1 | /* |
2 | * Copyright (c) Meta Platforms, Inc. and affiliates. |
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/allocations.h" /* ZSTD_customMalloc, ZSTD_customCalloc, ZSTD_customFree */ |
15 | #include "../common/zstd_deps.h" /* INT_MAX, ZSTD_memset, ZSTD_memcpy */ |
16 | #include "../common/mem.h" |
17 | #include "hist.h" /* HIST_countFast_wksp */ |
18 | #define FSE_STATIC_LINKING_ONLY /* FSE_encodeSymbol */ |
19 | #include "../common/fse.h" |
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 | #include "../common/bits.h" /* ZSTD_highbit32, ZSTD_rotateRight_U64 */ |
31 | |
32 | /* *************************************************************** |
33 | * Tuning parameters |
34 | *****************************************************************/ |
35 | /*! |
36 | * COMPRESS_HEAPMODE : |
37 | * Select how default decompression function ZSTD_compress() allocates its context, |
38 | * on stack (0, default), or into heap (1). |
39 | * Note that functions with explicit context such as ZSTD_compressCCtx() are unaffected. |
40 | */ |
41 | #ifndef ZSTD_COMPRESS_HEAPMODE |
42 | # define ZSTD_COMPRESS_HEAPMODE 0 |
43 | #endif |
44 | |
45 | /*! |
46 | * ZSTD_HASHLOG3_MAX : |
47 | * Maximum size of the hash table dedicated to find 3-bytes matches, |
48 | * in log format, aka 17 => 1 << 17 == 128Ki positions. |
49 | * This structure is only used in zstd_opt. |
50 | * Since allocation is centralized for all strategies, it has to be known here. |
51 | * The actual (selected) size of the hash table is then stored in ZSTD_matchState_t.hashLog3, |
52 | * so that zstd_opt.c doesn't need to know about this constant. |
53 | */ |
54 | #ifndef ZSTD_HASHLOG3_MAX |
55 | # define ZSTD_HASHLOG3_MAX 17 |
56 | #endif |
57 | |
58 | /*-************************************* |
59 | * Helper functions |
60 | ***************************************/ |
61 | /* ZSTD_compressBound() |
62 | * Note that the result from this function is only valid for |
63 | * the one-pass compression functions. |
64 | * When employing the streaming mode, |
65 | * if flushes are frequently altering the size of blocks, |
66 | * the overhead from block headers can make the compressed data larger |
67 | * than the return value of ZSTD_compressBound(). |
68 | */ |
69 | size_t ZSTD_compressBound(size_t srcSize) { |
70 | size_t const r = ZSTD_COMPRESSBOUND(srcSize); |
71 | if (r==0) return ERROR(srcSize_wrong); |
72 | return r; |
73 | } |
74 | |
75 | |
76 | /*-************************************* |
77 | * Context memory management |
78 | ***************************************/ |
79 | struct ZSTD_CDict_s { |
80 | const void* dictContent; |
81 | size_t dictContentSize; |
82 | ZSTD_dictContentType_e dictContentType; /* The dictContentType the CDict was created with */ |
83 | U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */ |
84 | ZSTD_cwksp workspace; |
85 | ZSTD_matchState_t matchState; |
86 | ZSTD_compressedBlockState_t cBlockState; |
87 | ZSTD_customMem customMem; |
88 | U32 dictID; |
89 | int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */ |
90 | ZSTD_paramSwitch_e useRowMatchFinder; /* Indicates whether the CDict was created with params that would use |
91 | * row-based matchfinder. Unless the cdict is reloaded, we will use |
92 | * the same greedy/lazy matchfinder at compression time. |
93 | */ |
94 | }; /* typedef'd to ZSTD_CDict within "zstd.h" */ |
95 | |
96 | ZSTD_CCtx* ZSTD_createCCtx(void) |
97 | { |
98 | return ZSTD_createCCtx_advanced(ZSTD_defaultCMem); |
99 | } |
100 | |
101 | static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager) |
102 | { |
103 | assert(cctx != NULL); |
104 | ZSTD_memset(cctx, 0, sizeof(*cctx)); |
105 | cctx->customMem = memManager; |
106 | cctx->bmi2 = ZSTD_cpuSupportsBmi2(); |
107 | { size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters); |
108 | assert(!ZSTD_isError(err)); |
109 | (void)err; |
110 | } |
111 | } |
112 | |
113 | ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem) |
114 | { |
115 | ZSTD_STATIC_ASSERT(zcss_init==0); |
116 | ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1)); |
117 | if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; |
118 | { ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_customMalloc(sizeof(ZSTD_CCtx), customMem); |
119 | if (!cctx) return NULL; |
120 | ZSTD_initCCtx(cctx, customMem); |
121 | return cctx; |
122 | } |
123 | } |
124 | |
125 | ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize) |
126 | { |
127 | ZSTD_cwksp ws; |
128 | ZSTD_CCtx* cctx; |
129 | if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL; /* minimum size */ |
130 | if ((size_t)workspace & 7) return NULL; /* must be 8-aligned */ |
131 | ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc); |
132 | |
133 | cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx)); |
134 | if (cctx == NULL) return NULL; |
135 | |
136 | ZSTD_memset(cctx, 0, sizeof(ZSTD_CCtx)); |
137 | ZSTD_cwksp_move(&cctx->workspace, &ws); |
138 | cctx->staticSize = workspaceSize; |
139 | |
140 | /* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */ |
141 | if (!ZSTD_cwksp_check_available(&cctx->workspace, ENTROPY_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL; |
142 | cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t)); |
143 | cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t)); |
144 | cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, ENTROPY_WORKSPACE_SIZE); |
145 | cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid()); |
146 | return cctx; |
147 | } |
148 | |
149 | /** |
150 | * Clears and frees all of the dictionaries in the CCtx. |
151 | */ |
152 | static void ZSTD_clearAllDicts(ZSTD_CCtx* cctx) |
153 | { |
154 | ZSTD_customFree(cctx->localDict.dictBuffer, cctx->customMem); |
155 | ZSTD_freeCDict(cctx->localDict.cdict); |
156 | ZSTD_memset(&cctx->localDict, 0, sizeof(cctx->localDict)); |
157 | ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); |
158 | cctx->cdict = NULL; |
159 | } |
160 | |
161 | static size_t ZSTD_sizeof_localDict(ZSTD_localDict dict) |
162 | { |
163 | size_t const bufferSize = dict.dictBuffer != NULL ? dict.dictSize : 0; |
164 | size_t const cdictSize = ZSTD_sizeof_CDict(dict.cdict); |
165 | return bufferSize + cdictSize; |
166 | } |
167 | |
168 | static void ZSTD_freeCCtxContent(ZSTD_CCtx* cctx) |
169 | { |
170 | assert(cctx != NULL); |
171 | assert(cctx->staticSize == 0); |
172 | ZSTD_clearAllDicts(cctx); |
173 | #ifdef ZSTD_MULTITHREAD |
174 | ZSTDMT_freeCCtx(cctx->mtctx); cctx->mtctx = NULL; |
175 | #endif |
176 | ZSTD_cwksp_free(&cctx->workspace, cctx->customMem); |
177 | } |
178 | |
179 | size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx) |
180 | { |
181 | if (cctx==NULL) return 0; /* support free on NULL */ |
182 | RETURN_ERROR_IF(cctx->staticSize, memory_allocation, |
183 | "not compatible with static CCtx" ); |
184 | { int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx); |
185 | ZSTD_freeCCtxContent(cctx); |
186 | if (!cctxInWorkspace) ZSTD_customFree(cctx, cctx->customMem); |
187 | } |
188 | return 0; |
189 | } |
190 | |
191 | |
192 | static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx) |
193 | { |
194 | #ifdef ZSTD_MULTITHREAD |
195 | return ZSTDMT_sizeof_CCtx(cctx->mtctx); |
196 | #else |
197 | (void)cctx; |
198 | return 0; |
199 | #endif |
200 | } |
201 | |
202 | |
203 | size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx) |
204 | { |
205 | if (cctx==NULL) return 0; /* support sizeof on NULL */ |
206 | /* cctx may be in the workspace */ |
207 | return (cctx->workspace.workspace == cctx ? 0 : sizeof(*cctx)) |
208 | + ZSTD_cwksp_sizeof(&cctx->workspace) |
209 | + ZSTD_sizeof_localDict(cctx->localDict) |
210 | + ZSTD_sizeof_mtctx(cctx); |
211 | } |
212 | |
213 | size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs) |
214 | { |
215 | return ZSTD_sizeof_CCtx(zcs); /* same object */ |
216 | } |
217 | |
218 | /* private API call, for dictBuilder only */ |
219 | const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); } |
220 | |
221 | /* Returns true if the strategy supports using a row based matchfinder */ |
222 | static int ZSTD_rowMatchFinderSupported(const ZSTD_strategy strategy) { |
223 | return (strategy >= ZSTD_greedy && strategy <= ZSTD_lazy2); |
224 | } |
225 | |
226 | /* Returns true if the strategy and useRowMatchFinder mode indicate that we will use the row based matchfinder |
227 | * for this compression. |
228 | */ |
229 | static int ZSTD_rowMatchFinderUsed(const ZSTD_strategy strategy, const ZSTD_paramSwitch_e mode) { |
230 | assert(mode != ZSTD_ps_auto); |
231 | return ZSTD_rowMatchFinderSupported(strategy) && (mode == ZSTD_ps_enable); |
232 | } |
233 | |
234 | /* Returns row matchfinder usage given an initial mode and cParams */ |
235 | static ZSTD_paramSwitch_e ZSTD_resolveRowMatchFinderMode(ZSTD_paramSwitch_e mode, |
236 | const ZSTD_compressionParameters* const cParams) { |
237 | #if defined(ZSTD_ARCH_X86_SSE2) || defined(ZSTD_ARCH_ARM_NEON) |
238 | int const kHasSIMD128 = 1; |
239 | #else |
240 | int const kHasSIMD128 = 0; |
241 | #endif |
242 | if (mode != ZSTD_ps_auto) return mode; /* if requested enabled, but no SIMD, we still will use row matchfinder */ |
243 | mode = ZSTD_ps_disable; |
244 | if (!ZSTD_rowMatchFinderSupported(cParams->strategy)) return mode; |
245 | if (kHasSIMD128) { |
246 | if (cParams->windowLog > 14) mode = ZSTD_ps_enable; |
247 | } else { |
248 | if (cParams->windowLog > 17) mode = ZSTD_ps_enable; |
249 | } |
250 | return mode; |
251 | } |
252 | |
253 | /* Returns block splitter usage (generally speaking, when using slower/stronger compression modes) */ |
254 | static ZSTD_paramSwitch_e ZSTD_resolveBlockSplitterMode(ZSTD_paramSwitch_e mode, |
255 | const ZSTD_compressionParameters* const cParams) { |
256 | if (mode != ZSTD_ps_auto) return mode; |
257 | return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 17) ? ZSTD_ps_enable : ZSTD_ps_disable; |
258 | } |
259 | |
260 | /* Returns 1 if the arguments indicate that we should allocate a chainTable, 0 otherwise */ |
261 | static int ZSTD_allocateChainTable(const ZSTD_strategy strategy, |
262 | const ZSTD_paramSwitch_e useRowMatchFinder, |
263 | const U32 forDDSDict) { |
264 | assert(useRowMatchFinder != ZSTD_ps_auto); |
265 | /* We always should allocate a chaintable if we are allocating a matchstate for a DDS dictionary matchstate. |
266 | * We do not allocate a chaintable if we are using ZSTD_fast, or are using the row-based matchfinder. |
267 | */ |
268 | return forDDSDict || ((strategy != ZSTD_fast) && !ZSTD_rowMatchFinderUsed(strategy, useRowMatchFinder)); |
269 | } |
270 | |
271 | /* Returns ZSTD_ps_enable if compression parameters are such that we should |
272 | * enable long distance matching (wlog >= 27, strategy >= btopt). |
273 | * Returns ZSTD_ps_disable otherwise. |
274 | */ |
275 | static ZSTD_paramSwitch_e ZSTD_resolveEnableLdm(ZSTD_paramSwitch_e mode, |
276 | const ZSTD_compressionParameters* const cParams) { |
277 | if (mode != ZSTD_ps_auto) return mode; |
278 | return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 27) ? ZSTD_ps_enable : ZSTD_ps_disable; |
279 | } |
280 | |
281 | static int ZSTD_resolveExternalSequenceValidation(int mode) { |
282 | return mode; |
283 | } |
284 | |
285 | /* Resolves maxBlockSize to the default if no value is present. */ |
286 | static size_t ZSTD_resolveMaxBlockSize(size_t maxBlockSize) { |
287 | if (maxBlockSize == 0) { |
288 | return ZSTD_BLOCKSIZE_MAX; |
289 | } else { |
290 | return maxBlockSize; |
291 | } |
292 | } |
293 | |
294 | static ZSTD_paramSwitch_e ZSTD_resolveExternalRepcodeSearch(ZSTD_paramSwitch_e value, int cLevel) { |
295 | if (value != ZSTD_ps_auto) return value; |
296 | if (cLevel < 10) { |
297 | return ZSTD_ps_disable; |
298 | } else { |
299 | return ZSTD_ps_enable; |
300 | } |
301 | } |
302 | |
303 | /* Returns 1 if compression parameters are such that CDict hashtable and chaintable indices are tagged. |
304 | * If so, the tags need to be removed in ZSTD_resetCCtx_byCopyingCDict. */ |
305 | static int ZSTD_CDictIndicesAreTagged(const ZSTD_compressionParameters* const cParams) { |
306 | return cParams->strategy == ZSTD_fast || cParams->strategy == ZSTD_dfast; |
307 | } |
308 | |
309 | static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams( |
310 | ZSTD_compressionParameters cParams) |
311 | { |
312 | ZSTD_CCtx_params cctxParams; |
313 | /* should not matter, as all cParams are presumed properly defined */ |
314 | ZSTD_CCtxParams_init(&cctxParams, ZSTD_CLEVEL_DEFAULT); |
315 | cctxParams.cParams = cParams; |
316 | |
317 | /* Adjust advanced params according to cParams */ |
318 | cctxParams.ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams.ldmParams.enableLdm, &cParams); |
319 | if (cctxParams.ldmParams.enableLdm == ZSTD_ps_enable) { |
320 | ZSTD_ldm_adjustParameters(&cctxParams.ldmParams, &cParams); |
321 | assert(cctxParams.ldmParams.hashLog >= cctxParams.ldmParams.bucketSizeLog); |
322 | assert(cctxParams.ldmParams.hashRateLog < 32); |
323 | } |
324 | cctxParams.useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams.useBlockSplitter, &cParams); |
325 | cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams); |
326 | cctxParams.validateSequences = ZSTD_resolveExternalSequenceValidation(cctxParams.validateSequences); |
327 | cctxParams.maxBlockSize = ZSTD_resolveMaxBlockSize(cctxParams.maxBlockSize); |
328 | cctxParams.searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(cctxParams.searchForExternalRepcodes, |
329 | cctxParams.compressionLevel); |
330 | assert(!ZSTD_checkCParams(cParams)); |
331 | return cctxParams; |
332 | } |
333 | |
334 | static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced( |
335 | ZSTD_customMem customMem) |
336 | { |
337 | ZSTD_CCtx_params* params; |
338 | if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; |
339 | params = (ZSTD_CCtx_params*)ZSTD_customCalloc( |
340 | sizeof(ZSTD_CCtx_params), customMem); |
341 | if (!params) { return NULL; } |
342 | ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT); |
343 | params->customMem = customMem; |
344 | return params; |
345 | } |
346 | |
347 | ZSTD_CCtx_params* ZSTD_createCCtxParams(void) |
348 | { |
349 | return ZSTD_createCCtxParams_advanced(ZSTD_defaultCMem); |
350 | } |
351 | |
352 | size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params) |
353 | { |
354 | if (params == NULL) { return 0; } |
355 | ZSTD_customFree(params, params->customMem); |
356 | return 0; |
357 | } |
358 | |
359 | size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params) |
360 | { |
361 | return ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT); |
362 | } |
363 | |
364 | size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) { |
365 | RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!" ); |
366 | ZSTD_memset(cctxParams, 0, sizeof(*cctxParams)); |
367 | cctxParams->compressionLevel = compressionLevel; |
368 | cctxParams->fParams.contentSizeFlag = 1; |
369 | return 0; |
370 | } |
371 | |
372 | #define ZSTD_NO_CLEVEL 0 |
373 | |
374 | /** |
375 | * Initializes `cctxParams` from `params` and `compressionLevel`. |
376 | * @param compressionLevel If params are derived from a compression level then that compression level, otherwise ZSTD_NO_CLEVEL. |
377 | */ |
378 | static void |
379 | ZSTD_CCtxParams_init_internal(ZSTD_CCtx_params* cctxParams, |
380 | const ZSTD_parameters* params, |
381 | int compressionLevel) |
382 | { |
383 | assert(!ZSTD_checkCParams(params->cParams)); |
384 | ZSTD_memset(cctxParams, 0, sizeof(*cctxParams)); |
385 | cctxParams->cParams = params->cParams; |
386 | cctxParams->fParams = params->fParams; |
387 | /* Should not matter, as all cParams are presumed properly defined. |
388 | * But, set it for tracing anyway. |
389 | */ |
390 | cctxParams->compressionLevel = compressionLevel; |
391 | cctxParams->useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams->useRowMatchFinder, ¶ms->cParams); |
392 | cctxParams->useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams->useBlockSplitter, ¶ms->cParams); |
393 | cctxParams->ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams->ldmParams.enableLdm, ¶ms->cParams); |
394 | cctxParams->validateSequences = ZSTD_resolveExternalSequenceValidation(cctxParams->validateSequences); |
395 | cctxParams->maxBlockSize = ZSTD_resolveMaxBlockSize(cctxParams->maxBlockSize); |
396 | cctxParams->searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(cctxParams->searchForExternalRepcodes, compressionLevel); |
397 | DEBUGLOG(4, "ZSTD_CCtxParams_init_internal: useRowMatchFinder=%d, useBlockSplitter=%d ldm=%d" , |
398 | cctxParams->useRowMatchFinder, cctxParams->useBlockSplitter, cctxParams->ldmParams.enableLdm); |
