1 /*
2 * Copyright (c) Yann Collet, Facebook, Inc.
3 * All rights reserved.
4 *
5 * This source code is licensed under both the BSD-style license (found in the
6 * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7 * in the COPYING file in the root directory of this source tree).
8 * You may select, at your option, one of the above-listed licenses.
9 */
10
11 /*-*************************************
12 * Dependencies
13 ***************************************/
14 #include "../common/zstd_deps.h" /* INT_MAX, ZSTD_memset, ZSTD_memcpy */
15 #include "../common/mem.h"
16 #include "hist.h" /* HIST_countFast_wksp */
17 #define FSE_STATIC_LINKING_ONLY /* FSE_encodeSymbol */
18 #include "../common/fse.h"
19 #define HUF_STATIC_LINKING_ONLY
20 #include "../common/huf.h"
21 #include "zstd_compress_internal.h"
22 #include "zstd_compress_sequences.h"
23 #include "zstd_compress_literals.h"
24 #include "zstd_fast.h"
25 #include "zstd_double_fast.h"
26 #include "zstd_lazy.h"
27 #include "zstd_opt.h"
28 #include "zstd_ldm.h"
29 #include "zstd_compress_superblock.h"
30
31 /* ***************************************************************
32 * Tuning parameters
33 *****************************************************************/
34 /*!
35 * COMPRESS_HEAPMODE :
36 * Select how default decompression function ZSTD_compress() allocates its context,
37 * on stack (0, default), or into heap (1).
38 * Note that functions with explicit context such as ZSTD_compressCCtx() are unaffected.
39 */
40 #ifndef ZSTD_COMPRESS_HEAPMODE
41 # define ZSTD_COMPRESS_HEAPMODE 0
42 #endif
43
44 /*!
45 * ZSTD_HASHLOG3_MAX :
46 * Maximum size of the hash table dedicated to find 3-bytes matches,
47 * in log format, aka 17 => 1 << 17 == 128Ki positions.
48 * This structure is only used in zstd_opt.
49 * Since allocation is centralized for all strategies, it has to be known here.
50 * The actual (selected) size of the hash table is then stored in ZSTD_matchState_t.hashLog3,
51 * so that zstd_opt.c doesn't need to know about this constant.
52 */
53 #ifndef ZSTD_HASHLOG3_MAX
54 # define ZSTD_HASHLOG3_MAX 17
55 #endif
56
57 /*-*************************************
58 * Helper functions
59 ***************************************/
60 /* ZSTD_compressBound()
61 * Note that the result from this function is only compatible with the "normal"
62 * full-block strategy.
63 * When there are a lot of small blocks due to frequent flush in streaming mode
64 * the overhead of headers can make the compressed data to be larger than the
65 * return value of ZSTD_compressBound().
66 */
ZSTD_compressBound(size_t srcSize)67 size_t ZSTD_compressBound(size_t srcSize) {
68 return ZSTD_COMPRESSBOUND(srcSize);
69 }
70
71
72 /*-*************************************
73 * Context memory management
74 ***************************************/
75 struct ZSTD_CDict_s {
76 const void* dictContent;
77 size_t dictContentSize;
78 ZSTD_dictContentType_e dictContentType; /* The dictContentType the CDict was created with */
79 U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */
80 ZSTD_cwksp workspace;
81 ZSTD_matchState_t matchState;
82 ZSTD_compressedBlockState_t cBlockState;
83 ZSTD_customMem customMem;
84 U32 dictID;
85 int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */
86 ZSTD_paramSwitch_e useRowMatchFinder; /* Indicates whether the CDict was created with params that would use
87 * row-based matchfinder. Unless the cdict is reloaded, we will use
88 * the same greedy/lazy matchfinder at compression time.
89 */
90 }; /* typedef'd to ZSTD_CDict within "zstd.h" */
91
ZSTD_createCCtx(void)92 ZSTD_CCtx* ZSTD_createCCtx(void)
93 {
94 return ZSTD_createCCtx_advanced(ZSTD_defaultCMem);
95 }
96
ZSTD_initCCtx(ZSTD_CCtx * cctx,ZSTD_customMem memManager)97 static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager)
98 {
99 assert(cctx != NULL);
100 ZSTD_memset(cctx, 0, sizeof(*cctx));
101 cctx->customMem = memManager;
102 cctx->bmi2 = ZSTD_cpuSupportsBmi2();
103 { size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
104 assert(!ZSTD_isError(err));
105 (void)err;
106 }
107 }
108
ZSTD_createCCtx_advanced(ZSTD_customMem customMem)109 ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
110 {
111 ZSTD_STATIC_ASSERT(zcss_init==0);
112 ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1));
113 if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
114 { ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_customMalloc(sizeof(ZSTD_CCtx), customMem);
115 if (!cctx) return NULL;
116 ZSTD_initCCtx(cctx, customMem);
117 return cctx;
118 }
119 }
120
ZSTD_initStaticCCtx(void * workspace,size_t workspaceSize)121 ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize)
122 {
123 ZSTD_cwksp ws;
124 ZSTD_CCtx* cctx;
125 if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL; /* minimum size */
126 if ((size_t)workspace & 7) return NULL; /* must be 8-aligned */
127 ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc);
128
129 cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));
130 if (cctx == NULL) return NULL;
131
132 ZSTD_memset(cctx, 0, sizeof(ZSTD_CCtx));
133 ZSTD_cwksp_move(&cctx->workspace, &ws);
134 cctx->staticSize = workspaceSize;
135
136 /* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */
137 if (!ZSTD_cwksp_check_available(&cctx->workspace, ENTROPY_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL;
138 cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
139 cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
140 cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, ENTROPY_WORKSPACE_SIZE);
141 cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
142 return cctx;
143 }
144
145 /**
146 * Clears and frees all of the dictionaries in the CCtx.
147 */
ZSTD_clearAllDicts(ZSTD_CCtx * cctx)148 static void ZSTD_clearAllDicts(ZSTD_CCtx* cctx)
149 {
150 ZSTD_customFree(cctx->localDict.dictBuffer, cctx->customMem);
151 ZSTD_freeCDict(cctx->localDict.cdict);
152 ZSTD_memset(&cctx->localDict, 0, sizeof(cctx->localDict));
153 ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));
154 cctx->cdict = NULL;
155 }
156
ZSTD_sizeof_localDict(ZSTD_localDict dict)157 static size_t ZSTD_sizeof_localDict(ZSTD_localDict dict)
158 {
159 size_t const bufferSize = dict.dictBuffer != NULL ? dict.dictSize : 0;
160 size_t const cdictSize = ZSTD_sizeof_CDict(dict.cdict);
161 return bufferSize + cdictSize;
162 }
163
ZSTD_freeCCtxContent(ZSTD_CCtx * cctx)164 static void ZSTD_freeCCtxContent(ZSTD_CCtx* cctx)
165 {
166 assert(cctx != NULL);
167 assert(cctx->staticSize == 0);
168 ZSTD_clearAllDicts(cctx);
169 #ifdef ZSTD_MULTITHREAD
170 ZSTDMT_freeCCtx(cctx->mtctx); cctx->mtctx = NULL;
171 #endif
172 ZSTD_cwksp_free(&cctx->workspace, cctx->customMem);
173 }
174
ZSTD_freeCCtx(ZSTD_CCtx * cctx)175 size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
176 {
177 if (cctx==NULL) return 0; /* support free on NULL */
178 RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
179 "not compatible with static CCtx");
180 {
181 int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx);
182 ZSTD_freeCCtxContent(cctx);
183 if (!cctxInWorkspace) {
184 ZSTD_customFree(cctx, cctx->customMem);
185 }
186 }
187 return 0;
188 }
189
190
ZSTD_sizeof_mtctx(const ZSTD_CCtx * cctx)191 static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
192 {
193 #ifdef ZSTD_MULTITHREAD
194 return ZSTDMT_sizeof_CCtx(cctx->mtctx);
195 #else
196 (void)cctx;
197 return 0;
198 #endif
199 }
200
201
ZSTD_sizeof_CCtx(const ZSTD_CCtx * cctx)202 size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)
203 {
204 if (cctx==NULL) return 0; /* support sizeof on NULL */
205 /* cctx may be in the workspace */
206 return (cctx->workspace.workspace == cctx ? 0 : sizeof(*cctx))
207 + ZSTD_cwksp_sizeof(&cctx->workspace)
208 + ZSTD_sizeof_localDict(cctx->localDict)
209 + ZSTD_sizeof_mtctx(cctx);
210 }
211
ZSTD_sizeof_CStream(const ZSTD_CStream * zcs)212 size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)
213 {
214 return ZSTD_sizeof_CCtx(zcs); /* same object */
215 }
216
217 /* private API call, for dictBuilder only */
ZSTD_getSeqStore(const ZSTD_CCtx * ctx)218 const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); }
219
220 /* Returns true if the strategy supports using a row based matchfinder */
ZSTD_rowMatchFinderSupported(const ZSTD_strategy strategy)221 static int ZSTD_rowMatchFinderSupported(const ZSTD_strategy strategy) {
222 return (strategy >= ZSTD_greedy && strategy <= ZSTD_lazy2);
223 }
224
225 /* Returns true if the strategy and useRowMatchFinder mode indicate that we will use the row based matchfinder
226 * for this compression.
227 */
ZSTD_rowMatchFinderUsed(const ZSTD_strategy strategy,const ZSTD_paramSwitch_e mode)228 static int ZSTD_rowMatchFinderUsed(const ZSTD_strategy strategy, const ZSTD_paramSwitch_e mode) {
229 assert(mode != ZSTD_ps_auto);
230 return ZSTD_rowMatchFinderSupported(strategy) && (mode == ZSTD_ps_enable);
231 }
232
233 /* Returns row matchfinder usage given an initial mode and cParams */
ZSTD_resolveRowMatchFinderMode(ZSTD_paramSwitch_e mode,const ZSTD_compressionParameters * const cParams)234 static ZSTD_paramSwitch_e ZSTD_resolveRowMatchFinderMode(ZSTD_paramSwitch_e mode,
235 const ZSTD_compressionParameters* const cParams) {
236 #if defined(ZSTD_ARCH_X86_SSE2) || defined(ZSTD_ARCH_ARM_NEON)
237 int const kHasSIMD128 = 1;
238 #else
239 int const kHasSIMD128 = 0;
240 #endif
241 if (mode != ZSTD_ps_auto) return mode; /* if requested enabled, but no SIMD, we still will use row matchfinder */
242 mode = ZSTD_ps_disable;
243 if (!ZSTD_rowMatchFinderSupported(cParams->strategy)) return mode;
244 if (kHasSIMD128) {
245 if (cParams->windowLog > 14) mode = ZSTD_ps_enable;
246 } else {
247 if (cParams->windowLog > 17) mode = ZSTD_ps_enable;
248 }
249 return mode;
250 }
251
252 /* Returns block splitter usage (generally speaking, when using slower/stronger compression modes) */
ZSTD_resolveBlockSplitterMode(ZSTD_paramSwitch_e mode,const ZSTD_compressionParameters * const cParams)253 static ZSTD_paramSwitch_e ZSTD_resolveBlockSplitterMode(ZSTD_paramSwitch_e mode,
254 const ZSTD_compressionParameters* const cParams) {
255 if (mode != ZSTD_ps_auto) return mode;
256 return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 17) ? ZSTD_ps_enable : ZSTD_ps_disable;
257 }
258
259 /* Returns 1 if the arguments indicate that we should allocate a chainTable, 0 otherwise */
ZSTD_allocateChainTable(const ZSTD_strategy strategy,const ZSTD_paramSwitch_e useRowMatchFinder,const U32 forDDSDict)260 static int ZSTD_allocateChainTable(const ZSTD_strategy strategy,
261 const ZSTD_paramSwitch_e useRowMatchFinder,
262 const U32 forDDSDict) {
263 assert(useRowMatchFinder != ZSTD_ps_auto);
264 /* We always should allocate a chaintable if we are allocating a matchstate for a DDS dictionary matchstate.
265 * We do not allocate a chaintable if we are using ZSTD_fast, or are using the row-based matchfinder.
266 */
267 return forDDSDict || ((strategy != ZSTD_fast) && !ZSTD_rowMatchFinderUsed(strategy, useRowMatchFinder));
268 }
269
270 /* Returns 1 if compression parameters are such that we should
271 * enable long distance matching (wlog >= 27, strategy >= btopt).
272 * Returns 0 otherwise.
273 */
ZSTD_resolveEnableLdm(ZSTD_paramSwitch_e mode,const ZSTD_compressionParameters * const cParams)274 static ZSTD_paramSwitch_e ZSTD_resolveEnableLdm(ZSTD_paramSwitch_e mode,
275 const ZSTD_compressionParameters* const cParams) {
276 if (mode != ZSTD_ps_auto) return mode;
277 return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 27) ? ZSTD_ps_enable : ZSTD_ps_disable;
278 }
279
ZSTD_makeCCtxParamsFromCParams(ZSTD_compressionParameters cParams)280 static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams(
281 ZSTD_compressionParameters cParams)
282 {
283 ZSTD_CCtx_params cctxParams;
284 /* should not matter, as all cParams are presumed properly defined */
285 ZSTD_CCtxParams_init(&cctxParams, ZSTD_CLEVEL_DEFAULT);
286 cctxParams.cParams = cParams;
287
288 /* Adjust advanced params according to cParams */
289 cctxParams.ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams.ldmParams.enableLdm, &cParams);
290 if (cctxParams.ldmParams.enableLdm == ZSTD_ps_enable) {
291 ZSTD_ldm_adjustParameters(&cctxParams.ldmParams, &cParams);
292 assert(cctxParams.ldmParams.hashLog >= cctxParams.ldmParams.bucketSizeLog);
293 assert(cctxParams.ldmParams.hashRateLog < 32);
294 }
295 cctxParams.useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams.useBlockSplitter, &cParams);
296 cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams);
297 assert(!ZSTD_checkCParams(cParams));
298 return cctxParams;
299 }
300
ZSTD_createCCtxParams_advanced(ZSTD_customMem customMem)301 static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced(
302 ZSTD_customMem customMem)
303 {
304 ZSTD_CCtx_params* params;
305 if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
306 params = (ZSTD_CCtx_params*)ZSTD_customCalloc(
307 sizeof(ZSTD_CCtx_params), customMem);
308 if (!params) { return NULL; }
309 ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT);
310 params->customMem = customMem;
311 return params;
312 }
313
ZSTD_createCCtxParams(void)314 ZSTD_CCtx_params* ZSTD_createCCtxParams(void)
315 {
316 return ZSTD_createCCtxParams_advanced(ZSTD_defaultCMem);
317 }
318
ZSTD_freeCCtxParams(ZSTD_CCtx_params * params)319 size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params)
320 {
321 if (params == NULL) { return 0; }
322 ZSTD_customFree(params, params->customMem);
323 return 0;
324 }
325
ZSTD_CCtxParams_reset(ZSTD_CCtx_params * params)326 size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params)
327 {
328 return ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT);
329 }
330
ZSTD_CCtxParams_init(ZSTD_CCtx_params * cctxParams,int compressionLevel)331 size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) {
332 RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
333 ZSTD_memset(cctxParams, 0, sizeof(*cctxParams));
334 cctxParams->compressionLevel = compressionLevel;
335 cctxParams->fParams.contentSizeFlag = 1;
336 return 0;
337 }
338
339 #define ZSTD_NO_CLEVEL 0
340
341 /**
342 * Initializes the cctxParams from params and compressionLevel.
343 * @param compressionLevel If params are derived from a compression level then that compression level, otherwise ZSTD_NO_CLEVEL.
344 */
ZSTD_CCtxParams_init_internal(ZSTD_CCtx_params * cctxParams,ZSTD_parameters const * params,int compressionLevel)345 static void ZSTD_CCtxParams_init_internal(ZSTD_CCtx_params* cctxParams, ZSTD_parameters const* params, int compressionLevel)
346 {
347 assert(!ZSTD_checkCParams(params->cParams));
348 ZSTD_memset(cctxParams, 0, sizeof(*cctxParams));
349 cctxParams->cParams = params->cParams;
350 cctxParams->fParams = params->fParams;
351 /* Should not matter, as all cParams are presumed properly defined.
352 * But, set it for tracing anyway.
353 */
354 cctxParams->compressionLevel = compressionLevel;
355 cctxParams->useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams->useRowMatchFinder, ¶ms->cParams);
356 cctxParams->useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams->useBlockSplitter, ¶ms->cParams);
357 cctxParams->ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams->ldmParams.enableLdm, ¶ms->cParams);
358 DEBUGLOG(4, "ZSTD_CCtxParams_init_internal: useRowMatchFinder=%d, useBlockSplitter=%d ldm=%d",
359 cctxParams->useRowMatchFinder, cctxParams->useBlockSplitter, cctxParams->ldmParams.enableLdm);
360 }
361
ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params * cctxParams,ZSTD_parameters params)362 size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)
363 {
364 RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");
365 FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
366 ZSTD_CCtxParams_init_internal(cctxParams, ¶ms, ZSTD_NO_CLEVEL);
367 return 0;
368 }
369
370 /**
371 * Sets cctxParams' cParams and fParams from params, but otherwise leaves them alone.
372 * @param param Validated zstd parameters.
373 */
ZSTD_CCtxParams_setZstdParams(ZSTD_CCtx_params * cctxParams,const ZSTD_parameters * params)374 static void ZSTD_CCtxParams_setZstdParams(
375 ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params)
376 {
377 assert(!ZSTD_checkCParams(params->cParams));
378 cctxParams->cParams = params->cParams;
379 cctxParams->fParams = params->fParams;
380 /* Should not matter, as all cParams are presumed properly defined.
381 * But, set it for tracing anyway.
382 */
383 cctxParams->compressionLevel = ZSTD_NO_CLEVEL;
384 }
385
ZSTD_cParam_getBounds(ZSTD_cParameter param)386 ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
387 {
388 ZSTD_bounds bounds = { 0, 0, 0 };
389
390 switch(param)
391 {
392 case ZSTD_c_compressionLevel:
393 bounds.lowerBound = ZSTD_minCLevel();
394 bounds.upperBound = ZSTD_maxCLevel();
395 return bounds;
396
397 case ZSTD_c_windowLog:
398 bounds.lowerBound = ZSTD_WINDOWLOG_MIN;
399 bounds.upperBound = ZSTD_WINDOWLOG_MAX;
400 return bounds;
401
402 case ZSTD_c_hashLog:
403 bounds.lowerBound = ZSTD_HASHLOG_MIN;
404 bounds.upperBound = ZSTD_HASHLOG_MAX;
405 return bounds;
406
407 case ZSTD_c_chainLog:
408 bounds.lowerBound = ZSTD_CHAINLOG_MIN;
409 bounds.upperBound = ZSTD_CHAINLOG_MAX;
410 return bounds;
411
412 case ZSTD_c_searchLog:
413 bounds.lowerBound = ZSTD_SEARCHLOG_MIN;
414 bounds.upperBound = ZSTD_SEARCHLOG_MAX;
415 return bounds;
416
417 case ZSTD_c_minMatch:
418 bounds.lowerBound = ZSTD_MINMATCH_MIN;
419 bounds.upperBound = ZSTD_MINMATCH_MAX;
420 return bounds;
421
422 case ZSTD_c_targetLength:
423 bounds.lowerBound = ZSTD_TARGETLENGTH_MIN;
424 bounds.upperBound = ZSTD_TARGETLENGTH_MAX;
425 return bounds;
426
427 case ZSTD_c_strategy:
428 bounds.lowerBound = ZSTD_STRATEGY_MIN;
429 bounds.upperBound = ZSTD_STRATEGY_MAX;
430 return bounds;
431
432 case ZSTD_c_contentSizeFlag:
433 bounds.lowerBound = 0;
434 bounds.upperBound = 1;
435 return bounds;
436
437 case ZSTD_c_checksumFlag:
438 bounds.lowerBound = 0;
439 bounds.upperBound = 1;
440 return bounds;
441
442 case ZSTD_c_dictIDFlag:
443 bounds.lowerBound = 0;
444 bounds.upperBound = 1;
445 return bounds;
446
447 case ZSTD_c_nbWorkers:
448 bounds.lowerBound = 0;
449 #ifdef ZSTD_MULTITHREAD
450 bounds.upperBound = ZSTDMT_NBWORKERS_MAX;
451 #else
452 bounds.upperBound = 0;
453 #endif
454 return bounds;
455
456 case ZSTD_c_jobSize:
457 bounds.lowerBound = 0;
458 #ifdef ZSTD_MULTITHREAD
459 bounds.upperBound = ZSTDMT_JOBSIZE_MAX;
460 #else
461 bounds.upperBound = 0;
462 #endif
463 return bounds;
464
465 case ZSTD_c_overlapLog:
466 #ifdef ZSTD_MULTITHREAD
467 bounds.lowerBound = ZSTD_OVERLAPLOG_MIN;
468 bounds.upperBound = ZSTD_OVERLAPLOG_MAX;
469 #else
470 bounds.lowerBound = 0;
471 bounds.upperBound = 0;
472 #endif
473 return bounds;
474
475 case ZSTD_c_enableDedicatedDictSearch:
476 bounds.lowerBound = 0;
477 bounds.upperBound = 1;
478 return bounds;
479
480 case ZSTD_c_enableLongDistanceMatching:
481 bounds.lowerBound = 0;
482 bounds.upperBound = 1;
483 return bounds;
484
485 case ZSTD_c_ldmHashLog:
486 bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN;
487 bounds.upperBound = ZSTD_LDM_HASHLOG_MAX;
488 return bounds;
489
490 case ZSTD_c_ldmMinMatch:
491 bounds.lowerBound = ZSTD_LDM_MINMATCH_MIN;
492 bounds.upperBound = ZSTD_LDM_MINMATCH_MAX;
493 return bounds;
494
495 case ZSTD_c_ldmBucketSizeLog:
496 bounds.lowerBound = ZSTD_LDM_BUCKETSIZELOG_MIN;
497 bounds.upperBound = ZSTD_LDM_BUCKETSIZELOG_MAX;
498 return bounds;
499
500 case ZSTD_c_ldmHashRateLog:
501 bounds.lowerBound = ZSTD_LDM_HASHRATELOG_MIN;
502 bounds.upperBound = ZSTD_LDM_HASHRATELOG_MAX;
503 return bounds;
504
505 /* experimental parameters */
506 case ZSTD_c_rsyncable:
507 bounds.lowerBound = 0;
508 bounds.upperBound = 1;
509 return bounds;
510
511 case ZSTD_c_forceMaxWindow :
512 bounds.lowerBound = 0;
513 bounds.upperBound = 1;
514 return bounds;
515
516 case ZSTD_c_format:
517 ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
518 bounds.lowerBound = ZSTD_f_zstd1;
519 bounds.upperBound = ZSTD_f_zstd1_magicless; /* note : how to ensure at compile time that this is the highest value enum ? */
520 return bounds;
521
522 case ZSTD_c_forceAttachDict:
523 ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceLoad);
524 bounds.lowerBound = ZSTD_dictDefaultAttach;
525 bounds.upperBound = ZSTD_dictForceLoad; /* note : how to ensure at compile time that this is the highest value enum ? */
526 return bounds;
527
528 case ZSTD_c_literalCompressionMode:
529 ZSTD_STATIC_ASSERT(ZSTD_ps_auto < ZSTD_ps_enable && ZSTD_ps_enable < ZSTD_ps_disable);
530 bounds.lowerBound = (int)ZSTD_ps_auto;
531 bounds.upperBound = (int)ZSTD_ps_disable;
532 return bounds;
533
534 case ZSTD_c_targetCBlockSize:
535 bounds.lowerBound = ZSTD_TARGETCBLOCKSIZE_MIN;
536 bounds.upperBound = ZSTD_TARGETCBLOCKSIZE_MAX;
537 return bounds;
538
539 case ZSTD_c_srcSizeHint:
540 bounds.lowerBound = ZSTD_SRCSIZEHINT_MIN;
541 bounds.upperBound = ZSTD_SRCSIZEHINT_MAX;
542 return bounds;
543
544 case ZSTD_c_stableInBuffer:
545 case ZSTD_c_stableOutBuffer:
546 bounds.lowerBound = (int)ZSTD_bm_buffered;
547 bounds.upperBound = (int)ZSTD_bm_stable;
548 return bounds;
549
550 case ZSTD_c_blockDelimiters:
551 bounds.lowerBound = (int)ZSTD_sf_noBlockDelimiters;
552 bounds.upperBound = (int)ZSTD_sf_explicitBlockDelimiters;
553 return bounds;
554
555 case ZSTD_c_validateSequences:
556 bounds.lowerBound = 0;
557 bounds.upperBound = 1;
558 return bounds;
559
560 case ZSTD_c_useBlockSplitter:
561 bounds.lowerBound = (int)ZSTD_ps_auto;
562 bounds.upperBound = (int)ZSTD_ps_disable;
563 return bounds;
564
565 case ZSTD_c_useRowMatchFinder:
566 bounds.lowerBound = (int)ZSTD_ps_auto;
567 bounds.upperBound = (int)ZSTD_ps_disable;
568 return bounds;
569
570 case ZSTD_c_deterministicRefPrefix:
571 bounds.lowerBound = 0;
572 bounds.upperBound = 1;
573 return bounds;
574
575 default:
576 bounds.error = ERROR(parameter_unsupported);
577 return bounds;
578 }
579 }
580
581 /* ZSTD_cParam_clampBounds:
582 * Clamps the value into the bounded range.
583 */
ZSTD_cParam_clampBounds(ZSTD_cParameter cParam,int * value)584 static size_t ZSTD_cParam_clampBounds(ZSTD_cParameter cParam, int* value)
585 {
586 ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);
587 if (ZSTD_isError(bounds.error)) return bounds.error;
588 if (*value < bounds.lowerBound) *value = bounds.lowerBound;
589 if (*value > bounds.upperBound) *value = bounds.upperBound;
590 return 0;
591 }
592
593 #define BOUNDCHECK(cParam, val) { \
594 RETURN_ERROR_IF(!ZSTD_cParam_withinBounds(cParam,val), \
595 parameter_outOfBound, "Param out of bounds"); \
596 }
597
598
ZSTD_isUpdateAuthorized(ZSTD_cParameter param)599 static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param)
600 {
601 switch(param)
602 {
603 case ZSTD_c_compressionLevel:
604 case ZSTD_c_hashLog:
605 case ZSTD_c_chainLog:
606 case ZSTD_c_searchLog:
607 case ZSTD_c_minMatch:
608 case ZSTD_c_targetLength:
609 case ZSTD_c_strategy:
610 return 1;
611
612 case ZSTD_c_format:
613 case ZSTD_c_windowLog:
614 case ZSTD_c_contentSizeFlag:
615 case ZSTD_c_checksumFlag:
616 case ZSTD_c_dictIDFlag:
617 case ZSTD_c_forceMaxWindow :
618 case ZSTD_c_nbWorkers:
619 case ZSTD_c_jobSize:
620 case ZSTD_c_overlapLog:
621 case ZSTD_c_rsyncable:
622 case ZSTD_c_enableDedicatedDictSearch:
623 case ZSTD_c_enableLongDistanceMatching:
624 case ZSTD_c_ldmHashLog:
625 case ZSTD_c_ldmMinMatch:
626 case ZSTD_c_ldmBucketSizeLog:
627 case ZSTD_c_ldmHashRateLog:
628 case ZSTD_c_forceAttachDict:
629 case ZSTD_c_literalCompressionMode:
630 case ZSTD_c_targetCBlockSize:
631 case ZSTD_c_srcSizeHint:
632 case ZSTD_c_stableInBuffer:
633 case ZSTD_c_stableOutBuffer:
634 case ZSTD_c_blockDelimiters:
635 case ZSTD_c_validateSequences:
636 case ZSTD_c_useBlockSplitter:
637 case ZSTD_c_useRowMatchFinder:
638 case ZSTD_c_deterministicRefPrefix:
639 default:
640 return 0;
641 }
642 }
643
ZSTD_CCtx_setParameter(ZSTD_CCtx * cctx,ZSTD_cParameter param,int value)644 size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
645 {
646 DEBUGLOG(4, "ZSTD_CCtx_setParameter (%i, %i)", (int)param, value);
647 if (cctx->streamStage != zcss_init) {
648 if (ZSTD_isUpdateAuthorized(param)) {
649 cctx->cParamsChanged = 1;
650 } else {
651 RETURN_ERROR(stage_wrong, "can only set params in ctx init stage");
652 } }
653
654 switch(param)
655 {
656 case ZSTD_c_nbWorkers:
657 RETURN_ERROR_IF((value!=0) && cctx->staticSize, parameter_unsupported,
658 "MT not compatible with static alloc");
659 break;
660
661 case ZSTD_c_compressionLevel:
662 case ZSTD_c_windowLog:
663 case ZSTD_c_hashLog:
664 case ZSTD_c_chainLog:
665 case ZSTD_c_searchLog:
666 case ZSTD_c_minMatch:
667 case ZSTD_c_targetLength:
668 case ZSTD_c_strategy:
669 case ZSTD_c_ldmHashRateLog:
670 case ZSTD_c_format:
671 case ZSTD_c_contentSizeFlag:
672 case ZSTD_c_checksumFlag:
673 case ZSTD_c_dictIDFlag:
674 case ZSTD_c_forceMaxWindow:
675 case ZSTD_c_forceAttachDict:
676 case ZSTD_c_literalCompressionMode:
677 case ZSTD_c_jobSize:
678 case ZSTD_c_overlapLog:
679 case ZSTD_c_rsyncable:
680 case ZSTD_c_enableDedicatedDictSearch:
681 case ZSTD_c_enableLongDistanceMatching:
682 case ZSTD_c_ldmHashLog:
683 case ZSTD_c_ldmMinMatch:
684 case ZSTD_c_ldmBucketSizeLog:
685 case ZSTD_c_targetCBlockSize:
686 case ZSTD_c_srcSizeHint:
687 case ZSTD_c_stableInBuffer:
688 case ZSTD_c_stableOutBuffer:
689 case ZSTD_c_blockDelimiters:
690 case ZSTD_c_validateSequences:
691 case ZSTD_c_useBlockSplitter:
692 case ZSTD_c_useRowMatchFinder:
693 case ZSTD_c_deterministicRefPrefix:
694 break;
695
696 default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
697 }
698 return ZSTD_CCtxParams_setParameter(&cctx->requestedParams, param, value);
699 }
700
ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params * CCtxParams,ZSTD_cParameter param,int value)701 size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
702 ZSTD_cParameter param, int value)
703 {
704 DEBUGLOG(4, "ZSTD_CCtxParams_setParameter (%i, %i)", (int)param, value);
705 switch(param)
706 {
707 case ZSTD_c_format :
708 BOUNDCHECK(ZSTD_c_format, value);
709 CCtxParams->format = (ZSTD_format_e)value;
710 return (size_t)CCtxParams->format;
711
712 case ZSTD_c_compressionLevel : {
713 FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
714 if (value == 0)
715 CCtxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
716 else
717 CCtxParams->compressionLevel = value;
718 if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel;
719 return 0; /* return type (size_t) cannot represent negative values */
720 }
721
722 case ZSTD_c_windowLog :
723 if (value!=0) /* 0 => use default */
724 BOUNDCHECK(ZSTD_c_windowLog, value);
725 CCtxParams->cParams.windowLog = (U32)value;
726 return CCtxParams->cParams.windowLog;
727
728 case ZSTD_c_hashLog :
729 if (value!=0) /* 0 => use default */
730 BOUNDCHECK(ZSTD_c_hashLog, value);
731 CCtxParams->cParams.hashLog = (U32)value;
732 return CCtxParams->cParams.hashLog;
733
734 case ZSTD_c_chainLog :
735 if (value!=0) /* 0 => use default */
736 BOUNDCHECK(ZSTD_c_chainLog, value);
737 CCtxParams->cParams.chainLog = (U32)value;
738 return CCtxParams->cParams.chainLog;
739
740 case ZSTD_c_searchLog :
741 if (value!=0) /* 0 => use default */
742 BOUNDCHECK(ZSTD_c_searchLog, value);
743 CCtxParams->cParams.searchLog = (U32)value;
744 return (size_t)value;
745
746 case ZSTD_c_minMatch :
747 if (value!=0) /* 0 => use default */
748 BOUNDCHECK(ZSTD_c_minMatch, value);
749 CCtxParams->cParams.minMatch = value;
750 return CCtxParams->cParams.minMatch;
751
752 case ZSTD_c_targetLength :
753 BOUNDCHECK(ZSTD_c_targetLength, value);
754 CCtxParams->cParams.targetLength = value;
755 return CCtxParams->cParams.targetLength;
756
757 case ZSTD_c_strategy :
758 if (value!=0) /* 0 => use default */
759 BOUNDCHECK(ZSTD_c_strategy, value);
760 CCtxParams->cParams.strategy = (ZSTD_strategy)value;
761 return (size_t)CCtxParams->cParams.strategy;
762
763 case ZSTD_c_contentSizeFlag :
764 /* Content size written in frame header _when known_ (default:1) */
765 DEBUGLOG(4, "set content size flag = %u", (value!=0));
766 CCtxParams->fParams.contentSizeFlag = value != 0;
767 return CCtxParams->fParams.contentSizeFlag;
768
769 case ZSTD_c_checksumFlag :
770 /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */
771 CCtxParams->fParams.checksumFlag = value != 0;
772 return CCtxParams->fParams.checksumFlag;
773
774 case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
775 DEBUGLOG(4, "set dictIDFlag = %u", (value!=0));
776 CCtxParams->fParams.noDictIDFlag = !value;
777 return !CCtxParams->fParams.noDictIDFlag;
778
779 case ZSTD_c_forceMaxWindow :
780 CCtxParams->forceWindow = (value != 0);
781 return CCtxParams->forceWindow;
782
783 case ZSTD_c_forceAttachDict : {
784 const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value;
785 BOUNDCHECK(ZSTD_c_forceAttachDict, pref);
786 CCtxParams->attachDictPref = pref;
787 return CCtxParams->attachDictPref;
788 }
789
790 case ZSTD_c_literalCompressionMode : {
791 const ZSTD_paramSwitch_e lcm = (ZSTD_paramSwitch_e)value;
792 BOUNDCHECK(ZSTD_c_literalCompressionMode, lcm);
793 CCtxParams->literalCompressionMode = lcm;
794 return CCtxParams->literalCompressionMode;
795 }
796
797 case ZSTD_c_nbWorkers :
798 #ifndef ZSTD_MULTITHREAD
799 RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
800 return 0;
801 #else
802 FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
803 CCtxParams->nbWorkers = value;
804 return CCtxParams->nbWorkers;
805 #endif
806
807 case ZSTD_c_jobSize :
808 #ifndef ZSTD_MULTITHREAD
809 RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
810 return 0;
811 #else
812 /* Adjust to the minimum non-default value. */
813 if (value != 0 && value < ZSTDMT_JOBSIZE_MIN)
814 value = ZSTDMT_JOBSIZE_MIN;
815 FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");
816 assert(value >= 0);
817 CCtxParams->jobSize = value;
818 return CCtxParams->jobSize;
819 #endif
820
821 case ZSTD_c_overlapLog :
822 #ifndef ZSTD_MULTITHREAD
823 RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
824 return 0;
825 #else
826 FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
827 CCtxParams->overlapLog = value;
828 return CCtxParams->overlapLog;
829 #endif
830
831 case ZSTD_c_rsyncable :
832 #ifndef ZSTD_MULTITHREAD
833 RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");
834 return 0;
835 #else
836 FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");
837 CCtxParams->rsyncable = value;
838 return CCtxParams->rsyncable;
839 #endif
840
841 case ZSTD_c_enableDedicatedDictSearch :
842 CCtxParams->enableDedicatedDictSearch = (value!=0);
843 return CCtxParams->enableDedicatedDictSearch;
844
845 case ZSTD_c_enableLongDistanceMatching :
846 CCtxParams->ldmParams.enableLdm = (ZSTD_paramSwitch_e)value;
847 return CCtxParams->ldmParams.enableLdm;
848
849 case ZSTD_c_ldmHashLog :
850 if (value!=0) /* 0 ==> auto */
851 BOUNDCHECK(ZSTD_c_ldmHashLog, value);
852 CCtxParams->ldmParams.hashLog = value;
853 return CCtxParams->ldmParams.hashLog;
854
855 case ZSTD_c_ldmMinMatch :
856 if (value!=0) /* 0 ==> default */
857 BOUNDCHECK(ZSTD_c_ldmMinMatch, value);
858 CCtxParams->ldmParams.minMatchLength = value;
859 return CCtxParams->ldmParams.minMatchLength;
860
861 case ZSTD_c_ldmBucketSizeLog :
862 if (value!=0) /* 0 ==> default */
863 BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value);
864 CCtxParams->ldmParams.bucketSizeLog = value;
865 return CCtxParams->ldmParams.bucketSizeLog;
866
867 case ZSTD_c_ldmHashRateLog :
868 if (value!=0) /* 0 ==> default */
869 BOUNDCHECK(ZSTD_c_ldmHashRateLog, value);
870 CCtxParams->ldmParams.hashRateLog = value;
871 return CCtxParams->ldmParams.hashRateLog;
872
873 case ZSTD_c_targetCBlockSize :
874 if (value!=0) /* 0 ==> default */
875 BOUNDCHECK(ZSTD_c_targetCBlockSize, value);
876 CCtxParams->targetCBlockSize = value;
877 return CCtxParams->targetCBlockSize;
878
879 case ZSTD_c_srcSizeHint :
880 if (value!=0) /* 0 ==> default */
881 BOUNDCHECK(ZSTD_c_srcSizeHint, value);
882 CCtxParams->srcSizeHint = value;
883 return CCtxParams->srcSizeHint;
884
885 case ZSTD_c_stableInBuffer:
886 BOUNDCHECK(ZSTD_c_stableInBuffer, value);
887 CCtxParams->inBufferMode = (ZSTD_bufferMode_e)value;
888 return CCtxParams->inBufferMode;
889
890 case ZSTD_c_stableOutBuffer:
891 BOUNDCHECK(ZSTD_c_stableOutBuffer, value);
892 CCtxParams->outBufferMode = (ZSTD_bufferMode_e)value;
893 return CCtxParams->outBufferMode;
894
895 case ZSTD_c_blockDelimiters:
896 BOUNDCHECK(ZSTD_c_blockDelimiters, value);
897 CCtxParams->blockDelimiters = (ZSTD_sequenceFormat_e)value;
898 return CCtxParams->blockDelimiters;
899
900 case ZSTD_c_validateSequences:
901 BOUNDCHECK(ZSTD_c_validateSequences, value);
902 CCtxParams->validateSequences = value;
903 return CCtxParams->validateSequences;
904
905 case ZSTD_c_useBlockSplitter:
906 BOUNDCHECK(ZSTD_c_useBlockSplitter, value);
907 CCtxParams->useBlockSplitter = (ZSTD_paramSwitch_e)value;
908 return CCtxParams->useBlockSplitter;
909
910 case ZSTD_c_useRowMatchFinder:
911 BOUNDCHECK(ZSTD_c_useRowMatchFinder, value);
912 CCtxParams->useRowMatchFinder = (ZSTD_paramSwitch_e)value;
913 return CCtxParams->useRowMatchFinder;
914
915 case ZSTD_c_deterministicRefPrefix:
916 BOUNDCHECK(ZSTD_c_deterministicRefPrefix, value);
917 CCtxParams->deterministicRefPrefix = !!value;
918 return CCtxParams->deterministicRefPrefix;
919
920 default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
921 }
922 }
923
ZSTD_CCtx_getParameter(ZSTD_CCtx const * cctx,ZSTD_cParameter param,int * value)924 size_t ZSTD_CCtx_getParameter(ZSTD_CCtx const* cctx, ZSTD_cParameter param, int* value)
925 {
926 return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value);
927 }
928
ZSTD_CCtxParams_getParameter(ZSTD_CCtx_params const * CCtxParams,ZSTD_cParameter param,int * value)929 size_t ZSTD_CCtxParams_getParameter(
930 ZSTD_CCtx_params const* CCtxParams, ZSTD_cParameter param, int* value)
931 {
932 switch(param)
933 {
934 case ZSTD_c_format :
935 *value = CCtxParams->format;
936 break;
937 case ZSTD_c_compressionLevel :
938 *value = CCtxParams->compressionLevel;
939 break;
940 case ZSTD_c_windowLog :
941 *value = (int)CCtxParams->cParams.windowLog;
942 break;
943 case ZSTD_c_hashLog :
944 *value = (int)CCtxParams->cParams.hashLog;
945 break;
946 case ZSTD_c_chainLog :
947 *value = (int)CCtxParams->cParams.chainLog;
948 break;
949 case ZSTD_c_searchLog :
950 *value = CCtxParams->cParams.searchLog;
951 break;
952 case ZSTD_c_minMatch :
953 *value = CCtxParams->cParams.minMatch;
954 break;
955 case ZSTD_c_targetLength :
956 *value = CCtxParams->cParams.targetLength;
957 break;
958 case ZSTD_c_strategy :
959 *value = (unsigned)CCtxParams->cParams.strategy;
960 break;
961 case ZSTD_c_contentSizeFlag :
962 *value = CCtxParams->fParams.contentSizeFlag;
963 break;
964 case ZSTD_c_checksumFlag :
965 *value = CCtxParams->fParams.checksumFlag;
966 break;
967 case ZSTD_c_dictIDFlag :
968 *value = !CCtxParams->fParams.noDictIDFlag;
969 break;
970 case ZSTD_c_forceMaxWindow :
971 *value = CCtxParams->forceWindow;
972 break;
973 case ZSTD_c_forceAttachDict :
974 *value = CCtxParams->attachDictPref;
975 break;
976 case ZSTD_c_literalCompressionMode :
977 *value = CCtxParams->literalCompressionMode;
978 break;
979 case ZSTD_c_nbWorkers :
980 #ifndef ZSTD_MULTITHREAD
981 assert(CCtxParams->nbWorkers == 0);
982 #endif
983 *value = CCtxParams->nbWorkers;
984 break;
985 case ZSTD_c_jobSize :
986 #ifndef ZSTD_MULTITHREAD
987 RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
988 #else
989 assert(CCtxParams->jobSize <= INT_MAX);
990 *value = (int)CCtxParams->jobSize;
991 break;
992 #endif
993 case ZSTD_c_overlapLog :
994 #ifndef ZSTD_MULTITHREAD
995 RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
996 #else
997 *value = CCtxParams->overlapLog;
998 break;
999 #endif
1000 case ZSTD_c_rsyncable :
1001 #ifndef ZSTD_MULTITHREAD
1002 RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");
1003 #else
1004 *value = CCtxParams->rsyncable;
1005 break;
1006 #endif
1007 case ZSTD_c_enableDedicatedDictSearch :
1008 *value = CCtxParams->enableDedicatedDictSearch;
1009 break;
1010 case ZSTD_c_enableLongDistanceMatching :
1011 *value = CCtxParams->ldmParams.enableLdm;
1012 break;
1013 case ZSTD_c_ldmHashLog :
1014 *value = CCtxParams->ldmParams.hashLog;
1015 break;
1016 case ZSTD_c_ldmMinMatch :
1017 *value = CCtxParams->ldmParams.minMatchLength;
1018 break;
1019 case ZSTD_c_ldmBucketSizeLog :
1020 *value = CCtxParams->ldmParams.bucketSizeLog;
1021 break;
1022 case ZSTD_c_ldmHashRateLog :
1023 *value = CCtxParams->ldmParams.hashRateLog;
1024 break;
1025 case ZSTD_c_targetCBlockSize :
1026 *value = (int)CCtxParams->targetCBlockSize;
1027 break;
1028 case ZSTD_c_srcSizeHint :
1029 *value = (int)CCtxParams->srcSizeHint;
1030 break;
1031 case ZSTD_c_stableInBuffer :
1032 *value = (int)CCtxParams->inBufferMode;
1033 break;
1034 case ZSTD_c_stableOutBuffer :
1035 *value = (int)CCtxParams->outBufferMode;
1036 break;
1037 case ZSTD_c_blockDelimiters :
1038 *value = (int)CCtxParams->blockDelimiters;
1039 break;
1040 case ZSTD_c_validateSequences :
1041 *value = (int)CCtxParams->validateSequences;
1042 break;
1043 case ZSTD_c_useBlockSplitter :
1044 *value = (int)CCtxParams->useBlockSplitter;
1045 break;
1046 case ZSTD_c_useRowMatchFinder :
1047 *value = (int)CCtxParams->useRowMatchFinder;
1048 break;
1049 case ZSTD_c_deterministicRefPrefix:
1050 *value = (int)CCtxParams->deterministicRefPrefix;
1051 break;
1052 default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
1053 }
1054 return 0;
1055 }
1056
1057 /** ZSTD_CCtx_setParametersUsingCCtxParams() :
1058 * just applies `params` into `cctx`
1059 * no action is performed, parameters are merely stored.
1060 * If ZSTDMT is enabled, parameters are pushed to cctx->mtctx.
1061 * This is possible even if a compression is ongoing.
1062 * In which case, new parameters will be applied on the fly, starting with next compression job.
1063 */
ZSTD_CCtx_setParametersUsingCCtxParams(ZSTD_CCtx * cctx,const ZSTD_CCtx_params * params)1064 size_t ZSTD_CCtx_setParametersUsingCCtxParams(
1065 ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params)
1066 {
1067 DEBUGLOG(4, "ZSTD_CCtx_setParametersUsingCCtxParams");
1068 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1069 "The context is in the wrong stage!");
1070 RETURN_ERROR_IF(cctx->cdict, stage_wrong,
1071 "Can't override parameters with cdict attached (some must "
1072 "be inherited from the cdict).");
1073
1074 cctx->requestedParams = *params;
1075 return 0;
1076 }
1077
ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx * cctx,unsigned long long pledgedSrcSize)1078 size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize)
1079 {
1080 DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %u bytes", (U32)pledgedSrcSize);
1081 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1082 "Can't set pledgedSrcSize when not in init stage.");
1083 cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
1084 return 0;
1085 }
1086
1087 static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(
1088 int const compressionLevel,
1089 size_t const dictSize);
1090 static int ZSTD_dedicatedDictSearch_isSupported(
1091 const ZSTD_compressionParameters* cParams);
1092 static void ZSTD_dedicatedDictSearch_revertCParams(
1093 ZSTD_compressionParameters* cParams);
1094
1095 /**
1096 * Initializes the local dict using the requested parameters.
1097 * NOTE: This does not use the pledged src size, because it may be used for more
1098 * than one compression.
1099 */
ZSTD_initLocalDict(ZSTD_CCtx * cctx)1100 static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx)
1101 {
1102 ZSTD_localDict* const dl = &cctx->localDict;
1103 if (dl->dict == NULL) {
1104 /* No local dictionary. */
1105 assert(dl->dictBuffer == NULL);
1106 assert(dl->cdict == NULL);
1107 assert(dl->dictSize == 0);
1108 return 0;
1109 }
1110 if (dl->cdict != NULL) {
1111 assert(cctx->cdict == dl->cdict);
1112 /* Local dictionary already initialized. */
1113 return 0;
1114 }
1115 assert(dl->dictSize > 0);
1116 assert(cctx->cdict == NULL);
1117 assert(cctx->prefixDict.dict == NULL);
1118
1119 dl->cdict = ZSTD_createCDict_advanced2(
1120 dl->dict,
1121 dl->dictSize,
1122 ZSTD_dlm_byRef,
1123 dl->dictContentType,
1124 &cctx->requestedParams,
1125 cctx->customMem);
1126 RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed");
1127 cctx->cdict = dl->cdict;
1128 return 0;
1129 }
1130
ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx * cctx,const void * dict,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType)1131 size_t ZSTD_CCtx_loadDictionary_advanced(
1132 ZSTD_CCtx* cctx, const void* dict, size_t dictSize,
1133 ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType)
1134 {
1135 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1136 "Can't load a dictionary when ctx is not in init stage.");
1137 DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)", (U32)dictSize);
1138 ZSTD_clearAllDicts(cctx); /* in case one already exists */
1139 if (dict == NULL || dictSize == 0) /* no dictionary mode */
1140 return 0;
1141 if (dictLoadMethod == ZSTD_dlm_byRef) {
1142 cctx->localDict.dict = dict;
1143 } else {
1144 void* dictBuffer;
1145 RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
1146 "no malloc for static CCtx");
1147 dictBuffer = ZSTD_customMalloc(dictSize, cctx->customMem);
1148 RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!");
1149 ZSTD_memcpy(dictBuffer, dict, dictSize);
1150 cctx->localDict.dictBuffer = dictBuffer;
1151 cctx->localDict.dict = dictBuffer;
1152 }
1153 cctx->localDict.dictSize = dictSize;
1154 cctx->localDict.dictContentType = dictContentType;
1155 return 0;
1156 }
1157
ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx * cctx,const void * dict,size_t dictSize)1158 size_t ZSTD_CCtx_loadDictionary_byReference(
1159 ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
1160 {
1161 return ZSTD_CCtx_loadDictionary_advanced(
1162 cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);
1163 }
1164
ZSTD_CCtx_loadDictionary(ZSTD_CCtx * cctx,const void * dict,size_t dictSize)1165 size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
1166 {
1167 return ZSTD_CCtx_loadDictionary_advanced(
1168 cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);
1169 }
1170
1171
ZSTD_CCtx_refCDict(ZSTD_CCtx * cctx,const ZSTD_CDict * cdict)1172 size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
1173 {
1174 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1175 "Can't ref a dict when ctx not in init stage.");
1176 /* Free the existing local cdict (if any) to save memory. */
1177 ZSTD_clearAllDicts(cctx);
1178 cctx->cdict = cdict;
1179 return 0;
1180 }
1181
ZSTD_CCtx_refThreadPool(ZSTD_CCtx * cctx,ZSTD_threadPool * pool)1182 size_t ZSTD_CCtx_refThreadPool(ZSTD_CCtx* cctx, ZSTD_threadPool* pool)
1183 {
1184 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1185 "Can't ref a pool when ctx not in init stage.");
1186 cctx->pool = pool;
1187 return 0;
1188 }
1189
ZSTD_CCtx_refPrefix(ZSTD_CCtx * cctx,const void * prefix,size_t prefixSize)1190 size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize)
1191 {
1192 return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dct_rawContent);
1193 }
1194
ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx * cctx,const void * prefix,size_t prefixSize,ZSTD_dictContentType_e dictContentType)1195 size_t ZSTD_CCtx_refPrefix_advanced(
1196 ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)
1197 {
1198 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1199 "Can't ref a prefix when ctx not in init stage.");
1200 ZSTD_clearAllDicts(cctx);
1201 if (prefix != NULL && prefixSize > 0) {
1202 cctx->prefixDict.dict = prefix;
1203 cctx->prefixDict.dictSize = prefixSize;
1204 cctx->prefixDict.dictContentType = dictContentType;
1205 }
1206 return 0;
1207 }
1208
1209 /*! ZSTD_CCtx_reset() :
1210 * Also dumps dictionary */
ZSTD_CCtx_reset(ZSTD_CCtx * cctx,ZSTD_ResetDirective reset)1211 size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)
1212 {
1213 if ( (reset == ZSTD_reset_session_only)
1214 || (reset == ZSTD_reset_session_and_parameters) ) {
1215 cctx->streamStage = zcss_init;
1216 cctx->pledgedSrcSizePlusOne = 0;
1217 }
1218 if ( (reset == ZSTD_reset_parameters)
1219 || (reset == ZSTD_reset_session_and_parameters) ) {
1220 RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,
1221 "Can't reset parameters only when not in init stage.");
1222 ZSTD_clearAllDicts(cctx);
1223 return ZSTD_CCtxParams_reset(&cctx->requestedParams);
1224 }
1225 return 0;
1226 }
1227
1228
1229 /** ZSTD_checkCParams() :
1230 control CParam values remain within authorized range.
1231 @return : 0, or an error code if one value is beyond authorized range */
ZSTD_checkCParams(ZSTD_compressionParameters cParams)1232 size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
1233 {
1234 BOUNDCHECK(ZSTD_c_windowLog, (int)cParams.windowLog);
1235 BOUNDCHECK(ZSTD_c_chainLog, (int)cParams.chainLog);
1236 BOUNDCHECK(ZSTD_c_hashLog, (int)cParams.hashLog);
1237 BOUNDCHECK(ZSTD_c_searchLog, (int)cParams.searchLog);
1238 BOUNDCHECK(ZSTD_c_minMatch, (int)cParams.minMatch);
1239 BOUNDCHECK(ZSTD_c_targetLength,(int)cParams.targetLength);
1240 BOUNDCHECK(ZSTD_c_strategy, cParams.strategy);
1241 return 0;
1242 }
1243
1244 /** ZSTD_clampCParams() :
1245 * make CParam values within valid range.
1246 * @return : valid CParams */
1247 static ZSTD_compressionParameters
ZSTD_clampCParams(ZSTD_compressionParameters cParams)1248 ZSTD_clampCParams(ZSTD_compressionParameters cParams)
1249 {
1250 # define CLAMP_TYPE(cParam, val, type) { \
1251 ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); \
1252 if ((int)val<bounds.lowerBound) val=(type)bounds.lowerBound; \
1253 else if ((int)val>bounds.upperBound) val=(type)bounds.upperBound; \
1254 }
1255 # define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, unsigned)
1256 CLAMP(ZSTD_c_windowLog, cParams.windowLog);
1257 CLAMP(ZSTD_c_chainLog, cParams.chainLog);
1258 CLAMP(ZSTD_c_hashLog, cParams.hashLog);
1259 CLAMP(ZSTD_c_searchLog, cParams.searchLog);
1260 CLAMP(ZSTD_c_minMatch, cParams.minMatch);
1261 CLAMP(ZSTD_c_targetLength,cParams.targetLength);
1262 CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy);
1263 return cParams;
1264 }
1265
1266 /** ZSTD_cycleLog() :
1267 * condition for correct operation : hashLog > 1 */
ZSTD_cycleLog(U32 hashLog,ZSTD_strategy strat)1268 U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
1269 {
1270 U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
1271 return hashLog - btScale;
1272 }
1273
1274 /** ZSTD_dictAndWindowLog() :
1275 * Returns an adjusted window log that is large enough to fit the source and the dictionary.
1276 * The zstd format says that the entire dictionary is valid if one byte of the dictionary
1277 * is within the window. So the hashLog and chainLog should be large enough to reference both
1278 * the dictionary and the window. So we must use this adjusted dictAndWindowLog when downsizing
1279 * the hashLog and windowLog.
1280 * NOTE: srcSize must not be ZSTD_CONTENTSIZE_UNKNOWN.
1281 */
ZSTD_dictAndWindowLog(U32 windowLog,U64 srcSize,U64 dictSize)1282 static U32 ZSTD_dictAndWindowLog(U32 windowLog, U64 srcSize, U64 dictSize)
1283 {
1284 const U64 maxWindowSize = 1ULL << ZSTD_WINDOWLOG_MAX;
1285 /* No dictionary ==> No change */
1286 if (dictSize == 0) {
1287 return windowLog;
1288 }
1289 assert(windowLog <= ZSTD_WINDOWLOG_MAX);
1290 assert(srcSize != ZSTD_CONTENTSIZE_UNKNOWN); /* Handled in ZSTD_adjustCParams_internal() */
1291 {
1292 U64 const windowSize = 1ULL << windowLog;
1293 U64 const dictAndWindowSize = dictSize + windowSize;
1294 /* If the window size is already large enough to fit both the source and the dictionary
1295 * then just use the window size. Otherwise adjust so that it fits the dictionary and
1296 * the window.
1297 */
1298 if (windowSize >= dictSize + srcSize) {
1299 return windowLog; /* Window size large enough already */
1300 } else if (dictAndWindowSize >= maxWindowSize) {
1301 return ZSTD_WINDOWLOG_MAX; /* Larger than max window log */
1302 } else {
1303 return ZSTD_highbit32((U32)dictAndWindowSize - 1) + 1;
1304 }
1305 }
1306 }
1307
1308 /** ZSTD_adjustCParams_internal() :
1309 * optimize `cPar` for a specified input (`srcSize` and `dictSize`).
1310 * mostly downsize to reduce memory consumption and initialization latency.
1311 * `srcSize` can be ZSTD_CONTENTSIZE_UNKNOWN when not known.
1312 * `mode` is the mode for parameter adjustment. See docs for `ZSTD_cParamMode_e`.
1313 * note : `srcSize==0` means 0!
1314 * condition : cPar is presumed validated (can be checked using ZSTD_checkCParams()). */
1315 static ZSTD_compressionParameters
ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,unsigned long long srcSize,size_t dictSize,ZSTD_cParamMode_e mode)1316 ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,
1317 unsigned long long srcSize,
1318 size_t dictSize,
1319 ZSTD_cParamMode_e mode)
1320 {
1321 const U64 minSrcSize = 513; /* (1<<9) + 1 */
1322 const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1);
1323 assert(ZSTD_checkCParams(cPar)==0);
1324
1325 switch (mode) {
1326 case ZSTD_cpm_unknown:
1327 case ZSTD_cpm_noAttachDict:
1328 /* If we don't know the source size, don't make any
1329 * assumptions about it. We will already have selected
1330 * smaller parameters if a dictionary is in use.
1331 */
1332 break;
1333 case ZSTD_cpm_createCDict:
1334 /* Assume a small source size when creating a dictionary
1335 * with an unknown source size.
1336 */
1337 if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN)
1338 srcSize = minSrcSize;
1339 break;
1340 case ZSTD_cpm_attachDict:
1341 /* Dictionary has its own dedicated parameters which have
1342 * already been selected. We are selecting parameters
1343 * for only the source.
1344 */
1345 dictSize = 0;
1346 break;
1347 default:
1348 assert(0);
1349 break;
1350 }
1351
1352 /* resize windowLog if input is small enough, to use less memory */
1353 if ( (srcSize < maxWindowResize)
1354 && (dictSize < maxWindowResize) ) {
1355 U32 const tSize = (U32)(srcSize + dictSize);
1356 static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN;
1357 U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN :
1358 ZSTD_highbit32(tSize-1) + 1;
1359 if (cPar.windowLog > srcLog) cPar.windowLog = srcLog;
1360 }
1361 if (srcSize != ZSTD_CONTENTSIZE_UNKNOWN) {
1362 U32 const dictAndWindowLog = ZSTD_dictAndWindowLog(cPar.windowLog, (U64)srcSize, (U64)dictSize);
1363 U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);
1364 if (cPar.hashLog > dictAndWindowLog+1) cPar.hashLog = dictAndWindowLog+1;
1365 if (cycleLog > dictAndWindowLog)
1366 cPar.chainLog -= (cycleLog - dictAndWindowLog);
1367 }
1368
1369 if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN)
1370 cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */
1371
1372 return cPar;
1373 }
1374
1375 ZSTD_compressionParameters
ZSTD_adjustCParams(ZSTD_compressionParameters cPar,unsigned long long srcSize,size_t dictSize)1376 ZSTD_adjustCParams(ZSTD_compressionParameters cPar,
1377 unsigned long long srcSize,
1378 size_t dictSize)
1379 {
1380 cPar = ZSTD_clampCParams(cPar); /* resulting cPar is necessarily valid (all parameters within range) */
1381 if (srcSize == 0) srcSize = ZSTD_CONTENTSIZE_UNKNOWN;
1382 return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize, ZSTD_cpm_unknown);
1383 }
1384
1385 static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode);
1386 static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode);
1387
ZSTD_overrideCParams(ZSTD_compressionParameters * cParams,const ZSTD_compressionParameters * overrides)1388 static void ZSTD_overrideCParams(
1389 ZSTD_compressionParameters* cParams,
1390 const ZSTD_compressionParameters* overrides)
1391 {
1392 if (overrides->windowLog) cParams->windowLog = overrides->windowLog;
1393 if (overrides->hashLog) cParams->hashLog = overrides->hashLog;
1394 if (overrides->chainLog) cParams->chainLog = overrides->chainLog;
1395 if (overrides->searchLog) cParams->searchLog = overrides->searchLog;
1396 if (overrides->minMatch) cParams->minMatch = overrides->minMatch;
1397 if (overrides->targetLength) cParams->targetLength = overrides->targetLength;
1398 if (overrides->strategy) cParams->strategy = overrides->strategy;
1399 }
1400
ZSTD_getCParamsFromCCtxParams(const ZSTD_CCtx_params * CCtxParams,U64 srcSizeHint,size_t dictSize,ZSTD_cParamMode_e mode)1401 ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
1402 const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
1403 {
1404 ZSTD_compressionParameters cParams;
1405 if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) {
1406 srcSizeHint = CCtxParams->srcSizeHint;
1407 }
1408 cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize, mode);
1409 if (CCtxParams->ldmParams.enableLdm == ZSTD_ps_enable) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;
1410 ZSTD_overrideCParams(&cParams, &CCtxParams->cParams);
1411 assert(!ZSTD_checkCParams(cParams));
1412 /* srcSizeHint == 0 means 0 */
1413 return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize, mode);
1414 }
1415
1416 static size_t
ZSTD_sizeof_matchState(const ZSTD_compressionParameters * const cParams,const ZSTD_paramSwitch_e useRowMatchFinder,const U32 enableDedicatedDictSearch,const U32 forCCtx)1417 ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
1418 const ZSTD_paramSwitch_e useRowMatchFinder,
1419 const U32 enableDedicatedDictSearch,
1420 const U32 forCCtx)
1421 {
1422 /* chain table size should be 0 for fast or row-hash strategies */
1423 size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, enableDedicatedDictSearch && !forCCtx)
1424 ? ((size_t)1 << cParams->chainLog)
1425 : 0;
1426 size_t const hSize = ((size_t)1) << cParams->hashLog;
1427 U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
1428 size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
1429 /* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't
1430 * surrounded by redzones in ASAN. */
1431 size_t const tableSpace = chainSize * sizeof(U32)
1432 + hSize * sizeof(U32)
1433 + h3Size * sizeof(U32);
1434 size_t const optPotentialSpace =
1435 ZSTD_cwksp_aligned_alloc_size((MaxML+1) * sizeof(U32))
1436 + ZSTD_cwksp_aligned_alloc_size((MaxLL+1) * sizeof(U32))
1437 + ZSTD_cwksp_aligned_alloc_size((MaxOff+1) * sizeof(U32))
1438 + ZSTD_cwksp_aligned_alloc_size((1<<Litbits) * sizeof(U32))
1439 + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t))
1440 + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
1441 size_t const lazyAdditionalSpace = ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)
1442 ? ZSTD_cwksp_aligned_alloc_size(hSize*sizeof(U16))
1443 : 0;
1444 size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
1445 ? optPotentialSpace
1446 : 0;
1447 size_t const slackSpace = ZSTD_cwksp_slack_space_required();
1448
1449 /* tables are guaranteed to be sized in multiples of 64 bytes (or 16 uint32_t) */
1450 ZSTD_STATIC_ASSERT(ZSTD_HASHLOG_MIN >= 4 && ZSTD_WINDOWLOG_MIN >= 4 && ZSTD_CHAINLOG_MIN >= 4);
1451 assert(useRowMatchFinder != ZSTD_ps_auto);
1452
1453 DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",
1454 (U32)chainSize, (U32)hSize, (U32)h3Size);
1455 return tableSpace + optSpace + slackSpace + lazyAdditionalSpace;
1456 }
1457
ZSTD_estimateCCtxSize_usingCCtxParams_internal(const ZSTD_compressionParameters * cParams,const ldmParams_t * ldmParams,const int isStatic,const ZSTD_paramSwitch_e useRowMatchFinder,const size_t buffInSize,const size_t buffOutSize,const U64 pledgedSrcSize)1458 static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal(
1459 const ZSTD_compressionParameters* cParams,
1460 const ldmParams_t* ldmParams,
1461 const int isStatic,
1462 const ZSTD_paramSwitch_e useRowMatchFinder,
1463 const size_t buffInSize,
1464 const size_t buffOutSize,
1465 const U64 pledgedSrcSize)
1466 {
1467 size_t const windowSize = (size_t) BOUNDED(1ULL, 1ULL << cParams->windowLog, pledgedSrcSize);
1468 size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
1469 U32 const divider = (cParams->minMatch==3) ? 3 : 4;
1470 size_t const maxNbSeq = blockSize / divider;
1471 size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
1472 + ZSTD_cwksp_aligned_alloc_size(maxNbSeq * sizeof(seqDef))
1473 + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
1474 size_t const entropySpace = ZSTD_cwksp_alloc_size(ENTROPY_WORKSPACE_SIZE);
1475 size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
1476 size_t const matchStateSize = ZSTD_sizeof_matchState(cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 0, /* forCCtx */ 1);
1477
1478 size_t const ldmSpace = ZSTD_ldm_getTableSize(*ldmParams);
1479 size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(*ldmParams, blockSize);
1480 size_t const ldmSeqSpace = ldmParams->enableLdm == ZSTD_ps_enable ?
1481 ZSTD_cwksp_aligned_alloc_size(maxNbLdmSeq * sizeof(rawSeq)) : 0;
1482
1483
1484 size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize)
1485 + ZSTD_cwksp_alloc_size(buffOutSize);
1486
1487 size_t const cctxSpace = isStatic ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
1488
1489 size_t const neededSpace =
1490 cctxSpace +
1491 entropySpace +
1492 blockStateSpace +
1493 ldmSpace +
1494 ldmSeqSpace +
1495 matchStateSize +
1496 tokenSpace +
1497 bufferSpace;
1498
1499 DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
1500 return neededSpace;
1501 }
1502
ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params * params)1503 size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
1504 {
1505 ZSTD_compressionParameters const cParams =
1506 ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
1507 ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder,
1508 &cParams);
1509
1510 RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
1511 /* estimateCCtxSize is for one-shot compression. So no buffers should
1512 * be needed. However, we still allocate two 0-sized buffers, which can
1513 * take space under ASAN. */
1514 return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
1515 &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, 0, 0, ZSTD_CONTENTSIZE_UNKNOWN);
1516 }
1517
ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams)1518 size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams)
1519 {
1520 ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams);
1521 if (ZSTD_rowMatchFinderSupported(cParams.strategy)) {
1522 /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */
1523 size_t noRowCCtxSize;
1524 size_t rowCCtxSize;
1525 initialParams.useRowMatchFinder = ZSTD_ps_disable;
1526 noRowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
1527 initialParams.useRowMatchFinder = ZSTD_ps_enable;
1528 rowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
1529 return MAX(noRowCCtxSize, rowCCtxSize);
1530 } else {
1531 return ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams);
1532 }
1533 }
1534
ZSTD_estimateCCtxSize_internal(int compressionLevel)1535 static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel)
1536 {
1537 int tier = 0;
1538 size_t largestSize = 0;
1539 static const unsigned long long srcSizeTiers[4] = {16 KB, 128 KB, 256 KB, ZSTD_CONTENTSIZE_UNKNOWN};
1540 for (; tier < 4; ++tier) {
1541 /* Choose the set of cParams for a given level across all srcSizes that give the largest cctxSize */
1542 ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeTiers[tier], 0, ZSTD_cpm_noAttachDict);
1543 largestSize = MAX(ZSTD_estimateCCtxSize_usingCParams(cParams), largestSize);
1544 }
1545 return largestSize;
1546 }
1547
ZSTD_estimateCCtxSize(int compressionLevel)1548 size_t ZSTD_estimateCCtxSize(int compressionLevel)
1549 {
1550 int level;
1551 size_t memBudget = 0;
1552 for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
1553 /* Ensure monotonically increasing memory usage as compression level increases */
1554 size_t const newMB = ZSTD_estimateCCtxSize_internal(level);
1555 if (newMB > memBudget) memBudget = newMB;
1556 }
1557 return memBudget;
1558 }
1559
ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params * params)1560 size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params)
1561 {
1562 RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
1563 { ZSTD_compressionParameters const cParams =
1564 ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
1565 size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
1566 size_t const inBuffSize = (params->inBufferMode == ZSTD_bm_buffered)
1567 ? ((size_t)1 << cParams.windowLog) + blockSize
1568 : 0;
1569 size_t const outBuffSize = (params->outBufferMode == ZSTD_bm_buffered)
1570 ? ZSTD_compressBound(blockSize) + 1
1571 : 0;
1572 ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, ¶ms->cParams);
1573
1574 return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
1575 &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, inBuffSize, outBuffSize,
1576 ZSTD_CONTENTSIZE_UNKNOWN);
1577 }
1578 }
1579
ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams)1580 size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams)
1581 {
1582 ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams);
1583 if (ZSTD_rowMatchFinderSupported(cParams.strategy)) {
1584 /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */
1585 size_t noRowCCtxSize;
1586 size_t rowCCtxSize;
1587 initialParams.useRowMatchFinder = ZSTD_ps_disable;
1588 noRowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
1589 initialParams.useRowMatchFinder = ZSTD_ps_enable;
1590 rowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
1591 return MAX(noRowCCtxSize, rowCCtxSize);
1592 } else {
1593 return ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams);
1594 }
1595 }
1596
ZSTD_estimateCStreamSize_internal(int compressionLevel)1597 static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel)
1598 {
1599 ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict);
1600 return ZSTD_estimateCStreamSize_usingCParams(cParams);
1601 }
1602
ZSTD_estimateCStreamSize(int compressionLevel)1603 size_t ZSTD_estimateCStreamSize(int compressionLevel)
1604 {
1605 int level;
1606 size_t memBudget = 0;
1607 for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
1608 size_t const newMB = ZSTD_estimateCStreamSize_internal(level);
1609 if (newMB > memBudget) memBudget = newMB;
1610 }
1611 return memBudget;
1612 }
1613
1614 /* ZSTD_getFrameProgression():
1615 * tells how much data has been consumed (input) and produced (output) for current frame.
1616 * able to count progression inside worker threads (non-blocking mode).
1617 */
ZSTD_getFrameProgression(const ZSTD_CCtx * cctx)1618 ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx)
1619 {
1620 #ifdef ZSTD_MULTITHREAD
1621 if (cctx->appliedParams.nbWorkers > 0) {
1622 return ZSTDMT_getFrameProgression(cctx->mtctx);
1623 }
1624 #endif
1625 { ZSTD_frameProgression fp;
1626 size_t const buffered = (cctx->inBuff == NULL) ? 0 :
1627 cctx->inBuffPos - cctx->inToCompress;
1628 if (buffered) assert(cctx->inBuffPos >= cctx->inToCompress);
1629 assert(buffered <= ZSTD_BLOCKSIZE_MAX);
1630 fp.ingested = cctx->consumedSrcSize + buffered;
1631 fp.consumed = cctx->consumedSrcSize;
1632 fp.produced = cctx->producedCSize;
1633 fp.flushed = cctx->producedCSize; /* simplified; some data might still be left within streaming output buffer */
1634 fp.currentJobID = 0;
1635 fp.nbActiveWorkers = 0;
1636 return fp;
1637 } }
1638
1639 /*! ZSTD_toFlushNow()
1640 * Only useful for multithreading scenarios currently (nbWorkers >= 1).
1641 */
ZSTD_toFlushNow(ZSTD_CCtx * cctx)1642 size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx)
1643 {
1644 #ifdef ZSTD_MULTITHREAD
1645 if (cctx->appliedParams.nbWorkers > 0) {
1646 return ZSTDMT_toFlushNow(cctx->mtctx);
1647 }
1648 #endif
1649 (void)cctx;
1650 return 0; /* over-simplification; could also check if context is currently running in streaming mode, and in which case, report how many bytes are left to be flushed within output buffer */
1651 }
1652
ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,ZSTD_compressionParameters cParams2)1653 static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
1654 ZSTD_compressionParameters cParams2)
1655 {
1656 (void)cParams1;
1657 (void)cParams2;
1658 assert(cParams1.windowLog == cParams2.windowLog);
1659 assert(cParams1.chainLog == cParams2.chainLog);
1660 assert(cParams1.hashLog == cParams2.hashLog);
1661 assert(cParams1.searchLog == cParams2.searchLog);
1662 assert(cParams1.minMatch == cParams2.minMatch);
1663 assert(cParams1.targetLength == cParams2.targetLength);
1664 assert(cParams1.strategy == cParams2.strategy);
1665 }
1666
ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t * bs)1667 void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs)
1668 {
1669 int i;
1670 for (i = 0; i < ZSTD_REP_NUM; ++i)
1671 bs->rep[i] = repStartValue[i];
1672 bs->entropy.huf.repeatMode = HUF_repeat_none;
1673 bs->entropy.fse.offcode_repeatMode = FSE_repeat_none;
1674 bs->entropy.fse.matchlength_repeatMode = FSE_repeat_none;
1675 bs->entropy.fse.litlength_repeatMode = FSE_repeat_none;
1676 }
1677
1678 /*! ZSTD_invalidateMatchState()
1679 * Invalidate all the matches in the match finder tables.
1680 * Requires nextSrc and base to be set (can be NULL).
1681 */
ZSTD_invalidateMatchState(ZSTD_matchState_t * ms)1682 static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms)
1683 {
1684 ZSTD_window_clear(&ms->window);
1685
1686 ms->nextToUpdate = ms->window.dictLimit;
1687 ms->loadedDictEnd = 0;
1688 ms->opt.litLengthSum = 0; /* force reset of btopt stats */
1689 ms->dictMatchState = NULL;
1690 }
1691
1692 /**
1693 * Controls, for this matchState reset, whether the tables need to be cleared /
1694 * prepared for the coming compression (ZSTDcrp_makeClean), or whether the
1695 * tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a
1696 * subsequent operation will overwrite the table space anyways (e.g., copying
1697 * the matchState contents in from a CDict).
1698 */
1699 typedef enum {
1700 ZSTDcrp_makeClean,
1701 ZSTDcrp_leaveDirty
1702 } ZSTD_compResetPolicy_e;
1703
1704 /**
1705 * Controls, for this matchState reset, whether indexing can continue where it
1706 * left off (ZSTDirp_continue), or whether it needs to be restarted from zero
1707 * (ZSTDirp_reset).
1708 */
1709 typedef enum {
1710 ZSTDirp_continue,
1711 ZSTDirp_reset
1712 } ZSTD_indexResetPolicy_e;
1713
1714 typedef enum {
1715 ZSTD_resetTarget_CDict,
1716 ZSTD_resetTarget_CCtx
1717 } ZSTD_resetTarget_e;
1718
1719
1720 static size_t
ZSTD_reset_matchState(ZSTD_matchState_t * ms,ZSTD_cwksp * ws,const ZSTD_compressionParameters * cParams,const ZSTD_paramSwitch_e useRowMatchFinder,const ZSTD_compResetPolicy_e crp,const ZSTD_indexResetPolicy_e forceResetIndex,const ZSTD_resetTarget_e forWho)1721 ZSTD_reset_matchState(ZSTD_matchState_t* ms,
1722 ZSTD_cwksp* ws,
1723 const ZSTD_compressionParameters* cParams,
1724 const ZSTD_paramSwitch_e useRowMatchFinder,
1725 const ZSTD_compResetPolicy_e crp,
1726 const ZSTD_indexResetPolicy_e forceResetIndex,
1727 const ZSTD_resetTarget_e forWho)
1728 {
1729 /* disable chain table allocation for fast or row-based strategies */
1730 size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder,
1731 ms->dedicatedDictSearch && (forWho == ZSTD_resetTarget_CDict))
1732 ? ((size_t)1 << cParams->chainLog)
1733 : 0;
1734 size_t const hSize = ((size_t)1) << cParams->hashLog;
1735 U32 const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
1736 size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
1737
1738 DEBUGLOG(4, "reset indices : %u", forceResetIndex == ZSTDirp_reset);
1739 assert(useRowMatchFinder != ZSTD_ps_auto);
1740 if (forceResetIndex == ZSTDirp_reset) {
1741 ZSTD_window_init(&ms->window);
1742 ZSTD_cwksp_mark_tables_dirty(ws);
1743 }
1744
1745 ms->hashLog3 = hashLog3;
1746
1747 ZSTD_invalidateMatchState(ms);
1748
1749 assert(!ZSTD_cwksp_reserve_failed(ws)); /* check that allocation hasn't already failed */
1750
1751 ZSTD_cwksp_clear_tables(ws);
1752
1753 DEBUGLOG(5, "reserving table space");
1754 /* table Space */
1755 ms->hashTable = (U32*)ZSTD_cwksp_reserve_table(ws, hSize * sizeof(U32));
1756 ms->chainTable = (U32*)ZSTD_cwksp_reserve_table(ws, chainSize * sizeof(U32));
1757 ms->hashTable3 = (U32*)ZSTD_cwksp_reserve_table(ws, h3Size * sizeof(U32));
1758 RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
1759 "failed a workspace allocation in ZSTD_reset_matchState");
1760
1761 DEBUGLOG(4, "reset table : %u", crp!=ZSTDcrp_leaveDirty);
1762 if (crp!=ZSTDcrp_leaveDirty) {
1763 /* reset tables only */
1764 ZSTD_cwksp_clean_tables(ws);
1765 }
1766
1767 /* opt parser space */
1768 if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) {
1769 DEBUGLOG(4, "reserving optimal parser space");
1770 ms->opt.litFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (1<<Litbits) * sizeof(unsigned));
1771 ms->opt.litLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxLL+1) * sizeof(unsigned));
1772 ms->opt.matchLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxML+1) * sizeof(unsigned));
1773 ms->opt.offCodeFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxOff+1) * sizeof(unsigned));
1774 ms->opt.matchTable = (ZSTD_match_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t));
1775 ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
1776 }
1777
1778 if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) {
1779 { /* Row match finder needs an additional table of hashes ("tags") */
1780 size_t const tagTableSize = hSize*sizeof(U16);
1781 ms->tagTable = (U16*)ZSTD_cwksp_reserve_aligned(ws, tagTableSize);
1782 if (ms->tagTable) ZSTD_memset(ms->tagTable, 0, tagTableSize);
1783 }
1784 { /* Switch to 32-entry rows if searchLog is 5 (or more) */
1785 U32 const rowLog = BOUNDED(4, cParams->searchLog, 6);
1786 assert(cParams->hashLog >= rowLog);
1787 ms->rowHashLog = cParams->hashLog - rowLog;
1788 }
1789 }
1790
1791 ms->cParams = *cParams;
1792
1793 RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
1794 "failed a workspace allocation in ZSTD_reset_matchState");
1795 return 0;
1796 }
1797
1798 /* ZSTD_indexTooCloseToMax() :
1799 * minor optimization : prefer memset() rather than reduceIndex()
1800 * which is measurably slow in some circumstances (reported for Visual Studio).
1801 * Works when re-using a context for a lot of smallish inputs :
1802 * if all inputs are smaller than ZSTD_INDEXOVERFLOW_MARGIN,
1803 * memset() will be triggered before reduceIndex().
1804 */
1805 #define ZSTD_INDEXOVERFLOW_MARGIN (16 MB)
ZSTD_indexTooCloseToMax(ZSTD_window_t w)1806 static int ZSTD_indexTooCloseToMax(ZSTD_window_t w)
1807 {
1808 return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN);
1809 }
1810
1811 /** ZSTD_dictTooBig():
1812 * When dictionaries are larger than ZSTD_CHUNKSIZE_MAX they can't be loaded in
1813 * one go generically. So we ensure that in that case we reset the tables to zero,
1814 * so that we can load as much of the dictionary as possible.
1815 */
ZSTD_dictTooBig(size_t const loadedDictSize)1816 static int ZSTD_dictTooBig(size_t const loadedDictSize)
1817 {
1818 return loadedDictSize > ZSTD_CHUNKSIZE_MAX;
1819 }
1820
1821 /*! ZSTD_resetCCtx_internal() :
1822 * @param loadedDictSize The size of the dictionary to be loaded
1823 * into the context, if any. If no dictionary is used, or the
1824 * dictionary is being attached / copied, then pass 0.
1825 * note : `params` are assumed fully validated at this stage.
1826 */
ZSTD_resetCCtx_internal(ZSTD_CCtx * zc,ZSTD_CCtx_params const * params,U64 const pledgedSrcSize,size_t const loadedDictSize,ZSTD_compResetPolicy_e const crp,ZSTD_buffered_policy_e const zbuff)1827 static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
1828 ZSTD_CCtx_params const* params,
1829 U64 const pledgedSrcSize,
1830 size_t const loadedDictSize,
1831 ZSTD_compResetPolicy_e const crp,
1832 ZSTD_buffered_policy_e const zbuff)
1833 {
1834 ZSTD_cwksp* const ws = &zc->workspace;
1835 DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u, useRowMatchFinder=%d useBlockSplitter=%d",
1836 (U32)pledgedSrcSize, params->cParams.windowLog, (int)params->useRowMatchFinder, (int)params->useBlockSplitter);
1837 assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
1838
1839 zc->isFirstBlock = 1;
1840
1841 /* Set applied params early so we can modify them for LDM,
1842 * and point params at the applied params.
1843 */
1844 zc->appliedParams = *params;
1845 params = &zc->appliedParams;
1846
1847 assert(params->useRowMatchFinder != ZSTD_ps_auto);
1848 assert(params->useBlockSplitter != ZSTD_ps_auto);
1849 assert(params->ldmParams.enableLdm != ZSTD_ps_auto);
1850 if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
1851 /* Adjust long distance matching parameters */
1852 ZSTD_ldm_adjustParameters(&zc->appliedParams.ldmParams, ¶ms->cParams);
1853 assert(params->ldmParams.hashLog >= params->ldmParams.bucketSizeLog);
1854 assert(params->ldmParams.hashRateLog < 32);
1855 }
1856
1857 { size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params->cParams.windowLog), pledgedSrcSize));
1858 size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
1859 U32 const divider = (params->cParams.minMatch==3) ? 3 : 4;
1860 size_t const maxNbSeq = blockSize / divider;
1861 size_t const buffOutSize = (zbuff == ZSTDb_buffered && params->outBufferMode == ZSTD_bm_buffered)
1862 ? ZSTD_compressBound(blockSize) + 1
1863 : 0;
1864 size_t const buffInSize = (zbuff == ZSTDb_buffered && params->inBufferMode == ZSTD_bm_buffered)
1865 ? windowSize + blockSize
1866 : 0;
1867 size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize);
1868
1869 int const indexTooClose = ZSTD_indexTooCloseToMax(zc->blockState.matchState.window);
1870 int const dictTooBig = ZSTD_dictTooBig(loadedDictSize);
1871 ZSTD_indexResetPolicy_e needsIndexReset =
1872 (indexTooClose || dictTooBig || !zc->initialized) ? ZSTDirp_reset : ZSTDirp_continue;
1873
1874 size_t const neededSpace =
1875 ZSTD_estimateCCtxSize_usingCCtxParams_internal(
1876 ¶ms->cParams, ¶ms->ldmParams, zc->staticSize != 0, params->useRowMatchFinder,
1877 buffInSize, buffOutSize, pledgedSrcSize);
1878 int resizeWorkspace;
1879
1880 FORWARD_IF_ERROR(neededSpace, "cctx size estimate failed!");
1881
1882 if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);
1883
1884 { /* Check if workspace is large enough, alloc a new one if needed */
1885 int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;
1886 int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);
1887 resizeWorkspace = workspaceTooSmall || workspaceWasteful;
1888 DEBUGLOG(4, "Need %zu B workspace", neededSpace);
1889 DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);
1890
1891 if (resizeWorkspace) {
1892 DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB",
1893 ZSTD_cwksp_sizeof(ws) >> 10,
1894 neededSpace >> 10);
1895
1896 RETURN_ERROR_IF(zc->staticSize, memory_allocation, "static cctx : no resize");
1897
1898 needsIndexReset = ZSTDirp_reset;
1899
1900 ZSTD_cwksp_free(ws, zc->customMem);
1901 FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem), "");
1902
1903 DEBUGLOG(5, "reserving object space");
1904 /* Statically sized space.
1905 * entropyWorkspace never moves,
1906 * though prev/next block swap places */
1907 assert(ZSTD_cwksp_check_available(ws, 2 * sizeof(ZSTD_compressedBlockState_t)));
1908 zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
1909 RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock");
1910 zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
1911 RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock");
1912 zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, ENTROPY_WORKSPACE_SIZE);
1913 RETURN_ERROR_IF(zc->entropyWorkspace == NULL, memory_allocation, "couldn't allocate entropyWorkspace");
1914 } }
1915
1916 ZSTD_cwksp_clear(ws);
1917
1918 /* init params */
1919 zc->blockState.matchState.cParams = params->cParams;
1920 zc->pledgedSrcSizePlusOne = pledgedSrcSize+1;
1921 zc->consumedSrcSize = 0;
1922 zc->producedCSize = 0;
1923 if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
1924 zc->appliedParams.fParams.contentSizeFlag = 0;
1925 DEBUGLOG(4, "pledged content size : %u ; flag : %u",
1926 (unsigned)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);
1927 zc->blockSize = blockSize;
1928
1929 XXH64_reset(&zc->xxhState, 0);
1930 zc->stage = ZSTDcs_init;
1931 zc->dictID = 0;
1932 zc->dictContentSize = 0;
1933
1934 ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock);
1935
1936 /* ZSTD_wildcopy() is used to copy into the literals buffer,
1937 * so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
1938 */
1939 zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);
1940 zc->seqStore.maxNbLit = blockSize;
1941
1942 /* buffers */
1943 zc->bufferedPolicy = zbuff;
1944 zc->inBuffSize = buffInSize;
1945 zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);
1946 zc->outBuffSize = buffOutSize;
1947 zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);
1948
1949 /* ldm bucketOffsets table */
1950 if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
1951 /* TODO: avoid memset? */
1952 size_t const numBuckets =
1953 ((size_t)1) << (params->ldmParams.hashLog -
1954 params->ldmParams.bucketSizeLog);
1955 zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets);
1956 ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets);
1957 }
1958
1959 /* sequences storage */
1960 ZSTD_referenceExternalSequences(zc, NULL, 0);
1961 zc->seqStore.maxNbSeq = maxNbSeq;
1962 zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
1963 zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
1964 zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
1965 zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));
1966
1967 FORWARD_IF_ERROR(ZSTD_reset_matchState(
1968 &zc->blockState.matchState,
1969 ws,
1970 ¶ms->cParams,
1971 params->useRowMatchFinder,
1972 crp,
1973 needsIndexReset,
1974 ZSTD_resetTarget_CCtx), "");
1975
1976 /* ldm hash table */
1977 if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
1978 /* TODO: avoid memset? */
1979 size_t const ldmHSize = ((size_t)1) << params->ldmParams.hashLog;
1980 zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t));
1981 ZSTD_memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t));
1982 zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq));
1983 zc->maxNbLdmSequences = maxNbLdmSeq;
1984
1985 ZSTD_window_init(&zc->ldmState.window);
1986 zc->ldmState.loadedDictEnd = 0;
1987 }
1988
1989 DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
1990 assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace, resizeWorkspace));
1991
1992 zc->initialized = 1;
1993
1994 return 0;
1995 }
1996 }
1997
1998 /* ZSTD_invalidateRepCodes() :
1999 * ensures next compression will not use repcodes from previous block.
2000 * Note : only works with regular variant;
2001 * do not use with extDict variant ! */
ZSTD_invalidateRepCodes(ZSTD_CCtx * cctx)2002 void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) {
2003 int i;
2004 for (i=0; i<ZSTD_REP_NUM; i++) cctx->blockState.prevCBlock->rep[i] = 0;
2005 assert(!ZSTD_window_hasExtDict(cctx->blockState.matchState.window));
2006 }
2007
2008 /* These are the approximate sizes for each strategy past which copying the
2009 * dictionary tables into the working context is faster than using them
2010 * in-place.
2011 */
2012 static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = {
2013 8 KB, /* unused */
2014 8 KB, /* ZSTD_fast */
2015 16 KB, /* ZSTD_dfast */
2016 32 KB, /* ZSTD_greedy */
2017 32 KB, /* ZSTD_lazy */
2018 32 KB, /* ZSTD_lazy2 */
2019 32 KB, /* ZSTD_btlazy2 */
2020 32 KB, /* ZSTD_btopt */
2021 8 KB, /* ZSTD_btultra */
2022 8 KB /* ZSTD_btultra2 */
2023 };
2024
ZSTD_shouldAttachDict(const ZSTD_CDict * cdict,const ZSTD_CCtx_params * params,U64 pledgedSrcSize)2025 static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict,
2026 const ZSTD_CCtx_params* params,
2027 U64 pledgedSrcSize)
2028 {
2029 size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy];
2030 int const dedicatedDictSearch = cdict->matchState.dedicatedDictSearch;
2031 return dedicatedDictSearch
2032 || ( ( pledgedSrcSize <= cutoff
2033 || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
2034 || params->attachDictPref == ZSTD_dictForceAttach )
2035 && params->attachDictPref != ZSTD_dictForceCopy
2036 && !params->forceWindow ); /* dictMatchState isn't correctly
2037 * handled in _enforceMaxDist */
2038 }
2039
2040 static size_t
ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx * cctx,const ZSTD_CDict * cdict,ZSTD_CCtx_params params,U64 pledgedSrcSize,ZSTD_buffered_policy_e zbuff)2041 ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,
2042 const ZSTD_CDict* cdict,
2043 ZSTD_CCtx_params params,
2044 U64 pledgedSrcSize,
2045 ZSTD_buffered_policy_e zbuff)
2046 {
2047 DEBUGLOG(4, "ZSTD_resetCCtx_byAttachingCDict() pledgedSrcSize=%llu",
2048 (unsigned long long)pledgedSrcSize);
2049 {
2050 ZSTD_compressionParameters adjusted_cdict_cParams = cdict->matchState.cParams;
2051 unsigned const windowLog = params.cParams.windowLog;
2052 assert(windowLog != 0);
2053 /* Resize working context table params for input only, since the dict
2054 * has its own tables. */
2055 /* pledgedSrcSize == 0 means 0! */
2056
2057 if (cdict->matchState.dedicatedDictSearch) {
2058 ZSTD_dedicatedDictSearch_revertCParams(&adjusted_cdict_cParams);
2059 }
2060
2061 params.cParams = ZSTD_adjustCParams_internal(adjusted_cdict_cParams, pledgedSrcSize,
2062 cdict->dictContentSize, ZSTD_cpm_attachDict);
2063 params.cParams.windowLog = windowLog;
2064 params.useRowMatchFinder = cdict->useRowMatchFinder; /* cdict overrides */
2065 FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize,
2066 /* loadedDictSize */ 0,
2067 ZSTDcrp_makeClean, zbuff), "");
2068 assert(cctx->appliedParams.cParams.strategy == adjusted_cdict_cParams.strategy);
2069 }
2070
2071 { const U32 cdictEnd = (U32)( cdict->matchState.window.nextSrc
2072 - cdict->matchState.window.base);
2073 const U32 cdictLen = cdictEnd - cdict->matchState.window.dictLimit;
2074 if (cdictLen == 0) {
2075 /* don't even attach dictionaries with no contents */
2076 DEBUGLOG(4, "skipping attaching empty dictionary");
2077 } else {
2078 DEBUGLOG(4, "attaching dictionary into context");
2079 cctx->blockState.matchState.dictMatchState = &cdict->matchState;
2080
2081 /* prep working match state so dict matches never have negative indices
2082 * when they are translated to the working context's index space. */
2083 if (cctx->blockState.matchState.window.dictLimit < cdictEnd) {
2084 cctx->blockState.matchState.window.nextSrc =
2085 cctx->blockState.matchState.window.base + cdictEnd;
2086 ZSTD_window_clear(&cctx->blockState.matchState.window);
2087 }
2088 /* loadedDictEnd is expressed within the referential of the active context */
2089 cctx->blockState.matchState.loadedDictEnd = cctx->blockState.matchState.window.dictLimit;
2090 } }
2091
2092 cctx->dictID = cdict->dictID;
2093 cctx->dictContentSize = cdict->dictContentSize;
2094
2095 /* copy block state */
2096 ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));
2097
2098 return 0;
2099 }
2100
ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx * cctx,const ZSTD_CDict * cdict,ZSTD_CCtx_params params,U64 pledgedSrcSize,ZSTD_buffered_policy_e zbuff)2101 static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,
2102 const ZSTD_CDict* cdict,
2103 ZSTD_CCtx_params params,
2104 U64 pledgedSrcSize,
2105 ZSTD_buffered_policy_e zbuff)
2106 {
2107 const ZSTD_compressionParameters *cdict_cParams = &cdict->matchState.cParams;
2108
2109 assert(!cdict->matchState.dedicatedDictSearch);
2110 DEBUGLOG(4, "ZSTD_resetCCtx_byCopyingCDict() pledgedSrcSize=%llu",
2111 (unsigned long long)pledgedSrcSize);
2112
2113 { unsigned const windowLog = params.cParams.windowLog;
2114 assert(windowLog != 0);
2115 /* Copy only compression parameters related to tables. */
2116 params.cParams = *cdict_cParams;
2117 params.cParams.windowLog = windowLog;
2118 params.useRowMatchFinder = cdict->useRowMatchFinder;
2119 FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize,
2120 /* loadedDictSize */ 0,
2121 ZSTDcrp_leaveDirty, zbuff), "");
2122 assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
2123 assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog);
2124 assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog);
2125 }
2126
2127 ZSTD_cwksp_mark_tables_dirty(&cctx->workspace);
2128 assert(params.useRowMatchFinder != ZSTD_ps_auto);
2129
2130 /* copy tables */
2131 { size_t const chainSize = ZSTD_allocateChainTable(cdict_cParams->strategy, cdict->useRowMatchFinder, 0 /* DDS guaranteed disabled */)
2132 ? ((size_t)1 << cdict_cParams->chainLog)
2133 : 0;
2134 size_t const hSize = (size_t)1 << cdict_cParams->hashLog;
2135
2136 ZSTD_memcpy(cctx->blockState.matchState.hashTable,
2137 cdict->matchState.hashTable,
2138 hSize * sizeof(U32));
2139 /* Do not copy cdict's chainTable if cctx has parameters such that it would not use chainTable */
2140 if (ZSTD_allocateChainTable(cctx->appliedParams.cParams.strategy, cctx->appliedParams.useRowMatchFinder, 0 /* forDDSDict */)) {
2141 ZSTD_memcpy(cctx->blockState.matchState.chainTable,
2142 cdict->matchState.chainTable,
2143 chainSize * sizeof(U32));
2144 }
2145 /* copy tag table */
2146 if (ZSTD_rowMatchFinderUsed(cdict_cParams->strategy, cdict->useRowMatchFinder)) {
2147 size_t const tagTableSize = hSize*sizeof(U16);
2148 ZSTD_memcpy(cctx->blockState.matchState.tagTable,
2149 cdict->matchState.tagTable,
2150 tagTableSize);
2151 }
2152 }
2153
2154 /* Zero the hashTable3, since the cdict never fills it */
2155 { int const h3log = cctx->blockState.matchState.hashLog3;
2156 size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
2157 assert(cdict->matchState.hashLog3 == 0);
2158 ZSTD_memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32));
2159 }
2160
2161 ZSTD_cwksp_mark_tables_clean(&cctx->workspace);
2162
2163 /* copy dictionary offsets */
2164 { ZSTD_matchState_t const* srcMatchState = &cdict->matchState;
2165 ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState;
2166 dstMatchState->window = srcMatchState->window;
2167 dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;
2168 dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;
2169 }
2170
2171 cctx->dictID = cdict->dictID;
2172 cctx->dictContentSize = cdict->dictContentSize;
2173
2174 /* copy block state */
2175 ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));
2176
2177 return 0;
2178 }
2179
2180 /* We have a choice between copying the dictionary context into the working
2181 * context, or referencing the dictionary context from the working context
2182 * in-place. We decide here which strategy to use. */
ZSTD_resetCCtx_usingCDict(ZSTD_CCtx * cctx,const ZSTD_CDict * cdict,const ZSTD_CCtx_params * params,U64 pledgedSrcSize,ZSTD_buffered_policy_e zbuff)2183 static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
2184 const ZSTD_CDict* cdict,
2185 const ZSTD_CCtx_params* params,
2186 U64 pledgedSrcSize,
2187 ZSTD_buffered_policy_e zbuff)
2188 {
2189
2190 DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)",
2191 (unsigned)pledgedSrcSize);
2192
2193 if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) {
2194 return ZSTD_resetCCtx_byAttachingCDict(
2195 cctx, cdict, *params, pledgedSrcSize, zbuff);
2196 } else {
2197 return ZSTD_resetCCtx_byCopyingCDict(
2198 cctx, cdict, *params, pledgedSrcSize, zbuff);
2199 }
2200 }
2201
2202 /*! ZSTD_copyCCtx_internal() :
2203 * Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
2204 * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
2205 * The "context", in this case, refers to the hash and chain tables,
2206 * entropy tables, and dictionary references.
2207 * `windowLog` value is enforced if != 0, otherwise value is copied from srcCCtx.
2208 * @return : 0, or an error code */
ZSTD_copyCCtx_internal(ZSTD_CCtx * dstCCtx,const ZSTD_CCtx * srcCCtx,ZSTD_frameParameters fParams,U64 pledgedSrcSize,ZSTD_buffered_policy_e zbuff)2209 static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
2210 const ZSTD_CCtx* srcCCtx,
2211 ZSTD_frameParameters fParams,
2212 U64 pledgedSrcSize,
2213 ZSTD_buffered_policy_e zbuff)
2214 {
2215 RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong,
2216 "Can't copy a ctx that's not in init stage.");
2217 DEBUGLOG(5, "ZSTD_copyCCtx_internal");
2218 ZSTD_memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));
2219 { ZSTD_CCtx_params params = dstCCtx->requestedParams;
2220 /* Copy only compression parameters related to tables. */
2221 params.cParams = srcCCtx->appliedParams.cParams;
2222 assert(srcCCtx->appliedParams.useRowMatchFinder != ZSTD_ps_auto);
2223 assert(srcCCtx->appliedParams.useBlockSplitter != ZSTD_ps_auto);
2224 assert(srcCCtx->appliedParams.ldmParams.enableLdm != ZSTD_ps_auto);
2225 params.useRowMatchFinder = srcCCtx->appliedParams.useRowMatchFinder;
2226 params.useBlockSplitter = srcCCtx->appliedParams.useBlockSplitter;
2227 params.ldmParams = srcCCtx->appliedParams.ldmParams;
2228 params.fParams = fParams;
2229 ZSTD_resetCCtx_internal(dstCCtx, ¶ms, pledgedSrcSize,
2230 /* loadedDictSize */ 0,
2231 ZSTDcrp_leaveDirty, zbuff);
2232 assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog);
2233 assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy);
2234 assert(dstCCtx->appliedParams.cParams.hashLog == srcCCtx->appliedParams.cParams.hashLog);
2235 assert(dstCCtx->appliedParams.cParams.chainLog == srcCCtx->appliedParams.cParams.chainLog);
2236 assert(dstCCtx->blockState.matchState.hashLog3 == srcCCtx->blockState.matchState.hashLog3);
2237 }
2238
2239 ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace);
2240
2241 /* copy tables */
2242 { size_t const chainSize = ZSTD_allocateChainTable(srcCCtx->appliedParams.cParams.strategy,
2243 srcCCtx->appliedParams.useRowMatchFinder,
2244 0 /* forDDSDict */)
2245 ? ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog)
2246 : 0;
2247 size_t const hSize = (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;
2248 int const h3log = srcCCtx->blockState.matchState.hashLog3;
2249 size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
2250
2251 ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable,
2252 srcCCtx->blockState.matchState.hashTable,
2253 hSize * sizeof(U32));
2254 ZSTD_memcpy(dstCCtx->blockState.matchState.chainTable,
2255 srcCCtx->blockState.matchState.chainTable,
2256 chainSize * sizeof(U32));
2257 ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable3,
2258 srcCCtx->blockState.matchState.hashTable3,
2259 h3Size * sizeof(U32));
2260 }
2261
2262 ZSTD_cwksp_mark_tables_clean(&dstCCtx->workspace);
2263
2264 /* copy dictionary offsets */
2265 {
2266 const ZSTD_matchState_t* srcMatchState = &srcCCtx->blockState.matchState;
2267 ZSTD_matchState_t* dstMatchState = &dstCCtx->blockState.matchState;
2268 dstMatchState->window = srcMatchState->window;
2269 dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;
2270 dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;
2271 }
2272 dstCCtx->dictID = srcCCtx->dictID;
2273 dstCCtx->dictContentSize = srcCCtx->dictContentSize;
2274
2275 /* copy block state */
2276 ZSTD_memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock));
2277
2278 return 0;
2279 }
2280
2281 /*! ZSTD_copyCCtx() :
2282 * Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
2283 * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
2284 * pledgedSrcSize==0 means "unknown".
2285 * @return : 0, or an error code */
ZSTD_copyCCtx(ZSTD_CCtx * dstCCtx,const ZSTD_CCtx * srcCCtx,unsigned long long pledgedSrcSize)2286 size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize)
2287 {
2288 ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
2289 ZSTD_buffered_policy_e const zbuff = srcCCtx->bufferedPolicy;
2290 ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1);
2291 if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
2292 fParams.contentSizeFlag = (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN);
2293
2294 return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx,
2295 fParams, pledgedSrcSize,
2296 zbuff);
2297 }
2298
2299
2300 #define ZSTD_ROWSIZE 16
2301 /*! ZSTD_reduceTable() :
2302 * reduce table indexes by `reducerValue`, or squash to zero.
2303 * PreserveMark preserves "unsorted mark" for btlazy2 strategy.
2304 * It must be set to a clear 0/1 value, to remove branch during inlining.
2305 * Presume table size is a multiple of ZSTD_ROWSIZE
2306 * to help auto-vectorization */
2307 FORCE_INLINE_TEMPLATE void
ZSTD_reduceTable_internal(U32 * const table,U32 const size,U32 const reducerValue,int const preserveMark)2308 ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerValue, int const preserveMark)
2309 {
2310 int const nbRows = (int)size / ZSTD_ROWSIZE;
2311 int cellNb = 0;
2312 int rowNb;
2313 /* Protect special index values < ZSTD_WINDOW_START_INDEX. */
2314 U32 const reducerThreshold = reducerValue + ZSTD_WINDOW_START_INDEX;
2315 assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */
2316 assert(size < (1U<<31)); /* can be casted to int */
2317
2318 #if ZSTD_MEMORY_SANITIZER && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
2319 /* To validate that the table re-use logic is sound, and that we don't
2320 * access table space that we haven't cleaned, we re-"poison" the table
2321 * space every time we mark it dirty.
2322 *
2323 * This function however is intended to operate on those dirty tables and
2324 * re-clean them. So when this function is used correctly, we can unpoison
2325 * the memory it operated on. This introduces a blind spot though, since
2326 * if we now try to operate on __actually__ poisoned memory, we will not
2327 * detect that. */
2328 __msan_unpoison(table, size * sizeof(U32));
2329 #endif
2330
2331 for (rowNb=0 ; rowNb < nbRows ; rowNb++) {
2332 int column;
2333 for (column=0; column<ZSTD_ROWSIZE; column++) {
2334 U32 newVal;
2335 if (preserveMark && table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) {
2336 /* This write is pointless, but is required(?) for the compiler
2337 * to auto-vectorize the loop. */
2338 newVal = ZSTD_DUBT_UNSORTED_MARK;
2339 } else if (table[cellNb] < reducerThreshold) {
2340 newVal = 0;
2341 } else {
2342 newVal = table[cellNb] - reducerValue;
2343 }
2344 table[cellNb] = newVal;
2345 cellNb++;
2346 } }
2347 }
2348
ZSTD_reduceTable(U32 * const table,U32 const size,U32 const reducerValue)2349 static void ZSTD_reduceTable(U32* const table, U32 const size, U32 const reducerValue)
2350 {
2351 ZSTD_reduceTable_internal(table, size, reducerValue, 0);
2352 }
2353
ZSTD_reduceTable_btlazy2(U32 * const table,U32 const size,U32 const reducerValue)2354 static void ZSTD_reduceTable_btlazy2(U32* const table, U32 const size, U32 const reducerValue)
2355 {
2356 ZSTD_reduceTable_internal(table, size, reducerValue, 1);
2357 }
2358
2359 /*! ZSTD_reduceIndex() :
2360 * rescale all indexes to avoid future overflow (indexes are U32) */
ZSTD_reduceIndex(ZSTD_matchState_t * ms,ZSTD_CCtx_params const * params,const U32 reducerValue)2361 static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* params, const U32 reducerValue)
2362 {
2363 { U32 const hSize = (U32)1 << params->cParams.hashLog;
2364 ZSTD_reduceTable(ms->hashTable, hSize, reducerValue);
2365 }
2366
2367 if (ZSTD_allocateChainTable(params->cParams.strategy, params->useRowMatchFinder, (U32)ms->dedicatedDictSearch)) {
2368 U32 const chainSize = (U32)1 << params->cParams.chainLog;
2369 if (params->cParams.strategy == ZSTD_btlazy2)
2370 ZSTD_reduceTable_btlazy2(ms->chainTable, chainSize, reducerValue);
2371 else
2372 ZSTD_reduceTable(ms->chainTable, chainSize, reducerValue);
2373 }
2374
2375 if (ms->hashLog3) {
2376 U32 const h3Size = (U32)1 << ms->hashLog3;
2377 ZSTD_reduceTable(ms->hashTable3, h3Size, reducerValue);
2378 }
2379 }
2380
2381
2382 /*-*******************************************************
2383 * Block entropic compression
2384 *********************************************************/
2385
2386 /* See doc/zstd_compression_format.md for detailed format description */
2387
ZSTD_seqToCodes(const seqStore_t * seqStorePtr)2388 void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
2389 {
2390 const seqDef* const sequences = seqStorePtr->sequencesStart;
2391 BYTE* const llCodeTable = seqStorePtr->llCode;
2392 BYTE* const ofCodeTable = seqStorePtr->ofCode;
2393 BYTE* const mlCodeTable = seqStorePtr->mlCode;
2394 U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
2395 U32 u;
2396 assert(nbSeq <= seqStorePtr->maxNbSeq);
2397 for (u=0; u<nbSeq; u++) {
2398 U32 const llv = sequences[u].litLength;
2399 U32 const mlv = sequences[u].mlBase;
2400 llCodeTable[u] = (BYTE)ZSTD_LLcode(llv);
2401 ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offBase);
2402 mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv);
2403 }
2404 if (seqStorePtr->longLengthType==ZSTD_llt_literalLength)
2405 llCodeTable[seqStorePtr->longLengthPos] = MaxLL;
2406 if (seqStorePtr->longLengthType==ZSTD_llt_matchLength)
2407 mlCodeTable[seqStorePtr->longLengthPos] = MaxML;
2408 }
2409
2410 /* ZSTD_useTargetCBlockSize():
2411 * Returns if target compressed block size param is being used.
2412 * If used, compression will do best effort to make a compressed block size to be around targetCBlockSize.
2413 * Returns 1 if true, 0 otherwise. */
ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params * cctxParams)2414 static int ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params* cctxParams)
2415 {
2416 DEBUGLOG(5, "ZSTD_useTargetCBlockSize (targetCBlockSize=%zu)", cctxParams->targetCBlockSize);
2417 return (cctxParams->targetCBlockSize != 0);
2418 }
2419
2420 /* ZSTD_blockSplitterEnabled():
2421 * Returns if block splitting param is being used
2422 * If used, compression will do best effort to split a block in order to improve compression ratio.
2423 * At the time this function is called, the parameter must be finalized.
2424 * Returns 1 if true, 0 otherwise. */
ZSTD_blockSplitterEnabled(ZSTD_CCtx_params * cctxParams)2425 static int ZSTD_blockSplitterEnabled(ZSTD_CCtx_params* cctxParams)
2426 {
2427 DEBUGLOG(5, "ZSTD_blockSplitterEnabled (useBlockSplitter=%d)", cctxParams->useBlockSplitter);
2428 assert(cctxParams->useBlockSplitter != ZSTD_ps_auto);
2429 return (cctxParams->useBlockSplitter == ZSTD_ps_enable);
2430 }
2431
2432 /* Type returned by ZSTD_buildSequencesStatistics containing finalized symbol encoding types
2433 * and size of the sequences statistics
2434 */
2435 typedef struct {
2436 U32 LLtype;
2437 U32 Offtype;
2438 U32 MLtype;
2439 size_t size;
2440 size_t lastCountSize; /* Accounts for bug in 1.3.4. More detail in ZSTD_entropyCompressSeqStore_internal() */
2441 } ZSTD_symbolEncodingTypeStats_t;
2442
2443 /* ZSTD_buildSequencesStatistics():
2444 * Returns a ZSTD_symbolEncodingTypeStats_t, or a zstd error code in the `size` field.
2445 * Modifies `nextEntropy` to have the appropriate values as a side effect.
2446 * nbSeq must be greater than 0.
