVector dst, QEMUIOVector src, size_t size)
674	{
675	int i;
676	size_t done;
677
678	assert(dst->nalloc != -1);
679
680	done = 0;
681	for (i = 0; (i < src->niov) && (done != size); i++) {
682	if (done + src->iov[i].iov_len > size) {
683	qemu_iovec_add(dst, src->iov[i].iov_base, size - done);
684	break;
685	} else {
686	qemu_iovec_add(dst, src->iov[i].iov_base, src->iov[i].iov_len);
687	}
688	done += src->iov[i].iov_len;
689	}
690	}
691
692	void qemu_iovec_destroy(QEMUIOVector *qiov)
693	{
694	assert(qiov->nalloc != -1);
695
696	qemu_free(qiov->iov);
697	}
698
699	void qemu_iovec_reset(QEMUIOVector *qiov)
700	{
701	assert(qiov->nalloc != -1);
702
703	qiov->niov = 0;
704	qiov->size = 0;
705	}
706
707	void qemu_iovec_to_buffer(QEMUIOVector qiov, void buf)
708	{
709	uint8_t p = (uint8_t )buf;
710	int i;
711
712	for (i = 0; i < qiov->niov; ++i) {
713	memcpy(p, qiov->iov[i].iov_base, qiov->iov[i].iov_len);
714	p += qiov->iov[i].iov_len;
715	}
716	}
717
718	void qemu_iovec_from_buffer(QEMUIOVector qiov, const void buf, size_t count)
719	{
720	const uint8_t p = (const uint8_t )buf;
721	size_t copy;
722	int i;
723
724	for (i = 0; i < qiov->niov && count; ++i) {
725	copy = count;
726	if (copy > qiov->iov[i].iov_len)
727	copy = qiov->iov[i].iov_len;
728	memcpy(qiov->iov[i].iov_base, p, copy);
729	p += copy;
730	count -= copy;
731	}
732	}
733
734	#ifndef _WIN32
735	/* Sets a specific flag */
736	int fcntl_setfl(int fd, int flag)
737	{
738	int flags;
739
740	flags = fcntl(fd, F_GETFL);
741	if (flags == -1)
742	return -errno;
743
744	if (fcntl(fd, F_SETFL, flags \| flag) == -1)
745	return -errno;
746
747	return 0;
748	}
749	#endif
750
751	#endif /* !VBOX */
752

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/recompiler/cutils.c@ 68389

Download in other formats: