1 : // Copyright 2011 Google Inc. All Rights Reserved.
2 : //
3 : // Redistribution and use in source and binary forms, with or without
4 : // modification, are permitted provided that the following conditions are
5 : // met:
6 : //
7 : // * Redistributions of source code must retain the above copyright
8 : // notice, this list of conditions and the following disclaimer.
9 : // * Redistributions in binary form must reproduce the above
10 : // copyright notice, this list of conditions and the following disclaimer
11 : // in the documentation and/or other materials provided with the
12 : // distribution.
13 : // * Neither the name of Google Inc. nor the names of its
14 : // contributors may be used to endorse or promote products derived from
15 : // this software without specific prior written permission.
16 : //
17 : // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
18 : // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
19 : // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
20 : // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
21 : // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
22 : // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
23 : // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 : // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 : // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 : // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
27 : // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
28 : //
29 : // Various stubs for the open-source version of Snappy.
30 :
31 : #ifndef UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_
32 : #define UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_
33 :
34 : #ifdef HAVE_CONFIG_H
35 : #include "config.h"
36 : #endif
37 :
38 : #include <iostream>
39 : #include <string>
40 :
41 : #include <assert.h>
42 : #include <stdlib.h>
43 : #include <string.h>
44 :
45 : #ifdef HAVE_SYS_MMAN_H
46 : #include <sys/mman.h>
47 : #endif
48 :
49 : #include "snappy-stubs-public.h"
50 :
51 : #if defined(__x86_64__)
52 :
53 : // Enable 64-bit optimized versions of some routines.
54 : #define ARCH_K8 1
55 :
56 : #endif
57 :
58 : // Needed by OS X, among others.
59 : #ifndef MAP_ANONYMOUS
60 : #define MAP_ANONYMOUS MAP_ANON
61 : #endif
62 :
63 : // Pull in std::min, std::ostream, and the likes. This is safe because this
64 : // header file is never used from any public header files.
65 : using namespace std;
66 :
67 : // The size of an array, if known at compile-time.
68 : // Will give unexpected results if used on a pointer.
69 : // We undefine it first, since some compilers already have a definition.
70 : #ifdef ARRAYSIZE
71 : #undef ARRAYSIZE
72 : #endif
73 : #define ARRAYSIZE(a) (sizeof(a) / sizeof(*(a)))
74 :
75 : // Static prediction hints.
76 : #ifdef HAVE_BUILTIN_EXPECT
77 : #define PREDICT_FALSE(x) (__builtin_expect(x, 0))
78 : #define PREDICT_TRUE(x) (__builtin_expect(!!(x), 1))
79 : #else
80 : #define PREDICT_FALSE(x) x
81 : #define PREDICT_TRUE(x) x
82 : #endif
83 :
84 : // This is only used for recomputing the tag byte table used during
85 : // decompression; for simplicity we just remove it from the open-source
86 : // version (anyone who wants to regenerate it can just do the call
87 : // themselves within main()).
88 : #define DEFINE_bool(flag_name, default_value, description) \
89 : bool FLAGS_ ## flag_name = default_value
90 : #define DECLARE_bool(flag_name) \
91 : extern bool FLAGS_ ## flag_name
92 :
93 : namespace snappy {
94 :
95 : static const uint32 kuint32max = static_cast<uint32>(0xFFFFFFFF);
96 : static const int64 kint64max = static_cast<int64>(0x7FFFFFFFFFFFFFFFLL);
97 :
98 : // Logging.
99 :
100 : #define LOG(level) LogMessage()
101 : #define VLOG(level) true ? (void)0 : \
102 : snappy::LogMessageVoidify() & snappy::LogMessage()
103 :
104 : class LogMessage {
105 : public:
106 0 : LogMessage() { }
107 : ~LogMessage() {
108 : cerr << endl;
109 : }
110 :
111 0 : LogMessage& operator<<(const std::string& msg) {
112 0 : cerr << msg;
113 0 : return *this;
114 : }
115 0 : LogMessage& operator<<(int x) {
116 0 : cerr << x;
117 0 : return *this;
118 : }
119 : };
120 :
121 : // Asserts, both versions activated in debug mode only,
122 : // and ones that are always active.
123 :
124 : #define CRASH_UNLESS(condition) \
125 : PREDICT_TRUE(condition) ? (void)0 : \
126 : snappy::LogMessageVoidify() & snappy::LogMessageCrash()
127 :
128 : class LogMessageCrash : public LogMessage {
129 : public:
130 0 : LogMessageCrash() { }
131 0 : ~LogMessageCrash() {
132 0 : cerr << endl;
133 0 : abort();
134 : }
135 : };
136 :
137 : // This class is used to explicitly ignore values in the conditional
138 : // logging macros. This avoids compiler warnings like "value computed
139 : // is not used" and "statement has no effect".
140 :
141 : class LogMessageVoidify {
142 : public:
143 0 : LogMessageVoidify() { }
144 : // This has to be an operator with a precedence lower than << but
145 : // higher than ?:
146 0 : void operator&(const LogMessage&) { }
147 : };
148 :
149 : #define CHECK(cond) CRASH_UNLESS(cond)
150 : #define CHECK_LE(a, b) CRASH_UNLESS((a) <= (b))
151 : #define CHECK_GE(a, b) CRASH_UNLESS((a) >= (b))
152 : #define CHECK_EQ(a, b) CRASH_UNLESS((a) == (b))
153 : #define CHECK_NE(a, b) CRASH_UNLESS((a) != (b))
154 : #define CHECK_LT(a, b) CRASH_UNLESS((a) < (b))
155 : #define CHECK_GT(a, b) CRASH_UNLESS((a) > (b))
156 :
157 : #ifdef NDEBUG
158 :
159 : #define DCHECK(cond) CRASH_UNLESS(true)
160 : #define DCHECK_LE(a, b) CRASH_UNLESS(true)
161 : #define DCHECK_GE(a, b) CRASH_UNLESS(true)
162 : #define DCHECK_EQ(a, b) CRASH_UNLESS(true)
163 : #define DCHECK_NE(a, b) CRASH_UNLESS(true)
164 : #define DCHECK_LT(a, b) CRASH_UNLESS(true)
165 : #define DCHECK_GT(a, b) CRASH_UNLESS(true)
166 :
167 : #else
168 :
169 : #define DCHECK(cond) CHECK(cond)
170 : #define DCHECK_LE(a, b) CHECK_LE(a, b)
171 : #define DCHECK_GE(a, b) CHECK_GE(a, b)
172 : #define DCHECK_EQ(a, b) CHECK_EQ(a, b)
173 : #define DCHECK_NE(a, b) CHECK_NE(a, b)
174 : #define DCHECK_LT(a, b) CHECK_LT(a, b)
175 : #define DCHECK_GT(a, b) CHECK_GT(a, b)
176 :
177 : #endif
178 :
179 : // Potentially unaligned loads and stores.
180 :
181 : #if defined(__i386__) || defined(__x86_64__) || defined(__powerpc__)
182 :
183 : #define UNALIGNED_LOAD16(_p) (*reinterpret_cast<const uint16 *>(_p))
184 : #define UNALIGNED_LOAD32(_p) (*reinterpret_cast<const uint32 *>(_p))
185 : #define UNALIGNED_LOAD64(_p) (*reinterpret_cast<const uint64 *>(_p))
186 :
187 : #define UNALIGNED_STORE16(_p, _val) (*reinterpret_cast<uint16 *>(_p) = (_val))
188 : #define UNALIGNED_STORE32(_p, _val) (*reinterpret_cast<uint32 *>(_p) = (_val))
189 : #define UNALIGNED_STORE64(_p, _val) (*reinterpret_cast<uint64 *>(_p) = (_val))
190 :
191 : #else
192 :
193 : // These functions are provided for architectures that don't support
194 : // unaligned loads and stores.
195 :
196 : inline uint16 UNALIGNED_LOAD16(const void *p) {
197 : uint16 t;
198 : memcpy(&t, p, sizeof t);
199 : return t;
200 : }
201 :
202 : inline uint32 UNALIGNED_LOAD32(const void *p) {
203 : uint32 t;
204 : memcpy(&t, p, sizeof t);
205 : return t;
206 : }
207 :
208 : inline uint64 UNALIGNED_LOAD64(const void *p) {
209 : uint64 t;
210 : memcpy(&t, p, sizeof t);
211 : return t;
212 : }
213 :
214 : inline void UNALIGNED_STORE16(void *p, uint16 v) {
215 : memcpy(p, &v, sizeof v);
216 : }
217 :
218 : inline void UNALIGNED_STORE32(void *p, uint32 v) {
219 : memcpy(p, &v, sizeof v);
220 : }
221 :
222 : inline void UNALIGNED_STORE64(void *p, uint64 v) {
223 : memcpy(p, &v, sizeof v);
224 : }
225 :
226 : #endif
227 :
228 : // The following guarantees declaration of the byte swap functions.
229 : #ifdef WORDS_BIGENDIAN
230 :
231 : #ifdef HAVE_SYS_BYTEORDER_H
232 : #include <sys/byteorder.h>
233 : #endif
234 :
235 : #ifdef HAVE_SYS_ENDIAN_H
236 : #include <sys/endian.h>
237 : #endif
238 :
239 : #ifdef _MSC_VER
240 : #include <stdlib.h>
241 : #define bswap_16(x) _byteswap_ushort(x)
242 : #define bswap_32(x) _byteswap_ulong(x)
243 : #define bswap_64(x) _byteswap_uint64(x)
244 :
245 : #elif defined(__APPLE__)
246 : // Mac OS X / Darwin features
247 : #include <libkern/OSByteOrder.h>
248 : #define bswap_16(x) OSSwapInt16(x)
249 : #define bswap_32(x) OSSwapInt32(x)
250 : #define bswap_64(x) OSSwapInt64(x)
251 :
252 : #elif defined(HAVE_BYTESWAP_H)
253 : #include <byteswap.h>
254 :
255 : #elif defined(bswap32)
256 : // FreeBSD defines bswap{16,32,64} in <sys/endian.h> (already #included).
257 : #define bswap_16(x) bswap16(x)
258 : #define bswap_32(x) bswap32(x)
259 : #define bswap_64(x) bswap64(x)
260 :
261 : #elif defined(BSWAP_64)
262 : // Solaris 10 defines BSWAP_{16,32,64} in <sys/byteorder.h> (already #included).
263 : #define bswap_16(x) BSWAP_16(x)
264 : #define bswap_32(x) BSWAP_32(x)
265 : #define bswap_64(x) BSWAP_64(x)
266 :
267 : #else
268 :
269 : inline uint16 bswap_16(uint16 x) {
270 : return (x << 8) | (x >> 8);
271 : }
272 :
273 : inline uint32 bswap_32(uint32 x) {
274 : x = ((x & 0xff00ff00UL) >> 8) | ((x & 0x00ff00ffUL) << 8);
275 : return (x >> 16) | (x << 16);
276 : }
277 :
278 : inline uint64 bswap_64(uint64 x) {
279 : x = ((x & 0xff00ff00ff00ff00ULL) >> 8) | ((x & 0x00ff00ff00ff00ffULL) << 8);
280 : x = ((x & 0xffff0000ffff0000ULL) >> 16) | ((x & 0x0000ffff0000ffffULL) << 16);
281 : return (x >> 32) | (x << 32);
282 : }
283 :
284 : #endif
285 :
286 : #endif // WORDS_BIGENDIAN
287 :
288 : // Convert to little-endian storage, opposite of network format.
289 : // Convert x from host to little endian: x = LittleEndian.FromHost(x);
290 : // convert x from little endian to host: x = LittleEndian.ToHost(x);
291 : //
292 : // Store values into unaligned memory converting to little endian order:
293 : // LittleEndian.Store16(p, x);
294 : //
295 : // Load unaligned values stored in little endian converting to host order:
296 : // x = LittleEndian.Load16(p);
297 : class LittleEndian {
298 : public:
299 : // Conversion functions.
300 : #ifdef WORDS_BIGENDIAN
301 :
302 : static uint16 FromHost16(uint16 x) { return bswap_16(x); }
303 : static uint16 ToHost16(uint16 x) { return bswap_16(x); }
304 :
305 : static uint32 FromHost32(uint32 x) { return bswap_32(x); }
306 : static uint32 ToHost32(uint32 x) { return bswap_32(x); }
307 :
308 : static bool IsLittleEndian() { return false; }
309 :
310 : #else // !defined(WORDS_BIGENDIAN)
311 :
312 57 : static uint16 FromHost16(uint16 x) { return x; }
313 : static uint16 ToHost16(uint16 x) { return x; }
314 :
315 : static uint32 FromHost32(uint32 x) { return x; }
316 4250 : static uint32 ToHost32(uint32 x) { return x; }
317 :
318 6477 : static bool IsLittleEndian() { return true; }
319 :
320 : #endif // !defined(WORDS_BIGENDIAN)
321 :
322 : // Functions to do unaligned loads and stores in little-endian order.
323 : static uint16 Load16(const void *p) {
324 : return ToHost16(UNALIGNED_LOAD16(p));
325 : }
326 :
327 57 : static void Store16(void *p, uint16 v) {
328 57 : UNALIGNED_STORE16(p, FromHost16(v));
329 57 : }
330 :
331 4250 : static uint32 Load32(const void *p) {
332 4250 : return ToHost32(UNALIGNED_LOAD32(p));
333 : }
334 :
335 : static void Store32(void *p, uint32 v) {
336 : UNALIGNED_STORE32(p, FromHost32(v));
337 : }
338 : };
339 :
340 : // Some bit-manipulation functions.
341 : class Bits {
342 : public:
343 : // Return floor(log2(n)) for positive integer n. Returns -1 iff n == 0.
344 : static int Log2Floor(uint32 n);
345 :
346 : // Return the first set least / most significant bit, 0-indexed. Returns an
347 : // undefined value if n == 0. FindLSBSetNonZero() is similar to ffs() except
348 : // that it's 0-indexed.
349 : static int FindLSBSetNonZero(uint32 n);
350 : static int FindLSBSetNonZero64(uint64 n);
351 :
352 : private:
353 : DISALLOW_COPY_AND_ASSIGN(Bits);
354 : };
355 :
356 : #ifdef HAVE_BUILTIN_CTZ
357 :
358 : inline int Bits::Log2Floor(uint32 n) {
359 : return n == 0 ? -1 : 31 ^ __builtin_clz(n);
360 : }
361 :
362 : inline int Bits::FindLSBSetNonZero(uint32 n) {
363 : return __builtin_ctz(n);
364 : }
365 :
366 : inline int Bits::FindLSBSetNonZero64(uint64 n) {
367 : return __builtin_ctzll(n);
368 : }
369 :
370 : #else // Portable versions.
371 :
372 1837 : inline int Bits::Log2Floor(uint32 n) {
373 1837 : if (n == 0)
374 0 : return -1;
375 1837 : int log = 0;
376 1837 : uint32 value = n;
377 11022 : for (int i = 4; i >= 0; --i) {
378 9185 : int shift = (1 << i);
379 9185 : uint32 x = value >> shift;
380 9185 : if (x != 0) {
381 1837 : value = x;
382 1837 : log += shift;
383 : }
384 : }
385 1837 : assert(value == 1);
386 1837 : return log;
387 : }
388 :
389 1657 : inline int Bits::FindLSBSetNonZero(uint32 n) {
390 1657 : int rc = 31;
391 9942 : for (int i = 4, shift = 1 << 4; i >= 0; --i) {
392 8285 : const uint32 x = n << shift;
393 8285 : if (x != 0) {
394 4985 : n = x;
395 4985 : rc -= shift;
396 : }
397 8285 : shift >>= 1;
398 : }
399 1657 : return rc;
400 : }
401 :
402 : // FindLSBSetNonZero64() is defined in terms of FindLSBSetNonZero().
403 : inline int Bits::FindLSBSetNonZero64(uint64 n) {
404 : const uint32 bottombits = static_cast<uint32>(n);
405 : if (bottombits == 0) {
406 : // Bottom bits are zero, so scan in top bits
407 : return 32 + FindLSBSetNonZero(static_cast<uint32>(n >> 32));
408 : } else {
409 : return FindLSBSetNonZero(bottombits);
410 : }
411 : }
412 :
413 : #endif // End portable versions.
414 :
415 : // Variable-length integer encoding.
416 : class Varint {
417 : public:
418 : // Maximum lengths of varint encoding of uint32.
419 : static const int kMax32 = 5;
420 :
421 : // Attempts to parse a varint32 from a prefix of the bytes in [ptr,limit-1].
422 : // Never reads a character at or beyond limit. If a valid/terminated varint32
423 : // was found in the range, stores it in *OUTPUT and returns a pointer just
424 : // past the last byte of the varint32. Else returns NULL. On success,
425 : // "result <= limit".