2447 *
2448 * entropyWkspSize must be of size at least ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1)*sizeof(U32)
2449 */
2450 static ZSTD_symbolEncodingTypeStats_t
ZSTD_buildSequencesStatistics(seqStore_t * seqStorePtr,size_t nbSeq,const ZSTD_fseCTables_t * prevEntropy,ZSTD_fseCTables_t * nextEntropy,BYTE * dst,const BYTE * const dstEnd,ZSTD_strategy strategy,unsigned * countWorkspace,void * entropyWorkspace,size_t entropyWkspSize)2451 ZSTD_buildSequencesStatistics(seqStore_t* seqStorePtr, size_t nbSeq,
2452 const ZSTD_fseCTables_t* prevEntropy, ZSTD_fseCTables_t* nextEntropy,
2453 BYTE* dst, const BYTE* const dstEnd,
2454 ZSTD_strategy strategy, unsigned* countWorkspace,
2455 void* entropyWorkspace, size_t entropyWkspSize) {
2456 BYTE* const ostart = dst;
2457 const BYTE* const oend = dstEnd;
2458 BYTE* op = ostart;
2459 FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable;
2460 FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable;
2461 FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable;
2462 const BYTE* const ofCodeTable = seqStorePtr->ofCode;
2463 const BYTE* const llCodeTable = seqStorePtr->llCode;
2464 const BYTE* const mlCodeTable = seqStorePtr->mlCode;
2465 ZSTD_symbolEncodingTypeStats_t stats;
2466
2467 stats.lastCountSize = 0;
2468 /* convert length/distances into codes */
2469 ZSTD_seqToCodes(seqStorePtr);
2470 assert(op <= oend);
2471 assert(nbSeq != 0); /* ZSTD_selectEncodingType() divides by nbSeq */
2472 /* build CTable for Literal Lengths */
2473 { unsigned max = MaxLL;
2474 size_t const mostFrequent = HIST_countFast_wksp(countWorkspace, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
2475 DEBUGLOG(5, "Building LL table");
2476 nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode;
2477 stats.LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode,
2478 countWorkspace, max, mostFrequent, nbSeq,
2479 LLFSELog, prevEntropy->litlengthCTable,
2480 LL_defaultNorm, LL_defaultNormLog,
2481 ZSTD_defaultAllowed, strategy);
2482 assert(set_basic < set_compressed && set_rle < set_compressed);
2483 assert(!(stats.LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
2484 { size_t const countSize = ZSTD_buildCTable(
2485 op, (size_t)(oend - op),
2486 CTable_LitLength, LLFSELog, (symbolEncodingType_e)stats.LLtype,
2487 countWorkspace, max, llCodeTable, nbSeq,
2488 LL_defaultNorm, LL_defaultNormLog, MaxLL,
2489 prevEntropy->litlengthCTable,
2490 sizeof(prevEntropy->litlengthCTable),
2491 entropyWorkspace, entropyWkspSize);
2492 if (ZSTD_isError(countSize)) {
2493 DEBUGLOG(3, "ZSTD_buildCTable for LitLens failed");
2494 stats.size = countSize;
2495 return stats;
2496 }
2497 if (stats.LLtype == set_compressed)
2498 stats.lastCountSize = countSize;
2499 op += countSize;
2500 assert(op <= oend);
2501 } }
2502 /* build CTable for Offsets */
2503 { unsigned max = MaxOff;
2504 size_t const mostFrequent = HIST_countFast_wksp(
2505 countWorkspace, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
2506 /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
2507 ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
2508 DEBUGLOG(5, "Building OF table");
2509 nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode;
2510 stats.Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode,
2511 countWorkspace, max, mostFrequent, nbSeq,
2512 OffFSELog, prevEntropy->offcodeCTable,
2513 OF_defaultNorm, OF_defaultNormLog,
2514 defaultPolicy, strategy);
2515 assert(!(stats.Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
2516 { size_t const countSize = ZSTD_buildCTable(
2517 op, (size_t)(oend - op),
2518 CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)stats.Offtype,
2519 countWorkspace, max, ofCodeTable, nbSeq,
2520 OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
2521 prevEntropy->offcodeCTable,
2522 sizeof(prevEntropy->offcodeCTable),
2523 entropyWorkspace, entropyWkspSize);
2524 if (ZSTD_isError(countSize)) {
2525 DEBUGLOG(3, "ZSTD_buildCTable for Offsets failed");
2526 stats.size = countSize;
2527 return stats;
2528 }
2529 if (stats.Offtype == set_compressed)
2530 stats.lastCountSize = countSize;
2531 op += countSize;
2532 assert(op <= oend);
2533 } }
2534 /* build CTable for MatchLengths */
2535 { unsigned max = MaxML;
2536 size_t const mostFrequent = HIST_countFast_wksp(
2537 countWorkspace, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */
2538 DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
2539 nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode;
2540 stats.MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode,
2541 countWorkspace, max, mostFrequent, nbSeq,
2542 MLFSELog, prevEntropy->matchlengthCTable,
2543 ML_defaultNorm, ML_defaultNormLog,
2544 ZSTD_defaultAllowed, strategy);
2545 assert(!(stats.MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
2546 { size_t const countSize = ZSTD_buildCTable(
2547 op, (size_t)(oend - op),
2548 CTable_MatchLength, MLFSELog, (symbolEncodingType_e)stats.MLtype,
2549 countWorkspace, max, mlCodeTable, nbSeq,
2550 ML_defaultNorm, ML_defaultNormLog, MaxML,
2551 prevEntropy->matchlengthCTable,
2552 sizeof(prevEntropy->matchlengthCTable),
2553 entropyWorkspace, entropyWkspSize);
2554 if (ZSTD_isError(countSize)) {
2555 DEBUGLOG(3, "ZSTD_buildCTable for MatchLengths failed");
2556 stats.size = countSize;
2557 return stats;
2558 }
2559 if (stats.MLtype == set_compressed)
2560 stats.lastCountSize = countSize;
2561 op += countSize;
2562 assert(op <= oend);
2563 } }
2564 stats.size = (size_t)(op-ostart);
2565 return stats;
2566 }
2567
2568 /* ZSTD_entropyCompressSeqStore_internal():
2569 * compresses both literals and sequences
2570 * Returns compressed size of block, or a zstd error.
2571 */
2572 #define SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO 20
2573 MEM_STATIC size_t
ZSTD_entropyCompressSeqStore_internal(seqStore_t * seqStorePtr,const ZSTD_entropyCTables_t * prevEntropy,ZSTD_entropyCTables_t * nextEntropy,const ZSTD_CCtx_params * cctxParams,void * dst,size_t dstCapacity,void * entropyWorkspace,size_t entropyWkspSize,const int bmi2)2574 ZSTD_entropyCompressSeqStore_internal(seqStore_t* seqStorePtr,
2575 const ZSTD_entropyCTables_t* prevEntropy,
2576 ZSTD_entropyCTables_t* nextEntropy,
2577 const ZSTD_CCtx_params* cctxParams,
2578 void* dst, size_t dstCapacity,
2579 void* entropyWorkspace, size_t entropyWkspSize,
2580 const int bmi2)
2581 {
2582 const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
2583 ZSTD_strategy const strategy = cctxParams->cParams.strategy;
2584 unsigned* count = (unsigned*)entropyWorkspace;
2585 FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable;
2586 FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable;
2587 FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable;
2588 const seqDef* const sequences = seqStorePtr->sequencesStart;
2589 const size_t nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
2590 const BYTE* const ofCodeTable = seqStorePtr->ofCode;
2591 const BYTE* const llCodeTable = seqStorePtr->llCode;
2592 const BYTE* const mlCodeTable = seqStorePtr->mlCode;
2593 BYTE* const ostart = (BYTE*)dst;
2594 BYTE* const oend = ostart + dstCapacity;
2595 BYTE* op = ostart;
2596 size_t lastCountSize;
2597
2598 entropyWorkspace = count + (MaxSeq + 1);
2599 entropyWkspSize -= (MaxSeq + 1) * sizeof(*count);
2600
2601 DEBUGLOG(4, "ZSTD_entropyCompressSeqStore_internal (nbSeq=%zu)", nbSeq);
2602 ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
2603 assert(entropyWkspSize >= HUF_WORKSPACE_SIZE);
2604
2605 /* Compress literals */
2606 { const BYTE* const literals = seqStorePtr->litStart;
2607 size_t const numSequences = seqStorePtr->sequences - seqStorePtr->sequencesStart;
2608 size_t const numLiterals = seqStorePtr->lit - seqStorePtr->litStart;
2609 /* Base suspicion of uncompressibility on ratio of literals to sequences */
2610 unsigned const suspectUncompressible = (numSequences == 0) || (numLiterals / numSequences >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO);
2611 size_t const litSize = (size_t)(seqStorePtr->lit - literals);
2612 size_t const cSize = ZSTD_compressLiterals(
2613 &prevEntropy->huf, &nextEntropy->huf,
2614 cctxParams->cParams.strategy,
2615 ZSTD_literalsCompressionIsDisabled(cctxParams),
2616 op, dstCapacity,
2617 literals, litSize,
2618 entropyWorkspace, entropyWkspSize,
2619 bmi2, suspectUncompressible);
2620 FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed");
2621 assert(cSize <= dstCapacity);
2622 op += cSize;
2623 }
2624
2625 /* Sequences Header */
2626 RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
2627 dstSize_tooSmall, "Can't fit seq hdr in output buf!");
2628 if (nbSeq < 128) {
2629 *op++ = (BYTE)nbSeq;
2630 } else if (nbSeq < LONGNBSEQ) {
2631 op[0] = (BYTE)((nbSeq>>8) + 0x80);
2632 op[1] = (BYTE)nbSeq;
2633 op+=2;
2634 } else {
2635 op[0]=0xFF;
2636 MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ));
2637 op+=3;
2638 }
2639 assert(op <= oend);
2640 if (nbSeq==0) {
2641 /* Copy the old tables over as if we repeated them */
2642 ZSTD_memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse));
2643 return (size_t)(op - ostart);
2644 }
2645 {
2646 ZSTD_symbolEncodingTypeStats_t stats;
2647 BYTE* seqHead = op++;
2648 /* build stats for sequences */
2649 stats = ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq,
2650 &prevEntropy->fse, &nextEntropy->fse,
2651 op, oend,
2652 strategy, count,
2653 entropyWorkspace, entropyWkspSize);
2654 FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!");
2655 *seqHead = (BYTE)((stats.LLtype<<6) + (stats.Offtype<<4) + (stats.MLtype<<2));
2656 lastCountSize = stats.lastCountSize;
2657 op += stats.size;
2658 }
2659
2660 { size_t const bitstreamSize = ZSTD_encodeSequences(
2661 op, (size_t)(oend - op),
2662 CTable_MatchLength, mlCodeTable,
2663 CTable_OffsetBits, ofCodeTable,
2664 CTable_LitLength, llCodeTable,
2665 sequences, nbSeq,
2666 longOffsets, bmi2);
2667 FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
2668 op += bitstreamSize;
2669 assert(op <= oend);
2670 /* zstd versions <= 1.3.4 mistakenly report corruption when
2671 * FSE_readNCount() receives a buffer < 4 bytes.
2672 * Fixed by https://github.com/facebook/zstd/pull/1146.
2673 * This can happen when the last set_compressed table present is 2
2674 * bytes and the bitstream is only one byte.
2675 * In this exceedingly rare case, we will simply emit an uncompressed
2676 * block, since it isn't worth optimizing.
2677 */
2678 if (lastCountSize && (lastCountSize + bitstreamSize) < 4) {
2679 /* lastCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
2680 assert(lastCountSize + bitstreamSize == 3);
2681 DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "
2682 "emitting an uncompressed block.");
2683 return 0;
2684 }
2685 }
2686
2687 DEBUGLOG(5, "compressed block size : %u", (unsigned)(op - ostart));
2688 return (size_t)(op - ostart);
2689 }
2690
2691 MEM_STATIC size_t
ZSTD_entropyCompressSeqStore(seqStore_t * seqStorePtr,const ZSTD_entropyCTables_t * prevEntropy,ZSTD_entropyCTables_t * nextEntropy,const ZSTD_CCtx_params * cctxParams,void * dst,size_t dstCapacity,size_t srcSize,void * entropyWorkspace,size_t entropyWkspSize,int bmi2)2692 ZSTD_entropyCompressSeqStore(seqStore_t* seqStorePtr,
2693 const ZSTD_entropyCTables_t* prevEntropy,
2694 ZSTD_entropyCTables_t* nextEntropy,
2695 const ZSTD_CCtx_params* cctxParams,
2696 void* dst, size_t dstCapacity,
2697 size_t srcSize,
2698 void* entropyWorkspace, size_t entropyWkspSize,
2699 int bmi2)
2700 {
2701 size_t const cSize = ZSTD_entropyCompressSeqStore_internal(
2702 seqStorePtr, prevEntropy, nextEntropy, cctxParams,
2703 dst, dstCapacity,
2704 entropyWorkspace, entropyWkspSize, bmi2);
2705 if (cSize == 0) return 0;
2706 /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
2707 * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
2708 */
2709 if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity))
2710 return 0; /* block not compressed */
2711 FORWARD_IF_ERROR(cSize, "ZSTD_entropyCompressSeqStore_internal failed");
2712
2713 /* Check compressibility */
2714 { size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy);
2715 if (cSize >= maxCSize) return 0; /* block not compressed */
2716 }
2717 DEBUGLOG(4, "ZSTD_entropyCompressSeqStore() cSize: %zu", cSize);
2718 return cSize;
2719 }
2720
2721 /* ZSTD_selectBlockCompressor() :
2722 * Not static, but internal use only (used by long distance matcher)
2723 * assumption : strat is a valid strategy */
ZSTD_selectBlockCompressor(ZSTD_strategy strat,ZSTD_paramSwitch_e useRowMatchFinder,ZSTD_dictMode_e dictMode)2724 ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_paramSwitch_e useRowMatchFinder, ZSTD_dictMode_e dictMode)
2725 {
2726 static const ZSTD_blockCompressor blockCompressor[4][ZSTD_STRATEGY_MAX+1] = {
2727 { ZSTD_compressBlock_fast /* default for 0 */,
2728 ZSTD_compressBlock_fast,
2729 ZSTD_compressBlock_doubleFast,
2730 ZSTD_compressBlock_greedy,
2731 ZSTD_compressBlock_lazy,
2732 ZSTD_compressBlock_lazy2,
2733 ZSTD_compressBlock_btlazy2,
2734 ZSTD_compressBlock_btopt,
2735 ZSTD_compressBlock_btultra,
2736 ZSTD_compressBlock_btultra2 },
2737 { ZSTD_compressBlock_fast_extDict /* default for 0 */,
2738 ZSTD_compressBlock_fast_extDict,
2739 ZSTD_compressBlock_doubleFast_extDict,
2740 ZSTD_compressBlock_greedy_extDict,
2741 ZSTD_compressBlock_lazy_extDict,
2742 ZSTD_compressBlock_lazy2_extDict,
2743 ZSTD_compressBlock_btlazy2_extDict,
2744 ZSTD_compressBlock_btopt_extDict,
2745 ZSTD_compressBlock_btultra_extDict,
2746 ZSTD_compressBlock_btultra_extDict },
2747 { ZSTD_compressBlock_fast_dictMatchState /* default for 0 */,
2748 ZSTD_compressBlock_fast_dictMatchState,
2749 ZSTD_compressBlock_doubleFast_dictMatchState,
2750 ZSTD_compressBlock_greedy_dictMatchState,
2751 ZSTD_compressBlock_lazy_dictMatchState,
2752 ZSTD_compressBlock_lazy2_dictMatchState,
2753 ZSTD_compressBlock_btlazy2_dictMatchState,
2754 ZSTD_compressBlock_btopt_dictMatchState,
2755 ZSTD_compressBlock_btultra_dictMatchState,
2756 ZSTD_compressBlock_btultra_dictMatchState },
2757 { NULL /* default for 0 */,
2758 NULL,
2759 NULL,
2760 ZSTD_compressBlock_greedy_dedicatedDictSearch,
2761 ZSTD_compressBlock_lazy_dedicatedDictSearch,
2762 ZSTD_compressBlock_lazy2_dedicatedDictSearch,
2763 NULL,
2764 NULL,
2765 NULL,
2766 NULL }
2767 };
2768 ZSTD_blockCompressor selectedCompressor;
2769 ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);
2770
2771 assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
2772 DEBUGLOG(4, "Selected block compressor: dictMode=%d strat=%d rowMatchfinder=%d", (int)dictMode, (int)strat, (int)useRowMatchFinder);
2773 if (ZSTD_rowMatchFinderUsed(strat, useRowMatchFinder)) {
2774 static const ZSTD_blockCompressor rowBasedBlockCompressors[4][3] = {
2775 { ZSTD_compressBlock_greedy_row,
2776 ZSTD_compressBlock_lazy_row,
2777 ZSTD_compressBlock_lazy2_row },
2778 { ZSTD_compressBlock_greedy_extDict_row,
2779 ZSTD_compressBlock_lazy_extDict_row,
2780 ZSTD_compressBlock_lazy2_extDict_row },
2781 { ZSTD_compressBlock_greedy_dictMatchState_row,
2782 ZSTD_compressBlock_lazy_dictMatchState_row,
2783 ZSTD_compressBlock_lazy2_dictMatchState_row },
2784 { ZSTD_compressBlock_greedy_dedicatedDictSearch_row,
2785 ZSTD_compressBlock_lazy_dedicatedDictSearch_row,
2786 ZSTD_compressBlock_lazy2_dedicatedDictSearch_row }
2787 };
2788 DEBUGLOG(4, "Selecting a row-based matchfinder");
2789 assert(useRowMatchFinder != ZSTD_ps_auto);
2790 selectedCompressor = rowBasedBlockCompressors[(int)dictMode][(int)strat - (int)ZSTD_greedy];
2791 } else {
2792 selectedCompressor = blockCompressor[(int)dictMode][(int)strat];
2793 }
2794 assert(selectedCompressor != NULL);
2795 return selectedCompressor;
2796 }
2797
ZSTD_storeLastLiterals(seqStore_t * seqStorePtr,const BYTE * anchor,size_t lastLLSize)2798 static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr,
2799 const BYTE* anchor, size_t lastLLSize)
2800 {
2801 ZSTD_memcpy(seqStorePtr->lit, anchor, lastLLSize);
2802 seqStorePtr->lit += lastLLSize;
2803 }
2804
ZSTD_resetSeqStore(seqStore_t * ssPtr)2805 void ZSTD_resetSeqStore(seqStore_t* ssPtr)
2806 {
2807 ssPtr->lit = ssPtr->litStart;
2808 ssPtr->sequences = ssPtr->sequencesStart;
2809 ssPtr->longLengthType = ZSTD_llt_none;
2810 }
2811
2812 typedef enum { ZSTDbss_compress, ZSTDbss_noCompress } ZSTD_buildSeqStore_e;
2813
ZSTD_buildSeqStore(ZSTD_CCtx * zc,const void * src,size_t srcSize)2814 static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)
2815 {
2816 ZSTD_matchState_t* const ms = &zc->blockState.matchState;
2817 DEBUGLOG(5, "ZSTD_buildSeqStore (srcSize=%zu)", srcSize);
2818 assert(srcSize <= ZSTD_BLOCKSIZE_MAX);
2819 /* Assert that we have correctly flushed the ctx params into the ms's copy */
2820 ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams);
2821 if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {
2822 if (zc->appliedParams.cParams.strategy >= ZSTD_btopt) {
2823 ZSTD_ldm_skipRawSeqStoreBytes(&zc->externSeqStore, srcSize);
2824 } else {
2825 ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch);
2826 }
2827 return ZSTDbss_noCompress; /* don't even attempt compression below a certain srcSize */
2828 }
2829 ZSTD_resetSeqStore(&(zc->seqStore));
2830 /* required for optimal parser to read stats from dictionary */
2831 ms->opt.symbolCosts = &zc->blockState.prevCBlock->entropy;
2832 /* tell the optimal parser how we expect to compress literals */
2833 ms->opt.literalCompressionMode = zc->appliedParams.literalCompressionMode;
2834 /* a gap between an attached dict and the current window is not safe,
2835 * they must remain adjacent,
2836 * and when that stops being the case, the dict must be unset */
2837 assert(ms->dictMatchState == NULL || ms->loadedDictEnd == ms->window.dictLimit);
2838
2839 /* limited update after a very long match */
2840 { const BYTE* const base = ms->window.base;
2841 const BYTE* const istart = (const BYTE*)src;
2842 const U32 curr = (U32)(istart-base);
2843 if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */
2844 if (curr > ms->nextToUpdate + 384)
2845 ms->nextToUpdate = curr - MIN(192, (U32)(curr - ms->nextToUpdate - 384));
2846 }
2847
2848 /* select and store sequences */
2849 { ZSTD_dictMode_e const dictMode = ZSTD_matchState_dictMode(ms);
2850 size_t lastLLSize;
2851 { int i;
2852 for (i = 0; i < ZSTD_REP_NUM; ++i)
2853 zc->blockState.nextCBlock->rep[i] = zc->blockState.prevCBlock->rep[i];
2854 }
2855 if (zc->externSeqStore.pos < zc->externSeqStore.size) {
2856 assert(zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_disable);
2857 /* Updates ldmSeqStore.pos */
2858 lastLLSize =
2859 ZSTD_ldm_blockCompress(&zc->externSeqStore,
2860 ms, &zc->seqStore,
2861 zc->blockState.nextCBlock->rep,
2862 zc->appliedParams.useRowMatchFinder,
2863 src, srcSize);
2864 assert(zc->externSeqStore.pos <= zc->externSeqStore.size);
2865 } else if (zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) {
2866 rawSeqStore_t ldmSeqStore = kNullRawSeqStore;
2867
2868 ldmSeqStore.seq = zc->ldmSequences;
2869 ldmSeqStore.capacity = zc->maxNbLdmSequences;
2870 /* Updates ldmSeqStore.size */
2871 FORWARD_IF_ERROR(ZSTD_ldm_generateSequences(&zc->ldmState, &ldmSeqStore,
2872 &zc->appliedParams.ldmParams,
2873 src, srcSize), "");
2874 /* Updates ldmSeqStore.pos */
2875 lastLLSize =
2876 ZSTD_ldm_blockCompress(&ldmSeqStore,
2877 ms, &zc->seqStore,
2878 zc->blockState.nextCBlock->rep,
2879 zc->appliedParams.useRowMatchFinder,
2880 src, srcSize);
2881 assert(ldmSeqStore.pos == ldmSeqStore.size);
2882 } else { /* not long range mode */
2883 ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy,
2884 zc->appliedParams.useRowMatchFinder,
2885 dictMode);
2886 ms->ldmSeqStore = NULL;
2887 lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize);
2888 }
2889 { const BYTE* const lastLiterals = (const BYTE*)src + srcSize - lastLLSize;
2890 ZSTD_storeLastLiterals(&zc->seqStore, lastLiterals, lastLLSize);
2891 } }
2892 return ZSTDbss_compress;
2893 }
2894
ZSTD_copyBlockSequences(ZSTD_CCtx * zc)2895 static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc)
2896 {
2897 const seqStore_t* seqStore = ZSTD_getSeqStore(zc);
2898 const seqDef* seqStoreSeqs = seqStore->sequencesStart;
2899 size_t seqStoreSeqSize = seqStore->sequences - seqStoreSeqs;
2900 size_t seqStoreLiteralsSize = (size_t)(seqStore->lit - seqStore->litStart);
2901 size_t literalsRead = 0;
2902 size_t lastLLSize;
2903
2904 ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];
2905 size_t i;
2906 repcodes_t updatedRepcodes;
2907
2908 assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);
2909 /* Ensure we have enough space for last literals "sequence" */
2910 assert(zc->seqCollector.maxSequences >= seqStoreSeqSize + 1);
2911 ZSTD_memcpy(updatedRepcodes.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
2912 for (i = 0; i < seqStoreSeqSize; ++i) {
2913 U32 rawOffset = seqStoreSeqs[i].offBase - ZSTD_REP_NUM;
2914 outSeqs[i].litLength = seqStoreSeqs[i].litLength;
2915 outSeqs[i].matchLength = seqStoreSeqs[i].mlBase + MINMATCH;
2916 outSeqs[i].rep = 0;
2917
2918 if (i == seqStore->longLengthPos) {
2919 if (seqStore->longLengthType == ZSTD_llt_literalLength) {
2920 outSeqs[i].litLength += 0x10000;
2921 } else if (seqStore->longLengthType == ZSTD_llt_matchLength) {
2922 outSeqs[i].matchLength += 0x10000;
2923 }
2924 }
2925
2926 if (seqStoreSeqs[i].offBase <= ZSTD_REP_NUM) {
2927 /* Derive the correct offset corresponding to a repcode */
2928 outSeqs[i].rep = seqStoreSeqs[i].offBase;
2929 if (outSeqs[i].litLength != 0) {
2930 rawOffset = updatedRepcodes.rep[outSeqs[i].rep - 1];
2931 } else {
2932 if (outSeqs[i].rep == 3) {
2933 rawOffset = updatedRepcodes.rep[0] - 1;
2934 } else {
2935 rawOffset = updatedRepcodes.rep[outSeqs[i].rep];
2936 }
2937 }
2938 }
2939 outSeqs[i].offset = rawOffset;
2940 /* seqStoreSeqs[i].offset == offCode+1, and ZSTD_updateRep() expects offCode
2941 so we provide seqStoreSeqs[i].offset - 1 */
2942 ZSTD_updateRep(updatedRepcodes.rep,
2943 seqStoreSeqs[i].offBase - 1,
2944 seqStoreSeqs[i].litLength == 0);
2945 literalsRead += outSeqs[i].litLength;
2946 }
2947 /* Insert last literals (if any exist) in the block as a sequence with ml == off == 0.
2948 * If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker
2949 * for the block boundary, according to the API.
2950 */
2951 assert(seqStoreLiteralsSize >= literalsRead);
2952 lastLLSize = seqStoreLiteralsSize - literalsRead;
2953 outSeqs[i].litLength = (U32)lastLLSize;
2954 outSeqs[i].matchLength = outSeqs[i].offset = outSeqs[i].rep = 0;
2955 seqStoreSeqSize++;
2956 zc->seqCollector.seqIndex += seqStoreSeqSize;
2957 }
2958
ZSTD_generateSequences(ZSTD_CCtx * zc,ZSTD_Sequence * outSeqs,size_t outSeqsSize,const void * src,size_t srcSize)2959 size_t ZSTD_generateSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
2960 size_t outSeqsSize, const void* src, size_t srcSize)
2961 {
2962 const size_t dstCapacity = ZSTD_compressBound(srcSize);
2963 void* dst = ZSTD_customMalloc(dstCapacity, ZSTD_defaultCMem);
2964 SeqCollector seqCollector;
2965
2966 RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!");
2967
2968 seqCollector.collectSequences = 1;
2969 seqCollector.seqStart = outSeqs;
2970 seqCollector.seqIndex = 0;
2971 seqCollector.maxSequences = outSeqsSize;
2972 zc->seqCollector = seqCollector;
2973
2974 ZSTD_compress2(zc, dst, dstCapacity, src, srcSize);
2975 ZSTD_customFree(dst, ZSTD_defaultCMem);
2976 return zc->seqCollector.seqIndex;
2977 }
2978
ZSTD_mergeBlockDelimiters(ZSTD_Sequence * sequences,size_t seqsSize)2979 size_t ZSTD_mergeBlockDelimiters(ZSTD_Sequence* sequences, size_t seqsSize) {
2980 size_t in = 0;
2981 size_t out = 0;
2982 for (; in < seqsSize; ++in) {
2983 if (sequences[in].offset == 0 && sequences[in].matchLength == 0) {
2984 if (in != seqsSize - 1) {
2985 sequences[in+1].litLength += sequences[in].litLength;
2986 }
2987 } else {
2988 sequences[out] = sequences[in];
2989 ++out;
2990 }
2991 }
2992 return out;
2993 }
2994
2995 /* Unrolled loop to read four size_ts of input at a time. Returns 1 if is RLE, 0 if not. */
ZSTD_isRLE(const BYTE * src,size_t length)2996 static int ZSTD_isRLE(const BYTE* src, size_t length) {
2997 const BYTE* ip = src;
2998 const BYTE value = ip[0];
2999 const size_t valueST = (size_t)((U64)value * 0x0101010101010101ULL);
3000 const size_t unrollSize = sizeof(size_t) * 4;
3001 const size_t unrollMask = unrollSize - 1;
3002 const size_t prefixLength = length & unrollMask;
3003 size_t i;
3004 size_t u;
3005 if (length == 1) return 1;
3006 /* Check if prefix is RLE first before using unrolled loop */
3007 if (prefixLength && ZSTD_count(ip+1, ip, ip+prefixLength) != prefixLength-1) {
3008 return 0;
3009 }
3010 for (i = prefixLength; i != length; i += unrollSize) {
3011 for (u = 0; u < unrollSize; u += sizeof(size_t)) {
3012 if (MEM_readST(ip + i + u) != valueST) {
3013 return 0;
3014 }
3015 }
3016 }
3017 return 1;
3018 }
3019
3020 /* Returns true if the given block may be RLE.
3021 * This is just a heuristic based on the compressibility.
3022 * It may return both false positives and false negatives.
3023 */
ZSTD_maybeRLE(seqStore_t const * seqStore)3024 static int ZSTD_maybeRLE(seqStore_t const* seqStore)
3025 {
3026 size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart);
3027 size_t const nbLits = (size_t)(seqStore->lit - seqStore->litStart);
3028
3029 return nbSeqs < 4 && nbLits < 10;
3030 }
3031
ZSTD_blockState_confirmRepcodesAndEntropyTables(ZSTD_blockState_t * const bs)3032 static void ZSTD_blockState_confirmRepcodesAndEntropyTables(ZSTD_blockState_t* const bs)
3033 {
3034 ZSTD_compressedBlockState_t* const tmp = bs->prevCBlock;
3035 bs->prevCBlock = bs->nextCBlock;
3036 bs->nextCBlock = tmp;
3037 }
3038
3039 /* Writes the block header */
writeBlockHeader(void * op,size_t cSize,size_t blockSize,U32 lastBlock)3040 static void writeBlockHeader(void* op, size_t cSize, size_t blockSize, U32 lastBlock) {
3041 U32 const cBlockHeader = cSize == 1 ?
3042 lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
3043 lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
3044 MEM_writeLE24(op, cBlockHeader);
3045 DEBUGLOG(3, "writeBlockHeader: cSize: %zu blockSize: %zu lastBlock: %u", cSize, blockSize, lastBlock);
3046 }
3047
3048 /** ZSTD_buildBlockEntropyStats_literals() :
3049 * Builds entropy for the literals.
3050 * Stores literals block type (raw, rle, compressed, repeat) and
3051 * huffman description table to hufMetadata.
3052 * Requires ENTROPY_WORKSPACE_SIZE workspace
3053 * @return : size of huffman description table or error code */
ZSTD_buildBlockEntropyStats_literals(void * const src,size_t srcSize,const ZSTD_hufCTables_t * prevHuf,ZSTD_hufCTables_t * nextHuf,ZSTD_hufCTablesMetadata_t * hufMetadata,const int literalsCompressionIsDisabled,void * workspace,size_t wkspSize)3054 static size_t ZSTD_buildBlockEntropyStats_literals(void* const src, size_t srcSize,
3055 const ZSTD_hufCTables_t* prevHuf,
3056 ZSTD_hufCTables_t* nextHuf,
3057 ZSTD_hufCTablesMetadata_t* hufMetadata,
3058 const int literalsCompressionIsDisabled,
3059 void* workspace, size_t wkspSize)
3060 {
3061 BYTE* const wkspStart = (BYTE*)workspace;
3062 BYTE* const wkspEnd = wkspStart + wkspSize;
3063 BYTE* const countWkspStart = wkspStart;
3064 unsigned* const countWksp = (unsigned*)workspace;
3065 const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned);
3066 BYTE* const nodeWksp = countWkspStart + countWkspSize;
3067 const size_t nodeWkspSize = wkspEnd-nodeWksp;
3068 unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
3069 unsigned huffLog = HUF_TABLELOG_DEFAULT;
3070 HUF_repeat repeat = prevHuf->repeatMode;
3071 DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_literals (srcSize=%zu)", srcSize);
3072
3073 /* Prepare nextEntropy assuming reusing the existing table */
3074 ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
3075
3076 if (literalsCompressionIsDisabled) {
3077 DEBUGLOG(5, "set_basic - disabled");
3078 hufMetadata->hType = set_basic;
3079 return 0;
3080 }
3081
3082 /* small ? don't even attempt compression (speed opt) */
3083 #ifndef COMPRESS_LITERALS_SIZE_MIN
3084 #define COMPRESS_LITERALS_SIZE_MIN 63
3085 #endif
3086 { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
3087 if (srcSize <= minLitSize) {
3088 DEBUGLOG(5, "set_basic - too small");
3089 hufMetadata->hType = set_basic;
3090 return 0;
3091 }
3092 }
3093
3094 /* Scan input and build symbol stats */
3095 { size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)src, srcSize, workspace, wkspSize);
3096 FORWARD_IF_ERROR(largest, "HIST_count_wksp failed");
3097 if (largest == srcSize) {
3098 DEBUGLOG(5, "set_rle");
3099 hufMetadata->hType = set_rle;
3100 return 0;
3101 }
3102 if (largest <= (srcSize >> 7)+4) {
3103 DEBUGLOG(5, "set_basic - no gain");
3104 hufMetadata->hType = set_basic;
3105 return 0;
3106 }
3107 }
3108
3109 /* Validate the previous Huffman table */
3110 if (repeat == HUF_repeat_check && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) {
3111 repeat = HUF_repeat_none;
3112 }
3113
3114 /* Build Huffman Tree */
3115 ZSTD_memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable));
3116 huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
3117 { size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp,
3118 maxSymbolValue, huffLog,
3119 nodeWksp, nodeWkspSize);
3120 FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp");
3121 huffLog = (U32)maxBits;
3122 { /* Build and write the CTable */
3123 size_t const newCSize = HUF_estimateCompressedSize(
3124 (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
3125 size_t const hSize = HUF_writeCTable_wksp(
3126 hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
3127 (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog,
3128 nodeWksp, nodeWkspSize);
3129 /* Check against repeating the previous CTable */
3130 if (repeat != HUF_repeat_none) {
3131 size_t const oldCSize = HUF_estimateCompressedSize(
3132 (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue);
3133 if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) {
3134 DEBUGLOG(5, "set_repeat - smaller");
3135 ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
3136 hufMetadata->hType = set_repeat;
3137 return 0;
3138 }
3139 }
3140 if (newCSize + hSize >= srcSize) {
3141 DEBUGLOG(5, "set_basic - no gains");
3142 ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
3143 hufMetadata->hType = set_basic;
3144 return 0;
3145 }
3146 DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize);
3147 hufMetadata->hType = set_compressed;
3148 nextHuf->repeatMode = HUF_repeat_check;
3149 return hSize;
3150 }
3151 }
3152 }
3153
3154
3155 /* ZSTD_buildDummySequencesStatistics():
3156 * Returns a ZSTD_symbolEncodingTypeStats_t with all encoding types as set_basic,
3157 * and updates nextEntropy to the appropriate repeatMode.
3158 */
3159 static ZSTD_symbolEncodingTypeStats_t
ZSTD_buildDummySequencesStatistics(ZSTD_fseCTables_t * nextEntropy)3160 ZSTD_buildDummySequencesStatistics(ZSTD_fseCTables_t* nextEntropy) {
3161 ZSTD_symbolEncodingTypeStats_t stats = {set_basic, set_basic, set_basic, 0, 0};
3162 nextEntropy->litlength_repeatMode = FSE_repeat_none;
3163 nextEntropy->offcode_repeatMode = FSE_repeat_none;
3164 nextEntropy->matchlength_repeatMode = FSE_repeat_none;
3165 return stats;
3166 }
3167
3168 /** ZSTD_buildBlockEntropyStats_sequences() :
3169 * Builds entropy for the sequences.
3170 * Stores symbol compression modes and fse table to fseMetadata.
3171 * Requires ENTROPY_WORKSPACE_SIZE wksp.