399 | } |
400 | |
401 | size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params) |
402 | { |
403 | RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!" ); |
404 | FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "" ); |
405 | ZSTD_CCtxParams_init_internal(cctxParams, ¶ms, ZSTD_NO_CLEVEL); |
406 | return 0; |
407 | } |
408 | |
409 | /** |
410 | * Sets cctxParams' cParams and fParams from params, but otherwise leaves them alone. |
411 | * @param params Validated zstd parameters. |
412 | */ |
413 | static void ZSTD_CCtxParams_setZstdParams( |
414 | ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params) |
415 | { |
416 | assert(!ZSTD_checkCParams(params->cParams)); |
417 | cctxParams->cParams = params->cParams; |
418 | cctxParams->fParams = params->fParams; |
419 | /* Should not matter, as all cParams are presumed properly defined. |
420 | * But, set it for tracing anyway. |
421 | */ |
422 | cctxParams->compressionLevel = ZSTD_NO_CLEVEL; |
423 | } |
424 | |
425 | ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param) |
426 | { |
427 | ZSTD_bounds bounds = { 0, 0, 0 }; |
428 | |
429 | switch(param) |
430 | { |
431 | case ZSTD_c_compressionLevel: |
432 | bounds.lowerBound = ZSTD_minCLevel(); |
433 | bounds.upperBound = ZSTD_maxCLevel(); |
434 | return bounds; |
435 | |
436 | case ZSTD_c_windowLog: |
437 | bounds.lowerBound = ZSTD_WINDOWLOG_MIN; |
438 | bounds.upperBound = ZSTD_WINDOWLOG_MAX; |
439 | return bounds; |
440 | |
441 | case ZSTD_c_hashLog: |
442 | bounds.lowerBound = ZSTD_HASHLOG_MIN; |
443 | bounds.upperBound = ZSTD_HASHLOG_MAX; |
444 | return bounds; |
445 | |
446 | case ZSTD_c_chainLog: |
447 | bounds.lowerBound = ZSTD_CHAINLOG_MIN; |
448 | bounds.upperBound = ZSTD_CHAINLOG_MAX; |
449 | return bounds; |
450 | |
451 | case ZSTD_c_searchLog: |
452 | bounds.lowerBound = ZSTD_SEARCHLOG_MIN; |
453 | bounds.upperBound = ZSTD_SEARCHLOG_MAX; |
454 | return bounds; |
455 | |
456 | case ZSTD_c_minMatch: |
457 | bounds.lowerBound = ZSTD_MINMATCH_MIN; |
458 | bounds.upperBound = ZSTD_MINMATCH_MAX; |
459 | return bounds; |
460 | |
461 | case ZSTD_c_targetLength: |
462 | bounds.lowerBound = ZSTD_TARGETLENGTH_MIN; |
463 | bounds.upperBound = ZSTD_TARGETLENGTH_MAX; |
464 | return bounds; |
465 | |
466 | case ZSTD_c_strategy: |
467 | bounds.lowerBound = ZSTD_STRATEGY_MIN; |
468 | bounds.upperBound = ZSTD_STRATEGY_MAX; |
469 | return bounds; |
470 | |
471 | case ZSTD_c_contentSizeFlag: |
472 | bounds.lowerBound = 0; |
473 | bounds.upperBound = 1; |
474 | return bounds; |
475 | |
476 | case ZSTD_c_checksumFlag: |
477 | bounds.lowerBound = 0; |
478 | bounds.upperBound = 1; |
479 | return bounds; |
480 | |
481 | case ZSTD_c_dictIDFlag: |
482 | bounds.lowerBound = 0; |
483 | bounds.upperBound = 1; |
484 | return bounds; |
485 | |
486 | case ZSTD_c_nbWorkers: |
487 | bounds.lowerBound = 0; |
488 | #ifdef ZSTD_MULTITHREAD |
489 | bounds.upperBound = ZSTDMT_NBWORKERS_MAX; |
490 | #else |
491 | bounds.upperBound = 0; |
492 | #endif |
493 | return bounds; |
494 | |
495 | case ZSTD_c_jobSize: |
496 | bounds.lowerBound = 0; |
497 | #ifdef ZSTD_MULTITHREAD |
498 | bounds.upperBound = ZSTDMT_JOBSIZE_MAX; |
499 | #else |
500 | bounds.upperBound = 0; |
501 | #endif |
502 | return bounds; |
503 | |
504 | case ZSTD_c_overlapLog: |
505 | #ifdef ZSTD_MULTITHREAD |
506 | bounds.lowerBound = ZSTD_OVERLAPLOG_MIN; |
507 | bounds.upperBound = ZSTD_OVERLAPLOG_MAX; |
508 | #else |
509 | bounds.lowerBound = 0; |
510 | bounds.upperBound = 0; |
511 | #endif |
512 | return bounds; |
513 | |
514 | case ZSTD_c_enableDedicatedDictSearch: |
515 | bounds.lowerBound = 0; |
516 | bounds.upperBound = 1; |
517 | return bounds; |
518 | |
519 | case ZSTD_c_enableLongDistanceMatching: |
520 | bounds.lowerBound = (int)ZSTD_ps_auto; |
521 | bounds.upperBound = (int)ZSTD_ps_disable; |
522 | return bounds; |
523 | |
524 | case ZSTD_c_ldmHashLog: |
525 | bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN; |
526 | bounds.upperBound = ZSTD_LDM_HASHLOG_MAX; |
527 | return bounds; |
528 | |
529 | case ZSTD_c_ldmMinMatch: |
530 | bounds.lowerBound = ZSTD_LDM_MINMATCH_MIN; |
531 | bounds.upperBound = ZSTD_LDM_MINMATCH_MAX; |
532 | return bounds; |
533 | |
534 | case ZSTD_c_ldmBucketSizeLog: |
535 | bounds.lowerBound = ZSTD_LDM_BUCKETSIZELOG_MIN; |
536 | bounds.upperBound = ZSTD_LDM_BUCKETSIZELOG_MAX; |
537 | return bounds; |
538 | |
539 | case ZSTD_c_ldmHashRateLog: |
540 | bounds.lowerBound = ZSTD_LDM_HASHRATELOG_MIN; |
541 | bounds.upperBound = ZSTD_LDM_HASHRATELOG_MAX; |
542 | return bounds; |
543 | |
544 | /* experimental parameters */ |
545 | case ZSTD_c_rsyncable: |
546 | bounds.lowerBound = 0; |
547 | bounds.upperBound = 1; |
548 | return bounds; |
549 | |
550 | case ZSTD_c_forceMaxWindow : |
551 | bounds.lowerBound = 0; |
552 | bounds.upperBound = 1; |
553 | return bounds; |
554 | |
555 | case ZSTD_c_format: |
556 | ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless); |
557 | bounds.lowerBound = ZSTD_f_zstd1; |
558 | bounds.upperBound = ZSTD_f_zstd1_magicless; /* note : how to ensure at compile time that this is the highest value enum ? */ |
559 | return bounds; |
560 | |
561 | case ZSTD_c_forceAttachDict: |
562 | ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceLoad); |
563 | bounds.lowerBound = ZSTD_dictDefaultAttach; |
564 | bounds.upperBound = ZSTD_dictForceLoad; /* note : how to ensure at compile time that this is the highest value enum ? */ |
565 | return bounds; |
566 | |
567 | case ZSTD_c_literalCompressionMode: |
568 | ZSTD_STATIC_ASSERT(ZSTD_ps_auto < ZSTD_ps_enable && ZSTD_ps_enable < ZSTD_ps_disable); |
569 | bounds.lowerBound = (int)ZSTD_ps_auto; |
570 | bounds.upperBound = (int)ZSTD_ps_disable; |
571 | return bounds; |
572 | |
573 | case ZSTD_c_targetCBlockSize: |
574 | bounds.lowerBound = ZSTD_TARGETCBLOCKSIZE_MIN; |
575 | bounds.upperBound = ZSTD_TARGETCBLOCKSIZE_MAX; |
576 | return bounds; |
577 | |
578 | case ZSTD_c_srcSizeHint: |
579 | bounds.lowerBound = ZSTD_SRCSIZEHINT_MIN; |
580 | bounds.upperBound = ZSTD_SRCSIZEHINT_MAX; |
581 | return bounds; |
582 | |
583 | case ZSTD_c_stableInBuffer: |
584 | case ZSTD_c_stableOutBuffer: |
585 | bounds.lowerBound = (int)ZSTD_bm_buffered; |
586 | bounds.upperBound = (int)ZSTD_bm_stable; |
587 | return bounds; |
588 | |
589 | case ZSTD_c_blockDelimiters: |
590 | bounds.lowerBound = (int)ZSTD_sf_noBlockDelimiters; |
591 | bounds.upperBound = (int)ZSTD_sf_explicitBlockDelimiters; |
592 | return bounds; |
593 | |
594 | case ZSTD_c_validateSequences: |
595 | bounds.lowerBound = 0; |
596 | bounds.upperBound = 1; |
597 | return bounds; |
598 | |
599 | case ZSTD_c_useBlockSplitter: |
600 | bounds.lowerBound = (int)ZSTD_ps_auto; |
601 | bounds.upperBound = (int)ZSTD_ps_disable; |
602 | return bounds; |
603 | |
604 | case ZSTD_c_useRowMatchFinder: |
605 | bounds.lowerBound = (int)ZSTD_ps_auto; |
606 | bounds.upperBound = (int)ZSTD_ps_disable; |
607 | return bounds; |
608 | |
609 | case ZSTD_c_deterministicRefPrefix: |
610 | bounds.lowerBound = 0; |
611 | bounds.upperBound = 1; |
612 | return bounds; |
613 | |
614 | case ZSTD_c_prefetchCDictTables: |
615 | bounds.lowerBound = (int)ZSTD_ps_auto; |
616 | bounds.upperBound = (int)ZSTD_ps_disable; |
617 | return bounds; |
618 | |
619 | case ZSTD_c_enableSeqProducerFallback: |
620 | bounds.lowerBound = 0; |
621 | bounds.upperBound = 1; |
622 | return bounds; |
623 | |
624 | case ZSTD_c_maxBlockSize: |
625 | bounds.lowerBound = ZSTD_BLOCKSIZE_MAX_MIN; |
626 | bounds.upperBound = ZSTD_BLOCKSIZE_MAX; |
627 | return bounds; |
628 | |
629 | case ZSTD_c_searchForExternalRepcodes: |
630 | bounds.lowerBound = (int)ZSTD_ps_auto; |
631 | bounds.upperBound = (int)ZSTD_ps_disable; |
632 | return bounds; |
633 | |
634 | default: |
635 | bounds.error = ERROR(parameter_unsupported); |
636 | return bounds; |
637 | } |
638 | } |
639 | |
640 | /* ZSTD_cParam_clampBounds: |
641 | * Clamps the value into the bounded range. |
642 | */ |
643 | static size_t ZSTD_cParam_clampBounds(ZSTD_cParameter cParam, int* value) |
644 | { |
645 | ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); |
646 | if (ZSTD_isError(bounds.error)) return bounds.error; |
647 | if (*value < bounds.lowerBound) *value = bounds.lowerBound; |
648 | if (*value > bounds.upperBound) *value = bounds.upperBound; |
649 | return 0; |
650 | } |
651 | |
652 | #define BOUNDCHECK(cParam, val) { \ |
653 | RETURN_ERROR_IF(!ZSTD_cParam_withinBounds(cParam,val), \ |
654 | parameter_outOfBound, "Param out of bounds"); \ |
655 | } |
656 | |
657 | |
658 | static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param) |
659 | { |
660 | switch(param) |
661 | { |
662 | case ZSTD_c_compressionLevel: |
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 | return 1; |
670 | |
671 | case ZSTD_c_format: |
672 | case ZSTD_c_windowLog: |
673 | case ZSTD_c_contentSizeFlag: |
674 | case ZSTD_c_checksumFlag: |
675 | case ZSTD_c_dictIDFlag: |
676 | case ZSTD_c_forceMaxWindow : |
677 | case ZSTD_c_nbWorkers: |
678 | case ZSTD_c_jobSize: |
679 | case ZSTD_c_overlapLog: |
680 | case ZSTD_c_rsyncable: |
681 | case ZSTD_c_enableDedicatedDictSearch: |
682 | case ZSTD_c_enableLongDistanceMatching: |
683 | case ZSTD_c_ldmHashLog: |
684 | case ZSTD_c_ldmMinMatch: |
685 | case ZSTD_c_ldmBucketSizeLog: |
686 | case ZSTD_c_ldmHashRateLog: |
687 | case ZSTD_c_forceAttachDict: |
688 | case ZSTD_c_literalCompressionMode: |
689 | case ZSTD_c_targetCBlockSize: |
690 | case ZSTD_c_srcSizeHint: |
691 | case ZSTD_c_stableInBuffer: |
692 | case ZSTD_c_stableOutBuffer: |
693 | case ZSTD_c_blockDelimiters: |
694 | case ZSTD_c_validateSequences: |
695 | case ZSTD_c_useBlockSplitter: |
696 | case ZSTD_c_useRowMatchFinder: |
697 | case ZSTD_c_deterministicRefPrefix: |
698 | case ZSTD_c_prefetchCDictTables: |
699 | case ZSTD_c_enableSeqProducerFallback: |
700 | case ZSTD_c_maxBlockSize: |
701 | case ZSTD_c_searchForExternalRepcodes: |
702 | default: |
703 | return 0; |
704 | } |
705 | } |
706 | |
707 | size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value) |
708 | { |
709 | DEBUGLOG(4, "ZSTD_CCtx_setParameter (%i, %i)" , (int)param, value); |
710 | if (cctx->streamStage != zcss_init) { |
711 | if (ZSTD_isUpdateAuthorized(param)) { |
712 | cctx->cParamsChanged = 1; |
713 | } else { |
714 | RETURN_ERROR(stage_wrong, "can only set params in cctx init stage" ); |
715 | } } |
716 | |
717 | switch(param) |
718 | { |
719 | case ZSTD_c_nbWorkers: |
720 | RETURN_ERROR_IF((value!=0) && cctx->staticSize, parameter_unsupported, |
721 | "MT not compatible with static alloc" ); |
722 | break; |
723 | |
724 | case ZSTD_c_compressionLevel: |
725 | case ZSTD_c_windowLog: |
726 | case ZSTD_c_hashLog: |
727 | case ZSTD_c_chainLog: |
728 | case ZSTD_c_searchLog: |
729 | case ZSTD_c_minMatch: |
730 | case ZSTD_c_targetLength: |
731 | case ZSTD_c_strategy: |
732 | case ZSTD_c_ldmHashRateLog: |
733 | case ZSTD_c_format: |
734 | case ZSTD_c_contentSizeFlag: |
735 | case ZSTD_c_checksumFlag: |
736 | case ZSTD_c_dictIDFlag: |
737 | case ZSTD_c_forceMaxWindow: |
738 | case ZSTD_c_forceAttachDict: |
739 | case ZSTD_c_literalCompressionMode: |
740 | case ZSTD_c_jobSize: |
741 | case ZSTD_c_overlapLog: |
742 | case ZSTD_c_rsyncable: |
743 | case ZSTD_c_enableDedicatedDictSearch: |
744 | case ZSTD_c_enableLongDistanceMatching: |
745 | case ZSTD_c_ldmHashLog: |
746 | case ZSTD_c_ldmMinMatch: |
747 | case ZSTD_c_ldmBucketSizeLog: |
748 | case ZSTD_c_targetCBlockSize: |
749 | case ZSTD_c_srcSizeHint: |
750 | case ZSTD_c_stableInBuffer: |
751 | case ZSTD_c_stableOutBuffer: |
752 | case ZSTD_c_blockDelimiters: |
753 | case ZSTD_c_validateSequences: |
754 | case ZSTD_c_useBlockSplitter: |
755 | case ZSTD_c_useRowMatchFinder: |
756 | case ZSTD_c_deterministicRefPrefix: |
757 | case ZSTD_c_prefetchCDictTables: |
758 | case ZSTD_c_enableSeqProducerFallback: |
759 | case ZSTD_c_maxBlockSize: |
760 | case ZSTD_c_searchForExternalRepcodes: |
761 | break; |
762 | |
763 | default: RETURN_ERROR(parameter_unsupported, "unknown parameter" ); |
764 | } |
765 | return ZSTD_CCtxParams_setParameter(&cctx->requestedParams, param, value); |
766 | } |
767 | |
768 | size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, |
769 | ZSTD_cParameter param, int value) |
770 | { |
771 | DEBUGLOG(4, "ZSTD_CCtxParams_setParameter (%i, %i)" , (int)param, value); |
772 | switch(param) |
773 | { |
774 | case ZSTD_c_format : |
775 | BOUNDCHECK(ZSTD_c_format, value); |
776 | CCtxParams->format = (ZSTD_format_e)value; |
777 | return (size_t)CCtxParams->format; |
778 | |
779 | case ZSTD_c_compressionLevel : { |
780 | FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "" ); |
781 | if (value == 0) |
782 | CCtxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* 0 == default */ |
783 | else |
784 | CCtxParams->compressionLevel = value; |
785 | if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel; |
786 | return 0; /* return type (size_t) cannot represent negative values */ |
787 | } |
788 | |
789 | case ZSTD_c_windowLog : |
790 | if (value!=0) /* 0 => use default */ |
791 | BOUNDCHECK(ZSTD_c_windowLog, value); |
792 | CCtxParams->cParams.windowLog = (U32)value; |
793 | return CCtxParams->cParams.windowLog; |
794 | |
795 | case ZSTD_c_hashLog : |
796 | if (value!=0) /* 0 => use default */ |
797 | BOUNDCHECK(ZSTD_c_hashLog, value); |
798 | CCtxParams->cParams.hashLog = (U32)value; |
799 | return CCtxParams->cParams.hashLog; |
800 | |
801 | case ZSTD_c_chainLog : |
802 | if (value!=0) /* 0 => use default */ |
803 | BOUNDCHECK(ZSTD_c_chainLog, value); |
804 | CCtxParams->cParams.chainLog = (U32)value; |
805 | return CCtxParams->cParams.chainLog; |
806 | |
807 | case ZSTD_c_searchLog : |
808 | if (value!=0) /* 0 => use default */ |
809 | BOUNDCHECK(ZSTD_c_searchLog, value); |
810 | CCtxParams->cParams.searchLog = (U32)value; |
811 | return (size_t)value; |
812 | |
813 | case ZSTD_c_minMatch : |
814 | if (value!=0) /* 0 => use default */ |
815 | BOUNDCHECK(ZSTD_c_minMatch, value); |
816 | CCtxParams->cParams.minMatch = (U32)value; |
817 | return CCtxParams->cParams.minMatch; |
818 | |
819 | case ZSTD_c_targetLength : |
820 | BOUNDCHECK(ZSTD_c_targetLength, value); |
821 | CCtxParams->cParams.targetLength = (U32)value; |
822 | return CCtxParams->cParams.targetLength; |
823 | |
824 | case ZSTD_c_strategy : |
825 | if (value!=0) /* 0 => use default */ |
826 | BOUNDCHECK(ZSTD_c_strategy, value); |
827 | CCtxParams->cParams.strategy = (ZSTD_strategy)value; |
828 | return (size_t)CCtxParams->cParams.strategy; |
829 | |
830 | case ZSTD_c_contentSizeFlag : |
831 | /* Content size written in frame header _when known_ (default:1) */ |
832 | DEBUGLOG(4, "set content size flag = %u" , (value!=0)); |
833 | CCtxParams->fParams.contentSizeFlag = value != 0; |
834 | return (size_t)CCtxParams->fParams.contentSizeFlag; |
835 | |
836 | case ZSTD_c_checksumFlag : |
837 | /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */ |
838 | CCtxParams->fParams.checksumFlag = value != 0; |
839 | return (size_t)CCtxParams->fParams.checksumFlag; |
840 | |
841 | case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */ |
842 | DEBUGLOG(4, "set dictIDFlag = %u" , (value!=0)); |
843 | CCtxParams->fParams.noDictIDFlag = !value; |
844 | return !CCtxParams->fParams.noDictIDFlag; |
845 | |
846 | case ZSTD_c_forceMaxWindow : |
847 | CCtxParams->forceWindow = (value != 0); |
848 | return (size_t)CCtxParams->forceWindow; |
849 | |
850 | case ZSTD_c_forceAttachDict : { |
851 | const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value; |
852 | BOUNDCHECK(ZSTD_c_forceAttachDict, (int)pref); |
853 | CCtxParams->attachDictPref = pref; |
854 | return CCtxParams->attachDictPref; |
855 | } |
856 | |
857 | case ZSTD_c_literalCompressionMode : { |
858 | const ZSTD_paramSwitch_e lcm = (ZSTD_paramSwitch_e)value; |
859 | BOUNDCHECK(ZSTD_c_literalCompressionMode, (int)lcm); |
860 | CCtxParams->literalCompressionMode = lcm; |
861 | return CCtxParams->literalCompressionMode; |
862 | } |
863 | |
864 | case ZSTD_c_nbWorkers : |
865 | #ifndef ZSTD_MULTITHREAD |
866 | RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading" ); |
867 | return 0; |
868 | #else |
869 | FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "" ); |
870 | CCtxParams->nbWorkers = value; |
871 | return CCtxParams->nbWorkers; |
872 | #endif |
873 | |
874 | case ZSTD_c_jobSize : |
875 | #ifndef ZSTD_MULTITHREAD |
876 | RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading" ); |
877 | return 0; |
878 | #else |
879 | /* Adjust to the minimum non-default value. */ |
880 | if (value != 0 && value < ZSTDMT_JOBSIZE_MIN) |
881 | value = ZSTDMT_JOBSIZE_MIN; |
882 | FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "" ); |
883 | assert(value >= 0); |
884 | CCtxParams->jobSize = value; |
885 | return CCtxParams->jobSize; |
886 | #endif |
887 | |
888 | case ZSTD_c_overlapLog : |
889 | #ifndef ZSTD_MULTITHREAD |
890 | RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading" ); |
891 | return 0; |
892 | #else |
893 | FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "" ); |
894 | CCtxParams->overlapLog = value; |
895 | return CCtxParams->overlapLog; |
896 | #endif |
897 | |
898 | case ZSTD_c_rsyncable : |
899 | #ifndef ZSTD_MULTITHREAD |
900 | RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading" ); |
901 | return 0; |
902 | #else |
903 | FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "" ); |
904 | CCtxParams->rsyncable = value; |
905 | return CCtxParams->rsyncable; |
906 | #endif |
907 | |
908 | case ZSTD_c_enableDedicatedDictSearch : |
909 | CCtxParams->enableDedicatedDictSearch = (value!=0); |
910 | return (size_t)CCtxParams->enableDedicatedDictSearch; |
911 | |
912 | case ZSTD_c_enableLongDistanceMatching : |
913 | BOUNDCHECK(ZSTD_c_enableLongDistanceMatching, value); |
914 | CCtxParams->ldmParams.enableLdm = (ZSTD_paramSwitch_e)value; |
915 | return CCtxParams->ldmParams.enableLdm; |
916 | |
917 | case ZSTD_c_ldmHashLog : |
918 | if (value!=0) /* 0 ==> auto */ |
919 | BOUNDCHECK(ZSTD_c_ldmHashLog, value); |
920 | CCtxParams->ldmParams.hashLog = (U32)value; |
921 | return CCtxParams->ldmParams.hashLog; |
922 | |
923 | case ZSTD_c_ldmMinMatch : |
924 | if (value!=0) /* 0 ==> default */ |
925 | BOUNDCHECK(ZSTD_c_ldmMinMatch, value); |
926 | CCtxParams->ldmParams.minMatchLength = (U32)value; |
927 | return CCtxParams->ldmParams.minMatchLength; |
928 | |
929 | case ZSTD_c_ldmBucketSizeLog : |
930 | if (value!=0) /* 0 ==> default */ |
931 | BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value); |
932 | CCtxParams->ldmParams.bucketSizeLog = (U32)value; |
933 | return CCtxParams->ldmParams.bucketSizeLog; |
934 | |
935 | case ZSTD_c_ldmHashRateLog : |
936 | if (value!=0) /* 0 ==> default */ |
937 | BOUNDCHECK(ZSTD_c_ldmHashRateLog, value); |
938 | CCtxParams->ldmParams.hashRateLog = (U32)value; |
939 | return CCtxParams->ldmParams.hashRateLog; |
940 | |
941 | case ZSTD_c_targetCBlockSize : |
942 | if (value!=0) /* 0 ==> default */ |
943 | BOUNDCHECK(ZSTD_c_targetCBlockSize, value); |
944 | CCtxParams->targetCBlockSize = (U32)value; |
945 | return CCtxParams->targetCBlockSize; |
946 | |
947 | case ZSTD_c_srcSizeHint : |
948 | if (value!=0) /* 0 ==> default */ |
949 | BOUNDCHECK(ZSTD_c_srcSizeHint, value); |
950 | CCtxParams->srcSizeHint = value; |
951 | return (size_t)CCtxParams->srcSizeHint; |
952 | |
953 | case ZSTD_c_stableInBuffer: |
954 | BOUNDCHECK(ZSTD_c_stableInBuffer, value); |
955 | CCtxParams->inBufferMode = (ZSTD_bufferMode_e)value; |
956 | return CCtxParams->inBufferMode; |
957 | |
958 | case ZSTD_c_stableOutBuffer: |
959 | BOUNDCHECK(ZSTD_c_stableOutBuffer, value); |
960 | CCtxParams->outBufferMode = (ZSTD_bufferMode_e)value; |
961 | return CCtxParams->outBufferMode; |
962 | |
963 | case ZSTD_c_blockDelimiters: |
964 | BOUNDCHECK(ZSTD_c_blockDelimiters, value); |
965 | CCtxParams->blockDelimiters = (ZSTD_sequenceFormat_e)value; |
966 | return CCtxParams->blockDelimiters; |
967 | |
968 | case ZSTD_c_validateSequences: |
969 | BOUNDCHECK(ZSTD_c_validateSequences, value); |
970 | CCtxParams->validateSequences = value; |
971 | return CCtxParams->validateSequences; |
972 | |
973 | case ZSTD_c_useBlockSplitter: |
974 | BOUNDCHECK(ZSTD_c_useBlockSplitter, value); |
975 | CCtxParams->useBlockSplitter = (ZSTD_paramSwitch_e)value; |
976 | return CCtxParams->useBlockSplitter; |
977 | |
978 | case ZSTD_c_useRowMatchFinder: |
979 | BOUNDCHECK(ZSTD_c_useRowMatchFinder, value); |
980 | CCtxParams->useRowMatchFinder = (ZSTD_paramSwitch_e)value; |
981 | return CCtxParams->useRowMatchFinder; |
982 | |
983 | case ZSTD_c_deterministicRefPrefix: |
984 | BOUNDCHECK(ZSTD_c_deterministicRefPrefix, value); |
985 | CCtxParams->deterministicRefPrefix = !!value; |
986 | return CCtxParams->deterministicRefPrefix; |
987 | |
988 | case ZSTD_c_prefetchCDictTables: |
989 | BOUNDCHECK(ZSTD_c_prefetchCDictTables, value); |
990 | CCtxParams->prefetchCDictTables = (ZSTD_paramSwitch_e)value; |
991 | return CCtxParams->prefetchCDictTables; |
992 | |
993 | case ZSTD_c_enableSeqProducerFallback: |
994 | BOUNDCHECK(ZSTD_c_enableSeqProducerFallback, value); |
995 | CCtxParams->enableMatchFinderFallback = value; |
996 | return CCtxParams->enableMatchFinderFallback; |
997 | |
998 | case ZSTD_c_maxBlockSize: |
999 | if (value!=0) /* 0 ==> default */ |
1000 | BOUNDCHECK(ZSTD_c_maxBlockSize, value); |
1001 | CCtxParams->maxBlockSize = value; |
1002 | return CCtxParams->maxBlockSize; |
1003 | |
1004 | case ZSTD_c_searchForExternalRepcodes: |
1005 | BOUNDCHECK(ZSTD_c_searchForExternalRepcodes, value); |
1006 | CCtxParams->searchForExternalRepcodes = (ZSTD_paramSwitch_e)value; |
1007 | return CCtxParams->searchForExternalRepcodes; |
1008 | |
1009 | default: RETURN_ERROR(parameter_unsupported, "unknown parameter" ); |
1010 | } |
1011 | } |
1012 | |
1013 | size_t ZSTD_CCtx_getParameter(ZSTD_CCtx const* cctx, ZSTD_cParameter param, int* value) |
1014 | { |
1015 | return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value); |
1016 | } |
1017 | |
1018 | size_t ZSTD_CCtxParams_getParameter( |
1019 | ZSTD_CCtx_params const* CCtxParams, ZSTD_cParameter param, int* value) |
1020 | { |
1021 | switch(param) |
1022 | { |
1023 | case ZSTD_c_format : |
1024 | *value = CCtxParams->format; |
1025 | break; |
1026 | case ZSTD_c_compressionLevel : |
1027 | *value = CCtxParams->compressionLevel; |
1028 | break; |
1029 | case ZSTD_c_windowLog : |
1030 | *value = (int)CCtxParams->cParams.windowLog; |
1031 | break; |
1032 | case ZSTD_c_hashLog : |
1033 | *value = (int)CCtxParams->cParams.hashLog; |
1034 | break; |
1035 | case ZSTD_c_chainLog : |
1036 | *value = (int)CCtxParams->cParams.chainLog; |
1037 | break; |
1038 | case ZSTD_c_searchLog : |
1039 | *value = CCtxParams->cParams.searchLog; |
1040 | break; |
1041 | case ZSTD_c_minMatch : |
1042 | *value = CCtxParams->cParams.minMatch; |
1043 | break; |
1044 | case ZSTD_c_targetLength : |
1045 | *value = CCtxParams->cParams.targetLength; |
1046 | break; |
1047 | case ZSTD_c_strategy : |
1048 | *value = (unsigned)CCtxParams->cParams.strategy; |
1049 | break; |
1050 | case ZSTD_c_contentSizeFlag : |
1051 | *value = CCtxParams->fParams.contentSizeFlag; |
1052 | break; |
1053 | case ZSTD_c_checksumFlag : |
1054 | *value = CCtxParams->fParams.checksumFlag; |
1055 | break; |
1056 | case ZSTD_c_dictIDFlag : |
1057 | *value = !CCtxParams->fParams.noDictIDFlag; |
1058 | break; |
1059 | case ZSTD_c_forceMaxWindow : |
1060 | *value = CCtxParams->forceWindow; |
1061 | break; |
1062 | case ZSTD_c_forceAttachDict : |
1063 | *value = CCtxParams->attachDictPref; |
1064 | break; |
1065 | case ZSTD_c_literalCompressionMode : |
1066 | *value = CCtxParams->literalCompressionMode; |
1067 | break; |
1068 | case ZSTD_c_nbWorkers : |
1069 | #ifndef ZSTD_MULTITHREAD |
1070 | assert(CCtxParams->nbWorkers == 0); |
1071 | #endif |
1072 | *value = CCtxParams->nbWorkers; |
1073 | break; |
1074 | case ZSTD_c_jobSize : |
1075 | #ifndef ZSTD_MULTITHREAD |
1076 | RETURN_ERROR(parameter_unsupported, "not compiled with multithreading" ); |
1077 | #else |
1078 | assert(CCtxParams->jobSize <= INT_MAX); |
1079 | *value = (int)CCtxParams->jobSize; |
1080 | break; |
1081 | #endif |
1082 | case ZSTD_c_overlapLog : |
1083 | #ifndef ZSTD_MULTITHREAD |
1084 | RETURN_ERROR(parameter_unsupported, "not compiled with multithreading" ); |
1085 | #else |
1086 | *value = CCtxParams->overlapLog; |
1087 | break; |
1088 | #endif |
1089 | case ZSTD_c_rsyncable : |
1090 | #ifndef ZSTD_MULTITHREAD |
1091 | RETURN_ERROR(parameter_unsupported, "not compiled with multithreading" ); |
1092 | #else |
1093 | *value = CCtxParams->rsyncable; |
1094 | break; |
1095 | #endif |
1096 | case ZSTD_c_enableDedicatedDictSearch : |
1097 | *value = CCtxParams->enableDedicatedDictSearch; |
1098 | break; |
1099 | case ZSTD_c_enableLongDistanceMatching : |
1100 | *value = CCtxParams->ldmParams.enableLdm; |
1101 | break; |
1102 | case ZSTD_c_ldmHashLog : |
1103 | *value = CCtxParams->ldmParams.hashLog; |
1104 | break; |
1105 | case ZSTD_c_ldmMinMatch : |
1106 | *value = CCtxParams->ldmParams.minMatchLength; |
1107 | break; |
1108 | case ZSTD_c_ldmBucketSizeLog : |
1109 | *value = CCtxParams->ldmParams.bucketSizeLog; |
1110 | break; |
1111 | case ZSTD_c_ldmHashRateLog : |
1112 | *value = CCtxParams->ldmParams.hashRateLog; |
1113 | break; |
1114 | case ZSTD_c_targetCBlockSize : |
1115 | *value = (int)CCtxParams->targetCBlockSize; |
1116 | break; |
1117 | case ZSTD_c_srcSizeHint : |
1118 | *value = (int)CCtxParams->srcSizeHint; |
1119 | break; |
1120 | case ZSTD_c_stableInBuffer : |
1121 | *value = (int)CCtxParams->inBufferMode; |
1122 | break; |
1123 | case ZSTD_c_stableOutBuffer : |
1124 | *value = (int)CCtxParams->outBufferMode; |
1125 | break; |
1126 | case ZSTD_c_blockDelimiters : |
1127 | *value = (int)CCtxParams->blockDelimiters; |
1128 | break; |
1129 | case ZSTD_c_validateSequences : |
1130 | *value = (int)CCtxParams->validateSequences; |
1131 | break; |
1132 | case ZSTD_c_useBlockSplitter : |
1133 | *value = (int)CCtxParams->useBlockSplitter; |
1134 | break; |
1135 | case ZSTD_c_useRowMatchFinder : |
1136 | *value = (int)CCtxParams->useRowMatchFinder; |
1137 | break; |
1138 | case ZSTD_c_deterministicRefPrefix: |
1139 | *value = (int)CCtxParams->deterministicRefPrefix; |
1140 | break; |
1141 | case ZSTD_c_prefetchCDictTables: |
1142 | *value = (int)CCtxParams->prefetchCDictTables; |
1143 | break; |
1144 | case ZSTD_c_enableSeqProducerFallback: |
1145 | *value = CCtxParams->enableMatchFinderFallback; |
1146 | break; |
1147 | case ZSTD_c_maxBlockSize: |
1148 | *value = (int)CCtxParams->maxBlockSize; |
1149 | break; |
1150 | case ZSTD_c_searchForExternalRepcodes: |
1151 | *value = (int)CCtxParams->searchForExternalRepcodes; |
1152 | break; |
1153 | default: RETURN_ERROR(parameter_unsupported, "unknown parameter" ); |
1154 | } |
1155 | return 0; |
1156 | } |
1157 | |
1158 | /** ZSTD_CCtx_setParametersUsingCCtxParams() : |
1159 | * just applies `params` into `cctx` |
1160 | * no action is performed, parameters are merely stored. |
1161 | * If ZSTDMT is enabled, parameters are pushed to cctx->mtctx. |
1162 | * This is possible even if a compression is ongoing. |
1163 | * In which case, new parameters will be applied on the fly, starting with next compression job. |
1164 | */ |
1165 | size_t ZSTD_CCtx_setParametersUsingCCtxParams( |
1166 | ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params) |
1167 | { |
1168 | DEBUGLOG(4, "ZSTD_CCtx_setParametersUsingCCtxParams" ); |
1169 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1170 | "The context is in the wrong stage!" ); |
1171 | RETURN_ERROR_IF(cctx->cdict, stage_wrong, |
1172 | "Can't override parameters with cdict attached (some must " |
1173 | "be inherited from the cdict)." ); |
1174 | |
1175 | cctx->requestedParams = *params; |
1176 | return 0; |
1177 | } |
1178 | |
1179 | size_t ZSTD_CCtx_setCParams(ZSTD_CCtx* cctx, ZSTD_compressionParameters cparams) |
1180 | { |
1181 | ZSTD_STATIC_ASSERT(sizeof(cparams) == 7 * 4 /* all params are listed below */); |
1182 | DEBUGLOG(4, "ZSTD_CCtx_setCParams" ); |
1183 | /* only update if all parameters are valid */ |
1184 | FORWARD_IF_ERROR(ZSTD_checkCParams(cparams), "" ); |
1185 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, cparams.windowLog), "" ); |
1186 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_chainLog, cparams.chainLog), "" ); |
1187 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_hashLog, cparams.hashLog), "" ); |
1188 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_searchLog, cparams.searchLog), "" ); |
1189 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_minMatch, cparams.minMatch), "" ); |
1190 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_targetLength, cparams.targetLength), "" ); |
1191 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_strategy, cparams.strategy), "" ); |
1192 | return 0; |
1193 | } |
1194 | |
1195 | size_t ZSTD_CCtx_setFParams(ZSTD_CCtx* cctx, ZSTD_frameParameters fparams) |
1196 | { |
1197 | ZSTD_STATIC_ASSERT(sizeof(fparams) == 3 * 4 /* all params are listed below */); |
1198 | DEBUGLOG(4, "ZSTD_CCtx_setFParams" ); |
1199 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_contentSizeFlag, fparams.contentSizeFlag != 0), "" ); |
1200 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_checksumFlag, fparams.checksumFlag != 0), "" ); |
1201 | FORWARD_IF_ERROR(ZSTD_CCtx_setParameter(cctx, ZSTD_c_dictIDFlag, fparams.noDictIDFlag == 0), "" ); |
1202 | return 0; |
1203 | } |
1204 | |
1205 | size_t ZSTD_CCtx_setParams(ZSTD_CCtx* cctx, ZSTD_parameters params) |
1206 | { |
1207 | DEBUGLOG(4, "ZSTD_CCtx_setParams" ); |
1208 | /* First check cParams, because we want to update all or none. */ |
1209 | FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "" ); |
1210 | /* Next set fParams, because this could fail if the cctx isn't in init stage. */ |
1211 | FORWARD_IF_ERROR(ZSTD_CCtx_setFParams(cctx, params.fParams), "" ); |
1212 | /* Finally set cParams, which should succeed. */ |
1213 | FORWARD_IF_ERROR(ZSTD_CCtx_setCParams(cctx, params.cParams), "" ); |
1214 | return 0; |
1215 | } |
1216 | |
1217 | size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize) |
1218 | { |
1219 | DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %llu bytes" , pledgedSrcSize); |
1220 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1221 | "Can't set pledgedSrcSize when not in init stage." ); |
1222 | cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1; |
1223 | return 0; |
1224 | } |
1225 | |
1226 | static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams( |
1227 | int const compressionLevel, |
1228 | size_t const dictSize); |
1229 | static int ZSTD_dedicatedDictSearch_isSupported( |
1230 | const ZSTD_compressionParameters* cParams); |
1231 | static void ZSTD_dedicatedDictSearch_revertCParams( |
1232 | ZSTD_compressionParameters* cParams); |
1233 | |
1234 | /** |
1235 | * Initializes the local dictionary using requested parameters. |
1236 | * NOTE: Initialization does not employ the pledged src size, |
1237 | * because the dictionary may be used for multiple compressions. |
1238 | */ |
1239 | static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx) |
1240 | { |
1241 | ZSTD_localDict* const dl = &cctx->localDict; |
1242 | if (dl->dict == NULL) { |
1243 | /* No local dictionary. */ |
1244 | assert(dl->dictBuffer == NULL); |
1245 | assert(dl->cdict == NULL); |
1246 | assert(dl->dictSize == 0); |
1247 | return 0; |
1248 | } |
1249 | if (dl->cdict != NULL) { |
1250 | /* Local dictionary already initialized. */ |
1251 | assert(cctx->cdict == dl->cdict); |
1252 | return 0; |
1253 | } |
1254 | assert(dl->dictSize > 0); |
1255 | assert(cctx->cdict == NULL); |
1256 | assert(cctx->prefixDict.dict == NULL); |
1257 | |
1258 | dl->cdict = ZSTD_createCDict_advanced2( |
1259 | dl->dict, |
1260 | dl->dictSize, |
1261 | ZSTD_dlm_byRef, |
1262 | dl->dictContentType, |
1263 | &cctx->requestedParams, |
1264 | cctx->customMem); |
1265 | RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed" ); |
1266 | cctx->cdict = dl->cdict; |
1267 | return 0; |
1268 | } |
1269 | |
1270 | size_t ZSTD_CCtx_loadDictionary_advanced( |
1271 | ZSTD_CCtx* cctx, |
1272 | const void* dict, size_t dictSize, |
1273 | ZSTD_dictLoadMethod_e dictLoadMethod, |
1274 | ZSTD_dictContentType_e dictContentType) |
1275 | { |
1276 | DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)" , (U32)dictSize); |
1277 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1278 | "Can't load a dictionary when cctx is not in init stage." ); |
1279 | ZSTD_clearAllDicts(cctx); /* erase any previously set dictionary */ |
1280 | if (dict == NULL || dictSize == 0) /* no dictionary */ |
1281 | return 0; |
1282 | if (dictLoadMethod == ZSTD_dlm_byRef) { |
1283 | cctx->localDict.dict = dict; |
1284 | } else { |
1285 | /* copy dictionary content inside CCtx to own its lifetime */ |
1286 | void* dictBuffer; |
1287 | RETURN_ERROR_IF(cctx->staticSize, memory_allocation, |
1288 | "static CCtx can't allocate for an internal copy of dictionary" ); |
1289 | dictBuffer = ZSTD_customMalloc(dictSize, cctx->customMem); |
1290 | RETURN_ERROR_IF(dictBuffer==NULL, memory_allocation, |
1291 | "allocation failed for dictionary content" ); |
1292 | ZSTD_memcpy(dictBuffer, dict, dictSize); |
1293 | cctx->localDict.dictBuffer = dictBuffer; /* owned ptr to free */ |
1294 | cctx->localDict.dict = dictBuffer; /* read-only reference */ |
1295 | } |
1296 | cctx->localDict.dictSize = dictSize; |
1297 | cctx->localDict.dictContentType = dictContentType; |
1298 | return 0; |
1299 | } |
1300 | |
1301 | size_t ZSTD_CCtx_loadDictionary_byReference( |
1302 | ZSTD_CCtx* cctx, const void* dict, size_t dictSize) |
1303 | { |
1304 | return ZSTD_CCtx_loadDictionary_advanced( |
1305 | cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto); |
1306 | } |
1307 | |
1308 | size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize) |
1309 | { |
1310 | return ZSTD_CCtx_loadDictionary_advanced( |
1311 | cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto); |
1312 | } |
1313 | |
1314 | |
1315 | size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) |
1316 | { |
1317 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1318 | "Can't ref a dict when ctx not in init stage." ); |
1319 | /* Free the existing local cdict (if any) to save memory. */ |
1320 | ZSTD_clearAllDicts(cctx); |
1321 | cctx->cdict = cdict; |
1322 | return 0; |
1323 | } |
1324 | |
1325 | size_t ZSTD_CCtx_refThreadPool(ZSTD_CCtx* cctx, ZSTD_threadPool* pool) |
1326 | { |
1327 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1328 | "Can't ref a pool when ctx not in init stage." ); |
1329 | cctx->pool = pool; |
1330 | return 0; |
1331 | } |
1332 | |
1333 | size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize) |
1334 | { |
1335 | return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dct_rawContent); |
1336 | } |
1337 | |
1338 | size_t ZSTD_CCtx_refPrefix_advanced( |
1339 | ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType) |
1340 | { |
1341 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1342 | "Can't ref a prefix when ctx not in init stage." ); |
1343 | ZSTD_clearAllDicts(cctx); |
1344 | if (prefix != NULL && prefixSize > 0) { |
1345 | cctx->prefixDict.dict = prefix; |
1346 | cctx->prefixDict.dictSize = prefixSize; |
1347 | cctx->prefixDict.dictContentType = dictContentType; |
1348 | } |
1349 | return 0; |
1350 | } |
1351 | |
1352 | /*! ZSTD_CCtx_reset() : |
1353 | * Also dumps dictionary */ |
1354 | size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset) |
1355 | { |
1356 | if ( (reset == ZSTD_reset_session_only) |
1357 | || (reset == ZSTD_reset_session_and_parameters) ) { |
1358 | cctx->streamStage = zcss_init; |
1359 | cctx->pledgedSrcSizePlusOne = 0; |
1360 | } |
1361 | if ( (reset == ZSTD_reset_parameters) |
1362 | || (reset == ZSTD_reset_session_and_parameters) ) { |
1363 | RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, |
1364 | "Reset parameters is only possible during init stage." ); |
1365 | ZSTD_clearAllDicts(cctx); |
1366 | ZSTD_memset(&cctx->externalMatchCtx, 0, sizeof(cctx->externalMatchCtx)); |
1367 | return ZSTD_CCtxParams_reset(&cctx->requestedParams); |
1368 | } |
1369 | return 0; |
1370 | } |
1371 | |
1372 | |
1373 | /** ZSTD_checkCParams() : |
1374 | control CParam values remain within authorized range. |
1375 | @return : 0, or an error code if one value is beyond authorized range */ |
1376 | size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams) |
1377 | { |
1378 | BOUNDCHECK(ZSTD_c_windowLog, (int)cParams.windowLog); |
1379 | BOUNDCHECK(ZSTD_c_chainLog, (int)cParams.chainLog); |
1380 | BOUNDCHECK(ZSTD_c_hashLog, (int)cParams.hashLog); |
1381 | BOUNDCHECK(ZSTD_c_searchLog, (int)cParams.searchLog); |
1382 | BOUNDCHECK(ZSTD_c_minMatch, (int)cParams.minMatch); |
1383 | BOUNDCHECK(ZSTD_c_targetLength,(int)cParams.targetLength); |
1384 | BOUNDCHECK(ZSTD_c_strategy, cParams.strategy); |
1385 | return 0; |
1386 | } |
1387 | |
1388 | /** ZSTD_clampCParams() : |
1389 | * make CParam values within valid range. |
1390 | * @return : valid CParams */ |
1391 | static ZSTD_compressionParameters |
1392 | ZSTD_clampCParams(ZSTD_compressionParameters cParams) |
1393 | { |
1394 | # define CLAMP_TYPE(cParam, val, type) { \ |
1395 | ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); \ |
1396 | if ((int)val<bounds.lowerBound) val=(type)bounds.lowerBound; \ |
1397 | else if ((int)val>bounds.upperBound) val=(type)bounds.upperBound; \ |
1398 | } |
1399 | # define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, unsigned) |
1400 | CLAMP(ZSTD_c_windowLog, cParams.windowLog); |
1401 | CLAMP(ZSTD_c_chainLog, cParams.chainLog); |
1402 | CLAMP(ZSTD_c_hashLog, cParams.hashLog); |
1403 | CLAMP(ZSTD_c_searchLog, cParams.searchLog); |
1404 | CLAMP(ZSTD_c_minMatch, cParams.minMatch); |
1405 | CLAMP(ZSTD_c_targetLength,cParams.targetLength); |
1406 | CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy); |
1407 | return cParams; |
1408 | } |
1409 | |
1410 | /** ZSTD_cycleLog() : |
1411 | * condition for correct operation : hashLog > 1 */ |
1412 | U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat) |
1413 | { |
1414 | U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2); |
1415 | return hashLog - btScale; |
1416 | } |
1417 | |
1418 | /** ZSTD_dictAndWindowLog() : |
1419 | * Returns an adjusted window log that is large enough to fit the source and the dictionary. |
1420 | * The zstd format says that the entire dictionary is valid if one byte of the dictionary |
1421 | * is within the window. So the hashLog and chainLog should be large enough to reference both |
1422 | * the dictionary and the window. So we must use this adjusted dictAndWindowLog when downsizing |
1423 | * the hashLog and windowLog. |
1424 | * NOTE: srcSize must not be ZSTD_CONTENTSIZE_UNKNOWN. |
1425 | */ |
1426 | static U32 ZSTD_dictAndWindowLog(U32 windowLog, U64 srcSize, U64 dictSize) |
1427 | { |
1428 | const U64 maxWindowSize = 1ULL << ZSTD_WINDOWLOG_MAX; |
1429 | /* No dictionary ==> No change */ |
1430 | if (dictSize == 0) { |
1431 | return windowLog; |
1432 | } |
1433 | assert(windowLog <= ZSTD_WINDOWLOG_MAX); |
1434 | assert(srcSize != ZSTD_CONTENTSIZE_UNKNOWN); /* Handled in ZSTD_adjustCParams_internal() */ |
1435 | { |
1436 | U64 const windowSize = 1ULL << windowLog; |
1437 | U64 const dictAndWindowSize = dictSize + windowSize; |
1438 | /* If the window size is already large enough to fit both the source and the dictionary |
1439 | * then just use the window size. Otherwise adjust so that it fits the dictionary and |
1440 | * the window. |
1441 | */ |
1442 | if (windowSize >= dictSize + srcSize) { |
1443 | return windowLog; /* Window size large enough already */ |
1444 | } else if (dictAndWindowSize >= maxWindowSize) { |
1445 | return ZSTD_WINDOWLOG_MAX; /* Larger than max window log */ |
1446 | } else { |
1447 | return ZSTD_highbit32((U32)dictAndWindowSize - 1) + 1; |
1448 | } |
1449 | } |
1450 | } |
1451 | |
1452 | /** ZSTD_adjustCParams_internal() : |
1453 | * optimize `cPar` for a specified input (`srcSize` and `dictSize`). |
1454 | * mostly downsize to reduce memory consumption and initialization latency. |
1455 | * `srcSize` can be ZSTD_CONTENTSIZE_UNKNOWN when not known. |
1456 | * `mode` is the mode for parameter adjustment. See docs for `ZSTD_cParamMode_e`. |
1457 | * note : `srcSize==0` means 0! |
1458 | * condition : cPar is presumed validated (can be checked using ZSTD_checkCParams()). */ |
1459 | static ZSTD_compressionParameters |
1460 | ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar, |
1461 | unsigned long long srcSize, |
1462 | size_t dictSize, |
1463 | ZSTD_cParamMode_e mode, |
1464 | ZSTD_paramSwitch_e useRowMatchFinder) |
1465 | { |
1466 | const U64 minSrcSize = 513; /* (1<<9) + 1 */ |
1467 | const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1); |
1468 | assert(ZSTD_checkCParams(cPar)==0); |
1469 | |
1470 | switch (mode) { |
1471 | case ZSTD_cpm_unknown: |
1472 | case ZSTD_cpm_noAttachDict: |
1473 | /* If we don't know the source size, don't make any |
1474 | * assumptions about it. We will already have selected |
1475 | * smaller parameters if a dictionary is in use. |
1476 | */ |
1477 | break; |
1478 | case ZSTD_cpm_createCDict: |
1479 | /* Assume a small source size when creating a dictionary |
1480 | * with an unknown source size. |
1481 | */ |
1482 | if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN) |
1483 | srcSize = minSrcSize; |
1484 | break; |
1485 | case ZSTD_cpm_attachDict: |
1486 | /* Dictionary has its own dedicated parameters which have |
1487 | * already been selected. We are selecting parameters |
1488 | * for only the source. |
1489 | */ |
1490 | dictSize = 0; |
1491 | break; |
1492 | default: |
1493 | assert(0); |
1494 | break; |
1495 | } |
1496 | |
1497 | /* resize windowLog if input is small enough, to use less memory */ |
1498 | if ( (srcSize <= maxWindowResize) |
1499 | && (dictSize <= maxWindowResize) ) { |
1500 | U32 const tSize = (U32)(srcSize + dictSize); |
1501 | static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN; |
1502 | U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN : |
1503 | ZSTD_highbit32(tSize-1) + 1; |
1504 | if (cPar.windowLog > srcLog) cPar.windowLog = srcLog; |
1505 | } |
1506 | if (srcSize != ZSTD_CONTENTSIZE_UNKNOWN) { |
1507 | U32 const dictAndWindowLog = ZSTD_dictAndWindowLog(cPar.windowLog, (U64)srcSize, (U64)dictSize); |
1508 | U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy); |
1509 | if (cPar.hashLog > dictAndWindowLog+1) cPar.hashLog = dictAndWindowLog+1; |
1510 | if (cycleLog > dictAndWindowLog) |
1511 | cPar.chainLog -= (cycleLog - dictAndWindowLog); |
1512 | } |
1513 | |
1514 | if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN) |
1515 | cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */ |
1516 | |
1517 | /* We can't use more than 32 bits of hash in total, so that means that we require: |
1518 | * (hashLog + 8) <= 32 && (chainLog + 8) <= 32 |
1519 | */ |
1520 | if (mode == ZSTD_cpm_createCDict && ZSTD_CDictIndicesAreTagged(&cPar)) { |
1521 | U32 const maxShortCacheHashLog = 32 - ZSTD_SHORT_CACHE_TAG_BITS; |
1522 | if (cPar.hashLog > maxShortCacheHashLog) { |
1523 | cPar.hashLog = maxShortCacheHashLog; |
1524 | } |
1525 | if (cPar.chainLog > maxShortCacheHashLog) { |
1526 | cPar.chainLog = maxShortCacheHashLog; |
1527 | } |
1528 | } |
1529 | |
1530 | |
1531 | /* At this point, we aren't 100% sure if we are using the row match finder. |
1532 | * Unless it is explicitly disabled, conservatively assume that it is enabled. |
1533 | * In this case it will only be disabled for small sources, so shrinking the |
1534 | * hash log a little bit shouldn't result in any ratio loss. |
1535 | */ |
1536 | if (useRowMatchFinder == ZSTD_ps_auto) |
1537 | useRowMatchFinder = ZSTD_ps_enable; |
1538 | |
1539 | /* We can't hash more than 32-bits in total. So that means that we require: |
1540 | * (hashLog - rowLog + 8) <= 32 |
1541 | */ |
1542 | if (ZSTD_rowMatchFinderUsed(cPar.strategy, useRowMatchFinder)) { |
1543 | /* Switch to 32-entry rows if searchLog is 5 (or more) */ |
1544 | U32 const rowLog = BOUNDED(4, cPar.searchLog, 6); |
1545 | U32 const maxRowHashLog = 32 - ZSTD_ROW_HASH_TAG_BITS; |
1546 | U32 const maxHashLog = maxRowHashLog + rowLog; |
1547 | assert(cPar.hashLog >= rowLog); |
1548 | if (cPar.hashLog > maxHashLog) { |
1549 | cPar.hashLog = maxHashLog; |
1550 | } |
1551 | } |
1552 | |
1553 | return cPar; |
1554 | } |
1555 | |
1556 | ZSTD_compressionParameters |
1557 | ZSTD_adjustCParams(ZSTD_compressionParameters cPar, |
1558 | unsigned long long srcSize, |
1559 | size_t dictSize) |
1560 | { |
1561 | cPar = ZSTD_clampCParams(cPar); /* resulting cPar is necessarily valid (all parameters within range) */ |
1562 | if (srcSize == 0) srcSize = ZSTD_CONTENTSIZE_UNKNOWN; |
1563 | return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize, ZSTD_cpm_unknown, ZSTD_ps_auto); |
1564 | } |
1565 | |
1566 | static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode); |
1567 | static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode); |
1568 | |
1569 | static void ZSTD_overrideCParams( |
1570 | ZSTD_compressionParameters* cParams, |
1571 | const ZSTD_compressionParameters* overrides) |
1572 | { |
1573 | if (overrides->windowLog) cParams->windowLog = overrides->windowLog; |
1574 | if (overrides->hashLog) cParams->hashLog = overrides->hashLog; |
1575 | if (overrides->chainLog) cParams->chainLog = overrides->chainLog; |
1576 | if (overrides->searchLog) cParams->searchLog = overrides->searchLog; |
1577 | if (overrides->minMatch) cParams->minMatch = overrides->minMatch; |
1578 | if (overrides->targetLength) cParams->targetLength = overrides->targetLength; |
1579 | if (overrides->strategy) cParams->strategy = overrides->strategy; |
1580 | } |
1581 | |
1582 | ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams( |
1583 | const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) |
1584 | { |
1585 | ZSTD_compressionParameters cParams; |
1586 | if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) { |
1587 | srcSizeHint = CCtxParams->srcSizeHint; |
1588 | } |
1589 | cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize, mode); |
1590 | if (CCtxParams->ldmParams.enableLdm == ZSTD_ps_enable) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG; |
1591 | ZSTD_overrideCParams(&cParams, &CCtxParams->cParams); |
1592 | assert(!ZSTD_checkCParams(cParams)); |
1593 | /* srcSizeHint == 0 means 0 */ |
1594 | return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize, mode, CCtxParams->useRowMatchFinder); |
1595 | } |
1596 | |
1597 | static size_t |
1598 | ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams, |
1599 | const ZSTD_paramSwitch_e useRowMatchFinder, |
1600 | const U32 enableDedicatedDictSearch, |
1601 | const U32 forCCtx) |
1602 | { |
1603 | /* chain table size should be 0 for fast or row-hash strategies */ |
1604 | size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, enableDedicatedDictSearch && !forCCtx) |
1605 | ? ((size_t)1 << cParams->chainLog) |
1606 | : 0; |
1607 | size_t const hSize = ((size_t)1) << cParams->hashLog; |
1608 | U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0; |
1609 | size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0; |
1610 | /* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't |
1611 | * surrounded by redzones in ASAN. */ |
1612 | size_t const tableSpace = chainSize * sizeof(U32) |
1613 | + hSize * sizeof(U32) |
1614 | + h3Size * sizeof(U32); |
1615 | size_t const optPotentialSpace = |
1616 | ZSTD_cwksp_aligned_alloc_size((MaxML+1) * sizeof(U32)) |
1617 | + ZSTD_cwksp_aligned_alloc_size((MaxLL+1) * sizeof(U32)) |
1618 | + ZSTD_cwksp_aligned_alloc_size((MaxOff+1) * sizeof(U32)) |
1619 | + ZSTD_cwksp_aligned_alloc_size((1<<Litbits) * sizeof(U32)) |
1620 | + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t)) |
1621 | + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t)); |
1622 | size_t const lazyAdditionalSpace = ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder) |
1623 | ? ZSTD_cwksp_aligned_alloc_size(hSize) |
1624 | : 0; |
1625 | size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt)) |
1626 | ? optPotentialSpace |
1627 | : 0; |
1628 | size_t const slackSpace = ZSTD_cwksp_slack_space_required(); |
1629 | |
1630 | /* tables are guaranteed to be sized in multiples of 64 bytes (or 16 uint32_t) */ |
1631 | ZSTD_STATIC_ASSERT(ZSTD_HASHLOG_MIN >= 4 && ZSTD_WINDOWLOG_MIN >= 4 && ZSTD_CHAINLOG_MIN >= 4); |
1632 | assert(useRowMatchFinder != ZSTD_ps_auto); |
1633 | |
1634 | DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u" , |
1635 | (U32)chainSize, (U32)hSize, (U32)h3Size); |
1636 | return tableSpace + optSpace + slackSpace + lazyAdditionalSpace; |
1637 | } |
1638 | |
1639 | /* Helper function for calculating memory requirements. |
1640 | * Gives a tighter bound than ZSTD_sequenceBound() by taking minMatch into account. */ |
1641 | static size_t ZSTD_maxNbSeq(size_t blockSize, unsigned minMatch, int useSequenceProducer) { |
1642 | U32 const divider = (minMatch==3 || useSequenceProducer) ? 3 : 4; |
1643 | return blockSize / divider; |
1644 | } |
1645 | |
1646 | static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal( |
1647 | const ZSTD_compressionParameters* cParams, |
1648 | const ldmParams_t* ldmParams, |
1649 | const int isStatic, |
1650 | const ZSTD_paramSwitch_e useRowMatchFinder, |
1651 | const size_t buffInSize, |
1652 | const size_t buffOutSize, |
1653 | const U64 pledgedSrcSize, |
1654 | int useSequenceProducer, |
1655 | size_t maxBlockSize) |
1656 | { |
1657 | size_t const windowSize = (size_t) BOUNDED(1ULL, 1ULL << cParams->windowLog, pledgedSrcSize); |
1658 | size_t const blockSize = MIN(ZSTD_resolveMaxBlockSize(maxBlockSize), windowSize); |
1659 | size_t const maxNbSeq = ZSTD_maxNbSeq(blockSize, cParams->minMatch, useSequenceProducer); |
1660 | size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize) |
1661 | + ZSTD_cwksp_aligned_alloc_size(maxNbSeq * sizeof(seqDef)) |
1662 | + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE)); |
1663 | size_t const entropySpace = ZSTD_cwksp_alloc_size(ENTROPY_WORKSPACE_SIZE); |
1664 | size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t)); |
1665 | size_t const matchStateSize = ZSTD_sizeof_matchState(cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 0, /* forCCtx */ 1); |
1666 | |
1667 | size_t const ldmSpace = ZSTD_ldm_getTableSize(*ldmParams); |
1668 | size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(*ldmParams, blockSize); |
1669 | size_t const ldmSeqSpace = ldmParams->enableLdm == ZSTD_ps_enable ? |
1670 | ZSTD_cwksp_aligned_alloc_size(maxNbLdmSeq * sizeof(rawSeq)) : 0; |
1671 | |
1672 | |
1673 | size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) |
1674 | + ZSTD_cwksp_alloc_size(buffOutSize); |
1675 | |
1676 | size_t const cctxSpace = isStatic ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0; |
1677 | |
1678 | size_t const maxNbExternalSeq = ZSTD_sequenceBound(blockSize); |
1679 | size_t const externalSeqSpace = useSequenceProducer |
1680 | ? ZSTD_cwksp_aligned_alloc_size(maxNbExternalSeq * sizeof(ZSTD_Sequence)) |
1681 | : 0; |
1682 | |
1683 | size_t const neededSpace = |
1684 | cctxSpace + |
1685 | entropySpace + |
1686 | blockStateSpace + |
1687 | ldmSpace + |
1688 | ldmSeqSpace + |
1689 | matchStateSize + |
1690 | tokenSpace + |
1691 | bufferSpace + |
1692 | externalSeqSpace; |
1693 | |
1694 | DEBUGLOG(5, "estimate workspace : %u" , (U32)neededSpace); |
1695 | return neededSpace; |
1696 | } |
1697 | |
1698 | size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params) |
1699 | { |
1700 | ZSTD_compressionParameters const cParams = |
1701 | ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); |
1702 | ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, |
1703 | &cParams); |
1704 | |
1705 | RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only." ); |
1706 | /* estimateCCtxSize is for one-shot compression. So no buffers should |
1707 | * be needed. However, we still allocate two 0-sized buffers, which can |
1708 | * take space under ASAN. */ |
1709 | return ZSTD_estimateCCtxSize_usingCCtxParams_internal( |
1710 | &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, 0, 0, ZSTD_CONTENTSIZE_UNKNOWN, params->useSequenceProducer, params->maxBlockSize); |
1711 | } |
1712 | |
1713 | size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams) |
1714 | { |
1715 | ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams); |
1716 | if (ZSTD_rowMatchFinderSupported(cParams.strategy)) { |
1717 | /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */ |
1718 | size_t noRowCCtxSize; |
1719 | size_t rowCCtxSize; |
1720 | initialParams.useRowMatchFinder = ZSTD_ps_disable; |
1721 | noRowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); |
1722 | initialParams.useRowMatchFinder = ZSTD_ps_enable; |
1723 | rowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); |
1724 | return MAX(noRowCCtxSize, rowCCtxSize); |
1725 | } else { |
1726 | return ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); |
1727 | } |
1728 | } |
1729 | |
1730 | static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel) |
1731 | { |
1732 | int tier = 0; |
1733 | size_t largestSize = 0; |
1734 | static const unsigned long long srcSizeTiers[4] = {16 KB, 128 KB, 256 KB, ZSTD_CONTENTSIZE_UNKNOWN}; |
1735 | for (; tier < 4; ++tier) { |
1736 | /* Choose the set of cParams for a given level across all srcSizes that give the largest cctxSize */ |
1737 | ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeTiers[tier], 0, ZSTD_cpm_noAttachDict); |
1738 | largestSize = MAX(ZSTD_estimateCCtxSize_usingCParams(cParams), largestSize); |
1739 | } |
1740 | return largestSize; |
1741 | } |
1742 | |
1743 | size_t ZSTD_estimateCCtxSize(int compressionLevel) |
1744 | { |
1745 | int level; |
1746 | size_t memBudget = 0; |
1747 | for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) { |
1748 | /* Ensure monotonically increasing memory usage as compression level increases */ |
1749 | size_t const newMB = ZSTD_estimateCCtxSize_internal(level); |
1750 | if (newMB > memBudget) memBudget = newMB; |
1751 | } |
1752 | return memBudget; |
1753 | } |
1754 | |
1755 | size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params) |
1756 | { |
1757 | RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only." ); |
1758 | { ZSTD_compressionParameters const cParams = |
1759 | ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); |
1760 | size_t const blockSize = MIN(ZSTD_resolveMaxBlockSize(params->maxBlockSize), (size_t)1 << cParams.windowLog); |
1761 | size_t const inBuffSize = (params->inBufferMode == ZSTD_bm_buffered) |
1762 | ? ((size_t)1 << cParams.windowLog) + blockSize |
1763 | : 0; |
1764 | size_t const outBuffSize = (params->outBufferMode == ZSTD_bm_buffered) |
1765 | ? ZSTD_compressBound(blockSize) + 1 |
1766 | : 0; |
1767 | ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, ¶ms->cParams); |
1768 | |
1769 | return ZSTD_estimateCCtxSize_usingCCtxParams_internal( |
1770 | &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, inBuffSize, outBuffSize, |
1771 | ZSTD_CONTENTSIZE_UNKNOWN, params->useSequenceProducer, params->maxBlockSize); |
1772 | } |
1773 | } |
1774 | |
1775 | size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams) |
1776 | { |
1777 | ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams); |
1778 | if (ZSTD_rowMatchFinderSupported(cParams.strategy)) { |
1779 | /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */ |
1780 | size_t noRowCCtxSize; |
1781 | size_t rowCCtxSize; |
1782 | initialParams.useRowMatchFinder = ZSTD_ps_disable; |
1783 | noRowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); |
1784 | initialParams.useRowMatchFinder = ZSTD_ps_enable; |
1785 | rowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); |
1786 | return MAX(noRowCCtxSize, rowCCtxSize); |
1787 | } else { |
1788 | return ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); |
1789 | } |
1790 | } |
1791 | |
1792 | static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel) |
1793 | { |
1794 | ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); |
1795 | return ZSTD_estimateCStreamSize_usingCParams(cParams); |
1796 | } |
1797 | |
1798 | size_t ZSTD_estimateCStreamSize(int compressionLevel) |
1799 | { |
1800 | int level; |
1801 | size_t memBudget = 0; |
1802 | for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) { |
1803 | size_t const newMB = ZSTD_estimateCStreamSize_internal(level); |
1804 | if (newMB > memBudget) memBudget = newMB; |
1805 | } |
1806 | return memBudget; |
1807 | } |
1808 | |
1809 | /* ZSTD_getFrameProgression(): |
1810 | * tells how much data has been consumed (input) and produced (output) for current frame. |
1811 | * able to count progression inside worker threads (non-blocking mode). |
1812 | */ |
1813 | ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx) |
1814 | { |
1815 | #ifdef ZSTD_MULTITHREAD |
1816 | if (cctx->appliedParams.nbWorkers > 0) { |
1817 | return ZSTDMT_getFrameProgression(cctx->mtctx); |
1818 | } |
1819 | #endif |
1820 | { ZSTD_frameProgression fp; |
1821 | size_t const buffered = (cctx->inBuff == NULL) ? 0 : |
1822 | cctx->inBuffPos - cctx->inToCompress; |
1823 | if (buffered) assert(cctx->inBuffPos >= cctx->inToCompress); |
1824 | assert(buffered <= ZSTD_BLOCKSIZE_MAX); |
1825 | fp.ingested = cctx->consumedSrcSize + buffered; |
1826 | fp.consumed = cctx->consumedSrcSize; |
1827 | fp.produced = cctx->producedCSize; |
1828 | fp.flushed = cctx->producedCSize; /* simplified; some data might still be left within streaming output buffer */ |
1829 | fp.currentJobID = 0; |
1830 | fp.nbActiveWorkers = 0; |
1831 | return fp; |
1832 | } } |
1833 | |
1834 | /*! ZSTD_toFlushNow() |
1835 | * Only useful for multithreading scenarios currently (nbWorkers >= 1). |
1836 | */ |
1837 | size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx) |
1838 | { |
1839 | #ifdef ZSTD_MULTITHREAD |
1840 | if (cctx->appliedParams.nbWorkers > 0) { |
1841 | return ZSTDMT_toFlushNow(cctx->mtctx); |
1842 | } |
1843 | #endif |
1844 | (void)cctx; |
1845 | 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 */ |
1846 | } |
1847 | |
1848 | static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1, |
1849 | ZSTD_compressionParameters cParams2) |
1850 | { |
1851 | (void)cParams1; |
1852 | (void)cParams2; |
1853 | assert(cParams1.windowLog == cParams2.windowLog); |
1854 | assert(cParams1.chainLog == cParams2.chainLog); |
1855 | assert(cParams1.hashLog == cParams2.hashLog); |
1856 | assert(cParams1.searchLog == cParams2.searchLog); |
1857 | assert(cParams1.minMatch == cParams2.minMatch); |
1858 | assert(cParams1.targetLength == cParams2.targetLength); |
1859 | assert(cParams1.strategy == cParams2.strategy); |
1860 | } |
1861 | |
1862 | void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs) |
1863 | { |
1864 | int i; |
1865 | for (i = 0; i < ZSTD_REP_NUM; ++i) |
1866 | bs->rep[i] = repStartValue[i]; |
1867 | bs->entropy.huf.repeatMode = HUF_repeat_none; |
1868 | bs->entropy.fse.offcode_repeatMode = FSE_repeat_none; |
1869 | bs->entropy.fse.matchlength_repeatMode = FSE_repeat_none; |
1870 | bs->entropy.fse.litlength_repeatMode = FSE_repeat_none; |
1871 | } |
1872 | |
1873 | /*! ZSTD_invalidateMatchState() |
1874 | * Invalidate all the matches in the match finder tables. |
1875 | * Requires nextSrc and base to be set (can be NULL). |
1876 | */ |
1877 | static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms) |
1878 | { |
1879 | ZSTD_window_clear(&ms->window); |
1880 | |
1881 | ms->nextToUpdate = ms->window.dictLimit; |
1882 | ms->loadedDictEnd = 0; |
1883 | ms->opt.litLengthSum = 0; /* force reset of btopt stats */ |
1884 | ms->dictMatchState = NULL; |
1885 | } |
1886 | |
1887 | /** |
1888 | * Controls, for this matchState reset, whether the tables need to be cleared / |
1889 | * prepared for the coming compression (ZSTDcrp_makeClean), or whether the |
1890 | * tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a |
1891 | * subsequent operation will overwrite the table space anyways (e.g., copying |
1892 | * the matchState contents in from a CDict). |
1893 | */ |
1894 | typedef enum { |
1895 | ZSTDcrp_makeClean, |
1896 | ZSTDcrp_leaveDirty |
1897 | } ZSTD_compResetPolicy_e; |
1898 | |
1899 | /** |
1900 | * Controls, for this matchState reset, whether indexing can continue where it |
1901 | * left off (ZSTDirp_continue), or whether it needs to be restarted from zero |
1902 | * (ZSTDirp_reset). |
1903 | */ |
1904 | typedef enum { |
1905 | ZSTDirp_continue, |
1906 | ZSTDirp_reset |
1907 | } ZSTD_indexResetPolicy_e; |
1908 | |
1909 | typedef enum { |
1910 | ZSTD_resetTarget_CDict, |
1911 | ZSTD_resetTarget_CCtx |
1912 | } ZSTD_resetTarget_e; |
1913 | |
1914 | /* Mixes bits in a 64 bits in a value, based on XXH3_rrmxmx */ |
1915 | static U64 ZSTD_bitmix(U64 val, U64 len) { |
1916 | val ^= ZSTD_rotateRight_U64(val, 49) ^ ZSTD_rotateRight_U64(val, 24); |
1917 | val *= 0x9FB21C651E98DF25ULL; |
1918 | val ^= (val >> 35) + len ; |
1919 | val *= 0x9FB21C651E98DF25ULL; |
1920 | return val ^ (val >> 28); |
1921 | } |
1922 | |
1923 | /* Mixes in the hashSalt and hashSaltEntropy to create a new hashSalt */ |
1924 | static void ZSTD_advanceHashSalt(ZSTD_matchState_t* ms) { |
1925 | ms->hashSalt = ZSTD_bitmix(ms->hashSalt, 8) ^ ZSTD_bitmix((U64) ms->hashSaltEntropy, 4); |
1926 | } |
1927 | |
1928 | static size_t |
1929 | ZSTD_reset_matchState(ZSTD_matchState_t* ms, |
1930 | ZSTD_cwksp* ws, |
1931 | const ZSTD_compressionParameters* cParams, |
1932 | const ZSTD_paramSwitch_e useRowMatchFinder, |
1933 | const ZSTD_compResetPolicy_e crp, |
1934 | const ZSTD_indexResetPolicy_e forceResetIndex, |
1935 | const ZSTD_resetTarget_e forWho) |
1936 | { |
1937 | /* disable chain table allocation for fast or row-based strategies */ |
1938 | size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, |
1939 | ms->dedicatedDictSearch && (forWho == ZSTD_resetTarget_CDict)) |
1940 | ? ((size_t)1 << cParams->chainLog) |
1941 | : 0; |
1942 | size_t const hSize = ((size_t)1) << cParams->hashLog; |
1943 | U32 const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0; |
1944 | size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0; |
1945 | |
1946 | DEBUGLOG(4, "reset indices : %u" , forceResetIndex == ZSTDirp_reset); |
1947 | assert(useRowMatchFinder != ZSTD_ps_auto); |
1948 | if (forceResetIndex == ZSTDirp_reset) { |
1949 | ZSTD_window_init(&ms->window); |
1950 | ZSTD_cwksp_mark_tables_dirty(ws); |
1951 | } |
1952 | |
1953 | ms->hashLog3 = hashLog3; |
1954 | ms->lazySkipping = 0; |
1955 | |
1956 | ZSTD_invalidateMatchState(ms); |
1957 | |
1958 | assert(!ZSTD_cwksp_reserve_failed(ws)); /* check that allocation hasn't already failed */ |
1959 | |
1960 | ZSTD_cwksp_clear_tables(ws); |
1961 | |
1962 | DEBUGLOG(5, "reserving table space" ); |
1963 | /* table Space */ |
1964 | ms->hashTable = (U32*)ZSTD_cwksp_reserve_table(ws, hSize * sizeof(U32)); |
1965 | ms->chainTable = (U32*)ZSTD_cwksp_reserve_table(ws, chainSize * sizeof(U32)); |
1966 | ms->hashTable3 = (U32*)ZSTD_cwksp_reserve_table(ws, h3Size * sizeof(U32)); |
1967 | RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation, |
1968 | "failed a workspace allocation in ZSTD_reset_matchState" ); |
1969 | |
1970 | DEBUGLOG(4, "reset table : %u" , crp!=ZSTDcrp_leaveDirty); |
1971 | if (crp!=ZSTDcrp_leaveDirty) { |
1972 | /* reset tables only */ |
1973 | ZSTD_cwksp_clean_tables(ws); |
1974 | } |
1975 | |
1976 | if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) { |
1977 | /* Row match finder needs an additional table of hashes ("tags") */ |
1978 | size_t const tagTableSize = hSize; |
1979 | /* We want to generate a new salt in case we reset a Cctx, but we always want to use |
1980 | * 0 when we reset a Cdict */ |
1981 | if(forWho == ZSTD_resetTarget_CCtx) { |
1982 | ms->tagTable = (BYTE*) ZSTD_cwksp_reserve_aligned_init_once(ws, tagTableSize); |
1983 | ZSTD_advanceHashSalt(ms); |
1984 | } else { |
1985 | /* When we are not salting we want to always memset the memory */ |
1986 | ms->tagTable = (BYTE*) ZSTD_cwksp_reserve_aligned(ws, tagTableSize); |
1987 | ZSTD_memset(ms->tagTable, 0, tagTableSize); |
1988 | ms->hashSalt = 0; |
1989 | } |
1990 | { /* Switch to 32-entry rows if searchLog is 5 (or more) */ |
1991 | U32 const rowLog = BOUNDED(4, cParams->searchLog, 6); |
1992 | assert(cParams->hashLog >= rowLog); |
1993 | ms->rowHashLog = cParams->hashLog - rowLog; |
1994 | } |
1995 | } |
1996 | |
1997 | /* opt parser space */ |
1998 | if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) { |
1999 | DEBUGLOG(4, "reserving optimal parser space" ); |
2000 | ms->opt.litFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (1<<Litbits) * sizeof(unsigned)); |
2001 | ms->opt.litLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxLL+1) * sizeof(unsigned)); |
2002 | ms->opt.matchLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxML+1) * sizeof(unsigned)); |
2003 | ms->opt.offCodeFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxOff+1) * sizeof(unsigned)); |
2004 | ms->opt.matchTable = (ZSTD_match_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t)); |
2005 | ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t)); |
2006 | } |
2007 | |
2008 | ms->cParams = *cParams; |
2009 | |
2010 | RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation, |
2011 | "failed a workspace allocation in ZSTD_reset_matchState" ); |
2012 | return 0; |
2013 | } |
2014 | |
2015 | /* ZSTD_indexTooCloseToMax() : |
2016 | * minor optimization : prefer memset() rather than reduceIndex() |
2017 | * which is measurably slow in some circumstances (reported for Visual Studio). |
2018 | * Works when re-using a context for a lot of smallish inputs : |
2019 | * if all inputs are smaller than ZSTD_INDEXOVERFLOW_MARGIN, |
2020 | * memset() will be triggered before reduceIndex(). |
2021 | */ |
2022 | #define ZSTD_INDEXOVERFLOW_MARGIN (16 MB) |
2023 | static int ZSTD_indexTooCloseToMax(ZSTD_window_t w) |
2024 | { |
2025 | return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN); |
2026 | } |
2027 | |
2028 | /** ZSTD_dictTooBig(): |
2029 | * When dictionaries are larger than ZSTD_CHUNKSIZE_MAX they can't be loaded in |
2030 | * one go generically. So we ensure that in that case we reset the tables to zero, |
2031 | * so that we can load as much of the dictionary as possible. |
2032 | */ |
2033 | static int ZSTD_dictTooBig(size_t const loadedDictSize) |
2034 | { |
2035 | return loadedDictSize > ZSTD_CHUNKSIZE_MAX; |
2036 | } |
2037 | |
2038 | /*! ZSTD_resetCCtx_internal() : |
2039 | * @param loadedDictSize The size of the dictionary to be loaded |
2040 | * into the context, if any. If no dictionary is used, or the |
2041 | * dictionary is being attached / copied, then pass 0. |
2042 | * note : `params` are assumed fully validated at this stage. |
2043 | */ |
2044 | static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, |
2045 | ZSTD_CCtx_params const* params, |
2046 | U64 const pledgedSrcSize, |
2047 | size_t const loadedDictSize, |
2048 | ZSTD_compResetPolicy_e const crp, |
2049 | ZSTD_buffered_policy_e const zbuff) |
2050 | { |
2051 | ZSTD_cwksp* const ws = &zc->workspace; |
2052 | DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u, useRowMatchFinder=%d useBlockSplitter=%d" , |
2053 | (U32)pledgedSrcSize, params->cParams.windowLog, (int)params->useRowMatchFinder, (int)params->useBlockSplitter); |
2054 | assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams))); |
2055 | |
2056 | zc->isFirstBlock = 1; |
2057 | |
2058 | /* Set applied params early so we can modify them for LDM, |
2059 | * and point params at the applied params. |
2060 | */ |
2061 | zc->appliedParams = *params; |
2062 | params = &zc->appliedParams; |
2063 | |
2064 | assert(params->useRowMatchFinder != ZSTD_ps_auto); |
2065 | assert(params->useBlockSplitter != ZSTD_ps_auto); |
2066 | assert(params->ldmParams.enableLdm != ZSTD_ps_auto); |
2067 | assert(params->maxBlockSize != 0); |
2068 | if (params->ldmParams.enableLdm == ZSTD_ps_enable) { |
2069 | /* Adjust long distance matching parameters */ |
2070 | ZSTD_ldm_adjustParameters(&zc->appliedParams.ldmParams, ¶ms->cParams); |
2071 | assert(params->ldmParams.hashLog >= params->ldmParams.bucketSizeLog); |
2072 | assert(params->ldmParams.hashRateLog < 32); |
2073 | } |
2074 | |
2075 | { size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params->cParams.windowLog), pledgedSrcSize)); |
2076 | size_t const blockSize = MIN(params->maxBlockSize, windowSize); |
2077 | size_t const maxNbSeq = ZSTD_maxNbSeq(blockSize, params->cParams.minMatch, params->useSequenceProducer); |
2078 | size_t const buffOutSize = (zbuff == ZSTDb_buffered && params->outBufferMode == ZSTD_bm_buffered) |
2079 | ? ZSTD_compressBound(blockSize) + 1 |
2080 | : 0; |
2081 | size_t const buffInSize = (zbuff == ZSTDb_buffered && params->inBufferMode == ZSTD_bm_buffered) |
2082 | ? windowSize + blockSize |
2083 | : 0; |
2084 | size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize); |
2085 | |
2086 | int const indexTooClose = ZSTD_indexTooCloseToMax(zc->blockState.matchState.window); |
2087 | int const dictTooBig = ZSTD_dictTooBig(loadedDictSize); |
2088 | ZSTD_indexResetPolicy_e needsIndexReset = |
2089 | (indexTooClose || dictTooBig || !zc->initialized) ? ZSTDirp_reset : ZSTDirp_continue; |
2090 | |
2091 | size_t const neededSpace = |
2092 | ZSTD_estimateCCtxSize_usingCCtxParams_internal( |
2093 | ¶ms->cParams, ¶ms->ldmParams, zc->staticSize != 0, params->useRowMatchFinder, |
2094 | buffInSize, buffOutSize, pledgedSrcSize, params->useSequenceProducer, params->maxBlockSize); |
2095 | int resizeWorkspace; |
2096 | |
2097 | FORWARD_IF_ERROR(neededSpace, "cctx size estimate failed!" ); |
2098 | |
2099 | if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0); |
2100 | |
2101 | { /* Check if workspace is large enough, alloc a new one if needed */ |
2102 | int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace; |
2103 | int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace); |
2104 | resizeWorkspace = workspaceTooSmall || workspaceWasteful; |
2105 | DEBUGLOG(4, "Need %zu B workspace" , neededSpace); |
2106 | DEBUGLOG(4, "windowSize: %zu - blockSize: %zu" , windowSize, blockSize); |
2107 | |
2108 | if (resizeWorkspace) { |
2109 | DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB" , |
2110 | ZSTD_cwksp_sizeof(ws) >> 10, |
2111 | neededSpace >> 10); |
2112 | |
2113 | RETURN_ERROR_IF(zc->staticSize, memory_allocation, "static cctx : no resize" ); |
2114 | |
2115 | needsIndexReset = ZSTDirp_reset; |
2116 | |
2117 | ZSTD_cwksp_free(ws, zc->customMem); |
2118 | FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem), "" ); |
2119 | |
2120 | DEBUGLOG(5, "reserving object space" ); |
2121 | /* Statically sized space. |
2122 | * entropyWorkspace never moves, |
2123 | * though prev/next block swap places */ |
2124 | assert(ZSTD_cwksp_check_available(ws, 2 * sizeof(ZSTD_compressedBlockState_t))); |
2125 | zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t)); |
2126 | RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock" ); |
2127 | zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t)); |
2128 | RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock" ); |
2129 | zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, ENTROPY_WORKSPACE_SIZE); |
2130 | RETURN_ERROR_IF(zc->entropyWorkspace == NULL, memory_allocation, "couldn't allocate entropyWorkspace" ); |
2131 | } } |
2132 | |
2133 | ZSTD_cwksp_clear(ws); |
2134 | |
2135 | /* init params */ |
2136 | zc->blockState.matchState.cParams = params->cParams; |
2137 | zc->blockState.matchState.prefetchCDictTables = params->prefetchCDictTables == ZSTD_ps_enable; |
2138 | zc->pledgedSrcSizePlusOne = pledgedSrcSize+1; |
2139 | zc->consumedSrcSize = 0; |
2140 | zc->producedCSize = 0; |
2141 | if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN) |
2142 | zc->appliedParams.fParams.contentSizeFlag = 0; |
2143 | DEBUGLOG(4, "pledged content size : %u ; flag : %u" , |
2144 | (unsigned)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag); |
2145 | zc->blockSize = blockSize; |
2146 | |
2147 | XXH64_reset(&zc->xxhState, 0); |
2148 | zc->stage = ZSTDcs_init; |
2149 | zc->dictID = 0; |
2150 | zc->dictContentSize = 0; |
2151 | |
2152 | ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock); |
2153 | |
2154 | FORWARD_IF_ERROR(ZSTD_reset_matchState( |
2155 | &zc->blockState.matchState, |
2156 | ws, |
2157 | ¶ms->cParams, |
2158 | params->useRowMatchFinder, |
2159 | crp, |
2160 | needsIndexReset, |
2161 | ZSTD_resetTarget_CCtx), "" ); |
2162 | |
2163 | zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef)); |
2164 | |
2165 | /* ldm hash table */ |
2166 | if (params->ldmParams.enableLdm == ZSTD_ps_enable) { |
2167 | /* TODO: avoid memset? */ |
2168 | size_t const ldmHSize = ((size_t)1) << params->ldmParams.hashLog; |
2169 | zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t)); |
2170 | ZSTD_memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t)); |
2171 | zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq)); |
2172 | zc->maxNbLdmSequences = maxNbLdmSeq; |
2173 | |
2174 | ZSTD_window_init(&zc->ldmState.window); |
2175 | zc->ldmState.loadedDictEnd = 0; |
2176 | } |
2177 | |
2178 | /* reserve space for block-level external sequences */ |
2179 | if (params->useSequenceProducer) { |
2180 | size_t const maxNbExternalSeq = ZSTD_sequenceBound(blockSize); |
2181 | zc->externalMatchCtx.seqBufferCapacity = maxNbExternalSeq; |
2182 | zc->externalMatchCtx.seqBuffer = |
2183 | (ZSTD_Sequence*)ZSTD_cwksp_reserve_aligned(ws, maxNbExternalSeq * sizeof(ZSTD_Sequence)); |
2184 | } |
2185 | |
2186 | /* buffers */ |
2187 | |
2188 | /* ZSTD_wildcopy() is used to copy into the literals buffer, |
2189 | * so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes. |
2190 | */ |
2191 | zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH); |
2192 | zc->seqStore.maxNbLit = blockSize; |
2193 | |
2194 | zc->bufferedPolicy = zbuff; |
2195 | zc->inBuffSize = buffInSize; |
2196 | zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize); |
2197 | zc->outBuffSize = buffOutSize; |
2198 | zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize); |
2199 | |
2200 | /* ldm bucketOffsets table */ |
2201 | if (params->ldmParams.enableLdm == ZSTD_ps_enable) { |
2202 | /* TODO: avoid memset? */ |
2203 | size_t const numBuckets = |
2204 | ((size_t)1) << (params->ldmParams.hashLog - |
2205 | params->ldmParams.bucketSizeLog); |
2206 | zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets); |
2207 | ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets); |
2208 | } |
2209 | |
2210 | /* sequences storage */ |
2211 | ZSTD_referenceExternalSequences(zc, NULL, 0); |
2212 | zc->seqStore.maxNbSeq = maxNbSeq; |
2213 | zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE)); |
2214 | zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE)); |
2215 | zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE)); |
2216 | |
2217 | DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available" , ZSTD_cwksp_available_space(ws)); |
2218 | assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace)); |
2219 | |
2220 | zc->initialized = 1; |
2221 | |
2222 | return 0; |
2223 | } |
2224 | } |
2225 | |
2226 | /* ZSTD_invalidateRepCodes() : |
2227 | * ensures next compression will not use repcodes from previous block. |
2228 | * Note : only works with regular variant; |
2229 | * do not use with extDict variant ! */ |
2230 | void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) { |
2231 | int i; |
2232 | for (i=0; i<ZSTD_REP_NUM; i++) cctx->blockState.prevCBlock->rep[i] = 0; |
2233 | assert(!ZSTD_window_hasExtDict(cctx->blockState.matchState.window)); |
2234 | } |
2235 | |
2236 | /* These are the approximate sizes for each strategy past which copying the |
2237 | * dictionary tables into the working context is faster than using them |
2238 | * in-place. |
2239 | */ |
2240 | static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = { |
2241 | 8 KB, /* unused */ |
2242 | 8 KB, /* ZSTD_fast */ |
2243 | 16 KB, /* ZSTD_dfast */ |
2244 | 32 KB, /* ZSTD_greedy */ |
2245 | 32 KB, /* ZSTD_lazy */ |
2246 | 32 KB, /* ZSTD_lazy2 */ |
2247 | 32 KB, /* ZSTD_btlazy2 */ |
2248 | 32 KB, /* ZSTD_btopt */ |
2249 | 8 KB, /* ZSTD_btultra */ |
2250 | 8 KB /* ZSTD_btultra2 */ |
2251 | }; |
2252 | |
2253 | static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict, |
2254 | const ZSTD_CCtx_params* params, |
2255 | U64 pledgedSrcSize) |
2256 | { |
2257 | size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy]; |
2258 | int const dedicatedDictSearch = cdict->matchState.dedicatedDictSearch; |
2259 | return dedicatedDictSearch |
2260 | || ( ( pledgedSrcSize <= cutoff |
2261 | || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN |
2262 | || params->attachDictPref == ZSTD_dictForceAttach ) |
2263 | && params->attachDictPref != ZSTD_dictForceCopy |
2264 | && !params->forceWindow ); /* dictMatchState isn't correctly |
2265 | * handled in _enforceMaxDist */ |
2266 | } |
2267 | |
2268 | static size_t |
2269 | ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx, |
2270 | const ZSTD_CDict* cdict, |
2271 | ZSTD_CCtx_params params, |
2272 | U64 pledgedSrcSize, |
2273 | ZSTD_buffered_policy_e zbuff) |
2274 | { |
2275 | DEBUGLOG(4, "ZSTD_resetCCtx_byAttachingCDict() pledgedSrcSize=%llu" , |
2276 | (unsigned long long)pledgedSrcSize); |
2277 | { |
2278 | ZSTD_compressionParameters adjusted_cdict_cParams = cdict->matchState.cParams; |
2279 | unsigned const windowLog = params.cParams.windowLog; |
2280 | assert(windowLog != 0); |
2281 | /* Resize working context table params for input only, since the dict |
2282 | * has its own tables. */ |
2283 | /* pledgedSrcSize == 0 means 0! */ |
2284 | |
2285 | if (cdict->matchState.dedicatedDictSearch) { |
2286 | ZSTD_dedicatedDictSearch_revertCParams(&adjusted_cdict_cParams); |
2287 | } |
2288 | |
2289 | params.cParams = ZSTD_adjustCParams_internal(adjusted_cdict_cParams, pledgedSrcSize, |
2290 | cdict->dictContentSize, ZSTD_cpm_attachDict, |
2291 | params.useRowMatchFinder); |
2292 | params.cParams.windowLog = windowLog; |
2293 | params.useRowMatchFinder = cdict->useRowMatchFinder; /* cdict overrides */ |
2294 | FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize, |
2295 | /* loadedDictSize */ 0, |
2296 | ZSTDcrp_makeClean, zbuff), "" ); |
2297 | assert(cctx->appliedParams.cParams.strategy == adjusted_cdict_cParams.strategy); |
2298 | } |
2299 | |
2300 | { const U32 cdictEnd = (U32)( cdict->matchState.window.nextSrc |
2301 | - cdict->matchState.window.base); |
2302 | const U32 cdictLen = cdictEnd - cdict->matchState.window.dictLimit; |
2303 | if (cdictLen == 0) { |
2304 | /* don't even attach dictionaries with no contents */ |
2305 | DEBUGLOG(4, "skipping attaching empty dictionary" ); |
2306 | } else { |
2307 | DEBUGLOG(4, "attaching dictionary into context" ); |
2308 | cctx->blockState.matchState.dictMatchState = &cdict->matchState; |
2309 | |
2310 | /* prep working match state so dict matches never have negative indices |
2311 | * when they are translated to the working context's index space. */ |
2312 | if (cctx->blockState.matchState.window.dictLimit < cdictEnd) { |
2313 | cctx->blockState.matchState.window.nextSrc = |
2314 | cctx->blockState.matchState.window.base + cdictEnd; |
2315 | ZSTD_window_clear(&cctx->blockState.matchState.window); |
2316 | } |
2317 | /* loadedDictEnd is expressed within the referential of the active context */ |
2318 | cctx->blockState.matchState.loadedDictEnd = cctx->blockState.matchState.window.dictLimit; |
2319 | } } |
2320 | |
2321 | cctx->dictID = cdict->dictID; |
2322 | cctx->dictContentSize = cdict->dictContentSize; |
2323 | |
2324 | /* copy block state */ |
2325 | ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); |
2326 | |
2327 | return 0; |
2328 | } |
2329 | |
2330 | static void ZSTD_copyCDictTableIntoCCtx(U32* dst, U32 const* src, size_t tableSize, |
2331 | ZSTD_compressionParameters const* cParams) { |
2332 | if (ZSTD_CDictIndicesAreTagged(cParams)){ |
2333 | /* Remove tags from the CDict table if they are present. |
2334 | * See docs on "short cache" in zstd_compress_internal.h for context. */ |
2335 | size_t i; |
2336 | for (i = 0; i < tableSize; i++) { |
2337 | U32 const taggedIndex = src[i]; |
2338 | U32 const index = taggedIndex >> ZSTD_SHORT_CACHE_TAG_BITS; |
2339 | dst[i] = index; |
2340 | } |
2341 | } else { |
2342 | ZSTD_memcpy(dst, src, tableSize * sizeof(U32)); |
2343 | } |
2344 | } |
2345 | |
2346 | static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx, |
2347 | const ZSTD_CDict* cdict, |
2348 | ZSTD_CCtx_params params, |
2349 | U64 pledgedSrcSize, |
2350 | ZSTD_buffered_policy_e zbuff) |
2351 | { |
2352 | const ZSTD_compressionParameters *cdict_cParams = &cdict->matchState.cParams; |
2353 | |
2354 | assert(!cdict->matchState.dedicatedDictSearch); |
2355 | DEBUGLOG(4, "ZSTD_resetCCtx_byCopyingCDict() pledgedSrcSize=%llu" , |
2356 | (unsigned long long)pledgedSrcSize); |
2357 | |
2358 | { unsigned const windowLog = params.cParams.windowLog; |
2359 | assert(windowLog != 0); |
2360 | /* Copy only compression parameters related to tables. */ |
2361 | params.cParams = *cdict_cParams; |
2362 | params.cParams.windowLog = windowLog; |
2363 | params.useRowMatchFinder = cdict->useRowMatchFinder; |
2364 | FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize, |
2365 | /* loadedDictSize */ 0, |
2366 | ZSTDcrp_leaveDirty, zbuff), "" ); |
2367 | assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy); |
2368 | assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog); |
2369 | assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog); |
2370 | } |
2371 | |
2372 | ZSTD_cwksp_mark_tables_dirty(&cctx->workspace); |
2373 | assert(params.useRowMatchFinder != ZSTD_ps_auto); |
2374 | |
2375 | /* copy tables */ |
2376 | { size_t const chainSize = ZSTD_allocateChainTable(cdict_cParams->strategy, cdict->useRowMatchFinder, 0 /* DDS guaranteed disabled */) |
2377 | ? ((size_t)1 << cdict_cParams->chainLog) |
2378 | : 0; |
2379 | size_t const hSize = (size_t)1 << cdict_cParams->hashLog; |
2380 | |
2381 | ZSTD_copyCDictTableIntoCCtx(cctx->blockState.matchState.hashTable, |
2382 | cdict->matchState.hashTable, |
2383 | hSize, cdict_cParams); |
2384 | |
2385 | /* Do not copy cdict's chainTable if cctx has parameters such that it would not use chainTable */ |
2386 | if (ZSTD_allocateChainTable(cctx->appliedParams.cParams.strategy, cctx->appliedParams.useRowMatchFinder, 0 /* forDDSDict */)) { |
2387 | ZSTD_copyCDictTableIntoCCtx(cctx->blockState.matchState.chainTable, |
2388 | cdict->matchState.chainTable, |
2389 | chainSize, cdict_cParams); |
2390 | } |
2391 | /* copy tag table */ |
2392 | if (ZSTD_rowMatchFinderUsed(cdict_cParams->strategy, cdict->useRowMatchFinder)) { |
2393 | size_t const tagTableSize = hSize; |
2394 | ZSTD_memcpy(cctx->blockState.matchState.tagTable, |
2395 | cdict->matchState.tagTable, |
2396 | tagTableSize); |
2397 | cctx->blockState.matchState.hashSalt = cdict->matchState.hashSalt; |
2398 | } |
2399 | } |
2400 | |
2401 | /* Zero the hashTable3, since the cdict never fills it */ |
2402 | { int const h3log = cctx->blockState.matchState.hashLog3; |