426 : static const char* Parse32WithLimit(const char* ptr, const char* limit,
427 : uint32* OUTPUT);
428 :
429 : // REQUIRES "ptr" points to a buffer of length sufficient to hold "v".
430 : // EFFECTS Encodes "v" into "ptr" and returns a pointer to the
431 : // byte just past the last encoded byte.
432 : static char* Encode32(char* ptr, uint32 v);
433 :
434 : // EFFECTS Appends the varint representation of "value" to "*s".
435 : static void Append32(string* s, uint32 value);
436 : };
437 :
438 1970 : inline const char* Varint::Parse32WithLimit(const char* p,
439 : const char* l,
440 : uint32* OUTPUT) {
441 1970 : const unsigned char* ptr = reinterpret_cast<const unsigned char*>(p);
442 1970 : const unsigned char* limit = reinterpret_cast<const unsigned char*>(l);
443 : uint32 b, result;
444 1970 : if (ptr >= limit) return NULL;
445 1970 : b = *(ptr++); result = b & 127; if (b < 128) goto done;
446 224 : if (ptr >= limit) return NULL;
447 224 : b = *(ptr++); result |= (b & 127) << 7; if (b < 128) goto done;
448 0 : if (ptr >= limit) return NULL;
449 0 : b = *(ptr++); result |= (b & 127) << 14; if (b < 128) goto done;
450 0 : if (ptr >= limit) return NULL;
451 0 : b = *(ptr++); result |= (b & 127) << 21; if (b < 128) goto done;
452 0 : if (ptr >= limit) return NULL;
453 0 : b = *(ptr++); result |= (b & 127) << 28; if (b < 16) goto done;
454 0 : return NULL; // Value is too long to be a varint32
455 : done:
456 1970 : *OUTPUT = result;
457 1970 : return reinterpret_cast<const char*>(ptr);
458 : }
459 :
460 1837 : inline char* Varint::Encode32(char* sptr, uint32 v) {
461 : // Operate on characters as unsigneds
462 1837 : unsigned char* ptr = reinterpret_cast<unsigned char*>(sptr);
463 : static const int B = 128;
464 1837 : if (v < (1<<7)) {
465 1792 : *(ptr++) = v;
466 45 : } else if (v < (1<<14)) {
467 45 : *(ptr++) = v | B;
468 45 : *(ptr++) = v>>7;
469 0 : } else if (v < (1<<21)) {
470 0 : *(ptr++) = v | B;
471 0 : *(ptr++) = (v>>7) | B;
472 0 : *(ptr++) = v>>14;
473 0 : } else if (v < (1<<28)) {
474 0 : *(ptr++) = v | B;
475 0 : *(ptr++) = (v>>7) | B;
476 0 : *(ptr++) = (v>>14) | B;
477 0 : *(ptr++) = v>>21;
478 : } else {
479 0 : *(ptr++) = v | B;
480 0 : *(ptr++) = (v>>7) | B;
481 0 : *(ptr++) = (v>>14) | B;
482 0 : *(ptr++) = (v>>21) | B;
483 0 : *(ptr++) = v>>28;
484 : }
485 1837 : return reinterpret_cast<char*>(ptr);
486 : }
487 :
488 : // If you know the internal layout of the std::string in use, you can
489 : // replace this function with one that resizes the string without
490 : // filling the new space with zeros (if applicable) --
491 : // it will be non-portable but faster.
492 0 : inline void STLStringResizeUninitialized(string* s, size_t new_size) {
493 0 : s->resize(new_size);
494 0 : }
495 :
496 : // Return a mutable char* pointing to a string's internal buffer,
497 : // which may not be null-terminated. Writing through this pointer will
498 : // modify the string.
499 : //
500 : // string_as_array(&str)[i] is valid for 0 <= i < str.size() until the
501 : // next call to a string method that invalidates iterators.
502 : //
503 : // As of 2006-04, there is no standard-blessed way of getting a
504 : // mutable reference to a string's internal buffer. However, issue 530
505 : // (http://www.open-std.org/JTC1/SC22/WG21/docs/lwg-defects.html#530)
506 : // proposes this as the method. It will officially be part of the standard
507 : // for C++0x. This should already work on all current implementations.
508 0 : inline char* string_as_array(string* str) {
509 0 : return str->empty() ? NULL : &*str->begin();
510 : }
511 :
512 : } // namespace snappy
513 :
514 : #endif // UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_
|