3172 * @return : size of fse tables or error code */
ZSTD_buildBlockEntropyStats_sequences(seqStore_t * seqStorePtr,const ZSTD_fseCTables_t * prevEntropy,ZSTD_fseCTables_t * nextEntropy,const ZSTD_CCtx_params * cctxParams,ZSTD_fseCTablesMetadata_t * fseMetadata,void * workspace,size_t wkspSize)3173 static size_t ZSTD_buildBlockEntropyStats_sequences(seqStore_t* seqStorePtr,
3174 const ZSTD_fseCTables_t* prevEntropy,
3175 ZSTD_fseCTables_t* nextEntropy,
3176 const ZSTD_CCtx_params* cctxParams,
3177 ZSTD_fseCTablesMetadata_t* fseMetadata,
3178 void* workspace, size_t wkspSize)
3179 {
3180 ZSTD_strategy const strategy = cctxParams->cParams.strategy;
3181 size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
3182 BYTE* const ostart = fseMetadata->fseTablesBuffer;
3183 BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer);
3184 BYTE* op = ostart;
3185 unsigned* countWorkspace = (unsigned*)workspace;
3186 unsigned* entropyWorkspace = countWorkspace + (MaxSeq + 1);
3187 size_t entropyWorkspaceSize = wkspSize - (MaxSeq + 1) * sizeof(*countWorkspace);
3188 ZSTD_symbolEncodingTypeStats_t stats;
3189
3190 DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_sequences (nbSeq=%zu)", nbSeq);
3191 stats = nbSeq != 0 ? ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq,
3192 prevEntropy, nextEntropy, op, oend,
3193 strategy, countWorkspace,
3194 entropyWorkspace, entropyWorkspaceSize)
3195 : ZSTD_buildDummySequencesStatistics(nextEntropy);
3196 FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!");
3197 fseMetadata->llType = (symbolEncodingType_e) stats.LLtype;
3198 fseMetadata->ofType = (symbolEncodingType_e) stats.Offtype;
3199 fseMetadata->mlType = (symbolEncodingType_e) stats.MLtype;
3200 fseMetadata->lastCountSize = stats.lastCountSize;
3201 return stats.size;
3202 }
3203
3204
3205 /** ZSTD_buildBlockEntropyStats() :
3206 * Builds entropy for the block.
3207 * Requires workspace size ENTROPY_WORKSPACE_SIZE
3208 *
3209 * @return : 0 on success or error code
3210 */
ZSTD_buildBlockEntropyStats(seqStore_t * seqStorePtr,const ZSTD_entropyCTables_t * prevEntropy,ZSTD_entropyCTables_t * nextEntropy,const ZSTD_CCtx_params * cctxParams,ZSTD_entropyCTablesMetadata_t * entropyMetadata,void * workspace,size_t wkspSize)3211 size_t ZSTD_buildBlockEntropyStats(seqStore_t* seqStorePtr,
3212 const ZSTD_entropyCTables_t* prevEntropy,
3213 ZSTD_entropyCTables_t* nextEntropy,
3214 const ZSTD_CCtx_params* cctxParams,
3215 ZSTD_entropyCTablesMetadata_t* entropyMetadata,
3216 void* workspace, size_t wkspSize)
3217 {
3218 size_t const litSize = seqStorePtr->lit - seqStorePtr->litStart;
3219 entropyMetadata->hufMetadata.hufDesSize =
3220 ZSTD_buildBlockEntropyStats_literals(seqStorePtr->litStart, litSize,
3221 &prevEntropy->huf, &nextEntropy->huf,
3222 &entropyMetadata->hufMetadata,
3223 ZSTD_literalsCompressionIsDisabled(cctxParams),
3224 workspace, wkspSize);
3225 FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildBlockEntropyStats_literals failed");
3226 entropyMetadata->fseMetadata.fseTablesSize =
3227 ZSTD_buildBlockEntropyStats_sequences(seqStorePtr,
3228 &prevEntropy->fse, &nextEntropy->fse,
3229 cctxParams,
3230 &entropyMetadata->fseMetadata,
3231 workspace, wkspSize);
3232 FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildBlockEntropyStats_sequences failed");
3233 return 0;
3234 }
3235
3236 /* Returns the size estimate for the literals section (header + content) of a block */
ZSTD_estimateBlockSize_literal(const BYTE * literals,size_t litSize,const ZSTD_hufCTables_t * huf,const ZSTD_hufCTablesMetadata_t * hufMetadata,void * workspace,size_t wkspSize,int writeEntropy)3237 static size_t ZSTD_estimateBlockSize_literal(const BYTE* literals, size_t litSize,
3238 const ZSTD_hufCTables_t* huf,
3239 const ZSTD_hufCTablesMetadata_t* hufMetadata,
3240 void* workspace, size_t wkspSize,
3241 int writeEntropy)
3242 {
3243 unsigned* const countWksp = (unsigned*)workspace;
3244 unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
3245 size_t literalSectionHeaderSize = 3 + (litSize >= 1 KB) + (litSize >= 16 KB);
3246 U32 singleStream = litSize < 256;
3247
3248 if (hufMetadata->hType == set_basic) return litSize;
3249 else if (hufMetadata->hType == set_rle) return 1;
3250 else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
3251 size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
3252 if (ZSTD_isError(largest)) return litSize;
3253 { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
3254 if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize;
3255 if (!singleStream) cLitSizeEstimate += 6; /* multi-stream huffman uses 6-byte jump table */
3256 return cLitSizeEstimate + literalSectionHeaderSize;
3257 } }
3258 assert(0); /* impossible */
3259 return 0;
3260 }
3261
3262 /* Returns the size estimate for the FSE-compressed symbols (of, ml, ll) of a block */
ZSTD_estimateBlockSize_symbolType(symbolEncodingType_e type,const BYTE * codeTable,size_t nbSeq,unsigned maxCode,const FSE_CTable * fseCTable,const U8 * additionalBits,short const * defaultNorm,U32 defaultNormLog,U32 defaultMax,void * workspace,size_t wkspSize)3263 static size_t ZSTD_estimateBlockSize_symbolType(symbolEncodingType_e type,
3264 const BYTE* codeTable, size_t nbSeq, unsigned maxCode,
3265 const FSE_CTable* fseCTable,
3266 const U8* additionalBits,
3267 short const* defaultNorm, U32 defaultNormLog, U32 defaultMax,
3268 void* workspace, size_t wkspSize)
3269 {
3270 unsigned* const countWksp = (unsigned*)workspace;
3271 const BYTE* ctp = codeTable;
3272 const BYTE* const ctStart = ctp;
3273 const BYTE* const ctEnd = ctStart + nbSeq;
3274 size_t cSymbolTypeSizeEstimateInBits = 0;
3275 unsigned max = maxCode;
3276
3277 HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */
3278 if (type == set_basic) {
3279 /* We selected this encoding type, so it must be valid. */
3280 assert(max <= defaultMax);
3281 (void)defaultMax;
3282 cSymbolTypeSizeEstimateInBits = ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max);
3283 } else if (type == set_rle) {
3284 cSymbolTypeSizeEstimateInBits = 0;
3285 } else if (type == set_compressed || type == set_repeat) {
3286 cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max);
3287 }
3288 if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) {
3289 return nbSeq * 10;
3290 }
3291 while (ctp < ctEnd) {
3292 if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp];
3293 else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */
3294 ctp++;
3295 }
3296 return cSymbolTypeSizeEstimateInBits >> 3;
3297 }
3298
3299 /* Returns the size estimate for the sequences section (header + content) of a block */
ZSTD_estimateBlockSize_sequences(const BYTE * ofCodeTable,const BYTE * llCodeTable,const BYTE * mlCodeTable,size_t nbSeq,const ZSTD_fseCTables_t * fseTables,const ZSTD_fseCTablesMetadata_t * fseMetadata,void * workspace,size_t wkspSize,int writeEntropy)3300 static size_t ZSTD_estimateBlockSize_sequences(const BYTE* ofCodeTable,
3301 const BYTE* llCodeTable,
3302 const BYTE* mlCodeTable,
3303 size_t nbSeq,
3304 const ZSTD_fseCTables_t* fseTables,
3305 const ZSTD_fseCTablesMetadata_t* fseMetadata,
3306 void* workspace, size_t wkspSize,
3307 int writeEntropy)
3308 {
3309 size_t sequencesSectionHeaderSize = 1 /* seqHead */ + 1 /* min seqSize size */ + (nbSeq >= 128) + (nbSeq >= LONGNBSEQ);
3310 size_t cSeqSizeEstimate = 0;
3311 cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, nbSeq, MaxOff,
3312 fseTables->offcodeCTable, NULL,
3313 OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
3314 workspace, wkspSize);
3315 cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->llType, llCodeTable, nbSeq, MaxLL,
3316 fseTables->litlengthCTable, LL_bits,
3317 LL_defaultNorm, LL_defaultNormLog, MaxLL,
3318 workspace, wkspSize);
3319 cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, nbSeq, MaxML,
3320 fseTables->matchlengthCTable, ML_bits,
3321 ML_defaultNorm, ML_defaultNormLog, MaxML,
3322 workspace, wkspSize);
3323 if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize;
3324 return cSeqSizeEstimate + sequencesSectionHeaderSize;
3325 }
3326
3327 /* Returns the size estimate for a given stream of literals, of, ll, ml */
ZSTD_estimateBlockSize(const BYTE * literals,size_t litSize,const BYTE * ofCodeTable,const BYTE * llCodeTable,const BYTE * mlCodeTable,size_t nbSeq,const ZSTD_entropyCTables_t * entropy,const ZSTD_entropyCTablesMetadata_t * entropyMetadata,void * workspace,size_t wkspSize,int writeLitEntropy,int writeSeqEntropy)3328 static size_t ZSTD_estimateBlockSize(const BYTE* literals, size_t litSize,
3329 const BYTE* ofCodeTable,
3330 const BYTE* llCodeTable,
3331 const BYTE* mlCodeTable,
3332 size_t nbSeq,
3333 const ZSTD_entropyCTables_t* entropy,
3334 const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
3335 void* workspace, size_t wkspSize,
3336 int writeLitEntropy, int writeSeqEntropy) {
3337 size_t const literalsSize = ZSTD_estimateBlockSize_literal(literals, litSize,
3338 &entropy->huf, &entropyMetadata->hufMetadata,
3339 workspace, wkspSize, writeLitEntropy);
3340 size_t const seqSize = ZSTD_estimateBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
3341 nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
3342 workspace, wkspSize, writeSeqEntropy);
3343 return seqSize + literalsSize + ZSTD_blockHeaderSize;
3344 }
3345
3346 /* Builds entropy statistics and uses them for blocksize estimation.
3347 *
3348 * Returns the estimated compressed size of the seqStore, or a zstd error.
3349 */
ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(seqStore_t * seqStore,ZSTD_CCtx * zc)3350 static size_t ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(seqStore_t* seqStore, ZSTD_CCtx* zc) {
3351 ZSTD_entropyCTablesMetadata_t* entropyMetadata = &zc->blockSplitCtx.entropyMetadata;
3352 DEBUGLOG(6, "ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize()");
3353 FORWARD_IF_ERROR(ZSTD_buildBlockEntropyStats(seqStore,
3354 &zc->blockState.prevCBlock->entropy,
3355 &zc->blockState.nextCBlock->entropy,
3356 &zc->appliedParams,
3357 entropyMetadata,
3358 zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
3359 return ZSTD_estimateBlockSize(seqStore->litStart, (size_t)(seqStore->lit - seqStore->litStart),
3360 seqStore->ofCode, seqStore->llCode, seqStore->mlCode,
3361 (size_t)(seqStore->sequences - seqStore->sequencesStart),
3362 &zc->blockState.nextCBlock->entropy, entropyMetadata, zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE,
3363 (int)(entropyMetadata->hufMetadata.hType == set_compressed), 1);
3364 }
3365
3366 /* Returns literals bytes represented in a seqStore */
ZSTD_countSeqStoreLiteralsBytes(const seqStore_t * const seqStore)3367 static size_t ZSTD_countSeqStoreLiteralsBytes(const seqStore_t* const seqStore) {
3368 size_t literalsBytes = 0;
3369 size_t const nbSeqs = seqStore->sequences - seqStore->sequencesStart;
3370 size_t i;
3371 for (i = 0; i < nbSeqs; ++i) {
3372 seqDef seq = seqStore->sequencesStart[i];
3373 literalsBytes += seq.litLength;
3374 if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_literalLength) {
3375 literalsBytes += 0x10000;
3376 }
3377 }
3378 return literalsBytes;
3379 }
3380
3381 /* Returns match bytes represented in a seqStore */
ZSTD_countSeqStoreMatchBytes(const seqStore_t * const seqStore)3382 static size_t ZSTD_countSeqStoreMatchBytes(const seqStore_t* const seqStore) {
3383 size_t matchBytes = 0;
3384 size_t const nbSeqs = seqStore->sequences - seqStore->sequencesStart;
3385 size_t i;
3386 for (i = 0; i < nbSeqs; ++i) {
3387 seqDef seq = seqStore->sequencesStart[i];
3388 matchBytes += seq.mlBase + MINMATCH;
3389 if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_matchLength) {
3390 matchBytes += 0x10000;
3391 }
3392 }
3393 return matchBytes;
3394 }
3395
3396 /* Derives the seqStore that is a chunk of the originalSeqStore from [startIdx, endIdx).
3397 * Stores the result in resultSeqStore.
3398 */
ZSTD_deriveSeqStoreChunk(seqStore_t * resultSeqStore,const seqStore_t * originalSeqStore,size_t startIdx,size_t endIdx)3399 static void ZSTD_deriveSeqStoreChunk(seqStore_t* resultSeqStore,
3400 const seqStore_t* originalSeqStore,
3401 size_t startIdx, size_t endIdx) {
3402 BYTE* const litEnd = originalSeqStore->lit;
3403 size_t literalsBytes;
3404 size_t literalsBytesPreceding = 0;
3405
3406 *resultSeqStore = *originalSeqStore;
3407 if (startIdx > 0) {
3408 resultSeqStore->sequences = originalSeqStore->sequencesStart + startIdx;
3409 literalsBytesPreceding = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore);
3410 }
3411
3412 /* Move longLengthPos into the correct position if necessary */
3413 if (originalSeqStore->longLengthType != ZSTD_llt_none) {
3414 if (originalSeqStore->longLengthPos < startIdx || originalSeqStore->longLengthPos > endIdx) {
3415 resultSeqStore->longLengthType = ZSTD_llt_none;
3416 } else {
3417 resultSeqStore->longLengthPos -= (U32)startIdx;
3418 }
3419 }
3420 resultSeqStore->sequencesStart = originalSeqStore->sequencesStart + startIdx;
3421 resultSeqStore->sequences = originalSeqStore->sequencesStart + endIdx;
3422 literalsBytes = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore);
3423 resultSeqStore->litStart += literalsBytesPreceding;
3424 if (endIdx == (size_t)(originalSeqStore->sequences - originalSeqStore->sequencesStart)) {
3425 /* This accounts for possible last literals if the derived chunk reaches the end of the block */
3426 resultSeqStore->lit = litEnd;
3427 } else {
3428 resultSeqStore->lit = resultSeqStore->litStart+literalsBytes;
3429 }
3430 resultSeqStore->llCode += startIdx;
3431 resultSeqStore->mlCode += startIdx;
3432 resultSeqStore->ofCode += startIdx;
3433 }
3434
3435 /**
3436 * Returns the raw offset represented by the combination of offCode, ll0, and repcode history.
3437 * offCode must represent a repcode in the numeric representation of ZSTD_storeSeq().
3438 */
3439 static U32
ZSTD_resolveRepcodeToRawOffset(const U32 rep[ZSTD_REP_NUM],const U32 offCode,const U32 ll0)3440 ZSTD_resolveRepcodeToRawOffset(const U32 rep[ZSTD_REP_NUM], const U32 offCode, const U32 ll0)
3441 {
3442 U32 const adjustedOffCode = STORED_REPCODE(offCode) - 1 + ll0; /* [ 0 - 3 ] */
3443 assert(STORED_IS_REPCODE(offCode));
3444 if (adjustedOffCode == ZSTD_REP_NUM) {
3445 /* litlength == 0 and offCode == 2 implies selection of first repcode - 1 */
3446 assert(rep[0] > 0);
3447 return rep[0] - 1;
3448 }
3449 return rep[adjustedOffCode];
3450 }
3451
3452 /**
3453 * ZSTD_seqStore_resolveOffCodes() reconciles any possible divergences in offset history that may arise
3454 * due to emission of RLE/raw blocks that disturb the offset history,
3455 * and replaces any repcodes within the seqStore that may be invalid.
3456 *
3457 * dRepcodes are updated as would be on the decompression side.
3458 * cRepcodes are updated exactly in accordance with the seqStore.
3459 *
3460 * Note : this function assumes seq->offBase respects the following numbering scheme :
3461 * 0 : invalid
3462 * 1-3 : repcode 1-3
3463 * 4+ : real_offset+3
3464 */
ZSTD_seqStore_resolveOffCodes(repcodes_t * const dRepcodes,repcodes_t * const cRepcodes,seqStore_t * const seqStore,U32 const nbSeq)3465 static void ZSTD_seqStore_resolveOffCodes(repcodes_t* const dRepcodes, repcodes_t* const cRepcodes,
3466 seqStore_t* const seqStore, U32 const nbSeq) {
3467 U32 idx = 0;
3468 for (; idx < nbSeq; ++idx) {
3469 seqDef* const seq = seqStore->sequencesStart + idx;
3470 U32 const ll0 = (seq->litLength == 0);
3471 U32 const offCode = OFFBASE_TO_STORED(seq->offBase);
3472 assert(seq->offBase > 0);
3473 if (STORED_IS_REPCODE(offCode)) {
3474 U32 const dRawOffset = ZSTD_resolveRepcodeToRawOffset(dRepcodes->rep, offCode, ll0);
3475 U32 const cRawOffset = ZSTD_resolveRepcodeToRawOffset(cRepcodes->rep, offCode, ll0);
3476 /* Adjust simulated decompression repcode history if we come across a mismatch. Replace
3477 * the repcode with the offset it actually references, determined by the compression
3478 * repcode history.
3479 */
3480 if (dRawOffset != cRawOffset) {
3481 seq->offBase = cRawOffset + ZSTD_REP_NUM;
3482 }
3483 }
3484 /* Compression repcode history is always updated with values directly from the unmodified seqStore.
3485 * Decompression repcode history may use modified seq->offset value taken from compression repcode history.
3486 */
3487 ZSTD_updateRep(dRepcodes->rep, OFFBASE_TO_STORED(seq->offBase), ll0);
3488 ZSTD_updateRep(cRepcodes->rep, offCode, ll0);
3489 }
3490 }
3491
3492 /* ZSTD_compressSeqStore_singleBlock():
3493 * Compresses a seqStore into a block with a block header, into the buffer dst.
3494 *
3495 * Returns the total size of that block (including header) or a ZSTD error code.
3496 */
3497 static size_t
ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx * zc,seqStore_t * const seqStore,repcodes_t * const dRep,repcodes_t * const cRep,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 lastBlock,U32 isPartition)3498 ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx* zc, seqStore_t* const seqStore,
3499 repcodes_t* const dRep, repcodes_t* const cRep,
3500 void* dst, size_t dstCapacity,
3501 const void* src, size_t srcSize,
3502 U32 lastBlock, U32 isPartition)
3503 {
3504 const U32 rleMaxLength = 25;
3505 BYTE* op = (BYTE*)dst;
3506 const BYTE* ip = (const BYTE*)src;
3507 size_t cSize;
3508 size_t cSeqsSize;
3509
3510 /* In case of an RLE or raw block, the simulated decompression repcode history must be reset */
3511 repcodes_t const dRepOriginal = *dRep;
3512 DEBUGLOG(5, "ZSTD_compressSeqStore_singleBlock");
3513 if (isPartition)
3514 ZSTD_seqStore_resolveOffCodes(dRep, cRep, seqStore, (U32)(seqStore->sequences - seqStore->sequencesStart));
3515
3516 RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "Block header doesn't fit");
3517 cSeqsSize = ZSTD_entropyCompressSeqStore(seqStore,
3518 &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
3519 &zc->appliedParams,
3520 op + ZSTD_blockHeaderSize, dstCapacity - ZSTD_blockHeaderSize,
3521 srcSize,
3522 zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
3523 zc->bmi2);
3524 FORWARD_IF_ERROR(cSeqsSize, "ZSTD_entropyCompressSeqStore failed!");
3525
3526 if (!zc->isFirstBlock &&
3527 cSeqsSize < rleMaxLength &&
3528 ZSTD_isRLE((BYTE const*)src, srcSize)) {
3529 /* We don't want to emit our first block as a RLE even if it qualifies because
3530 * doing so will cause the decoder (cli only) to throw a "should consume all input error."
3531 * This is only an issue for zstd <= v1.4.3
3532 */
3533 cSeqsSize = 1;
3534 }
3535
3536 if (zc->seqCollector.collectSequences) {
3537 ZSTD_copyBlockSequences(zc);
3538 ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
3539 return 0;
3540 }
3541
3542 if (cSeqsSize == 0) {
3543 cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock);
3544 FORWARD_IF_ERROR(cSize, "Nocompress block failed");
3545 DEBUGLOG(4, "Writing out nocompress block, size: %zu", cSize);
3546 *dRep = dRepOriginal; /* reset simulated decompression repcode history */
3547 } else if (cSeqsSize == 1) {
3548 cSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, srcSize, lastBlock);
3549 FORWARD_IF_ERROR(cSize, "RLE compress block failed");
3550 DEBUGLOG(4, "Writing out RLE block, size: %zu", cSize);
3551 *dRep = dRepOriginal; /* reset simulated decompression repcode history */
3552 } else {
3553 ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
3554 writeBlockHeader(op, cSeqsSize, srcSize, lastBlock);
3555 cSize = ZSTD_blockHeaderSize + cSeqsSize;
3556 DEBUGLOG(4, "Writing out compressed block, size: %zu", cSize);
3557 }
3558
3559 if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
3560 zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
3561
3562 return cSize;
3563 }
3564
3565 /* Struct to keep track of where we are in our recursive calls. */
3566 typedef struct {
3567 U32* splitLocations; /* Array of split indices */
3568 size_t idx; /* The current index within splitLocations being worked on */
3569 } seqStoreSplits;
3570
3571 #define MIN_SEQUENCES_BLOCK_SPLITTING 300
3572
3573 /* Helper function to perform the recursive search for block splits.
3574 * Estimates the cost of seqStore prior to split, and estimates the cost of splitting the sequences in half.
3575 * If advantageous to split, then we recurse down the two sub-blocks. If not, or if an error occurred in estimation, then
3576 * we do not recurse.
3577 *
3578 * Note: The recursion depth is capped by a heuristic minimum number of sequences, defined by MIN_SEQUENCES_BLOCK_SPLITTING.
3579 * In theory, this means the absolute largest recursion depth is 10 == log2(maxNbSeqInBlock/MIN_SEQUENCES_BLOCK_SPLITTING).
3580 * In practice, recursion depth usually doesn't go beyond 4.
3581 *
3582 * Furthermore, the number of splits is capped by ZSTD_MAX_NB_BLOCK_SPLITS. At ZSTD_MAX_NB_BLOCK_SPLITS == 196 with the current existing blockSize
3583 * maximum of 128 KB, this value is actually impossible to reach.
3584 */
3585 static void
ZSTD_deriveBlockSplitsHelper(seqStoreSplits * splits,size_t startIdx,size_t endIdx,ZSTD_CCtx * zc,const seqStore_t * origSeqStore)3586 ZSTD_deriveBlockSplitsHelper(seqStoreSplits* splits, size_t startIdx, size_t endIdx,
3587 ZSTD_CCtx* zc, const seqStore_t* origSeqStore)
3588 {
3589 seqStore_t* fullSeqStoreChunk = &zc->blockSplitCtx.fullSeqStoreChunk;
3590 seqStore_t* firstHalfSeqStore = &zc->blockSplitCtx.firstHalfSeqStore;
3591 seqStore_t* secondHalfSeqStore = &zc->blockSplitCtx.secondHalfSeqStore;
3592 size_t estimatedOriginalSize;
3593 size_t estimatedFirstHalfSize;
3594 size_t estimatedSecondHalfSize;
3595 size_t midIdx = (startIdx + endIdx)/2;
3596
3597 if (endIdx - startIdx < MIN_SEQUENCES_BLOCK_SPLITTING || splits->idx >= ZSTD_MAX_NB_BLOCK_SPLITS) {
3598 DEBUGLOG(6, "ZSTD_deriveBlockSplitsHelper: Too few sequences");
3599 return;
3600 }
3601 DEBUGLOG(4, "ZSTD_deriveBlockSplitsHelper: startIdx=%zu endIdx=%zu", startIdx, endIdx);
3602 ZSTD_deriveSeqStoreChunk(fullSeqStoreChunk, origSeqStore, startIdx, endIdx);
3603 ZSTD_deriveSeqStoreChunk(firstHalfSeqStore, origSeqStore, startIdx, midIdx);
3604 ZSTD_deriveSeqStoreChunk(secondHalfSeqStore, origSeqStore, midIdx, endIdx);
3605 estimatedOriginalSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(fullSeqStoreChunk, zc);
3606 estimatedFirstHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(firstHalfSeqStore, zc);
3607 estimatedSecondHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(secondHalfSeqStore, zc);
3608 DEBUGLOG(4, "Estimated original block size: %zu -- First half split: %zu -- Second half split: %zu",
3609 estimatedOriginalSize, estimatedFirstHalfSize, estimatedSecondHalfSize);
3610 if (ZSTD_isError(estimatedOriginalSize) || ZSTD_isError(estimatedFirstHalfSize) || ZSTD_isError(estimatedSecondHalfSize)) {
3611 return;
3612 }
3613 if (estimatedFirstHalfSize + estimatedSecondHalfSize < estimatedOriginalSize) {
3614 ZSTD_deriveBlockSplitsHelper(splits, startIdx, midIdx, zc, origSeqStore);
3615 splits->splitLocations[splits->idx] = (U32)midIdx;
3616 splits->idx++;
3617 ZSTD_deriveBlockSplitsHelper(splits, midIdx, endIdx, zc, origSeqStore);
3618 }
3619 }
3620
3621 /* Base recursive function. Populates a table with intra-block partition indices that can improve compression ratio.
3622 *
3623 * Returns the number of splits made (which equals the size of the partition table - 1).
3624 */
ZSTD_deriveBlockSplits(ZSTD_CCtx * zc,U32 partitions[],U32 nbSeq)3625 static size_t ZSTD_deriveBlockSplits(ZSTD_CCtx* zc, U32 partitions[], U32 nbSeq) {
3626 seqStoreSplits splits = {partitions, 0};
3627 if (nbSeq <= 4) {
3628 DEBUGLOG(4, "ZSTD_deriveBlockSplits: Too few sequences to split");
3629 /* Refuse to try and split anything with less than 4 sequences */
3630 return 0;
3631 }
3632 ZSTD_deriveBlockSplitsHelper(&splits, 0, nbSeq, zc, &zc->seqStore);
3633 splits.splitLocations[splits.idx] = nbSeq;
3634 DEBUGLOG(5, "ZSTD_deriveBlockSplits: final nb partitions: %zu", splits.idx+1);
3635 return splits.idx;
3636 }
3637
3638 /* ZSTD_compressBlock_splitBlock():
3639 * Attempts to split a given block into multiple blocks to improve compression ratio.
3640 *
3641 * Returns combined size of all blocks (which includes headers), or a ZSTD error code.
3642 */
3643 static size_t
ZSTD_compressBlock_splitBlock_internal(ZSTD_CCtx * zc,void * dst,size_t dstCapacity,const void * src,size_t blockSize,U32 lastBlock,U32 nbSeq)3644 ZSTD_compressBlock_splitBlock_internal(ZSTD_CCtx* zc, void* dst, size_t dstCapacity,
3645 const void* src, size_t blockSize, U32 lastBlock, U32 nbSeq)
3646 {
3647 size_t cSize = 0;
3648 const BYTE* ip = (const BYTE*)src;
3649 BYTE* op = (BYTE*)dst;
3650 size_t i = 0;
3651 size_t srcBytesTotal = 0;
3652 U32* partitions = zc->blockSplitCtx.partitions; /* size == ZSTD_MAX_NB_BLOCK_SPLITS */
3653 seqStore_t* nextSeqStore = &zc->blockSplitCtx.nextSeqStore;
3654 seqStore_t* currSeqStore = &zc->blockSplitCtx.currSeqStore;
3655 size_t numSplits = ZSTD_deriveBlockSplits(zc, partitions, nbSeq);
3656
3657 /* If a block is split and some partitions are emitted as RLE/uncompressed, then repcode history
3658 * may become invalid. In order to reconcile potentially invalid repcodes, we keep track of two
3659 * separate repcode histories that simulate repcode history on compression and decompression side,
3660 * and use the histories to determine whether we must replace a particular repcode with its raw offset.
3661 *
3662 * 1) cRep gets updated for each partition, regardless of whether the block was emitted as uncompressed
3663 * or RLE. This allows us to retrieve the offset value that an invalid repcode references within
3664 * a nocompress/RLE block.
3665 * 2) dRep gets updated only for compressed partitions, and when a repcode gets replaced, will use
3666 * the replacement offset value rather than the original repcode to update the repcode history.
3667 * dRep also will be the final repcode history sent to the next block.
3668 *
3669 * See ZSTD_seqStore_resolveOffCodes() for more details.
3670 */
3671 repcodes_t dRep;
3672 repcodes_t cRep;
3673 ZSTD_memcpy(dRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
3674 ZSTD_memcpy(cRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t));
3675 ZSTD_memset(nextSeqStore, 0, sizeof(seqStore_t));
3676
3677 DEBUGLOG(4, "ZSTD_compressBlock_splitBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
3678 (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
3679 (unsigned)zc->blockState.matchState.nextToUpdate);
3680
3681 if (numSplits == 0) {
3682 size_t cSizeSingleBlock = ZSTD_compressSeqStore_singleBlock(zc, &zc->seqStore,
3683 &dRep, &cRep,
3684 op, dstCapacity,
3685 ip, blockSize,
3686 lastBlock, 0 /* isPartition */);
3687 FORWARD_IF_ERROR(cSizeSingleBlock, "Compressing single block from splitBlock_internal() failed!");
3688 DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal: No splits");
3689 assert(cSizeSingleBlock <= ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize);
3690 return cSizeSingleBlock;
3691 }
3692
3693 ZSTD_deriveSeqStoreChunk(currSeqStore, &zc->seqStore, 0, partitions[0]);
3694 for (i = 0; i <= numSplits; ++i) {
3695 size_t srcBytes;
3696 size_t cSizeChunk;
3697 U32 const lastPartition = (i == numSplits);
3698 U32 lastBlockEntireSrc = 0;
3699
3700 srcBytes = ZSTD_countSeqStoreLiteralsBytes(currSeqStore) + ZSTD_countSeqStoreMatchBytes(currSeqStore);
3701 srcBytesTotal += srcBytes;
3702 if (lastPartition) {
3703 /* This is the final partition, need to account for possible last literals */
3704 srcBytes += blockSize - srcBytesTotal;
3705 lastBlockEntireSrc = lastBlock;
3706 } else {
3707 ZSTD_deriveSeqStoreChunk(nextSeqStore, &zc->seqStore, partitions[i], partitions[i+1]);
3708 }
3709
3710 cSizeChunk = ZSTD_compressSeqStore_singleBlock(zc, currSeqStore,
3711 &dRep, &cRep,
3712 op, dstCapacity,
3713 ip, srcBytes,
3714 lastBlockEntireSrc, 1 /* isPartition */);
3715 DEBUGLOG(5, "Estimated size: %zu actual size: %zu", ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(currSeqStore, zc), cSizeChunk);
3716 FORWARD_IF_ERROR(cSizeChunk, "Compressing chunk failed!");
3717
3718 ip += srcBytes;
3719 op += cSizeChunk;
3720 dstCapacity -= cSizeChunk;
3721 cSize += cSizeChunk;
3722 *currSeqStore = *nextSeqStore;
3723 assert(cSizeChunk <= ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize);
3724 }
3725 /* cRep and dRep may have diverged during the compression. If so, we use the dRep repcodes
3726 * for the next block.
3727 */
3728 ZSTD_memcpy(zc->blockState.prevCBlock->rep, dRep.rep, sizeof(repcodes_t));
3729 return cSize;
3730 }
3731
3732 static size_t
ZSTD_compressBlock_splitBlock(ZSTD_CCtx * zc,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 lastBlock)3733 ZSTD_compressBlock_splitBlock(ZSTD_CCtx* zc,
3734 void* dst, size_t dstCapacity,
3735 const void* src, size_t srcSize, U32 lastBlock)
3736 {
3737 const BYTE* ip = (const BYTE*)src;
3738 BYTE* op = (BYTE*)dst;
3739 U32 nbSeq;
3740 size_t cSize;
3741 DEBUGLOG(4, "ZSTD_compressBlock_splitBlock");
3742 assert(zc->appliedParams.useBlockSplitter == ZSTD_ps_enable);
3743
3744 { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
3745 FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
3746 if (bss == ZSTDbss_noCompress) {
3747 if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
3748 zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
3749 cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock);
3750 FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
3751 DEBUGLOG(4, "ZSTD_compressBlock_splitBlock: Nocompress block");
3752 return cSize;
3753 }
3754 nbSeq = (U32)(zc->seqStore.sequences - zc->seqStore.sequencesStart);
3755 }
3756
3757 cSize = ZSTD_compressBlock_splitBlock_internal(zc, dst, dstCapacity, src, srcSize, lastBlock, nbSeq);
3758 FORWARD_IF_ERROR(cSize, "Splitting blocks failed!");
3759 return cSize;
3760 }
3761
3762 static size_t
ZSTD_compressBlock_internal(ZSTD_CCtx * zc,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 frame)3763 ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
3764 void* dst, size_t dstCapacity,
3765 const void* src, size_t srcSize, U32 frame)
3766 {
3767 /* This the upper bound for the length of an rle block.
3768 * This isn't the actual upper bound. Finding the real threshold
3769 * needs further investigation.
3770 */
3771 const U32 rleMaxLength = 25;
3772 size_t cSize;
3773 const BYTE* ip = (const BYTE*)src;
3774 BYTE* op = (BYTE*)dst;
3775 DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
3776 (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
3777 (unsigned)zc->blockState.matchState.nextToUpdate);
3778
3779 { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
3780 FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
3781 if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }
3782 }
3783
3784 if (zc->seqCollector.collectSequences) {
3785 ZSTD_copyBlockSequences(zc);
3786 ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
3787 return 0;
3788 }
3789
3790 /* encode sequences and literals */
3791 cSize = ZSTD_entropyCompressSeqStore(&zc->seqStore,
3792 &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
3793 &zc->appliedParams,
3794 dst, dstCapacity,
3795 srcSize,
3796 zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
3797 zc->bmi2);
3798
3799 if (frame &&
3800 /* We don't want to emit our first block as a RLE even if it qualifies because
3801 * doing so will cause the decoder (cli only) to throw a "should consume all input error."
3802 * This is only an issue for zstd <= v1.4.3
3803 */
3804 !zc->isFirstBlock &&
3805 cSize < rleMaxLength &&
3806 ZSTD_isRLE(ip, srcSize))
3807 {
3808 cSize = 1;
3809 op[0] = ip[0];
3810 }
3811
3812 out:
3813 if (!ZSTD_isError(cSize) && cSize > 1) {
3814 ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
3815 }
3816 /* We check that dictionaries have offset codes available for the first
3817 * block. After the first block, the offcode table might not have large
3818 * enough codes to represent the offsets in the data.
3819 */
3820 if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
3821 zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
3822
3823 return cSize;
3824 }
3825
ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx * zc,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const size_t bss,U32 lastBlock)3826 static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc,
3827 void* dst, size_t dstCapacity,
3828 const void* src, size_t srcSize,
3829 const size_t bss, U32 lastBlock)
3830 {
3831 DEBUGLOG(6, "Attempting ZSTD_compressSuperBlock()");
3832 if (bss == ZSTDbss_compress) {
3833 if (/* We don't want to emit our first block as a RLE even if it qualifies because
3834 * doing so will cause the decoder (cli only) to throw a "should consume all input error."
3835 * This is only an issue for zstd <= v1.4.3
3836 */
3837 !zc->isFirstBlock &&
3838 ZSTD_maybeRLE(&zc->seqStore) &&
3839 ZSTD_isRLE((BYTE const*)src, srcSize))
3840 {
3841 return ZSTD_rleCompressBlock(dst, dstCapacity, *(BYTE const*)src, srcSize, lastBlock);
3842 }
3843 /* Attempt superblock compression.
3844 *
3845 * Note that compressed size of ZSTD_compressSuperBlock() is not bound by the
3846 * standard ZSTD_compressBound(). This is a problem, because even if we have
3847 * space now, taking an extra byte now could cause us to run out of space later
3848 * and violate ZSTD_compressBound().
3849 *
3850 * Define blockBound(blockSize) = blockSize + ZSTD_blockHeaderSize.