2403 | size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0; |
2404 | assert(cdict->matchState.hashLog3 == 0); |
2405 | ZSTD_memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32)); |
2406 | } |
2407 | |
2408 | ZSTD_cwksp_mark_tables_clean(&cctx->workspace); |
2409 | |
2410 | /* copy dictionary offsets */ |
2411 | { ZSTD_matchState_t const* srcMatchState = &cdict->matchState; |
2412 | ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState; |
2413 | dstMatchState->window = srcMatchState->window; |
2414 | dstMatchState->nextToUpdate = srcMatchState->nextToUpdate; |
2415 | dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd; |
2416 | } |
2417 | |
2418 | cctx->dictID = cdict->dictID; |
2419 | cctx->dictContentSize = cdict->dictContentSize; |
2420 | |
2421 | /* copy block state */ |
2422 | ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); |
2423 | |
2424 | return 0; |
2425 | } |
2426 | |
2427 | /* We have a choice between copying the dictionary context into the working |
2428 | * context, or referencing the dictionary context from the working context |
2429 | * in-place. We decide here which strategy to use. */ |
2430 | static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx, |
2431 | const ZSTD_CDict* cdict, |
2432 | const ZSTD_CCtx_params* params, |
2433 | U64 pledgedSrcSize, |
2434 | ZSTD_buffered_policy_e zbuff) |
2435 | { |
2436 | |
2437 | DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)" , |
2438 | (unsigned)pledgedSrcSize); |
2439 | |
2440 | if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) { |
2441 | return ZSTD_resetCCtx_byAttachingCDict( |
2442 | cctx, cdict, *params, pledgedSrcSize, zbuff); |
2443 | } else { |
2444 | return ZSTD_resetCCtx_byCopyingCDict( |
2445 | cctx, cdict, *params, pledgedSrcSize, zbuff); |
2446 | } |
2447 | } |
2448 | |
2449 | /*! ZSTD_copyCCtx_internal() : |
2450 | * Duplicate an existing context `srcCCtx` into another one `dstCCtx`. |
2451 | * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()). |
2452 | * The "context", in this case, refers to the hash and chain tables, |
2453 | * entropy tables, and dictionary references. |
2454 | * `windowLog` value is enforced if != 0, otherwise value is copied from srcCCtx. |
2455 | * @return : 0, or an error code */ |
2456 | static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, |
2457 | const ZSTD_CCtx* srcCCtx, |
2458 | ZSTD_frameParameters fParams, |
2459 | U64 pledgedSrcSize, |
2460 | ZSTD_buffered_policy_e zbuff) |
2461 | { |
2462 | RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong, |
2463 | "Can't copy a ctx that's not in init stage." ); |
2464 | DEBUGLOG(5, "ZSTD_copyCCtx_internal" ); |
2465 | ZSTD_memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem)); |
2466 | { ZSTD_CCtx_params params = dstCCtx->requestedParams; |
2467 | /* Copy only compression parameters related to tables. */ |
2468 | params.cParams = srcCCtx->appliedParams.cParams; |
2469 | assert(srcCCtx->appliedParams.useRowMatchFinder != ZSTD_ps_auto); |
2470 | assert(srcCCtx->appliedParams.useBlockSplitter != ZSTD_ps_auto); |
2471 | assert(srcCCtx->appliedParams.ldmParams.enableLdm != ZSTD_ps_auto); |
2472 | params.useRowMatchFinder = srcCCtx->appliedParams.useRowMatchFinder; |
2473 | params.useBlockSplitter = srcCCtx->appliedParams.useBlockSplitter; |
2474 | params.ldmParams = srcCCtx->appliedParams.ldmParams; |
2475 | params.fParams = fParams; |
2476 | params.maxBlockSize = srcCCtx->appliedParams.maxBlockSize; |
2477 | ZSTD_resetCCtx_internal(dstCCtx, ¶ms, pledgedSrcSize, |
2478 | /* loadedDictSize */ 0, |
2479 | ZSTDcrp_leaveDirty, zbuff); |
2480 | assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog); |
2481 | assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy); |
2482 | assert(dstCCtx->appliedParams.cParams.hashLog == srcCCtx->appliedParams.cParams.hashLog); |
2483 | assert(dstCCtx->appliedParams.cParams.chainLog == srcCCtx->appliedParams.cParams.chainLog); |
2484 | assert(dstCCtx->blockState.matchState.hashLog3 == srcCCtx->blockState.matchState.hashLog3); |
2485 | } |
2486 | |
2487 | ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace); |
2488 | |
2489 | /* copy tables */ |
2490 | { size_t const chainSize = ZSTD_allocateChainTable(srcCCtx->appliedParams.cParams.strategy, |
2491 | srcCCtx->appliedParams.useRowMatchFinder, |
2492 | 0 /* forDDSDict */) |
2493 | ? ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog) |
2494 | : 0; |
2495 | size_t const hSize = (size_t)1 << srcCCtx->appliedParams.cParams.hashLog; |
2496 | int const h3log = srcCCtx->blockState.matchState.hashLog3; |
2497 | size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0; |
2498 | |
2499 | ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable, |
2500 | srcCCtx->blockState.matchState.hashTable, |
2501 | hSize * sizeof(U32)); |
2502 | ZSTD_memcpy(dstCCtx->blockState.matchState.chainTable, |
2503 | srcCCtx->blockState.matchState.chainTable, |
2504 | chainSize * sizeof(U32)); |
2505 | ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable3, |
2506 | srcCCtx->blockState.matchState.hashTable3, |
2507 | h3Size * sizeof(U32)); |
2508 | } |
2509 | |
2510 | ZSTD_cwksp_mark_tables_clean(&dstCCtx->workspace); |
2511 | |
2512 | /* copy dictionary offsets */ |
2513 | { |
2514 | const ZSTD_matchState_t* srcMatchState = &srcCCtx->blockState.matchState; |
2515 | ZSTD_matchState_t* dstMatchState = &dstCCtx->blockState.matchState; |
2516 | dstMatchState->window = srcMatchState->window; |
2517 | dstMatchState->nextToUpdate = srcMatchState->nextToUpdate; |
2518 | dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd; |
2519 | } |
2520 | dstCCtx->dictID = srcCCtx->dictID; |
2521 | dstCCtx->dictContentSize = srcCCtx->dictContentSize; |
2522 | |
2523 | /* copy block state */ |
2524 | ZSTD_memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock)); |
2525 | |
2526 | return 0; |
2527 | } |
2528 | |
2529 | /*! ZSTD_copyCCtx() : |
2530 | * Duplicate an existing context `srcCCtx` into another one `dstCCtx`. |
2531 | * Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()). |
2532 | * pledgedSrcSize==0 means "unknown". |
2533 | * @return : 0, or an error code */ |
2534 | size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize) |
2535 | { |
2536 | ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; |
2537 | ZSTD_buffered_policy_e const zbuff = srcCCtx->bufferedPolicy; |
2538 | ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1); |
2539 | if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN; |
2540 | fParams.contentSizeFlag = (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN); |
2541 | |
2542 | return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx, |
2543 | fParams, pledgedSrcSize, |
2544 | zbuff); |
2545 | } |
2546 | |
2547 | |
2548 | #define ZSTD_ROWSIZE 16 |
2549 | /*! ZSTD_reduceTable() : |
2550 | * reduce table indexes by `reducerValue`, or squash to zero. |
2551 | * PreserveMark preserves "unsorted mark" for btlazy2 strategy. |
2552 | * It must be set to a clear 0/1 value, to remove branch during inlining. |
2553 | * Presume table size is a multiple of ZSTD_ROWSIZE |
2554 | * to help auto-vectorization */ |
2555 | FORCE_INLINE_TEMPLATE void |
2556 | ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerValue, int const preserveMark) |
2557 | { |
2558 | int const nbRows = (int)size / ZSTD_ROWSIZE; |
2559 | int cellNb = 0; |
2560 | int rowNb; |
2561 | /* Protect special index values < ZSTD_WINDOW_START_INDEX. */ |
2562 | U32 const reducerThreshold = reducerValue + ZSTD_WINDOW_START_INDEX; |
2563 | assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */ |
2564 | assert(size < (1U<<31)); /* can be casted to int */ |
2565 | |
2566 | #if ZSTD_MEMORY_SANITIZER && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE) |
2567 | /* To validate that the table re-use logic is sound, and that we don't |
2568 | * access table space that we haven't cleaned, we re-"poison" the table |
2569 | * space every time we mark it dirty. |
2570 | * |
2571 | * This function however is intended to operate on those dirty tables and |
2572 | * re-clean them. So when this function is used correctly, we can unpoison |
2573 | * the memory it operated on. This introduces a blind spot though, since |
2574 | * if we now try to operate on __actually__ poisoned memory, we will not |
2575 | * detect that. */ |
2576 | __msan_unpoison(table, size * sizeof(U32)); |
2577 | #endif |
2578 | |
2579 | for (rowNb=0 ; rowNb < nbRows ; rowNb++) { |
2580 | int column; |
2581 | for (column=0; column<ZSTD_ROWSIZE; column++) { |
2582 | U32 newVal; |
2583 | if (preserveMark && table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) { |
2584 | /* This write is pointless, but is required(?) for the compiler |
2585 | * to auto-vectorize the loop. */ |
2586 | newVal = ZSTD_DUBT_UNSORTED_MARK; |
2587 | } else if (table[cellNb] < reducerThreshold) { |
2588 | newVal = 0; |
2589 | } else { |
2590 | newVal = table[cellNb] - reducerValue; |
2591 | } |
2592 | table[cellNb] = newVal; |
2593 | cellNb++; |
2594 | } } |
2595 | } |
2596 | |
2597 | static void ZSTD_reduceTable(U32* const table, U32 const size, U32 const reducerValue) |
2598 | { |
2599 | ZSTD_reduceTable_internal(table, size, reducerValue, 0); |
2600 | } |
2601 | |
2602 | static void ZSTD_reduceTable_btlazy2(U32* const table, U32 const size, U32 const reducerValue) |
2603 | { |
2604 | ZSTD_reduceTable_internal(table, size, reducerValue, 1); |
2605 | } |
2606 | |
2607 | /*! ZSTD_reduceIndex() : |
2608 | * rescale all indexes to avoid future overflow (indexes are U32) */ |
2609 | static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* params, const U32 reducerValue) |
2610 | { |
2611 | { U32 const hSize = (U32)1 << params->cParams.hashLog; |
2612 | ZSTD_reduceTable(ms->hashTable, hSize, reducerValue); |
2613 | } |
2614 | |
2615 | if (ZSTD_allocateChainTable(params->cParams.strategy, params->useRowMatchFinder, (U32)ms->dedicatedDictSearch)) { |
2616 | U32 const chainSize = (U32)1 << params->cParams.chainLog; |
2617 | if (params->cParams.strategy == ZSTD_btlazy2) |
2618 | ZSTD_reduceTable_btlazy2(ms->chainTable, chainSize, reducerValue); |
2619 | else |
2620 | ZSTD_reduceTable(ms->chainTable, chainSize, reducerValue); |
2621 | } |
2622 | |
2623 | if (ms->hashLog3) { |
2624 | U32 const h3Size = (U32)1 << ms->hashLog3; |
2625 | ZSTD_reduceTable(ms->hashTable3, h3Size, reducerValue); |
2626 | } |
2627 | } |
2628 | |
2629 | |
2630 | /*-******************************************************* |
2631 | * Block entropic compression |
2632 | *********************************************************/ |
2633 | |
2634 | /* See doc/zstd_compression_format.md for detailed format description */ |
2635 | |
2636 | int ZSTD_seqToCodes(const seqStore_t* seqStorePtr) |
2637 | { |
2638 | const seqDef* const sequences = seqStorePtr->sequencesStart; |
2639 | BYTE* const llCodeTable = seqStorePtr->llCode; |
2640 | BYTE* const ofCodeTable = seqStorePtr->ofCode; |
2641 | BYTE* const mlCodeTable = seqStorePtr->mlCode; |
2642 | U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart); |
2643 | U32 u; |
2644 | int longOffsets = 0; |
2645 | assert(nbSeq <= seqStorePtr->maxNbSeq); |
2646 | for (u=0; u<nbSeq; u++) { |
2647 | U32 const llv = sequences[u].litLength; |
2648 | U32 const ofCode = ZSTD_highbit32(sequences[u].offBase); |
2649 | U32 const mlv = sequences[u].mlBase; |
2650 | llCodeTable[u] = (BYTE)ZSTD_LLcode(llv); |
2651 | ofCodeTable[u] = (BYTE)ofCode; |
2652 | mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv); |
2653 | assert(!(MEM_64bits() && ofCode >= STREAM_ACCUMULATOR_MIN)); |
2654 | if (MEM_32bits() && ofCode >= STREAM_ACCUMULATOR_MIN) |
2655 | longOffsets = 1; |
2656 | } |
2657 | if (seqStorePtr->longLengthType==ZSTD_llt_literalLength) |
2658 | llCodeTable[seqStorePtr->longLengthPos] = MaxLL; |
2659 | if (seqStorePtr->longLengthType==ZSTD_llt_matchLength) |
2660 | mlCodeTable[seqStorePtr->longLengthPos] = MaxML; |
2661 | return longOffsets; |
2662 | } |
2663 | |
2664 | /* ZSTD_useTargetCBlockSize(): |
2665 | * Returns if target compressed block size param is being used. |
2666 | * If used, compression will do best effort to make a compressed block size to be around targetCBlockSize. |
2667 | * Returns 1 if true, 0 otherwise. */ |
2668 | static int ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params* cctxParams) |
2669 | { |
2670 | DEBUGLOG(5, "ZSTD_useTargetCBlockSize (targetCBlockSize=%zu)" , cctxParams->targetCBlockSize); |
2671 | return (cctxParams->targetCBlockSize != 0); |
2672 | } |
2673 | |
2674 | /* ZSTD_blockSplitterEnabled(): |
2675 | * Returns if block splitting param is being used |
2676 | * If used, compression will do best effort to split a block in order to improve compression ratio. |
2677 | * At the time this function is called, the parameter must be finalized. |
2678 | * Returns 1 if true, 0 otherwise. */ |
2679 | static int ZSTD_blockSplitterEnabled(ZSTD_CCtx_params* cctxParams) |
2680 | { |
2681 | DEBUGLOG(5, "ZSTD_blockSplitterEnabled (useBlockSplitter=%d)" , cctxParams->useBlockSplitter); |
2682 | assert(cctxParams->useBlockSplitter != ZSTD_ps_auto); |
2683 | return (cctxParams->useBlockSplitter == ZSTD_ps_enable); |
2684 | } |
2685 | |
2686 | /* Type returned by ZSTD_buildSequencesStatistics containing finalized symbol encoding types |
2687 | * and size of the sequences statistics |
2688 | */ |
2689 | typedef struct { |
2690 | U32 LLtype; |
2691 | U32 Offtype; |
2692 | U32 MLtype; |
2693 | size_t size; |
2694 | size_t lastCountSize; /* Accounts for bug in 1.3.4. More detail in ZSTD_entropyCompressSeqStore_internal() */ |
2695 | int longOffsets; |
2696 | } ZSTD_symbolEncodingTypeStats_t; |
2697 | |
2698 | /* ZSTD_buildSequencesStatistics(): |
2699 | * Returns a ZSTD_symbolEncodingTypeStats_t, or a zstd error code in the `size` field. |
2700 | * Modifies `nextEntropy` to have the appropriate values as a side effect. |
2701 | * nbSeq must be greater than 0. |
2702 | * |
2703 | * entropyWkspSize must be of size at least ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1)*sizeof(U32) |
2704 | */ |
2705 | static ZSTD_symbolEncodingTypeStats_t |
2706 | ZSTD_buildSequencesStatistics( |
2707 | const seqStore_t* seqStorePtr, size_t nbSeq, |
2708 | const ZSTD_fseCTables_t* prevEntropy, ZSTD_fseCTables_t* nextEntropy, |
2709 | BYTE* dst, const BYTE* const dstEnd, |
2710 | ZSTD_strategy strategy, unsigned* countWorkspace, |
2711 | void* entropyWorkspace, size_t entropyWkspSize) |
2712 | { |
2713 | BYTE* const ostart = dst; |
2714 | const BYTE* const oend = dstEnd; |
2715 | BYTE* op = ostart; |
2716 | FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable; |
2717 | FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable; |
2718 | FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable; |
2719 | const BYTE* const ofCodeTable = seqStorePtr->ofCode; |
2720 | const BYTE* const llCodeTable = seqStorePtr->llCode; |
2721 | const BYTE* const mlCodeTable = seqStorePtr->mlCode; |
2722 | ZSTD_symbolEncodingTypeStats_t stats; |
2723 | |
2724 | stats.lastCountSize = 0; |
2725 | /* convert length/distances into codes */ |
2726 | stats.longOffsets = ZSTD_seqToCodes(seqStorePtr); |
2727 | assert(op <= oend); |
2728 | assert(nbSeq != 0); /* ZSTD_selectEncodingType() divides by nbSeq */ |
2729 | /* build CTable for Literal Lengths */ |
2730 | { unsigned max = MaxLL; |
2731 | size_t const mostFrequent = HIST_countFast_wksp(countWorkspace, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ |
2732 | DEBUGLOG(5, "Building LL table" ); |
2733 | nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode; |
2734 | stats.LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode, |
2735 | countWorkspace, max, mostFrequent, nbSeq, |
2736 | LLFSELog, prevEntropy->litlengthCTable, |
2737 | LL_defaultNorm, LL_defaultNormLog, |
2738 | ZSTD_defaultAllowed, strategy); |
2739 | assert(set_basic < set_compressed && set_rle < set_compressed); |
2740 | assert(!(stats.LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ |
2741 | { size_t const countSize = ZSTD_buildCTable( |
2742 | op, (size_t)(oend - op), |
2743 | CTable_LitLength, LLFSELog, (symbolEncodingType_e)stats.LLtype, |
2744 | countWorkspace, max, llCodeTable, nbSeq, |
2745 | LL_defaultNorm, LL_defaultNormLog, MaxLL, |
2746 | prevEntropy->litlengthCTable, |
2747 | sizeof(prevEntropy->litlengthCTable), |
2748 | entropyWorkspace, entropyWkspSize); |
2749 | if (ZSTD_isError(countSize)) { |
2750 | DEBUGLOG(3, "ZSTD_buildCTable for LitLens failed" ); |
2751 | stats.size = countSize; |
2752 | return stats; |
2753 | } |
2754 | if (stats.LLtype == set_compressed) |
2755 | stats.lastCountSize = countSize; |
2756 | op += countSize; |