3851 *
3852 * In order to respect ZSTD_compressBound() we must attempt to emit a raw
3853 * uncompressed block in these cases:
3854 * * cSize == 0: Return code for an uncompressed block.
3855 * * cSize == dstSize_tooSmall: We may have expanded beyond blockBound(srcSize).
3856 * ZSTD_noCompressBlock() will return dstSize_tooSmall if we are really out of
3857 * output space.
3858 * * cSize >= blockBound(srcSize): We have expanded the block too much so
3859 * emit an uncompressed block.
3860 */
3861 {
3862 size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock);
3863 if (cSize != ERROR(dstSize_tooSmall)) {
3864 size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy);
3865 FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed");
3866 if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) {
3867 ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState);
3868 return cSize;
3869 }
3870 }
3871 }
3872 }
3873
3874 DEBUGLOG(6, "Resorting to ZSTD_noCompressBlock()");
3875 /* Superblock compression failed, attempt to emit a single no compress block.
3876 * The decoder will be able to stream this block since it is uncompressed.
3877 */
3878 return ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock);
3879 }
3880
ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx * zc,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 lastBlock)3881 static size_t ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx* zc,
3882 void* dst, size_t dstCapacity,
3883 const void* src, size_t srcSize,
3884 U32 lastBlock)
3885 {
3886 size_t cSize = 0;
3887 const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);
3888 DEBUGLOG(5, "ZSTD_compressBlock_targetCBlockSize (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u, srcSize=%zu)",
3889 (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate, srcSize);
3890 FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");
3891
3892 cSize = ZSTD_compressBlock_targetCBlockSize_body(zc, dst, dstCapacity, src, srcSize, bss, lastBlock);
3893 FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize_body failed");
3894
3895 if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
3896 zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
3897
3898 return cSize;
3899 }
3900
ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t * ms,ZSTD_cwksp * ws,ZSTD_CCtx_params const * params,void const * ip,void const * iend)3901 static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms,
3902 ZSTD_cwksp* ws,
3903 ZSTD_CCtx_params const* params,
3904 void const* ip,
3905 void const* iend)
3906 {
3907 U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy);
3908 U32 const maxDist = (U32)1 << params->cParams.windowLog;
3909 if (ZSTD_window_needOverflowCorrection(ms->window, cycleLog, maxDist, ms->loadedDictEnd, ip, iend)) {
3910 U32 const correction = ZSTD_window_correctOverflow(&ms->window, cycleLog, maxDist, ip);
3911 ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);
3912 ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);
3913 ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);
3914 ZSTD_cwksp_mark_tables_dirty(ws);
3915 ZSTD_reduceIndex(ms, params, correction);
3916 ZSTD_cwksp_mark_tables_clean(ws);
3917 if (ms->nextToUpdate < correction) ms->nextToUpdate = 0;
3918 else ms->nextToUpdate -= correction;
3919 /* invalidate dictionaries on overflow correction */
3920 ms->loadedDictEnd = 0;
3921 ms->dictMatchState = NULL;
3922 }
3923 }
3924
3925 /*! ZSTD_compress_frameChunk() :
3926 * Compress a chunk of data into one or multiple blocks.
3927 * All blocks will be terminated, all input will be consumed.
3928 * Function will issue an error if there is not enough `dstCapacity` to hold the compressed content.
3929 * Frame is supposed already started (header already produced)
3930 * @return : compressed size, or an error code
3931 */
ZSTD_compress_frameChunk(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 lastFrameChunk)3932 static size_t ZSTD_compress_frameChunk(ZSTD_CCtx* cctx,
3933 void* dst, size_t dstCapacity,
3934 const void* src, size_t srcSize,
3935 U32 lastFrameChunk)
3936 {
3937 size_t blockSize = cctx->blockSize;
3938 size_t remaining = srcSize;
3939 const BYTE* ip = (const BYTE*)src;
3940 BYTE* const ostart = (BYTE*)dst;
3941 BYTE* op = ostart;
3942 U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog;
3943
3944 assert(cctx->appliedParams.cParams.windowLog <= ZSTD_WINDOWLOG_MAX);
3945
3946 DEBUGLOG(4, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize);
3947 if (cctx->appliedParams.fParams.checksumFlag && srcSize)
3948 XXH64_update(&cctx->xxhState, src, srcSize);
3949
3950 while (remaining) {
3951 ZSTD_matchState_t* const ms = &cctx->blockState.matchState;
3952 U32 const lastBlock = lastFrameChunk & (blockSize >= remaining);
3953
3954 RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE,
3955 dstSize_tooSmall,
3956 "not enough space to store compressed block");
3957 if (remaining < blockSize) blockSize = remaining;
3958
3959 ZSTD_overflowCorrectIfNeeded(
3960 ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize);
3961 ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);
3962 ZSTD_window_enforceMaxDist(&ms->window, ip, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);
3963
3964 /* Ensure hash/chain table insertion resumes no sooner than lowlimit */
3965 if (ms->nextToUpdate < ms->window.lowLimit) ms->nextToUpdate = ms->window.lowLimit;
3966
3967 { size_t cSize;
3968 if (ZSTD_useTargetCBlockSize(&cctx->appliedParams)) {
3969 cSize = ZSTD_compressBlock_targetCBlockSize(cctx, op, dstCapacity, ip, blockSize, lastBlock);
3970 FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed");
3971 assert(cSize > 0);
3972 assert(cSize <= blockSize + ZSTD_blockHeaderSize);
3973 } else if (ZSTD_blockSplitterEnabled(&cctx->appliedParams)) {
3974 cSize = ZSTD_compressBlock_splitBlock(cctx, op, dstCapacity, ip, blockSize, lastBlock);
3975 FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_splitBlock failed");
3976 assert(cSize > 0 || cctx->seqCollector.collectSequences == 1);
3977 } else {
3978 cSize = ZSTD_compressBlock_internal(cctx,
3979 op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,
3980 ip, blockSize, 1 /* frame */);
3981 FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_internal failed");
3982
3983 if (cSize == 0) { /* block is not compressible */
3984 cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
3985 FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
3986 } else {
3987 U32 const cBlockHeader = cSize == 1 ?
3988 lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
3989 lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
3990 MEM_writeLE24(op, cBlockHeader);
3991 cSize += ZSTD_blockHeaderSize;
3992 }
3993 }
3994
3995
3996 ip += blockSize;
3997 assert(remaining >= blockSize);
3998 remaining -= blockSize;
3999 op += cSize;
4000 assert(dstCapacity >= cSize);
4001 dstCapacity -= cSize;
4002 cctx->isFirstBlock = 0;
4003 DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u",
4004 (unsigned)cSize);
4005 } }
4006
4007 if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending;
4008 return (size_t)(op-ostart);
4009 }
4010
4011
ZSTD_writeFrameHeader(void * dst,size_t dstCapacity,const ZSTD_CCtx_params * params,U64 pledgedSrcSize,U32 dictID)4012 static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
4013 const ZSTD_CCtx_params* params, U64 pledgedSrcSize, U32 dictID)
4014 { BYTE* const op = (BYTE*)dst;
4015 U32 const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536); /* 0-3 */
4016 U32 const dictIDSizeCode = params->fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength; /* 0-3 */
4017 U32 const checksumFlag = params->fParams.checksumFlag>0;
4018 U32 const windowSize = (U32)1 << params->cParams.windowLog;
4019 U32 const singleSegment = params->fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
4020 BYTE const windowLogByte = (BYTE)((params->cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
4021 U32 const fcsCode = params->fParams.contentSizeFlag ?
4022 (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0; /* 0-3 */
4023 BYTE const frameHeaderDescriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );
4024 size_t pos=0;
4025
4026 assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
4027 RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall,
4028 "dst buf is too small to fit worst-case frame header size.");
4029 DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
4030 !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
4031 if (params->format == ZSTD_f_zstd1) {
4032 MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
4033 pos = 4;
4034 }
4035 op[pos++] = frameHeaderDescriptionByte;
4036 if (!singleSegment) op[pos++] = windowLogByte;
4037 switch(dictIDSizeCode)
4038 {
4039 default:
4040 assert(0); /* impossible */
4041 ZSTD_FALLTHROUGH;
4042 case 0 : break;
4043 case 1 : op[pos] = (BYTE)(dictID); pos++; break;
4044 case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break;
4045 case 3 : MEM_writeLE32(op+pos, dictID); pos+=4; break;
4046 }
4047 switch(fcsCode)
4048 {
4049 default:
4050 assert(0); /* impossible */
4051 ZSTD_FALLTHROUGH;
4052 case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break;
4053 case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break;
4054 case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break;
4055 case 3 : MEM_writeLE64(op+pos, (U64)(pledgedSrcSize)); pos+=8; break;
4056 }
4057 return pos;
4058 }
4059
4060 /* ZSTD_writeSkippableFrame_advanced() :
4061 * Writes out a skippable frame with the specified magic number variant (16 are supported),
4062 * from ZSTD_MAGIC_SKIPPABLE_START to ZSTD_MAGIC_SKIPPABLE_START+15, and the desired source data.
4063 *
4064 * Returns the total number of bytes written, or a ZSTD error code.
4065 */
ZSTD_writeSkippableFrame(void * dst,size_t dstCapacity,const void * src,size_t srcSize,unsigned magicVariant)4066 size_t ZSTD_writeSkippableFrame(void* dst, size_t dstCapacity,
4067 const void* src, size_t srcSize, unsigned magicVariant) {
4068 BYTE* op = (BYTE*)dst;
4069 RETURN_ERROR_IF(dstCapacity < srcSize + ZSTD_SKIPPABLEHEADERSIZE /* Skippable frame overhead */,
4070 dstSize_tooSmall, "Not enough room for skippable frame");
4071 RETURN_ERROR_IF(srcSize > (unsigned)0xFFFFFFFF, srcSize_wrong, "Src size too large for skippable frame");
4072 RETURN_ERROR_IF(magicVariant > 15, parameter_outOfBound, "Skippable frame magic number variant not supported");
4073
4074 MEM_writeLE32(op, (U32)(ZSTD_MAGIC_SKIPPABLE_START + magicVariant));
4075 MEM_writeLE32(op+4, (U32)srcSize);
4076 ZSTD_memcpy(op+8, src, srcSize);
4077 return srcSize + ZSTD_SKIPPABLEHEADERSIZE;
4078 }
4079
4080 /* ZSTD_writeLastEmptyBlock() :
4081 * output an empty Block with end-of-frame mark to complete a frame
4082 * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))
4083 * or an error code if `dstCapacity` is too small (<ZSTD_blockHeaderSize)
4084 */
ZSTD_writeLastEmptyBlock(void * dst,size_t dstCapacity)4085 size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity)
4086 {
4087 RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall,
4088 "dst buf is too small to write frame trailer empty block.");
4089 { U32 const cBlockHeader24 = 1 /*lastBlock*/ + (((U32)bt_raw)<<1); /* 0 size */
4090 MEM_writeLE24(dst, cBlockHeader24);
4091 return ZSTD_blockHeaderSize;
4092 }
4093 }
4094
ZSTD_referenceExternalSequences(ZSTD_CCtx * cctx,rawSeq * seq,size_t nbSeq)4095 size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq)
4096 {
4097 RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong,
4098 "wrong cctx stage");
4099 RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable,
4100 parameter_unsupported,
4101 "incompatible with ldm");
4102 cctx->externSeqStore.seq = seq;
4103 cctx->externSeqStore.size = nbSeq;
4104 cctx->externSeqStore.capacity = nbSeq;
4105 cctx->externSeqStore.pos = 0;
4106 cctx->externSeqStore.posInSequence = 0;
4107 return 0;
4108 }
4109
4110
ZSTD_compressContinue_internal(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,U32 frame,U32 lastFrameChunk)4111 static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
4112 void* dst, size_t dstCapacity,
4113 const void* src, size_t srcSize,
4114 U32 frame, U32 lastFrameChunk)
4115 {
4116 ZSTD_matchState_t* const ms = &cctx->blockState.matchState;
4117 size_t fhSize = 0;
4118
4119 DEBUGLOG(5, "ZSTD_compressContinue_internal, stage: %u, srcSize: %u",
4120 cctx->stage, (unsigned)srcSize);
4121 RETURN_ERROR_IF(cctx->stage==ZSTDcs_created, stage_wrong,
4122 "missing init (ZSTD_compressBegin)");
4123
4124 if (frame && (cctx->stage==ZSTDcs_init)) {
4125 fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,
4126 cctx->pledgedSrcSizePlusOne-1, cctx->dictID);
4127 FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
4128 assert(fhSize <= dstCapacity);
4129 dstCapacity -= fhSize;
4130 dst = (char*)dst + fhSize;
4131 cctx->stage = ZSTDcs_ongoing;
4132 }
4133
4134 if (!srcSize) return fhSize; /* do not generate an empty block if no input */
4135
4136 if (!ZSTD_window_update(&ms->window, src, srcSize, ms->forceNonContiguous)) {
4137 ms->forceNonContiguous = 0;
4138 ms->nextToUpdate = ms->window.dictLimit;
4139 }
4140 if (cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) {
4141 ZSTD_window_update(&cctx->ldmState.window, src, srcSize, /* forceNonContiguous */ 0);
4142 }
4143
4144 if (!frame) {
4145 /* overflow check and correction for block mode */
4146 ZSTD_overflowCorrectIfNeeded(
4147 ms, &cctx->workspace, &cctx->appliedParams,
4148 src, (BYTE const*)src + srcSize);
4149 }
4150
4151 DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (unsigned)cctx->blockSize);
4152 { size_t const cSize = frame ?
4153 ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
4154 ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */);
4155 FORWARD_IF_ERROR(cSize, "%s", frame ? "ZSTD_compress_frameChunk failed" : "ZSTD_compressBlock_internal failed");
4156 cctx->consumedSrcSize += srcSize;
4157 cctx->producedCSize += (cSize + fhSize);
4158 assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
4159 if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */
4160 ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
4161 RETURN_ERROR_IF(
4162 cctx->consumedSrcSize+1 > cctx->pledgedSrcSizePlusOne,
4163 srcSize_wrong,
4164 "error : pledgedSrcSize = %u, while realSrcSize >= %u",
4165 (unsigned)cctx->pledgedSrcSizePlusOne-1,
4166 (unsigned)cctx->consumedSrcSize);
4167 }
4168 return cSize + fhSize;
4169 }
4170 }
4171
ZSTD_compressContinue(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)4172 size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
4173 void* dst, size_t dstCapacity,
4174 const void* src, size_t srcSize)
4175 {
4176 DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)", (unsigned)srcSize);
4177 return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1 /* frame mode */, 0 /* last chunk */);
4178 }
4179
4180
ZSTD_getBlockSize(const ZSTD_CCtx * cctx)4181 size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)
4182 {
4183 ZSTD_compressionParameters const cParams = cctx->appliedParams.cParams;
4184 assert(!ZSTD_checkCParams(cParams));
4185 return MIN (ZSTD_BLOCKSIZE_MAX, (U32)1 << cParams.windowLog);
4186 }
4187
ZSTD_compressBlock(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)4188 size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
4189 {
4190 DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u", (unsigned)srcSize);
4191 { size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
4192 RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong, "input is larger than a block"); }
4193
4194 return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);
4195 }
4196
4197 /*! ZSTD_loadDictionaryContent() :
4198 * @return : 0, or an error code
4199 */
ZSTD_loadDictionaryContent(ZSTD_matchState_t * ms,ldmState_t * ls,ZSTD_cwksp * ws,ZSTD_CCtx_params const * params,const void * src,size_t srcSize,ZSTD_dictTableLoadMethod_e dtlm)4200 static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
4201 ldmState_t* ls,
4202 ZSTD_cwksp* ws,
4203 ZSTD_CCtx_params const* params,
4204 const void* src, size_t srcSize,
4205 ZSTD_dictTableLoadMethod_e dtlm)
4206 {
4207 const BYTE* ip = (const BYTE*) src;
4208 const BYTE* const iend = ip + srcSize;
4209 int const loadLdmDict = params->ldmParams.enableLdm == ZSTD_ps_enable && ls != NULL;
4210
4211 /* Assert that we the ms params match the params we're being given */
4212 ZSTD_assertEqualCParams(params->cParams, ms->cParams);
4213
4214 if (srcSize > ZSTD_CHUNKSIZE_MAX) {
4215 /* Allow the dictionary to set indices up to exactly ZSTD_CURRENT_MAX.
4216 * Dictionaries right at the edge will immediately trigger overflow
4217 * correction, but I don't want to insert extra constraints here.
4218 */
4219 U32 const maxDictSize = ZSTD_CURRENT_MAX - 1;
4220 /* We must have cleared our windows when our source is this large. */
4221 assert(ZSTD_window_isEmpty(ms->window));
4222 if (loadLdmDict)
4223 assert(ZSTD_window_isEmpty(ls->window));
4224 /* If the dictionary is too large, only load the suffix of the dictionary. */
4225 if (srcSize > maxDictSize) {
4226 ip = iend - maxDictSize;
4227 src = ip;
4228 srcSize = maxDictSize;
4229 }
4230 }
4231
4232 DEBUGLOG(4, "ZSTD_loadDictionaryContent(): useRowMatchFinder=%d", (int)params->useRowMatchFinder);
4233 ZSTD_window_update(&ms->window, src, srcSize, /* forceNonContiguous */ 0);
4234 ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base);
4235 ms->forceNonContiguous = params->deterministicRefPrefix;
4236
4237 if (loadLdmDict) {
4238 ZSTD_window_update(&ls->window, src, srcSize, /* forceNonContiguous */ 0);
4239 ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base);
4240 }
4241
4242 if (srcSize <= HASH_READ_SIZE) return 0;
4243
4244 ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, iend);
4245
4246 if (loadLdmDict)
4247 ZSTD_ldm_fillHashTable(ls, ip, iend, ¶ms->ldmParams);
4248
4249 switch(params->cParams.strategy)
4250 {
4251 case ZSTD_fast:
4252 ZSTD_fillHashTable(ms, iend, dtlm);
4253 break;
4254 case ZSTD_dfast:
4255 ZSTD_fillDoubleHashTable(ms, iend, dtlm);
4256 break;
4257
4258 case ZSTD_greedy:
4259 case ZSTD_lazy:
4260 case ZSTD_lazy2:
4261 assert(srcSize >= HASH_READ_SIZE);
4262 if (ms->dedicatedDictSearch) {
4263 assert(ms->chainTable != NULL);
4264 ZSTD_dedicatedDictSearch_lazy_loadDictionary(ms, iend-HASH_READ_SIZE);
4265 } else {
4266 assert(params->useRowMatchFinder != ZSTD_ps_auto);
4267 if (params->useRowMatchFinder == ZSTD_ps_enable) {
4268 size_t const tagTableSize = ((size_t)1 << params->cParams.hashLog) * sizeof(U16);
4269 ZSTD_memset(ms->tagTable, 0, tagTableSize);
4270 ZSTD_row_update(ms, iend-HASH_READ_SIZE);
4271 DEBUGLOG(4, "Using row-based hash table for lazy dict");
4272 } else {
4273 ZSTD_insertAndFindFirstIndex(ms, iend-HASH_READ_SIZE);
4274 DEBUGLOG(4, "Using chain-based hash table for lazy dict");
4275 }
4276 }
4277 break;
4278
4279 case ZSTD_btlazy2: /* we want the dictionary table fully sorted */
4280 case ZSTD_btopt:
4281 case ZSTD_btultra:
4282 case ZSTD_btultra2:
4283 assert(srcSize >= HASH_READ_SIZE);
4284 ZSTD_updateTree(ms, iend-HASH_READ_SIZE, iend);
4285 break;
4286
4287 default:
4288 assert(0); /* not possible : not a valid strategy id */
4289 }
4290
4291 ms->nextToUpdate = (U32)(iend - ms->window.base);
4292 return 0;
4293 }
4294
4295
4296 /* Dictionaries that assign zero probability to symbols that show up causes problems
4297 * when FSE encoding. Mark dictionaries with zero probability symbols as FSE_repeat_check
4298 * and only dictionaries with 100% valid symbols can be assumed valid.
4299 */
ZSTD_dictNCountRepeat(short * normalizedCounter,unsigned dictMaxSymbolValue,unsigned maxSymbolValue)4300 static FSE_repeat ZSTD_dictNCountRepeat(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue)
4301 {
4302 U32 s;
4303 if (dictMaxSymbolValue < maxSymbolValue) {
4304 return FSE_repeat_check;
4305 }
4306 for (s = 0; s <= maxSymbolValue; ++s) {
4307 if (normalizedCounter[s] == 0) {
4308 return FSE_repeat_check;
4309 }
4310 }
4311 return FSE_repeat_valid;
4312 }
4313
ZSTD_loadCEntropy(ZSTD_compressedBlockState_t * bs,void * workspace,const void * const dict,size_t dictSize)4314 size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
4315 const void* const dict, size_t dictSize)
4316 {
4317 short offcodeNCount[MaxOff+1];
4318 unsigned offcodeMaxValue = MaxOff;
4319 const BYTE* dictPtr = (const BYTE*)dict; /* skip magic num and dict ID */
4320 const BYTE* const dictEnd = dictPtr + dictSize;
4321 dictPtr += 8;
4322 bs->entropy.huf.repeatMode = HUF_repeat_check;
4323
4324 { unsigned maxSymbolValue = 255;
4325 unsigned hasZeroWeights = 1;
4326 size_t const hufHeaderSize = HUF_readCTable((HUF_CElt*)bs->entropy.huf.CTable, &maxSymbolValue, dictPtr,
4327 dictEnd-dictPtr, &hasZeroWeights);
4328
4329 /* We only set the loaded table as valid if it contains all non-zero
4330 * weights. Otherwise, we set it to check */
4331 if (!hasZeroWeights)
4332 bs->entropy.huf.repeatMode = HUF_repeat_valid;
4333
4334 RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted, "");
4335 RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted, "");
4336 dictPtr += hufHeaderSize;
4337 }
4338
4339 { unsigned offcodeLog;
4340 size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
4341 RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");
4342 RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");
4343 /* fill all offset symbols to avoid garbage at end of table */
4344 RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
4345 bs->entropy.fse.offcodeCTable,
4346 offcodeNCount, MaxOff, offcodeLog,
4347 workspace, HUF_WORKSPACE_SIZE)),
4348 dictionary_corrupted, "");
4349 /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
4350 dictPtr += offcodeHeaderSize;
4351 }
4352
4353 { short matchlengthNCount[MaxML+1];
4354 unsigned matchlengthMaxValue = MaxML, matchlengthLog;
4355 size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);
4356 RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");
4357 RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");
4358 RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
4359 bs->entropy.fse.matchlengthCTable,
4360 matchlengthNCount, matchlengthMaxValue, matchlengthLog,
4361 workspace, HUF_WORKSPACE_SIZE)),
4362 dictionary_corrupted, "");
4363 bs->entropy.fse.matchlength_repeatMode = ZSTD_dictNCountRepeat(matchlengthNCount, matchlengthMaxValue, MaxML);
4364 dictPtr += matchlengthHeaderSize;
4365 }
4366
4367 { short litlengthNCount[MaxLL+1];
4368 unsigned litlengthMaxValue = MaxLL, litlengthLog;
4369 size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);
4370 RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");
4371 RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");
4372 RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(
4373 bs->entropy.fse.litlengthCTable,
4374 litlengthNCount, litlengthMaxValue, litlengthLog,
4375 workspace, HUF_WORKSPACE_SIZE)),
4376 dictionary_corrupted, "");
4377 bs->entropy.fse.litlength_repeatMode = ZSTD_dictNCountRepeat(litlengthNCount, litlengthMaxValue, MaxLL);
4378 dictPtr += litlengthHeaderSize;
4379 }
4380
4381 RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");
4382 bs->rep[0] = MEM_readLE32(dictPtr+0);
4383 bs->rep[1] = MEM_readLE32(dictPtr+4);
4384 bs->rep[2] = MEM_readLE32(dictPtr+8);
4385 dictPtr += 12;
4386
4387 { size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
4388 U32 offcodeMax = MaxOff;
4389 if (dictContentSize <= ((U32)-1) - 128 KB) {
4390 U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */
4391 offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */
4392 }
4393 /* All offset values <= dictContentSize + 128 KB must be representable for a valid table */
4394 bs->entropy.fse.offcode_repeatMode = ZSTD_dictNCountRepeat(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff));
4395
4396 /* All repCodes must be <= dictContentSize and != 0 */
4397 { U32 u;
4398 for (u=0; u<3; u++) {
4399 RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, "");
4400 RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, "");
4401 } } }
4402
4403 return dictPtr - (const BYTE*)dict;
4404 }
4405
4406 /* Dictionary format :
4407 * See :
4408 * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#dictionary-format
4409 */
4410 /*! ZSTD_loadZstdDictionary() :
4411 * @return : dictID, or an error code
4412 * assumptions : magic number supposed already checked
4413 * dictSize supposed >= 8
4414 */
ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t * bs,ZSTD_matchState_t * ms,ZSTD_cwksp * ws,ZSTD_CCtx_params const * params,const void * dict,size_t dictSize,ZSTD_dictTableLoadMethod_e dtlm,void * workspace)4415 static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
4416 ZSTD_matchState_t* ms,
4417 ZSTD_cwksp* ws,
4418 ZSTD_CCtx_params const* params,
4419 const void* dict, size_t dictSize,
4420 ZSTD_dictTableLoadMethod_e dtlm,
4421 void* workspace)
4422 {
4423 const BYTE* dictPtr = (const BYTE*)dict;
4424 const BYTE* const dictEnd = dictPtr + dictSize;
4425 size_t dictID;
4426 size_t eSize;
4427 ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
4428 assert(dictSize >= 8);
4429 assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY);
4430
4431 dictID = params->fParams.noDictIDFlag ? 0 : MEM_readLE32(dictPtr + 4 /* skip magic number */ );
4432 eSize = ZSTD_loadCEntropy(bs, workspace, dict, dictSize);
4433 FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed");
4434 dictPtr += eSize;
4435
4436 {
4437 size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
4438 FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(
4439 ms, NULL, ws, params, dictPtr, dictContentSize, dtlm), "");
4440 }
4441 return dictID;
4442 }
4443
4444 /** ZSTD_compress_insertDictionary() :
4445 * @return : dictID, or an error code */
4446 static size_t
ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t * bs,ZSTD_matchState_t * ms,ldmState_t * ls,ZSTD_cwksp * ws,const ZSTD_CCtx_params * params,const void * dict,size_t dictSize,ZSTD_dictContentType_e dictContentType,ZSTD_dictTableLoadMethod_e dtlm,void * workspace)4447 ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
4448 ZSTD_matchState_t* ms,
4449 ldmState_t* ls,
4450 ZSTD_cwksp* ws,
4451 const ZSTD_CCtx_params* params,
4452 const void* dict, size_t dictSize,
4453 ZSTD_dictContentType_e dictContentType,
4454 ZSTD_dictTableLoadMethod_e dtlm,
4455 void* workspace)
4456 {
4457 DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize);
4458 if ((dict==NULL) || (dictSize<8)) {
4459 RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
4460 return 0;
4461 }
4462
4463 ZSTD_reset_compressedBlockState(bs);
4464
4465 /* dict restricted modes */
4466 if (dictContentType == ZSTD_dct_rawContent)
4467 return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm);
4468
4469 if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {
4470 if (dictContentType == ZSTD_dct_auto) {
4471 DEBUGLOG(4, "raw content dictionary detected");
4472 return ZSTD_loadDictionaryContent(
4473 ms, ls, ws, params, dict, dictSize, dtlm);
4474 }
4475 RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");
4476 assert(0); /* impossible */
4477 }
4478
4479 /* dict as full zstd dictionary */
4480 return ZSTD_loadZstdDictionary(
4481 bs, ms, ws, params, dict, dictSize, dtlm, workspace);
4482 }
4483
4484 #define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB)
4485 #define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6ULL)
4486
4487 /*! ZSTD_compressBegin_internal() :
4488 * @return : 0, or an error code */
ZSTD_compressBegin_internal(ZSTD_CCtx * cctx,const void * dict,size_t dictSize,ZSTD_dictContentType_e dictContentType,ZSTD_dictTableLoadMethod_e dtlm,const ZSTD_CDict * cdict,const ZSTD_CCtx_params * params,U64 pledgedSrcSize,ZSTD_buffered_policy_e zbuff)4489 static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
4490 const void* dict, size_t dictSize,
4491 ZSTD_dictContentType_e dictContentType,
4492 ZSTD_dictTableLoadMethod_e dtlm,
4493 const ZSTD_CDict* cdict,
4494 const ZSTD_CCtx_params* params, U64 pledgedSrcSize,
4495 ZSTD_buffered_policy_e zbuff)
4496 {
4497 size_t const dictContentSize = cdict ? cdict->dictContentSize : dictSize;
4498 #if ZSTD_TRACE
4499 cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0;
4500 #endif
4501 DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params->cParams.windowLog);
4502 /* params are supposed to be fully validated at this point */
4503 assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
4504 assert(!((dict) && (cdict))); /* either dict or cdict, not both */
4505 if ( (cdict)
4506 && (cdict->dictContentSize > 0)
4507 && ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
4508 || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
4509 || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
4510 || cdict->compressionLevel == 0)
4511 && (params->attachDictPref != ZSTD_dictForceLoad) ) {
4512 return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff);
4513 }
4514
4515 FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
4516 dictContentSize,
4517 ZSTDcrp_makeClean, zbuff) , "");
4518 { size_t const dictID = cdict ?
4519 ZSTD_compress_insertDictionary(
4520 cctx->blockState.prevCBlock, &cctx->blockState.matchState,
4521 &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent,
4522 cdict->dictContentSize, cdict->dictContentType, dtlm,
4523 cctx->entropyWorkspace)
4524 : ZSTD_compress_insertDictionary(
4525 cctx->blockState.prevCBlock, &cctx->blockState.matchState,
4526 &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize,
4527 dictContentType, dtlm, cctx->entropyWorkspace);
4528 FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
4529 assert(dictID <= UINT_MAX);
4530 cctx->dictID = (U32)dictID;
4531 cctx->dictContentSize = dictContentSize;
4532 }
4533 return 0;
4534 }
4535
ZSTD_compressBegin_advanced_internal(ZSTD_CCtx * cctx,const void * dict,size_t dictSize,ZSTD_dictContentType_e dictContentType,ZSTD_dictTableLoadMethod_e dtlm,const ZSTD_CDict * cdict,const ZSTD_CCtx_params * params,unsigned long long pledgedSrcSize)4536 size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
4537 const void* dict, size_t dictSize,
4538 ZSTD_dictContentType_e dictContentType,
4539 ZSTD_dictTableLoadMethod_e dtlm,
4540 const ZSTD_CDict* cdict,
4541 const ZSTD_CCtx_params* params,
4542 unsigned long long pledgedSrcSize)
4543 {
4544 DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params->cParams.windowLog);
4545 /* compression parameters verification and optimization */
4546 FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) , "");
4547 return ZSTD_compressBegin_internal(cctx,
4548 dict, dictSize, dictContentType, dtlm,
4549 cdict,
4550 params, pledgedSrcSize,
4551 ZSTDb_not_buffered);
4552 }
4553
4554 /*! ZSTD_compressBegin_advanced() :
4555 * @return : 0, or an error code */
ZSTD_compressBegin_advanced(ZSTD_CCtx * cctx,const void * dict,size_t dictSize,ZSTD_parameters params,unsigned long long pledgedSrcSize)4556 size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,
4557 const void* dict, size_t dictSize,
4558 ZSTD_parameters params, unsigned long long pledgedSrcSize)
4559 {
4560 ZSTD_CCtx_params cctxParams;
4561 ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, ZSTD_NO_CLEVEL);
4562 return ZSTD_compressBegin_advanced_internal(cctx,
4563 dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast,
4564 NULL /*cdict*/,
4565 &cctxParams, pledgedSrcSize);
4566 }
4567
ZSTD_compressBegin_usingDict(ZSTD_CCtx * cctx,const void * dict,size_t dictSize,int compressionLevel)4568 size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)
4569 {
4570 ZSTD_CCtx_params cctxParams;
4571 {
4572 ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_noAttachDict);
4573 ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel);
4574 }
4575 DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize);
4576 return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
4577 &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);
4578 }
4579
ZSTD_compressBegin(ZSTD_CCtx * cctx,int compressionLevel)4580 size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel)
4581 {
4582 return ZSTD_compressBegin_usingDict(cctx, NULL, 0, compressionLevel);
4583 }
4584
4585
4586 /*! ZSTD_writeEpilogue() :
4587 * Ends a frame.
4588 * @return : nb of bytes written into dst (or an error code) */
ZSTD_writeEpilogue(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity)4589 static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
4590 {
4591 BYTE* const ostart = (BYTE*)dst;
4592 BYTE* op = ostart;
4593 size_t fhSize = 0;
4594
4595 DEBUGLOG(4, "ZSTD_writeEpilogue");
4596 RETURN_ERROR_IF(cctx->stage == ZSTDcs_created, stage_wrong, "init missing");
4597
4598 /* special case : empty frame */
4599 if (cctx->stage == ZSTDcs_init) {
4600 fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0);
4601 FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");
4602 dstCapacity -= fhSize;
4603 op += fhSize;
4604 cctx->stage = ZSTDcs_ongoing;
4605 }
4606
4607 if (cctx->stage != ZSTDcs_ending) {
4608 /* write one last empty block, make it the "last" block */
4609 U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1) + 0;
4610 RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for epilogue");
4611 MEM_writeLE32(op, cBlockHeader24);
4612 op += ZSTD_blockHeaderSize;
4613 dstCapacity -= ZSTD_blockHeaderSize;
4614 }
4615
4616 if (cctx->appliedParams.fParams.checksumFlag) {
4617 U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
4618 RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
4619 DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", (unsigned)checksum);
4620 MEM_writeLE32(op, checksum);
4621 op += 4;
4622 }
4623
4624 cctx->stage = ZSTDcs_created; /* return to "created but no init" status */
4625 return op-ostart;
4626 }
4627
ZSTD_CCtx_trace(ZSTD_CCtx * cctx,size_t extraCSize)4628 void ZSTD_CCtx_trace(ZSTD_CCtx* cctx, size_t extraCSize)
4629 {
4630 #if ZSTD_TRACE
4631 if (cctx->traceCtx && ZSTD_trace_compress_end != NULL) {
4632 int const streaming = cctx->inBuffSize > 0 || cctx->outBuffSize > 0 || cctx->appliedParams.nbWorkers > 0;
4633 ZSTD_Trace trace;
4634 ZSTD_memset(&trace, 0, sizeof(trace));
4635 trace.version = ZSTD_VERSION_NUMBER;
4636 trace.streaming = streaming;
4637 trace.dictionaryID = cctx->dictID;
4638 trace.dictionarySize = cctx->dictContentSize;
4639 trace.uncompressedSize = cctx->consumedSrcSize;
4640 trace.compressedSize = cctx->producedCSize + extraCSize;
4641 trace.params = &cctx->appliedParams;
4642 trace.cctx = cctx;
4643 ZSTD_trace_compress_end(cctx->traceCtx, &trace);
4644 }
4645 cctx->traceCtx = 0;
4646 #else
4647 (void)cctx;
4648 (void)extraCSize;
4649 #endif
4650 }
4651
ZSTD_compressEnd(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)4652 size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
4653 void* dst, size_t dstCapacity,
4654 const void* src, size_t srcSize)
4655 {
4656 size_t endResult;
4657 size_t const cSize = ZSTD_compressContinue_internal(cctx,
4658 dst, dstCapacity, src, srcSize,
4659 1 /* frame mode */, 1 /* last chunk */);
4660 FORWARD_IF_ERROR(cSize, "ZSTD_compressContinue_internal failed");
4661 endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);
4662 FORWARD_IF_ERROR(endResult, "ZSTD_writeEpilogue failed");
4663 assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));
4664 if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */
4665 ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
4666 DEBUGLOG(4, "end of frame : controlling src size");
4667 RETURN_ERROR_IF(
4668 cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1,
4669 srcSize_wrong,
4670 "error : pledgedSrcSize = %u, while realSrcSize = %u",
4671 (unsigned)cctx->pledgedSrcSizePlusOne-1,
4672 (unsigned)cctx->consumedSrcSize);
4673 }
4674 ZSTD_CCtx_trace(cctx, endResult);
4675 return cSize + endResult;
4676 }
4677
ZSTD_compress_advanced(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const void * dict,size_t dictSize,ZSTD_parameters params)4678 size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
4679 void* dst, size_t dstCapacity,
4680 const void* src, size_t srcSize,
4681 const void* dict,size_t dictSize,
4682 ZSTD_parameters params)
4683 {
4684 DEBUGLOG(4, "ZSTD_compress_advanced");
4685 FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "");
4686 ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, ZSTD_NO_CLEVEL);
4687 return ZSTD_compress_advanced_internal(cctx,
4688 dst, dstCapacity,
4689 src, srcSize,
4690 dict, dictSize,
4691 &cctx->simpleApiParams);
4692 }
4693
4694 /* Internal */
ZSTD_compress_advanced_internal(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const void * dict,size_t dictSize,const ZSTD_CCtx_params * params)4695 size_t ZSTD_compress_advanced_internal(
4696 ZSTD_CCtx* cctx,
4697 void* dst, size_t dstCapacity,
4698 const void* src, size_t srcSize,
4699 const void* dict,size_t dictSize,
4700 const ZSTD_CCtx_params* params)
4701 {
4702 DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);
4703 FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
4704 dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
4705 params, srcSize, ZSTDb_not_buffered) , "");
4706 return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
4707 }
4708
ZSTD_compress_usingDict(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const void * dict,size_t dictSize,int compressionLevel)4709 size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx,
4710 void* dst, size_t dstCapacity,
4711 const void* src, size_t srcSize,
4712 const void* dict, size_t dictSize,
4713 int compressionLevel)
4714 {
4715 {
4716 ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0, ZSTD_cpm_noAttachDict);
4717 assert(params.fParams.contentSizeFlag == 1);
4718 ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT: compressionLevel);
4719 }
4720 DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize);
4721 return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctx->simpleApiParams);
4722 }
4723
ZSTD_compressCCtx(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,int compressionLevel)4724 size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
4725 void* dst, size_t dstCapacity,
4726 const void* src, size_t srcSize,
4727 int compressionLevel)
4728 {
4729 DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)", (unsigned)srcSize);
4730 assert(cctx != NULL);
4731 return ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel);
4732 }
4733
ZSTD_compress(void * dst,size_t dstCapacity,const void * src,size_t srcSize,int compressionLevel)4734 size_t ZSTD_compress(void* dst, size_t dstCapacity,
4735 const void* src, size_t srcSize,
4736 int compressionLevel)
4737 {
4738 size_t result;
4739 #if ZSTD_COMPRESS_HEAPMODE
4740 ZSTD_CCtx* cctx = ZSTD_createCCtx();
4741 RETURN_ERROR_IF(!cctx, memory_allocation, "ZSTD_createCCtx failed");
4742 result = ZSTD_compressCCtx(cctx, dst, dstCapacity, src, srcSize, compressionLevel);
4743 ZSTD_freeCCtx(cctx);
4744 #else
4745 ZSTD_CCtx ctxBody;
4746 ZSTD_initCCtx(&ctxBody, ZSTD_defaultCMem);
4747 result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel);
4748 ZSTD_freeCCtxContent(&ctxBody); /* can't free ctxBody itself, as it's on stack; free only heap content */
4749 #endif
4750 return result;
4751 }
4752
4753
4754 /* ===== Dictionary API ===== */
4755
4756 /*! ZSTD_estimateCDictSize_advanced() :
4757 * Estimate amount of memory that will be needed to create a dictionary with following arguments */
ZSTD_estimateCDictSize_advanced(size_t dictSize,ZSTD_compressionParameters cParams,ZSTD_dictLoadMethod_e dictLoadMethod)4758 size_t ZSTD_estimateCDictSize_advanced(
4759 size_t dictSize, ZSTD_compressionParameters cParams,
4760 ZSTD_dictLoadMethod_e dictLoadMethod)
4761 {
4762 DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));
4763 return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
4764 + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
4765 /* enableDedicatedDictSearch == 1 ensures that CDict estimation will not be too small
4766 * in case we are using DDS with row-hash. */
4767 + ZSTD_sizeof_matchState(&cParams, ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams),
4768 /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0)
4769 + (dictLoadMethod == ZSTD_dlm_byRef ? 0
4770 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *))));
4771 }
4772
ZSTD_estimateCDictSize(size_t dictSize,int compressionLevel)4773 size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
4774 {
4775 ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
4776 return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy);
4777 }
4778
ZSTD_sizeof_CDict(const ZSTD_CDict * cdict)4779 size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
4780 {
4781 if (cdict==NULL) return 0; /* support sizeof on NULL */
4782 DEBUGLOG(5, "sizeof(*cdict) : %u", (unsigned)sizeof(*cdict));
4783 /* cdict may be in the workspace */
4784 return (cdict->workspace.workspace == cdict ? 0 : sizeof(*cdict))
4785 + ZSTD_cwksp_sizeof(&cdict->workspace);
4786 }
4787
ZSTD_initCDict_internal(ZSTD_CDict * cdict,const void * dictBuffer,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType,ZSTD_CCtx_params params)4788 static size_t ZSTD_initCDict_internal(
4789 ZSTD_CDict* cdict,
4790 const void* dictBuffer, size_t dictSize,
4791 ZSTD_dictLoadMethod_e dictLoadMethod,
4792 ZSTD_dictContentType_e dictContentType,
4793 ZSTD_CCtx_params params)
4794 {
4795 DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType);
4796 assert(!ZSTD_checkCParams(params.cParams));
4797 cdict->matchState.cParams = params.cParams;
4798 cdict->matchState.dedicatedDictSearch = params.enableDedicatedDictSearch;
4799 if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {
4800 cdict->dictContent = dictBuffer;
4801 } else {
4802 void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*)));
4803 RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!");
4804 cdict->dictContent = internalBuffer;
4805 ZSTD_memcpy(internalBuffer, dictBuffer, dictSize);
4806 }
4807 cdict->dictContentSize = dictSize;
4808 cdict->dictContentType = dictContentType;
4809
4810 cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE);
4811
4812
4813 /* Reset the state to no dictionary */
4814 ZSTD_reset_compressedBlockState(&cdict->cBlockState);
4815 FORWARD_IF_ERROR(ZSTD_reset_matchState(
4816 &cdict->matchState,
4817 &cdict->workspace,
4818 ¶ms.cParams,
4819 params.useRowMatchFinder,
4820 ZSTDcrp_makeClean,
4821 ZSTDirp_reset,
4822 ZSTD_resetTarget_CDict), "");
4823 /* (Maybe) load the dictionary
4824 * Skips loading the dictionary if it is < 8 bytes.
4825 */
4826 { params.compressionLevel = ZSTD_CLEVEL_DEFAULT;
4827 params.fParams.contentSizeFlag = 1;
4828 { size_t const dictID = ZSTD_compress_insertDictionary(
4829 &cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace,
4830 ¶ms, cdict->dictContent, cdict->dictContentSize,
4831 dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace);
4832 FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");
4833 assert(dictID <= (size_t)(U32)-1);
4834 cdict->dictID = (U32)dictID;
4835 }
4836 }
4837
4838 return 0;
4839 }
4840
ZSTD_createCDict_advanced_internal(size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_compressionParameters cParams,ZSTD_paramSwitch_e useRowMatchFinder,U32 enableDedicatedDictSearch,ZSTD_customMem customMem)4841 static ZSTD_CDict* ZSTD_createCDict_advanced_internal(size_t dictSize,
4842 ZSTD_dictLoadMethod_e dictLoadMethod,
4843 ZSTD_compressionParameters cParams,
4844 ZSTD_paramSwitch_e useRowMatchFinder,
4845 U32 enableDedicatedDictSearch,
4846 ZSTD_customMem customMem)
4847 {
4848 if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL;
4849
4850 { size_t const workspaceSize =
4851 ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) +
4852 ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) +
4853 ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, enableDedicatedDictSearch, /* forCCtx */ 0) +
4854 (dictLoadMethod == ZSTD_dlm_byRef ? 0
4855 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))));
4856 void* const workspace = ZSTD_customMalloc(workspaceSize, customMem);
4857 ZSTD_cwksp ws;
4858 ZSTD_CDict* cdict;
4859
4860 if (!workspace) {
4861 ZSTD_customFree(workspace, customMem);
4862 return NULL;
4863 }
4864
4865 ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_dynamic_alloc);
4866
4867 cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
4868 assert(cdict != NULL);
4869 ZSTD_cwksp_move(&cdict->workspace, &ws);
4870 cdict->customMem = customMem;
4871 cdict->compressionLevel = ZSTD_NO_CLEVEL; /* signals advanced API usage */
4872 cdict->useRowMatchFinder = useRowMatchFinder;
4873 return cdict;
4874 }
4875 }
4876
ZSTD_createCDict_advanced(const void * dictBuffer,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType,ZSTD_compressionParameters cParams,ZSTD_customMem customMem)4877 ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
4878 ZSTD_dictLoadMethod_e dictLoadMethod,
4879 ZSTD_dictContentType_e dictContentType,
4880 ZSTD_compressionParameters cParams,
4881 ZSTD_customMem customMem)
4882 {
4883 ZSTD_CCtx_params cctxParams;
4884 ZSTD_memset(&cctxParams, 0, sizeof(cctxParams));
4885 ZSTD_CCtxParams_init(&cctxParams, 0);
4886 cctxParams.cParams = cParams;
4887 cctxParams.customMem = customMem;
4888 return ZSTD_createCDict_advanced2(
4889 dictBuffer, dictSize,
4890 dictLoadMethod, dictContentType,
4891 &cctxParams, customMem);
4892 }
4893
ZSTD_createCDict_advanced2(const void * dict,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType,const ZSTD_CCtx_params * originalCctxParams,ZSTD_customMem customMem)4894 ZSTD_CDict* ZSTD_createCDict_advanced2(
4895 const void* dict, size_t dictSize,
4896 ZSTD_dictLoadMethod_e dictLoadMethod,
4897 ZSTD_dictContentType_e dictContentType,
4898 const ZSTD_CCtx_params* originalCctxParams,
4899 ZSTD_customMem customMem)
4900 {
4901 ZSTD_CCtx_params cctxParams = *originalCctxParams;
4902 ZSTD_compressionParameters cParams;
4903 ZSTD_CDict* cdict;
4904
4905 DEBUGLOG(3, "ZSTD_createCDict_advanced2, mode %u", (unsigned)dictContentType);
4906 if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
4907
4908 if (cctxParams.enableDedicatedDictSearch) {
4909 cParams = ZSTD_dedicatedDictSearch_getCParams(
4910 cctxParams.compressionLevel, dictSize);
4911 ZSTD_overrideCParams(&cParams, &cctxParams.cParams);
4912 } else {
4913 cParams = ZSTD_getCParamsFromCCtxParams(
4914 &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
4915 }
4916
4917 if (!ZSTD_dedicatedDictSearch_isSupported(&cParams)) {
4918 /* Fall back to non-DDSS params */
4919 cctxParams.enableDedicatedDictSearch = 0;
4920 cParams = ZSTD_getCParamsFromCCtxParams(
4921 &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
4922 }
4923
4924 DEBUGLOG(3, "ZSTD_createCDict_advanced2: DDS: %u", cctxParams.enableDedicatedDictSearch);
4925 cctxParams.cParams = cParams;
4926 cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams);
4927
4928 cdict = ZSTD_createCDict_advanced_internal(dictSize,
4929 dictLoadMethod, cctxParams.cParams,
4930 cctxParams.useRowMatchFinder, cctxParams.enableDedicatedDictSearch,
4931 customMem);
4932
4933 if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
4934 dict, dictSize,
4935 dictLoadMethod, dictContentType,
4936 cctxParams) )) {
4937 ZSTD_freeCDict(cdict);
4938 return NULL;
4939 }
4940
4941 return cdict;
4942 }
4943
ZSTD_createCDict(const void * dict,size_t dictSize,int compressionLevel)4944 ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)
4945 {
4946 ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
4947 ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize,
4948 ZSTD_dlm_byCopy, ZSTD_dct_auto,
4949 cParams, ZSTD_defaultCMem);
4950 if (cdict)
4951 cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel;
4952 return cdict;
4953 }
4954
ZSTD_createCDict_byReference(const void * dict,size_t dictSize,int compressionLevel)4955 ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel)
4956 {
4957 ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict);
4958 ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize,
4959 ZSTD_dlm_byRef, ZSTD_dct_auto,
4960 cParams, ZSTD_defaultCMem);
4961 if (cdict)
4962 cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel;
4963 return cdict;
4964 }
4965
ZSTD_freeCDict(ZSTD_CDict * cdict)4966 size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
4967 {
4968 if (cdict==NULL) return 0; /* support free on NULL */
4969 { ZSTD_customMem const cMem = cdict->customMem;
4970 int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict);
4971 ZSTD_cwksp_free(&cdict->workspace, cMem);
4972 if (!cdictInWorkspace) {
4973 ZSTD_customFree(cdict, cMem);
4974 }
4975 return 0;
4976 }
4977 }
4978
4979 /*! ZSTD_initStaticCDict_advanced() :
4980 * Generate a digested dictionary in provided memory area.
4981 * workspace: The memory area to emplace the dictionary into.
4982 * Provided pointer must 8-bytes aligned.
4983 * It must outlive dictionary usage.
4984 * workspaceSize: Use ZSTD_estimateCDictSize()
4985 * to determine how large workspace must be.
4986 * cParams : use ZSTD_getCParams() to transform a compression level
4987 * into its relevants cParams.
4988 * @return : pointer to ZSTD_CDict*, or NULL if error (size too small)
4989 * Note : there is no corresponding "free" function.
4990 * Since workspace was allocated externally, it must be freed externally.
4991 */
ZSTD_initStaticCDict(void * workspace,size_t workspaceSize,const void * dict,size_t dictSize,ZSTD_dictLoadMethod_e dictLoadMethod,ZSTD_dictContentType_e dictContentType,ZSTD_compressionParameters cParams)4992 const ZSTD_CDict* ZSTD_initStaticCDict(
4993 void* workspace, size_t workspaceSize,
4994 const void* dict, size_t dictSize,
4995 ZSTD_dictLoadMethod_e dictLoadMethod,
4996 ZSTD_dictContentType_e dictContentType,
4997 ZSTD_compressionParameters cParams)
4998 {
4999 ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams);
5000 /* enableDedicatedDictSearch == 1 ensures matchstate is not too small in case this CDict will be used for DDS + row hash */
5001 size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0);
5002 size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
5003 + (dictLoadMethod == ZSTD_dlm_byRef ? 0
5004 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))))
5005 + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
5006 + matchStateSize;
5007 ZSTD_CDict* cdict;
5008 ZSTD_CCtx_params params;
5009
5010 if ((size_t)workspace & 7) return NULL; /* 8-aligned */
5011
5012 {
5013 ZSTD_cwksp ws;
5014 ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc);
5015 cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
5016 if (cdict == NULL) return NULL;
5017 ZSTD_cwksp_move(&cdict->workspace, &ws);
5018 }
5019
5020 DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u",
5021 (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize));
5022 if (workspaceSize < neededSize) return NULL;
5023
5024 ZSTD_CCtxParams_init(¶ms, 0);
5025 params.cParams = cParams;
5026 params.useRowMatchFinder = useRowMatchFinder;
5027 cdict->useRowMatchFinder = useRowMatchFinder;
5028
5029 if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
5030 dict, dictSize,
5031 dictLoadMethod, dictContentType,
5032 params) ))
5033 return NULL;
5034
5035 return cdict;
5036 }
5037
ZSTD_getCParamsFromCDict(const ZSTD_CDict * cdict)5038 ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict)
5039 {
5040 assert(cdict != NULL);
5041 return cdict->matchState.cParams;
5042 }
5043
5044 /*! ZSTD_getDictID_fromCDict() :
5045 * Provides the dictID of the dictionary loaded into `cdict`.
5046 * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
5047 * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
ZSTD_getDictID_fromCDict(const ZSTD_CDict * cdict)5048 unsigned ZSTD_getDictID_fromCDict(const ZSTD_CDict* cdict)
5049 {
5050 if (cdict==NULL) return 0;
5051 return cdict->dictID;
5052 }
5053
5054 /* ZSTD_compressBegin_usingCDict_internal() :
5055 * Implementation of various ZSTD_compressBegin_usingCDict* functions.
5056 */
ZSTD_compressBegin_usingCDict_internal(ZSTD_CCtx * const cctx,const ZSTD_CDict * const cdict,ZSTD_frameParameters const fParams,unsigned long long const pledgedSrcSize)5057 static size_t ZSTD_compressBegin_usingCDict_internal(
5058 ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,
5059 ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
5060 {
5061 ZSTD_CCtx_params cctxParams;
5062 DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_internal");
5063 RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!");
5064 /* Initialize the cctxParams from the cdict */
5065 {
5066 ZSTD_parameters params;
5067 params.fParams = fParams;
5068 params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
5069 || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
5070 || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
5071 || cdict->compressionLevel == 0 ) ?
5072 ZSTD_getCParamsFromCDict(cdict)
5073 : ZSTD_getCParams(cdict->compressionLevel,
5074 pledgedSrcSize,
5075 cdict->dictContentSize);
5076 ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, cdict->compressionLevel);
5077 }
5078 /* Increase window log to fit the entire dictionary and source if the
5079 * source size is known. Limit the increase to 19, which is the
5080 * window log for compression level 1 with the largest source size.
5081 */
5082 if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) {
5083 U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19);
5084 U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1;
5085 cctxParams.cParams.windowLog = MAX(cctxParams.cParams.windowLog, limitedSrcLog);
5086 }
5087 return ZSTD_compressBegin_internal(cctx,
5088 NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast,
5089 cdict,
5090 &cctxParams, pledgedSrcSize,
5091 ZSTDb_not_buffered);
5092 }
5093
5094
5095 /* ZSTD_compressBegin_usingCDict_advanced() :
5096 * This function is DEPRECATED.
5097 * cdict must be != NULL */
ZSTD_compressBegin_usingCDict_advanced(ZSTD_CCtx * const cctx,const ZSTD_CDict * const cdict,ZSTD_frameParameters const fParams,unsigned long long const pledgedSrcSize)5098 size_t ZSTD_compressBegin_usingCDict_advanced(
5099 ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,
5100 ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
5101 {
5102 return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, pledgedSrcSize);
5103 }
5104
5105 /* ZSTD_compressBegin_usingCDict() :
5106 * cdict must be != NULL */
ZSTD_compressBegin_usingCDict(ZSTD_CCtx * cctx,const ZSTD_CDict * cdict)5107 size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
5108 {
5109 ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
5110 return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN);
5111 }
5112
5113 /*! ZSTD_compress_usingCDict_internal():
5114 * Implementation of various ZSTD_compress_usingCDict* functions.
5115 */
ZSTD_compress_usingCDict_internal(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const ZSTD_CDict * cdict,ZSTD_frameParameters fParams)5116 static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx,
5117 void* dst, size_t dstCapacity,
5118 const void* src, size_t srcSize,
5119 const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)
5120 {
5121 FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */
5122 return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
5123 }
5124
5125 /*! ZSTD_compress_usingCDict_advanced():
5126 * This function is DEPRECATED.
5127 */
ZSTD_compress_usingCDict_advanced(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const ZSTD_CDict * cdict,ZSTD_frameParameters fParams)5128 size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
5129 void* dst, size_t dstCapacity,
5130 const void* src, size_t srcSize,
5131 const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)
5132 {
5133 return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams);
5134 }
5135
5136 /*! ZSTD_compress_usingCDict() :
5137 * Compression using a digested Dictionary.
5138 * Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
5139 * Note that compression parameters are decided at CDict creation time
5140 * while frame parameters are hardcoded */
ZSTD_compress_usingCDict(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize,const ZSTD_CDict * cdict)5141 size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
5142 void* dst, size_t dstCapacity,
5143 const void* src, size_t srcSize,
5144 const ZSTD_CDict* cdict)
5145 {
5146 ZSTD_frameParameters const fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
5147 return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams);
5148 }
5149
5150
5151
5152 /* ******************************************************************
5153 * Streaming
5154 ********************************************************************/
5155
ZSTD_createCStream(void)5156 ZSTD_CStream* ZSTD_createCStream(void)
5157 {
5158 DEBUGLOG(3, "ZSTD_createCStream");
5159 return ZSTD_createCStream_advanced(ZSTD_defaultCMem);
5160 }
5161
ZSTD_initStaticCStream(void * workspace,size_t workspaceSize)5162 ZSTD_CStream* ZSTD_initStaticCStream(void *workspace, size_t workspaceSize)
5163 {
5164 return ZSTD_initStaticCCtx(workspace, workspaceSize);
5165 }
5166
ZSTD_createCStream_advanced(ZSTD_customMem customMem)5167 ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)
5168 { /* CStream and CCtx are now same object */
5169 return ZSTD_createCCtx_advanced(customMem);
5170 }
5171
ZSTD_freeCStream(ZSTD_CStream * zcs)5172 size_t ZSTD_freeCStream(ZSTD_CStream* zcs)
5173 {
5174 return ZSTD_freeCCtx(zcs); /* same object */
5175 }
5176
5177
5178
5179 /*====== Initialization ======*/
5180
ZSTD_CStreamInSize(void)5181 size_t ZSTD_CStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX; }
5182
ZSTD_CStreamOutSize(void)5183 size_t ZSTD_CStreamOutSize(void)
5184 {
5185 return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ;
5186 }
5187
ZSTD_getCParamMode(ZSTD_CDict const * cdict,ZSTD_CCtx_params const * params,U64 pledgedSrcSize)5188 static ZSTD_cParamMode_e ZSTD_getCParamMode(ZSTD_CDict const* cdict, ZSTD_CCtx_params const* params, U64 pledgedSrcSize)
5189 {
5190 if (cdict != NULL && ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize))
5191 return ZSTD_cpm_attachDict;
5192 else
5193 return ZSTD_cpm_noAttachDict;
5194 }
5195
5196 /* ZSTD_resetCStream():
5197 * pledgedSrcSize == 0 means "unknown" */
ZSTD_resetCStream(ZSTD_CStream * zcs,unsigned long long pss)5198 size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss)
5199 {
5200 /* temporary : 0 interpreted as "unknown" during transition period.
5201 * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.
5202 * 0 will be interpreted as "empty" in the future.
5203 */
5204 U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
5205 DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (unsigned)pledgedSrcSize);
5206 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5207 FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
5208 return 0;
5209 }
5210
5211 /*! ZSTD_initCStream_internal() :
5212 * Note : for lib/compress only. Used by zstdmt_compress.c.
5213 * Assumption 1 : params are valid
5214 * Assumption 2 : either dict, or cdict, is defined, not both */
ZSTD_initCStream_internal(ZSTD_CStream * zcs,const void * dict,size_t dictSize,const ZSTD_CDict * cdict,const ZSTD_CCtx_params * params,unsigned long long pledgedSrcSize)5215 size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
5216 const void* dict, size_t dictSize, const ZSTD_CDict* cdict,
5217 const ZSTD_CCtx_params* params,
5218 unsigned long long pledgedSrcSize)
5219 {
5220 DEBUGLOG(4, "ZSTD_initCStream_internal");
5221 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5222 FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
5223 assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
5224 zcs->requestedParams = *params;
5225 assert(!((dict) && (cdict))); /* either dict or cdict, not both */
5226 if (dict) {
5227 FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
5228 } else {
5229 /* Dictionary is cleared if !cdict */
5230 FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
5231 }
5232 return 0;
5233 }
5234
5235 /* ZSTD_initCStream_usingCDict_advanced() :
5236 * same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream * zcs,const ZSTD_CDict * cdict,ZSTD_frameParameters fParams,unsigned long long pledgedSrcSize)5237 size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
5238 const ZSTD_CDict* cdict,
5239 ZSTD_frameParameters fParams,
5240 unsigned long long pledgedSrcSize)
5241 {
5242 DEBUGLOG(4, "ZSTD_initCStream_usingCDict_advanced");
5243 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5244 FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
5245 zcs->requestedParams.fParams = fParams;
5246 FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
5247 return 0;
5248 }
5249
5250 /* note : cdict must outlive compression session */
ZSTD_initCStream_usingCDict(ZSTD_CStream * zcs,const ZSTD_CDict * cdict)5251 size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
5252 {
5253 DEBUGLOG(4, "ZSTD_initCStream_usingCDict");
5254 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5255 FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");
5256 return 0;
5257 }
5258
5259
5260 /* ZSTD_initCStream_advanced() :
5261 * pledgedSrcSize must be exact.
5262 * if srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN.
5263 * dict is loaded with default parameters ZSTD_dct_auto and ZSTD_dlm_byCopy. */
ZSTD_initCStream_advanced(ZSTD_CStream * zcs,const void * dict,size_t dictSize,ZSTD_parameters params,unsigned long long pss)5264 size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
5265 const void* dict, size_t dictSize,
5266 ZSTD_parameters params, unsigned long long pss)
5267 {
5268 /* for compatibility with older programs relying on this behavior.
5269 * Users should now specify ZSTD_CONTENTSIZE_UNKNOWN.
5270 * This line will be removed in the future.
5271 */
5272 U64 const pledgedSrcSize = (pss==0 && params.fParams.contentSizeFlag==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
5273 DEBUGLOG(4, "ZSTD_initCStream_advanced");
5274 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5275 FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
5276 FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");
5277 ZSTD_CCtxParams_setZstdParams(&zcs->requestedParams, ¶ms);
5278 FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
5279 return 0;
5280 }
5281
ZSTD_initCStream_usingDict(ZSTD_CStream * zcs,const void * dict,size_t dictSize,int compressionLevel)5282 size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)
5283 {
5284 DEBUGLOG(4, "ZSTD_initCStream_usingDict");
5285 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5286 FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
5287 FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");
5288 return 0;
5289 }
5290
ZSTD_initCStream_srcSize(ZSTD_CStream * zcs,int compressionLevel,unsigned long long pss)5291 size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pss)
5292 {
5293 /* temporary : 0 interpreted as "unknown" during transition period.
5294 * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.
5295 * 0 will be interpreted as "empty" in the future.
5296 */
5297 U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;
5298 DEBUGLOG(4, "ZSTD_initCStream_srcSize");
5299 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5300 FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
5301 FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
5302 FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");
5303 return 0;
5304 }
5305
ZSTD_initCStream(ZSTD_CStream * zcs,int compressionLevel)5306 size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
5307 {
5308 DEBUGLOG(4, "ZSTD_initCStream");
5309 FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");
5310 FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");
5311 FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");
5312 return 0;
5313 }
5314
5315 /*====== Compression ======*/
5316
ZSTD_nextInputSizeHint(const ZSTD_CCtx * cctx)5317 static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx)
5318 {
5319 size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos;
5320 if (hintInSize==0) hintInSize = cctx->blockSize;
5321 return hintInSize;
5322 }
5323
5324 /** ZSTD_compressStream_generic():
5325 * internal function for all *compressStream*() variants
5326 * non-static, because can be called from zstdmt_compress.c
5327 * @return : hint size for next input */
ZSTD_compressStream_generic(ZSTD_CStream * zcs,ZSTD_outBuffer * output,ZSTD_inBuffer * input,ZSTD_EndDirective const flushMode)5328 static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
5329 ZSTD_outBuffer* output,
5330 ZSTD_inBuffer* input,
5331 ZSTD_EndDirective const flushMode)
5332 {
5333 const char* const istart = (const char*)input->src;
5334 const char* const iend = input->size != 0 ? istart + input->size : istart;
5335 const char* ip = input->pos != 0 ? istart + input->pos : istart;
5336 char* const ostart = (char*)output->dst;
5337 char* const oend = output->size != 0 ? ostart + output->size : ostart;
5338 char* op = output->pos != 0 ? ostart + output->pos : ostart;
5339 U32 someMoreWork = 1;
5340
5341 /* check expectations */
5342 DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (unsigned)flushMode);
5343 if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) {
5344 assert(zcs->inBuff != NULL);
5345 assert(zcs->inBuffSize > 0);
5346 }
5347 if (zcs->appliedParams.outBufferMode == ZSTD_bm_buffered) {
5348 assert(zcs->outBuff != NULL);
5349 assert(zcs->outBuffSize > 0);
5350 }
5351 assert(output->pos <= output->size);
5352 assert(input->pos <= input->size);
5353 assert((U32)flushMode <= (U32)ZSTD_e_end);
5354
5355 while (someMoreWork) {
5356 switch(zcs->streamStage)
5357 {
5358 case zcss_init:
5359 RETURN_ERROR(init_missing, "call ZSTD_initCStream() first!");
5360
5361 case zcss_load:
5362 if ( (flushMode == ZSTD_e_end)
5363 && ( (size_t)(oend-op) >= ZSTD_compressBound(iend-ip) /* Enough output space */
5364 || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) /* OR we are allowed to return dstSizeTooSmall */
5365 && (zcs->inBuffPos == 0) ) {
5366 /* shortcut to compression pass directly into output buffer */
5367 size_t const cSize = ZSTD_compressEnd(zcs,
5368 op, oend-op, ip, iend-ip);
5369 DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize);
5370 FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed");
5371 ip = iend;
5372 op += cSize;
5373 zcs->frameEnded = 1;
5374 ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
5375 someMoreWork = 0; break;
5376 }
5377 /* complete loading into inBuffer in buffered mode */
5378 if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) {
5379 size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos;
5380 size_t const loaded = ZSTD_limitCopy(
5381 zcs->inBuff + zcs->inBuffPos, toLoad,
5382 ip, iend-ip);
5383 zcs->inBuffPos += loaded;
5384 if (loaded != 0)
5385 ip += loaded;
5386 if ( (flushMode == ZSTD_e_continue)
5387 && (zcs->inBuffPos < zcs->inBuffTarget) ) {
5388 /* not enough input to fill full block : stop here */
5389 someMoreWork = 0; break;
5390 }
5391 if ( (flushMode == ZSTD_e_flush)
5392 && (zcs->inBuffPos == zcs->inToCompress) ) {
5393 /* empty */
5394 someMoreWork = 0; break;
5395 }
5396 }
5397 /* compress current block (note : this stage cannot be stopped in the middle) */
5398 DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode);
5399 { int const inputBuffered = (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered);
5400 void* cDst;
5401 size_t cSize;
5402 size_t oSize = oend-op;
5403 size_t const iSize = inputBuffered
5404 ? zcs->inBuffPos - zcs->inToCompress
5405 : MIN((size_t)(iend - ip), zcs->blockSize);
5406 if (oSize >= ZSTD_compressBound(iSize) || zcs->appliedParams.outBufferMode == ZSTD_bm_stable)
5407 cDst = op; /* compress into output buffer, to skip flush stage */
5408 else
5409 cDst = zcs->outBuff, oSize = zcs->outBuffSize;
5410 if (inputBuffered) {
5411 unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend);
5412 cSize = lastBlock ?
5413 ZSTD_compressEnd(zcs, cDst, oSize,
5414 zcs->inBuff + zcs->inToCompress, iSize) :
5415 ZSTD_compressContinue(zcs, cDst, oSize,
5416 zcs->inBuff + zcs->inToCompress, iSize);
5417 FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");
5418 zcs->frameEnded = lastBlock;
5419 /* prepare next block */
5420 zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;
5421 if (zcs->inBuffTarget > zcs->inBuffSize)
5422 zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;
5423 DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u",
5424 (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize);
5425 if (!lastBlock)
5426 assert(zcs->inBuffTarget <= zcs->inBuffSize);
5427 zcs->inToCompress = zcs->inBuffPos;
5428 } else {
5429 unsigned const lastBlock = (ip + iSize == iend);
5430 assert(flushMode == ZSTD_e_end /* Already validated */);
5431 cSize = lastBlock ?
5432 ZSTD_compressEnd(zcs, cDst, oSize, ip, iSize) :
5433 ZSTD_compressContinue(zcs, cDst, oSize, ip, iSize);
5434 /* Consume the input prior to error checking to mirror buffered mode. */
5435 if (iSize > 0)
5436 ip += iSize;
5437 FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");
5438 zcs->frameEnded = lastBlock;
5439 if (lastBlock)
5440 assert(ip == iend);
5441 }
5442 if (cDst == op) { /* no need to flush */
5443 op += cSize;
5444 if (zcs->frameEnded) {
5445 DEBUGLOG(5, "Frame completed directly in outBuffer");
5446 someMoreWork = 0;
5447 ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
5448 }
5449 break;
5450 }
5451 zcs->outBuffContentSize = cSize;
5452 zcs->outBuffFlushedSize = 0;
5453 zcs->streamStage = zcss_flush; /* pass-through to flush stage */
5454 }
5455 ZSTD_FALLTHROUGH;
5456 case zcss_flush:
5457 DEBUGLOG(5, "flush stage");
5458 assert(zcs->appliedParams.outBufferMode == ZSTD_bm_buffered);
5459 { size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
5460 size_t const flushed = ZSTD_limitCopy(op, (size_t)(oend-op),
5461 zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
5462 DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u",
5463 (unsigned)toFlush, (unsigned)(oend-op), (unsigned)flushed);
5464 if (flushed)
5465 op += flushed;
5466 zcs->outBuffFlushedSize += flushed;
5467 if (toFlush!=flushed) {
5468 /* flush not fully completed, presumably because dst is too small */
5469 assert(op==oend);
5470 someMoreWork = 0;
5471 break;
5472 }
5473 zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
5474 if (zcs->frameEnded) {
5475 DEBUGLOG(5, "Frame completed on flush");
5476 someMoreWork = 0;
5477 ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
5478 break;
5479 }
5480 zcs->streamStage = zcss_load;
5481 break;
5482 }
5483
5484 default: /* impossible */
5485 assert(0);
5486 }
5487 }
5488
5489 input->pos = ip - istart;
5490 output->pos = op - ostart;
5491 if (zcs->frameEnded) return 0;
5492 return ZSTD_nextInputSizeHint(zcs);
5493 }
5494
ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx * cctx)5495 static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx)
5496 {
5497 #ifdef ZSTD_MULTITHREAD
5498 if (cctx->appliedParams.nbWorkers >= 1) {
5499 assert(cctx->mtctx != NULL);
5500 return ZSTDMT_nextInputSizeHint(cctx->mtctx);
5501 }
5502 #endif
5503 return ZSTD_nextInputSizeHint(cctx);
5504
5505 }
5506
ZSTD_compressStream(ZSTD_CStream * zcs,ZSTD_outBuffer * output,ZSTD_inBuffer * input)5507 size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
5508 {
5509 FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) , "");
5510 return ZSTD_nextInputSizeHint_MTorST(zcs);
5511 }
5512
5513 /* After a compression call set the expected input/output buffer.
5514 * This is validated at the start of the next compression call.
5515 */
ZSTD_setBufferExpectations(ZSTD_CCtx * cctx,ZSTD_outBuffer const * output,ZSTD_inBuffer const * input)5516 static void ZSTD_setBufferExpectations(ZSTD_CCtx* cctx, ZSTD_outBuffer const* output, ZSTD_inBuffer const* input)
5517 {
5518 if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) {
5519 cctx->expectedInBuffer = *input;
5520 }
5521 if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) {
5522 cctx->expectedOutBufferSize = output->size - output->pos;
5523 }
5524 }
5525
5526 /* Validate that the input/output buffers match the expectations set by
5527 * ZSTD_setBufferExpectations.
5528 */
ZSTD_checkBufferStability(ZSTD_CCtx const * cctx,ZSTD_outBuffer const * output,ZSTD_inBuffer const * input,ZSTD_EndDirective endOp)5529 static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx,
5530 ZSTD_outBuffer const* output,
5531 ZSTD_inBuffer const* input,
5532 ZSTD_EndDirective endOp)
5533 {
5534 if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) {
5535 ZSTD_inBuffer const expect = cctx->expectedInBuffer;
5536 if (expect.src != input->src || expect.pos != input->pos || expect.size != input->size)
5537 RETURN_ERROR(srcBuffer_wrong, "ZSTD_c_stableInBuffer enabled but input differs!");
5538 if (endOp != ZSTD_e_end)
5539 RETURN_ERROR(srcBuffer_wrong, "ZSTD_c_stableInBuffer can only be used with ZSTD_e_end!");
5540 }
5541 if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) {
5542 size_t const outBufferSize = output->size - output->pos;
5543 if (cctx->expectedOutBufferSize != outBufferSize)
5544 RETURN_ERROR(dstBuffer_wrong, "ZSTD_c_stableOutBuffer enabled but output size differs!");
5545 }
5546 return 0;
5547 }
5548
ZSTD_CCtx_init_compressStream2(ZSTD_CCtx * cctx,ZSTD_EndDirective endOp,size_t inSize)5549 static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx,
5550 ZSTD_EndDirective endOp,
5551 size_t inSize) {
5552 ZSTD_CCtx_params params = cctx->requestedParams;
5553 ZSTD_prefixDict const prefixDict = cctx->prefixDict;
5554 FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */
5555 ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */
5556 assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */
5557 if (cctx->cdict && !cctx->localDict.cdict) {
5558 /* Let the cdict's compression level take priority over the requested params.
5559 * But do not take the cdict's compression level if the "cdict" is actually a localDict
5560 * generated from ZSTD_initLocalDict().
5561 */
5562 params.compressionLevel = cctx->cdict->compressionLevel;
5563 }
5564 DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");
5565 if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = inSize + 1; /* auto-fix pledgedSrcSize */
5566 {
5567 size_t const dictSize = prefixDict.dict
5568 ? prefixDict.dictSize
5569 : (cctx->cdict ? cctx->cdict->dictContentSize : 0);
5570 ZSTD_cParamMode_e const mode = ZSTD_getCParamMode(cctx->cdict, ¶ms, cctx->pledgedSrcSizePlusOne - 1);
5571 params.cParams = ZSTD_getCParamsFromCCtxParams(
5572 ¶ms, cctx->pledgedSrcSizePlusOne-1,
5573 dictSize, mode);
5574 }
5575
5576 params.useBlockSplitter = ZSTD_resolveBlockSplitterMode(params.useBlockSplitter, ¶ms.cParams);
5577 params.ldmParams.enableLdm = ZSTD_resolveEnableLdm(params.ldmParams.enableLdm, ¶ms.cParams);
5578 params.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params.useRowMatchFinder, ¶ms.cParams);
5579
5580 #ifdef ZSTD_MULTITHREAD
5581 if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) {
5582 params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */
5583 }
5584 if (params.nbWorkers > 0) {
5585 #if ZSTD_TRACE
5586 cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0;
5587 #endif
5588 /* mt context creation */
5589 if (cctx->mtctx == NULL) {
5590 DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",
5591 params.nbWorkers);
5592 cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem, cctx->pool);
5593 RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!");
5594 }
5595 /* mt compression */
5596 DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers);
5597 FORWARD_IF_ERROR( ZSTDMT_initCStream_internal(
5598 cctx->mtctx,
5599 prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,
5600 cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , "");
5601 cctx->dictID = cctx->cdict ? cctx->cdict->dictID : 0;
5602 cctx->dictContentSize = cctx->cdict ? cctx->cdict->dictContentSize : prefixDict.dictSize;
5603 cctx->consumedSrcSize = 0;
5604 cctx->producedCSize = 0;
5605 cctx->streamStage = zcss_load;
5606 cctx->appliedParams = params;
5607 } else
5608 #endif
5609 { U64 const pledgedSrcSize = cctx->pledgedSrcSizePlusOne - 1;
5610 assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
5611 FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
5612 prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, ZSTD_dtlm_fast,
5613 cctx->cdict,
5614 ¶ms, pledgedSrcSize,
5615 ZSTDb_buffered) , "");
5616 assert(cctx->appliedParams.nbWorkers == 0);
5617 cctx->inToCompress = 0;
5618 cctx->inBuffPos = 0;
5619 if (cctx->appliedParams.inBufferMode == ZSTD_bm_buffered) {
5620 /* for small input: avoid automatic flush on reaching end of block, since
5621 * it would require to add a 3-bytes null block to end frame
5622 */
5623 cctx->inBuffTarget = cctx->blockSize + (cctx->blockSize == pledgedSrcSize);
5624 } else {
5625 cctx->inBuffTarget = 0;
5626 }
5627 cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0;
5628 cctx->streamStage = zcss_load;
5629 cctx->frameEnded = 0;
5630 }
5631 return 0;
5632 }
5633
ZSTD_compressStream2(ZSTD_CCtx * cctx,ZSTD_outBuffer * output,ZSTD_inBuffer * input,ZSTD_EndDirective endOp)5634 size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
5635 ZSTD_outBuffer* output,
5636 ZSTD_inBuffer* input,
5637 ZSTD_EndDirective endOp)
5638 {
5639 DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp);
5640 /* check conditions */
5641 RETURN_ERROR_IF(output->pos > output->size, dstSize_tooSmall, "invalid output buffer");
5642 RETURN_ERROR_IF(input->pos > input->size, srcSize_wrong, "invalid input buffer");
5643 RETURN_ERROR_IF((U32)endOp > (U32)ZSTD_e_end, parameter_outOfBound, "invalid endDirective");
5644 assert(cctx != NULL);
5645
5646 /* transparent initialization stage */
5647 if (cctx->streamStage == zcss_init) {
5648 FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, endOp, input->size), "CompressStream2 initialization failed");
5649 ZSTD_setBufferExpectations(cctx, output, input); /* Set initial buffer expectations now that we've initialized */
5650 }
5651 /* end of transparent initialization stage */
5652
5653 FORWARD_IF_ERROR(ZSTD_checkBufferStability(cctx, output, input, endOp), "invalid buffers");
5654 /* compression stage */
5655 #ifdef ZSTD_MULTITHREAD
5656 if (cctx->appliedParams.nbWorkers > 0) {
5657 size_t flushMin;
5658 if (cctx->cParamsChanged) {
5659 ZSTDMT_updateCParams_whileCompressing(cctx->mtctx, &cctx->requestedParams);
5660 cctx->cParamsChanged = 0;
5661 }
5662 for (;;) {
5663 size_t const ipos = input->pos;
5664 size_t const opos = output->pos;
5665 flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp);
5666 cctx->consumedSrcSize += (U64)(input->pos - ipos);
5667 cctx->producedCSize += (U64)(output->pos - opos);
5668 if ( ZSTD_isError(flushMin)
5669 || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */
5670 if (flushMin == 0)
5671 ZSTD_CCtx_trace(cctx, 0);
5672 ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
5673 }
5674 FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed");
5675
5676 if (endOp == ZSTD_e_continue) {
5677 /* We only require some progress with ZSTD_e_continue, not maximal progress.
5678 * We're done if we've consumed or produced any bytes, or either buffer is
5679 * full.
5680 */
5681 if (input->pos != ipos || output->pos != opos || input->pos == input->size || output->pos == output->size)
5682 break;
5683 } else {
5684 assert(endOp == ZSTD_e_flush || endOp == ZSTD_e_end);
5685 /* We require maximal progress. We're done when the flush is complete or the
5686 * output buffer is full.
5687 */
5688 if (flushMin == 0 || output->pos == output->size)
5689 break;
5690 }
5691 }
5692 DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic");
5693 /* Either we don't require maximum forward progress, we've finished the
5694 * flush, or we are out of output space.
5695 */
5696 assert(endOp == ZSTD_e_continue || flushMin == 0 || output->pos == output->size);
5697 ZSTD_setBufferExpectations(cctx, output, input);
5698 return flushMin;
5699 }
5700 #endif
5701 FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "");
5702 DEBUGLOG(5, "completed ZSTD_compressStream2");
5703 ZSTD_setBufferExpectations(cctx, output, input);
5704 return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */
5705 }
5706
ZSTD_compressStream2_simpleArgs(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,size_t * dstPos,const void * src,size_t srcSize,size_t * srcPos,ZSTD_EndDirective endOp)5707 size_t ZSTD_compressStream2_simpleArgs (
5708 ZSTD_CCtx* cctx,
5709 void* dst, size_t dstCapacity, size_t* dstPos,
5710 const void* src, size_t srcSize, size_t* srcPos,
5711 ZSTD_EndDirective endOp)
5712 {
5713 ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
5714 ZSTD_inBuffer input = { src, srcSize, *srcPos };
5715 /* ZSTD_compressStream2() will check validity of dstPos and srcPos */
5716 size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp);
5717 *dstPos = output.pos;
5718 *srcPos = input.pos;
5719 return cErr;
5720 }
5721
ZSTD_compress2(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const void * src,size_t srcSize)5722 size_t ZSTD_compress2(ZSTD_CCtx* cctx,
5723 void* dst, size_t dstCapacity,
5724 const void* src, size_t srcSize)
5725 {
5726 ZSTD_bufferMode_e const originalInBufferMode = cctx->requestedParams.inBufferMode;
5727 ZSTD_bufferMode_e const originalOutBufferMode = cctx->requestedParams.outBufferMode;
5728 DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize);
5729 ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
5730 /* Enable stable input/output buffers. */
5731 cctx->requestedParams.inBufferMode = ZSTD_bm_stable;
5732 cctx->requestedParams.outBufferMode = ZSTD_bm_stable;
5733 { size_t oPos = 0;
5734 size_t iPos = 0;
5735 size_t const result = ZSTD_compressStream2_simpleArgs(cctx,
5736 dst, dstCapacity, &oPos,
5737 src, srcSize, &iPos,
5738 ZSTD_e_end);
5739 /* Reset to the original values. */
5740 cctx->requestedParams.inBufferMode = originalInBufferMode;
5741 cctx->requestedParams.outBufferMode = originalOutBufferMode;
5742 FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed");
5743 if (result != 0) { /* compression not completed, due to lack of output space */
5744 assert(oPos == dstCapacity);
5745 RETURN_ERROR(dstSize_tooSmall, "");
5746 }
5747 assert(iPos == srcSize); /* all input is expected consumed */
5748 return oPos;
5749 }
5750 }
5751
5752 typedef struct {
5753 U32 idx; /* Index in array of ZSTD_Sequence */
5754 U32 posInSequence; /* Position within sequence at idx */
5755 size_t posInSrc; /* Number of bytes given by sequences provided so far */
5756 } ZSTD_sequencePosition;
5757
5758 /* ZSTD_validateSequence() :
5759 * @offCode : is presumed to follow format required by ZSTD_storeSeq()
5760 * @returns a ZSTD error code if sequence is not valid
5761 */
5762 static size_t
ZSTD_validateSequence(U32 offCode,U32 matchLength,size_t posInSrc,U32 windowLog,size_t dictSize)5763 ZSTD_validateSequence(U32 offCode, U32 matchLength,
5764 size_t posInSrc, U32 windowLog, size_t dictSize)
5765 {
5766 U32 const windowSize = 1 << windowLog;
5767 /* posInSrc represents the amount of data the the decoder would decode up to this point.
5768 * As long as the amount of data decoded is less than or equal to window size, offsets may be
5769 * larger than the total length of output decoded in order to reference the dict, even larger than
5770 * window size. After output surpasses windowSize, we're limited to windowSize offsets again.
5771 */
5772 size_t const offsetBound = posInSrc > windowSize ? (size_t)windowSize : posInSrc + (size_t)dictSize;
5773 RETURN_ERROR_IF(offCode > STORE_OFFSET(offsetBound), corruption_detected, "Offset too large!");
5774 RETURN_ERROR_IF(matchLength < MINMATCH, corruption_detected, "Matchlength too small");
5775 return 0;
5776 }
5777
5778 /* Returns an offset code, given a sequence's raw offset, the ongoing repcode array, and whether litLength == 0 */
ZSTD_finalizeOffCode(U32 rawOffset,const U32 rep[ZSTD_REP_NUM],U32 ll0)5779 static U32 ZSTD_finalizeOffCode(U32 rawOffset, const U32 rep[ZSTD_REP_NUM], U32 ll0)
5780 {
5781 U32 offCode = STORE_OFFSET(rawOffset);
5782
5783 if (!ll0 && rawOffset == rep[0]) {
5784 offCode = STORE_REPCODE_1;
5785 } else if (rawOffset == rep[1]) {
5786 offCode = STORE_REPCODE(2 - ll0);
5787 } else if (rawOffset == rep[2]) {
5788 offCode = STORE_REPCODE(3 - ll0);
5789 } else if (ll0 && rawOffset == rep[0] - 1) {
5790 offCode = STORE_REPCODE_3;
5791 }
5792 return offCode;
5793 }
5794
5795 /* Returns 0 on success, and a ZSTD_error otherwise. This function scans through an array of
5796 * ZSTD_Sequence, storing the sequences it finds, until it reaches a block delimiter.
5797 */
5798 static size_t
ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx * cctx,ZSTD_sequencePosition * seqPos,const ZSTD_Sequence * const inSeqs,size_t inSeqsSize,const void * src,size_t blockSize)5799 ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx,
5800 ZSTD_sequencePosition* seqPos,
5801 const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
5802 const void* src, size_t blockSize)
5803 {
5804 U32 idx = seqPos->idx;
5805 BYTE const* ip = (BYTE const*)(src);
5806 const BYTE* const iend = ip + blockSize;
5807 repcodes_t updatedRepcodes;
5808 U32 dictSize;
5809
5810 if (cctx->cdict) {
5811 dictSize = (U32)cctx->cdict->dictContentSize;
5812 } else if (cctx->prefixDict.dict) {
5813 dictSize = (U32)cctx->prefixDict.dictSize;
5814 } else {
5815 dictSize = 0;
5816 }
5817 ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t));
5818 for (; (inSeqs[idx].matchLength != 0 || inSeqs[idx].offset != 0) && idx < inSeqsSize; ++idx) {
5819 U32 const litLength = inSeqs[idx].litLength;
5820 U32 const ll0 = (litLength == 0);
5821 U32 const matchLength = inSeqs[idx].matchLength;
5822 U32 const offCode = ZSTD_finalizeOffCode(inSeqs[idx].offset, updatedRepcodes.rep, ll0);
5823 ZSTD_updateRep(updatedRepcodes.rep, offCode, ll0);
5824
5825 DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offCode, matchLength, litLength);
5826 if (cctx->appliedParams.validateSequences) {
5827 seqPos->posInSrc += litLength + matchLength;
5828 FORWARD_IF_ERROR(ZSTD_validateSequence(offCode, matchLength, seqPos->posInSrc,
5829 cctx->appliedParams.cParams.windowLog, dictSize),
5830 "Sequence validation failed");
5831 }
5832 RETURN_ERROR_IF(idx - seqPos->idx > cctx->seqStore.maxNbSeq, memory_allocation,
5833 "Not enough memory allocated. Try adjusting ZSTD_c_minMatch.");
5834 ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offCode, matchLength);
5835 ip += matchLength + litLength;
5836 }
5837 ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t));
5838
5839 if (inSeqs[idx].litLength) {
5840 DEBUGLOG(6, "Storing last literals of size: %u", inSeqs[idx].litLength);
5841 ZSTD_storeLastLiterals(&cctx->seqStore, ip, inSeqs[idx].litLength);
5842 ip += inSeqs[idx].litLength;
5843 seqPos->posInSrc += inSeqs[idx].litLength;
5844 }
5845 RETURN_ERROR_IF(ip != iend, corruption_detected, "Blocksize doesn't agree with block delimiter!");
5846 seqPos->idx = idx+1;
5847 return 0;
5848 }
5849
5850 /* Returns the number of bytes to move the current read position back by. Only non-zero
5851 * if we ended up splitting a sequence. Otherwise, it may return a ZSTD error if something
5852 * went wrong.
5853 *
5854 * This function will attempt to scan through blockSize bytes represented by the sequences
5855 * in inSeqs, storing any (partial) sequences.
5856 *
5857 * Occasionally, we may want to change the actual number of bytes we consumed from inSeqs to
5858 * avoid splitting a match, or to avoid splitting a match such that it would produce a match
5859 * smaller than MINMATCH. In this case, we return the number of bytes that we didn't read from this block.
5860 */
5861 static size_t
ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx * cctx,ZSTD_sequencePosition * seqPos,const ZSTD_Sequence * const inSeqs,size_t inSeqsSize,const void * src,size_t blockSize)5862 ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos,
5863 const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
5864 const void* src, size_t blockSize)
5865 {
5866 U32 idx = seqPos->idx;
5867 U32 startPosInSequence = seqPos->posInSequence;
5868 U32 endPosInSequence = seqPos->posInSequence + (U32)blockSize;
5869 size_t dictSize;
5870 BYTE const* ip = (BYTE const*)(src);
5871 BYTE const* iend = ip + blockSize; /* May be adjusted if we decide to process fewer than blockSize bytes */
5872 repcodes_t updatedRepcodes;
5873 U32 bytesAdjustment = 0;
5874 U32 finalMatchSplit = 0;
5875
5876 if (cctx->cdict) {
5877 dictSize = cctx->cdict->dictContentSize;
5878 } else if (cctx->prefixDict.dict) {
5879 dictSize = cctx->prefixDict.dictSize;
5880 } else {
5881 dictSize = 0;
5882 }
5883 DEBUGLOG(5, "ZSTD_copySequencesToSeqStore: idx: %u PIS: %u blockSize: %zu", idx, startPosInSequence, blockSize);
5884 DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength);
5885 ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t));
5886 while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) {
5887 const ZSTD_Sequence currSeq = inSeqs[idx];
5888 U32 litLength = currSeq.litLength;
5889 U32 matchLength = currSeq.matchLength;
5890 U32 const rawOffset = currSeq.offset;
5891 U32 offCode;
5892
5893 /* Modify the sequence depending on where endPosInSequence lies */
5894 if (endPosInSequence >= currSeq.litLength + currSeq.matchLength) {
5895 if (startPosInSequence >= litLength) {
5896 startPosInSequence -= litLength;
5897 litLength = 0;
5898 matchLength -= startPosInSequence;
5899 } else {
5900 litLength -= startPosInSequence;
5901 }
5902 /* Move to the next sequence */
5903 endPosInSequence -= currSeq.litLength + currSeq.matchLength;
5904 startPosInSequence = 0;
5905 idx++;
5906 } else {
5907 /* This is the final (partial) sequence we're adding from inSeqs, and endPosInSequence
5908 does not reach the end of the match. So, we have to split the sequence */
5909 DEBUGLOG(6, "Require a split: diff: %u, idx: %u PIS: %u",
5910 currSeq.litLength + currSeq.matchLength - endPosInSequence, idx, endPosInSequence);
5911 if (endPosInSequence > litLength) {
5912 U32 firstHalfMatchLength;
5913 litLength = startPosInSequence >= litLength ? 0 : litLength - startPosInSequence;
5914 firstHalfMatchLength = endPosInSequence - startPosInSequence - litLength;
5915 if (matchLength > blockSize && firstHalfMatchLength >= cctx->appliedParams.cParams.minMatch) {
5916 /* Only ever split the match if it is larger than the block size */
5917 U32 secondHalfMatchLength = currSeq.matchLength + currSeq.litLength - endPosInSequence;
5918 if (secondHalfMatchLength < cctx->appliedParams.cParams.minMatch) {
5919 /* Move the endPosInSequence backward so that it creates match of minMatch length */
5920 endPosInSequence -= cctx->appliedParams.cParams.minMatch - secondHalfMatchLength;
5921 bytesAdjustment = cctx->appliedParams.cParams.minMatch - secondHalfMatchLength;
5922 firstHalfMatchLength -= bytesAdjustment;
5923 }
5924 matchLength = firstHalfMatchLength;
5925 /* Flag that we split the last match - after storing the sequence, exit the loop,
5926 but keep the value of endPosInSequence */
5927 finalMatchSplit = 1;
5928 } else {
5929 /* Move the position in sequence backwards so that we don't split match, and break to store
5930 * the last literals. We use the original currSeq.litLength as a marker for where endPosInSequence
5931 * should go. We prefer to do this whenever it is not necessary to split the match, or if doing so
5932 * would cause the first half of the match to be too small
5933 */
5934 bytesAdjustment = endPosInSequence - currSeq.litLength;
5935 endPosInSequence = currSeq.litLength;
5936 break;
5937 }
5938 } else {
5939 /* This sequence ends inside the literals, break to store the last literals */
5940 break;
5941 }
5942 }
5943 /* Check if this offset can be represented with a repcode */
5944 { U32 const ll0 = (litLength == 0);
5945 offCode = ZSTD_finalizeOffCode(rawOffset, updatedRepcodes.rep, ll0);
5946 ZSTD_updateRep(updatedRepcodes.rep, offCode, ll0);
5947 }
5948
5949 if (cctx->appliedParams.validateSequences) {
5950 seqPos->posInSrc += litLength + matchLength;
5951 FORWARD_IF_ERROR(ZSTD_validateSequence(offCode, matchLength, seqPos->posInSrc,
5952 cctx->appliedParams.cParams.windowLog, dictSize),
5953 "Sequence validation failed");
5954 }
5955 DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offCode, matchLength, litLength);
5956 RETURN_ERROR_IF(idx - seqPos->idx > cctx->seqStore.maxNbSeq, memory_allocation,
5957 "Not enough memory allocated. Try adjusting ZSTD_c_minMatch.");
5958 ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offCode, matchLength);
5959 ip += matchLength + litLength;
5960 }
5961 DEBUGLOG(5, "Ending seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength);
5962 assert(idx == inSeqsSize || endPosInSequence <= inSeqs[idx].litLength + inSeqs[idx].matchLength);
5963 seqPos->idx = idx;
5964 seqPos->posInSequence = endPosInSequence;
5965 ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t));
5966
5967 iend -= bytesAdjustment;
5968 if (ip != iend) {
5969 /* Store any last literals */
5970 U32 lastLLSize = (U32)(iend - ip);
5971 assert(ip <= iend);
5972 DEBUGLOG(6, "Storing last literals of size: %u", lastLLSize);
5973 ZSTD_storeLastLiterals(&cctx->seqStore, ip, lastLLSize);
5974 seqPos->posInSrc += lastLLSize;
5975 }
5976
5977 return bytesAdjustment;
5978 }
5979
5980 typedef size_t (*ZSTD_sequenceCopier) (ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos,
5981 const ZSTD_Sequence* const inSeqs, size_t inSeqsSize,
5982 const void* src, size_t blockSize);
ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode)5983 static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode)
5984 {
5985 ZSTD_sequenceCopier sequenceCopier = NULL;
5986 assert(ZSTD_cParam_withinBounds(ZSTD_c_blockDelimiters, mode));
5987 if (mode == ZSTD_sf_explicitBlockDelimiters) {
5988 return ZSTD_copySequencesToSeqStoreExplicitBlockDelim;
5989 } else if (mode == ZSTD_sf_noBlockDelimiters) {
5990 return ZSTD_copySequencesToSeqStoreNoBlockDelim;
5991 }
5992 assert(sequenceCopier != NULL);
5993 return sequenceCopier;
5994 }
5995
5996 /* Compress, block-by-block, all of the sequences given.
5997 *
5998 * Returns the cumulative size of all compressed blocks (including their headers),
5999 * otherwise a ZSTD error.
6000 */
6001 static size_t
ZSTD_compressSequences_internal(ZSTD_CCtx * cctx,void * dst,size_t dstCapacity,const ZSTD_Sequence * inSeqs,size_t inSeqsSize,const void * src,size_t srcSize)6002 ZSTD_compressSequences_internal(ZSTD_CCtx* cctx,
6003 void* dst, size_t dstCapacity,
6004 const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
6005 const void* src, size_t srcSize)
6006 {
6007 size_t cSize = 0;
6008 U32 lastBlock;
6009 size_t blockSize;
6010 size_t compressedSeqsSize;
6011 size_t remaining = srcSize;
6012 ZSTD_sequencePosition seqPos = {0, 0, 0};
6013
6014 BYTE const* ip = (BYTE const*)src;
6015 BYTE* op = (BYTE*)dst;
6016 ZSTD_sequenceCopier const sequenceCopier = ZSTD_selectSequenceCopier(cctx->appliedParams.blockDelimiters);
6017
6018 DEBUGLOG(4, "ZSTD_compressSequences_internal srcSize: %zu, inSeqsSize: %zu", srcSize, inSeqsSize);
6019 /* Special case: empty frame */
6020 if (remaining == 0) {
6021 U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1);
6022 RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "No room for empty frame block header");
6023 MEM_writeLE32(op, cBlockHeader24);
6024 op += ZSTD_blockHeaderSize;
6025 dstCapacity -= ZSTD_blockHeaderSize;
6026 cSize += ZSTD_blockHeaderSize;
6027 }
6028
6029 while (remaining) {
6030 size_t cBlockSize;
6031 size_t additionalByteAdjustment;
6032 lastBlock = remaining <= cctx->blockSize;
6033 blockSize = lastBlock ? (U32)remaining : (U32)cctx->blockSize;
6034 ZSTD_resetSeqStore(&cctx->seqStore);
6035 DEBUGLOG(4, "Working on new block. Blocksize: %zu", blockSize);
6036
6037 additionalByteAdjustment = sequenceCopier(cctx, &seqPos, inSeqs, inSeqsSize, ip, blockSize);
6038 FORWARD_IF_ERROR(additionalByteAdjustment, "Bad sequence copy");
6039 blockSize -= additionalByteAdjustment;
6040
6041 /* If blocks are too small, emit as a nocompress block */
6042 if (blockSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {
6043 cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
6044 FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed");
6045 DEBUGLOG(4, "Block too small, writing out nocompress block: cSize: %zu", cBlockSize);
6046 cSize += cBlockSize;
6047 ip += blockSize;
6048 op += cBlockSize;
6049 remaining -= blockSize;
6050 dstCapacity -= cBlockSize;
6051 continue;
6052 }
6053
6054 compressedSeqsSize = ZSTD_entropyCompressSeqStore(&cctx->seqStore,
6055 &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy,
6056 &cctx->appliedParams,
6057 op + ZSTD_blockHeaderSize /* Leave space for block header */, dstCapacity - ZSTD_blockHeaderSize,
6058 blockSize,
6059 cctx->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
6060 cctx->bmi2);
6061 FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed");
6062 DEBUGLOG(4, "Compressed sequences size: %zu", compressedSeqsSize);
6063
6064 if (!cctx->isFirstBlock &&
6065 ZSTD_maybeRLE(&cctx->seqStore) &&
6066 ZSTD_isRLE((BYTE const*)src, srcSize)) {
6067 /* We don't want to emit our first block as a RLE even if it qualifies because
6068 * doing so will cause the decoder (cli only) to throw a "should consume all input error."
6069 * This is only an issue for zstd <= v1.4.3
6070 */
6071 compressedSeqsSize = 1;
6072 }
6073
6074 if (compressedSeqsSize == 0) {
6075 /* ZSTD_noCompressBlock writes the block header as well */
6076 cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
6077 FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed");
6078 DEBUGLOG(4, "Writing out nocompress block, size: %zu", cBlockSize);
6079 } else if (compressedSeqsSize == 1) {
6080 cBlockSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, blockSize, lastBlock);
6081 FORWARD_IF_ERROR(cBlockSize, "RLE compress block failed");
6082 DEBUGLOG(4, "Writing out RLE block, size: %zu", cBlockSize);
6083 } else {
6084 U32 cBlockHeader;
6085 /* Error checking and repcodes update */
6086 ZSTD_blockState_confirmRepcodesAndEntropyTables(&cctx->blockState);
6087 if (cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
6088 cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
6089
6090 /* Write block header into beginning of block*/
6091 cBlockHeader = lastBlock + (((U32)bt_compressed)<<1) + (U32)(compressedSeqsSize << 3);
6092 MEM_writeLE24(op, cBlockHeader);
6093 cBlockSize = ZSTD_blockHeaderSize + compressedSeqsSize;
6094 DEBUGLOG(4, "Writing out compressed block, size: %zu", cBlockSize);
6095 }
6096
6097 cSize += cBlockSize;
6098 DEBUGLOG(4, "cSize running total: %zu", cSize);
6099
6100 if (lastBlock) {
6101 break;
6102 } else {
6103 ip += blockSize;
6104 op += cBlockSize;
6105 remaining -= blockSize;
6106 dstCapacity -= cBlockSize;
6107 cctx->isFirstBlock = 0;
6108 }
6109 }
6110
6111 return cSize;
6112 }
6113
ZSTD_compressSequences(ZSTD_CCtx * const cctx,void * dst,size_t dstCapacity,const ZSTD_Sequence * inSeqs,size_t inSeqsSize,const void * src,size_t srcSize)6114 size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstCapacity,
6115 const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
6116 const void* src, size_t srcSize)
6117 {
6118 BYTE* op = (BYTE*)dst;
6119 size_t cSize = 0;
6120 size_t compressedBlocksSize = 0;
6121 size_t frameHeaderSize = 0;
6122
6123 /* Transparent initialization stage, same as compressStream2() */
6124 DEBUGLOG(3, "ZSTD_compressSequences()");
6125 assert(cctx != NULL);
6126 FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed");
6127 /* Begin writing output, starting with frame header */
6128 frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity, &cctx->appliedParams, srcSize, cctx->dictID);
6129 op += frameHeaderSize;
6130 dstCapacity -= frameHeaderSize;
6131 cSize += frameHeaderSize;
6132 if (cctx->appliedParams.fParams.checksumFlag && srcSize) {
6133 XXH64_update(&cctx->xxhState, src, srcSize);
6134 }
6135 /* cSize includes block header size and compressed sequences size */
6136 compressedBlocksSize = ZSTD_compressSequences_internal(cctx,
6137 op, dstCapacity,
6138 inSeqs, inSeqsSize,
6139 src, srcSize);
6140 FORWARD_IF_ERROR(compressedBlocksSize, "Compressing blocks failed!");
6141 cSize += compressedBlocksSize;
6142 dstCapacity -= compressedBlocksSize;
6143
6144 if (cctx->appliedParams.fParams.checksumFlag) {
6145 U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
6146 RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");
6147 DEBUGLOG(4, "Write checksum : %08X", (unsigned)checksum);
6148 MEM_writeLE32((char*)dst + cSize, checksum);
6149 cSize += 4;
6150 }
6151
6152 DEBUGLOG(3, "Final compressed size: %zu", cSize);
6153 return cSize;
6154 }
6155
6156 /*====== Finalize ======*/
6157
6158 /*! ZSTD_flushStream() :
6159 * @return : amount of data remaining to flush */
ZSTD_flushStream(ZSTD_CStream * zcs,ZSTD_outBuffer * output)6160 size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
6161 {
6162 ZSTD_inBuffer input = { NULL, 0, 0 };
6163 return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush);
6164 }
6165
6166
ZSTD_endStream(ZSTD_CStream * zcs,ZSTD_outBuffer * output)6167 size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
6168 {
6169 ZSTD_inBuffer input = { NULL, 0, 0 };
6170 size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end);
6171 FORWARD_IF_ERROR( remainingToFlush , "ZSTD_compressStream2 failed");
6172 if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */
6173 /* single thread mode : attempt to calculate remaining to flush more precisely */
6174 { size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;
6175 size_t const checksumSize = (size_t)(zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4);
6176 size_t const toFlush = remainingToFlush + lastBlockSize + checksumSize;
6177 DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u", (unsigned)toFlush);
6178 return toFlush;
6179 }
6180 }
6181
6182
6183 /*-===== Pre-defined compression levels =====-*/
6184 #include "clevels.h"
6185
ZSTD_maxCLevel(void)6186 int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
ZSTD_minCLevel(void)6187 int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; }
ZSTD_defaultCLevel(void)6188 int ZSTD_defaultCLevel(void) { return ZSTD_CLEVEL_DEFAULT; }
6189
ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel,size_t const dictSize)6190 static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel, size_t const dictSize)
6191 {
6192 ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, 0, dictSize, ZSTD_cpm_createCDict);
6193 switch (cParams.strategy) {
6194 case ZSTD_fast:
6195 case ZSTD_dfast:
6196 break;
6197 case ZSTD_greedy:
6198 case ZSTD_lazy:
6199 case ZSTD_lazy2:
6200 cParams.hashLog += ZSTD_LAZY_DDSS_BUCKET_LOG;
6201 break;
6202 case ZSTD_btlazy2:
6203 case ZSTD_btopt:
6204 case ZSTD_btultra:
6205 case ZSTD_btultra2:
6206 break;
6207 }
6208 return cParams;
6209 }
6210
ZSTD_dedicatedDictSearch_isSupported(ZSTD_compressionParameters const * cParams)6211 static int ZSTD_dedicatedDictSearch_isSupported(
6212 ZSTD_compressionParameters const* cParams)
6213 {
6214 return (cParams->strategy >= ZSTD_greedy)
6215 && (cParams->strategy <= ZSTD_lazy2)
6216 && (cParams->hashLog > cParams->chainLog)
6217 && (cParams->chainLog <= 24);
6218 }
6219
6220 /**
6221 * Reverses the adjustment applied to cparams when enabling dedicated dict
6222 * search. This is used to recover the params set to be used in the working
6223 * context. (Otherwise, those tables would also grow.)
6224 */
ZSTD_dedicatedDictSearch_revertCParams(ZSTD_compressionParameters * cParams)6225 static void ZSTD_dedicatedDictSearch_revertCParams(
6226 ZSTD_compressionParameters* cParams) {
6227 switch (cParams->strategy) {
6228 case ZSTD_fast:
6229 case ZSTD_dfast:
6230 break;
6231 case ZSTD_greedy:
6232 case ZSTD_lazy:
6233 case ZSTD_lazy2:
6234 cParams->hashLog -= ZSTD_LAZY_DDSS_BUCKET_LOG;
6235 if (cParams->hashLog < ZSTD_HASHLOG_MIN) {
6236 cParams->hashLog = ZSTD_HASHLOG_MIN;
6237 }
6238 break;
6239 case ZSTD_btlazy2:
6240 case ZSTD_btopt:
6241 case ZSTD_btultra:
6242 case ZSTD_btultra2:
6243 break;
6244 }
6245 }
6246
ZSTD_getCParamRowSize(U64 srcSizeHint,size_t dictSize,ZSTD_cParamMode_e mode)6247 static U64 ZSTD_getCParamRowSize(U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
6248 {
6249 switch (mode) {
6250 case ZSTD_cpm_unknown:
6251 case ZSTD_cpm_noAttachDict:
6252 case ZSTD_cpm_createCDict:
6253 break;
6254 case ZSTD_cpm_attachDict:
6255 dictSize = 0;
6256 break;
6257 default:
6258 assert(0);
6259 break;
6260 }
6261 { int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN;
6262 size_t const addedSize = unknown && dictSize > 0 ? 500 : 0;
6263 return unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize;
6264 }
6265 }
6266
6267 /*! ZSTD_getCParams_internal() :
6268 * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.
6269 * Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown.
6270 * Use dictSize == 0 for unknown or unused.
6271 * Note: `mode` controls how we treat the `dictSize`. See docs for `ZSTD_cParamMode_e`. */
ZSTD_getCParams_internal(int compressionLevel,unsigned long long srcSizeHint,size_t dictSize,ZSTD_cParamMode_e mode)6272 static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode)
6273 {
6274 U64 const rSize = ZSTD_getCParamRowSize(srcSizeHint, dictSize, mode);
6275 U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);
6276 int row;
6277 DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel);
6278
6279 /* row */
6280 if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
6281 else if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */
6282 else if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;
6283 else row = compressionLevel;
6284
6285 { ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];
6286 DEBUGLOG(5, "ZSTD_getCParams_internal selected tableID: %u row: %u strat: %u", tableID, row, (U32)cp.strategy);
6287 /* acceleration factor */
6288 if (compressionLevel < 0) {
6289 int const clampedCompressionLevel = MAX(ZSTD_minCLevel(), compressionLevel);
6290 cp.targetLength = (unsigned)(-clampedCompressionLevel);
6291 }
6292 /* refine parameters based on srcSize & dictSize */
6293 return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode);
6294 }
6295 }
6296
6297 /*! ZSTD_getCParams() :
6298 * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.
6299 * Size values are optional, provide 0 if not known or unused */
ZSTD_getCParams(int compressionLevel,unsigned long long srcSizeHint,size_t dictSize)6300 ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)
6301 {
6302 if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;
6303 return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown);
6304 }
6305
6306 /*! ZSTD_getParams() :
6307 * same idea as ZSTD_getCParams()
6308 * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).
6309 * Fields of `ZSTD_frameParameters` are set to default values */
ZSTD_getParams_internal(int compressionLevel,unsigned long long srcSizeHint,size_t dictSize,ZSTD_cParamMode_e mode)6310 static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) {
6311 ZSTD_parameters params;
6312 ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, mode);
6313 DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel);
6314 ZSTD_memset(¶ms, 0, sizeof(params));
6315 params.cParams = cParams;
6316 params.fParams.contentSizeFlag = 1;
6317 return params;
6318 }
6319
6320 /*! ZSTD_getParams() :
6321 * same idea as ZSTD_getCParams()
6322 * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).
6323 * Fields of `ZSTD_frameParameters` are set to default values */
ZSTD_getParams(int compressionLevel,unsigned long long srcSizeHint,size_t dictSize)6324 ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) {
6325 if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;
6326 return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown);
6327 }
6328