2757 | assert(op <= oend); |
2758 | } } |
2759 | /* build CTable for Offsets */ |
2760 | { unsigned max = MaxOff; |
2761 | size_t const mostFrequent = HIST_countFast_wksp( |
2762 | countWorkspace, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ |
2763 | /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */ |
2764 | ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed; |
2765 | DEBUGLOG(5, "Building OF table" ); |
2766 | nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode; |
2767 | stats.Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode, |
2768 | countWorkspace, max, mostFrequent, nbSeq, |
2769 | OffFSELog, prevEntropy->offcodeCTable, |
2770 | OF_defaultNorm, OF_defaultNormLog, |
2771 | defaultPolicy, strategy); |
2772 | assert(!(stats.Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */ |
2773 | { size_t const countSize = ZSTD_buildCTable( |
2774 | op, (size_t)(oend - op), |
2775 | CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)stats.Offtype, |
2776 | countWorkspace, max, ofCodeTable, nbSeq, |
2777 | OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff, |
2778 | prevEntropy->offcodeCTable, |
2779 | sizeof(prevEntropy->offcodeCTable), |
2780 | entropyWorkspace, entropyWkspSize); |
2781 | if (ZSTD_isError(countSize)) { |
2782 | DEBUGLOG(3, "ZSTD_buildCTable for Offsets failed" ); |
2783 | stats.size = countSize; |
2784 | return stats; |
2785 | } |
2786 | if (stats.Offtype == set_compressed) |
2787 | stats.lastCountSize = countSize; |
2788 | op += countSize; |
2789 | assert(op <= oend); |
2790 | } } |
2791 | /* build CTable for MatchLengths */ |
2792 | { unsigned max = MaxML; |
2793 | size_t const mostFrequent = HIST_countFast_wksp( |
2794 | countWorkspace, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ |
2795 | DEBUGLOG(5, "Building ML table (remaining space : %i)" , (int)(oend-op)); |
2796 | nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode; |
2797 | stats.MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode, |
2798 | countWorkspace, max, mostFrequent, nbSeq, |
2799 | MLFSELog, prevEntropy->matchlengthCTable, |
2800 | ML_defaultNorm, ML_defaultNormLog, |
2801 | ZSTD_defaultAllowed, strategy); |
2802 | assert(!(stats.MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ |
2803 | { size_t const countSize = ZSTD_buildCTable( |
2804 | op, (size_t)(oend - op), |
2805 | CTable_MatchLength, MLFSELog, (symbolEncodingType_e)stats.MLtype, |
2806 | countWorkspace, max, mlCodeTable, nbSeq, |
2807 | ML_defaultNorm, ML_defaultNormLog, MaxML, |
2808 | prevEntropy->matchlengthCTable, |
2809 | sizeof(prevEntropy->matchlengthCTable), |
2810 | entropyWorkspace, entropyWkspSize); |
2811 | if (ZSTD_isError(countSize)) { |
2812 | DEBUGLOG(3, "ZSTD_buildCTable for MatchLengths failed" ); |
2813 | stats.size = countSize; |
2814 | return stats; |
2815 | } |
2816 | if (stats.MLtype == set_compressed) |
2817 | stats.lastCountSize = countSize; |
2818 | op += countSize; |
2819 | assert(op <= oend); |
2820 | } } |
2821 | stats.size = (size_t)(op-ostart); |
2822 | return stats; |
2823 | } |
2824 | |
2825 | /* ZSTD_entropyCompressSeqStore_internal(): |
2826 | * compresses both literals and sequences |
2827 | * Returns compressed size of block, or a zstd error. |
2828 | */ |
2829 | #define SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO 20 |
2830 | MEM_STATIC size_t |
2831 | ZSTD_entropyCompressSeqStore_internal( |
2832 | const seqStore_t* seqStorePtr, |
2833 | const ZSTD_entropyCTables_t* prevEntropy, |
2834 | ZSTD_entropyCTables_t* nextEntropy, |
2835 | const ZSTD_CCtx_params* cctxParams, |
2836 | void* dst, size_t dstCapacity, |
2837 | void* entropyWorkspace, size_t entropyWkspSize, |
2838 | const int bmi2) |
2839 | { |
2840 | ZSTD_strategy const strategy = cctxParams->cParams.strategy; |
2841 | unsigned* count = (unsigned*)entropyWorkspace; |
2842 | FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable; |
2843 | FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable; |
2844 | FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable; |
2845 | const seqDef* const sequences = seqStorePtr->sequencesStart; |
2846 | const size_t nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); |
2847 | const BYTE* const ofCodeTable = seqStorePtr->ofCode; |
2848 | const BYTE* const llCodeTable = seqStorePtr->llCode; |
2849 | const BYTE* const mlCodeTable = seqStorePtr->mlCode; |
2850 | BYTE* const ostart = (BYTE*)dst; |
2851 | BYTE* const oend = ostart + dstCapacity; |
2852 | BYTE* op = ostart; |
2853 | size_t lastCountSize; |
2854 | int longOffsets = 0; |
2855 | |
2856 | entropyWorkspace = count + (MaxSeq + 1); |
2857 | entropyWkspSize -= (MaxSeq + 1) * sizeof(*count); |
2858 | |
2859 | DEBUGLOG(5, "ZSTD_entropyCompressSeqStore_internal (nbSeq=%zu, dstCapacity=%zu)" , nbSeq, dstCapacity); |
2860 | ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog))); |
2861 | assert(entropyWkspSize >= HUF_WORKSPACE_SIZE); |
2862 | |
2863 | /* Compress literals */ |
2864 | { const BYTE* const literals = seqStorePtr->litStart; |
2865 | size_t const numSequences = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); |
2866 | size_t const numLiterals = (size_t)(seqStorePtr->lit - seqStorePtr->litStart); |
2867 | /* Base suspicion of uncompressibility on ratio of literals to sequences */ |
2868 | unsigned const suspectUncompressible = (numSequences == 0) || (numLiterals / numSequences >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO); |
2869 | size_t const litSize = (size_t)(seqStorePtr->lit - literals); |
2870 | |
2871 | size_t const cSize = ZSTD_compressLiterals( |
2872 | op, dstCapacity, |
2873 | literals, litSize, |
2874 | entropyWorkspace, entropyWkspSize, |
2875 | &prevEntropy->huf, &nextEntropy->huf, |
2876 | cctxParams->cParams.strategy, |
2877 | ZSTD_literalsCompressionIsDisabled(cctxParams), |
2878 | suspectUncompressible, bmi2); |
2879 | FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed" ); |
2880 | assert(cSize <= dstCapacity); |
2881 | op += cSize; |
2882 | } |
2883 | |
2884 | /* Sequences Header */ |
2885 | RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/, |
2886 | dstSize_tooSmall, "Can't fit seq hdr in output buf!" ); |
2887 | if (nbSeq < 128) { |
2888 | *op++ = (BYTE)nbSeq; |
2889 | } else if (nbSeq < LONGNBSEQ) { |
2890 | op[0] = (BYTE)((nbSeq>>8) + 0x80); |
2891 | op[1] = (BYTE)nbSeq; |
2892 | op+=2; |
2893 | } else { |
2894 | op[0]=0xFF; |
2895 | MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)); |
2896 | op+=3; |
2897 | } |
2898 | assert(op <= oend); |
2899 | if (nbSeq==0) { |
2900 | /* Copy the old tables over as if we repeated them */ |
2901 | ZSTD_memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse)); |
2902 | return (size_t)(op - ostart); |
2903 | } |
2904 | { BYTE* const seqHead = op++; |
2905 | /* build stats for sequences */ |
2906 | const ZSTD_symbolEncodingTypeStats_t stats = |
2907 | ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq, |
2908 | &prevEntropy->fse, &nextEntropy->fse, |
2909 | op, oend, |
2910 | strategy, count, |
2911 | entropyWorkspace, entropyWkspSize); |
2912 | FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!" ); |
2913 | *seqHead = (BYTE)((stats.LLtype<<6) + (stats.Offtype<<4) + (stats.MLtype<<2)); |
2914 | lastCountSize = stats.lastCountSize; |
2915 | op += stats.size; |
2916 | longOffsets = stats.longOffsets; |
2917 | } |
2918 | |
2919 | { size_t const bitstreamSize = ZSTD_encodeSequences( |
2920 | op, (size_t)(oend - op), |
2921 | CTable_MatchLength, mlCodeTable, |
2922 | CTable_OffsetBits, ofCodeTable, |
2923 | CTable_LitLength, llCodeTable, |
2924 | sequences, nbSeq, |
2925 | longOffsets, bmi2); |
2926 | FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed" ); |
2927 | op += bitstreamSize; |
2928 | assert(op <= oend); |
2929 | /* zstd versions <= 1.3.4 mistakenly report corruption when |
2930 | * FSE_readNCount() receives a buffer < 4 bytes. |
2931 | * Fixed by https://github.com/facebook/zstd/pull/1146. |
2932 | * This can happen when the last set_compressed table present is 2 |
2933 | * bytes and the bitstream is only one byte. |
2934 | * In this exceedingly rare case, we will simply emit an uncompressed |
2935 | * block, since it isn't worth optimizing. |
2936 | */ |
2937 | if (lastCountSize && (lastCountSize + bitstreamSize) < 4) { |
2938 | /* lastCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */ |
2939 | assert(lastCountSize + bitstreamSize == 3); |
2940 | DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by " |
2941 | "emitting an uncompressed block." ); |
2942 | return 0; |
2943 | } |
2944 | } |
2945 | |
2946 | DEBUGLOG(5, "compressed block size : %u" , (unsigned)(op - ostart)); |
2947 | return (size_t)(op - ostart); |
2948 | } |
2949 | |
2950 | MEM_STATIC size_t |
2951 | ZSTD_entropyCompressSeqStore( |
2952 | const seqStore_t* seqStorePtr, |
2953 | const ZSTD_entropyCTables_t* prevEntropy, |
2954 | ZSTD_entropyCTables_t* nextEntropy, |
2955 | const ZSTD_CCtx_params* cctxParams, |
2956 | void* dst, size_t dstCapacity, |
2957 | size_t srcSize, |
2958 | void* entropyWorkspace, size_t entropyWkspSize, |
2959 | int bmi2) |
2960 | { |
2961 | size_t const cSize = ZSTD_entropyCompressSeqStore_internal( |
2962 | seqStorePtr, prevEntropy, nextEntropy, cctxParams, |
2963 | dst, dstCapacity, |
2964 | entropyWorkspace, entropyWkspSize, bmi2); |
2965 | if (cSize == 0) return 0; |
2966 | /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block. |
2967 | * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block. |
2968 | */ |
2969 | if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity)) { |
2970 | DEBUGLOG(4, "not enough dstCapacity (%zu) for ZSTD_entropyCompressSeqStore_internal()=> do not compress block" , dstCapacity); |
2971 | return 0; /* block not compressed */ |
2972 | } |
2973 | FORWARD_IF_ERROR(cSize, "ZSTD_entropyCompressSeqStore_internal failed" ); |
2974 | |
2975 | /* Check compressibility */ |
2976 | { size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy); |
2977 | if (cSize >= maxCSize) return 0; /* block not compressed */ |
2978 | } |
2979 | DEBUGLOG(5, "ZSTD_entropyCompressSeqStore() cSize: %zu" , cSize); |
2980 | /* libzstd decoder before > v1.5.4 is not compatible with compressed blocks of size ZSTD_BLOCKSIZE_MAX exactly. |
2981 | * This restriction is indirectly already fulfilled by respecting ZSTD_minGain() condition above. |
2982 | */ |
2983 | assert(cSize < ZSTD_BLOCKSIZE_MAX); |
2984 | return cSize; |
2985 | } |
2986 | |
2987 | /* ZSTD_selectBlockCompressor() : |
2988 | * Not static, but internal use only (used by long distance matcher) |
2989 | * assumption : strat is a valid strategy */ |
2990 | ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_paramSwitch_e useRowMatchFinder, ZSTD_dictMode_e dictMode) |
2991 | { |
2992 | static const ZSTD_blockCompressor blockCompressor[4][ZSTD_STRATEGY_MAX+1] = { |
2993 | { ZSTD_compressBlock_fast /* default for 0 */, |
2994 | ZSTD_compressBlock_fast, |
2995 | ZSTD_compressBlock_doubleFast, |
2996 | ZSTD_compressBlock_greedy, |
2997 | ZSTD_compressBlock_lazy, |
2998 | ZSTD_compressBlock_lazy2, |
2999 | ZSTD_compressBlock_btlazy2, |
3000 | ZSTD_compressBlock_btopt, |
3001 | ZSTD_compressBlock_btultra, |
3002 | ZSTD_compressBlock_btultra2 }, |
3003 | { ZSTD_compressBlock_fast_extDict /* default for 0 */, |
3004 | ZSTD_compressBlock_fast_extDict, |
3005 | ZSTD_compressBlock_doubleFast_extDict, |
3006 | ZSTD_compressBlock_greedy_extDict, |
3007 | ZSTD_compressBlock_lazy_extDict, |
3008 | ZSTD_compressBlock_lazy2_extDict, |
3009 | ZSTD_compressBlock_btlazy2_extDict, |
3010 | ZSTD_compressBlock_btopt_extDict, |
3011 | ZSTD_compressBlock_btultra_extDict, |
3012 | ZSTD_compressBlock_btultra_extDict }, |
3013 | { ZSTD_compressBlock_fast_dictMatchState /* default for 0 */, |
3014 | ZSTD_compressBlock_fast_dictMatchState, |
3015 | ZSTD_compressBlock_doubleFast_dictMatchState, |
3016 | ZSTD_compressBlock_greedy_dictMatchState, |
3017 | ZSTD_compressBlock_lazy_dictMatchState, |
3018 | ZSTD_compressBlock_lazy2_dictMatchState, |
3019 | ZSTD_compressBlock_btlazy2_dictMatchState, |
3020 | ZSTD_compressBlock_btopt_dictMatchState, |
3021 | ZSTD_compressBlock_btultra_dictMatchState, |
3022 | ZSTD_compressBlock_btultra_dictMatchState }, |
3023 | { NULL /* default for 0 */, |
3024 | NULL, |
3025 | NULL, |
3026 | ZSTD_compressBlock_greedy_dedicatedDictSearch, |
3027 | ZSTD_compressBlock_lazy_dedicatedDictSearch, |
3028 | ZSTD_compressBlock_lazy2_dedicatedDictSearch, |
3029 | NULL, |
3030 | NULL, |
3031 | NULL, |
3032 | NULL } |
3033 | }; |
3034 | ZSTD_blockCompressor selectedCompressor; |
3035 | ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1); |
3036 | |
3037 | assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat)); |
3038 | DEBUGLOG(4, "Selected block compressor: dictMode=%d strat=%d rowMatchfinder=%d" , (int)dictMode, (int)strat, (int)useRowMatchFinder); |
3039 | if (ZSTD_rowMatchFinderUsed(strat, useRowMatchFinder)) { |
3040 | static const ZSTD_blockCompressor rowBasedBlockCompressors[4][3] = { |
3041 | { ZSTD_compressBlock_greedy_row, |
3042 | ZSTD_compressBlock_lazy_row, |
3043 | ZSTD_compressBlock_lazy2_row }, |
3044 | { ZSTD_compressBlock_greedy_extDict_row, |
3045 | ZSTD_compressBlock_lazy_extDict_row, |
3046 | ZSTD_compressBlock_lazy2_extDict_row }, |
3047 | { ZSTD_compressBlock_greedy_dictMatchState_row, |
3048 | ZSTD_compressBlock_lazy_dictMatchState_row, |
3049 | ZSTD_compressBlock_lazy2_dictMatchState_row }, |
3050 | { ZSTD_compressBlock_greedy_dedicatedDictSearch_row, |
3051 | ZSTD_compressBlock_lazy_dedicatedDictSearch_row, |
3052 | ZSTD_compressBlock_lazy2_dedicatedDictSearch_row } |
3053 | }; |
3054 | DEBUGLOG(4, "Selecting a row-based matchfinder" ); |
3055 | assert(useRowMatchFinder != ZSTD_ps_auto); |
3056 | selectedCompressor = rowBasedBlockCompressors[(int)dictMode][(int)strat - (int)ZSTD_greedy]; |
3057 | } else { |
3058 | selectedCompressor = blockCompressor[(int)dictMode][(int)strat]; |
3059 | } |
3060 | assert(selectedCompressor != NULL); |
3061 | return selectedCompressor; |
3062 | } |
3063 | |
3064 | static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr, |
3065 | const BYTE* anchor, size_t lastLLSize) |
3066 | { |
3067 | ZSTD_memcpy(seqStorePtr->lit, anchor, lastLLSize); |
3068 | seqStorePtr->lit += lastLLSize; |
3069 | } |
3070 | |
3071 | void ZSTD_resetSeqStore(seqStore_t* ssPtr) |
3072 | { |
3073 | ssPtr->lit = ssPtr->litStart; |
3074 | ssPtr->sequences = ssPtr->sequencesStart; |
3075 | ssPtr->longLengthType = ZSTD_llt_none; |
3076 | } |
3077 | |
3078 | /* ZSTD_postProcessSequenceProducerResult() : |
3079 | * Validates and post-processes sequences obtained through the external matchfinder API: |
3080 | * - Checks whether nbExternalSeqs represents an error condition. |
3081 | * - Appends a block delimiter to outSeqs if one is not already present. |
3082 | * See zstd.h for context regarding block delimiters. |
3083 | * Returns the number of sequences after post-processing, or an error code. */ |
3084 | static size_t ZSTD_postProcessSequenceProducerResult( |
3085 | ZSTD_Sequence* outSeqs, size_t nbExternalSeqs, size_t outSeqsCapacity, size_t srcSize |
3086 | ) { |
3087 | RETURN_ERROR_IF( |
3088 | nbExternalSeqs > outSeqsCapacity, |
3089 | sequenceProducer_failed, |
3090 | "External sequence producer returned error code %lu" , |
3091 | (unsigned long)nbExternalSeqs |
3092 | ); |
3093 | |
3094 | RETURN_ERROR_IF( |
3095 | nbExternalSeqs == 0 && srcSize > 0, |
3096 | sequenceProducer_failed, |
3097 | "Got zero sequences from external sequence producer for a non-empty src buffer!" |
3098 | ); |
3099 | |
3100 | if (srcSize == 0) { |
3101 | ZSTD_memset(&outSeqs[0], 0, sizeof(ZSTD_Sequence)); |
3102 | return 1; |
3103 | } |
3104 | |
3105 | { |
3106 | ZSTD_Sequence const lastSeq = outSeqs[nbExternalSeqs - 1]; |
3107 | |
3108 | /* We can return early if lastSeq is already a block delimiter. */ |
3109 | if (lastSeq.offset == 0 && lastSeq.matchLength == 0) { |
3110 | return nbExternalSeqs; |
3111 | } |
3112 | |
3113 | /* This error condition is only possible if the external matchfinder |
3114 | * produced an invalid parse, by definition of ZSTD_sequenceBound(). */ |
3115 | RETURN_ERROR_IF( |
3116 | nbExternalSeqs == outSeqsCapacity, |
3117 | sequenceProducer_failed, |
3118 | "nbExternalSeqs == outSeqsCapacity but lastSeq is not a block delimiter!" |
3119 | ); |
3120 | |
3121 | /* lastSeq is not a block delimiter, so we need to append one. */ |
3122 | ZSTD_memset(&outSeqs[nbExternalSeqs], 0, sizeof(ZSTD_Sequence)); |
3123 | return nbExternalSeqs + 1; |
3124 | } |
3125 | } |
3126 | |
3127 | /* ZSTD_fastSequenceLengthSum() : |
3128 | * Returns sum(litLen) + sum(matchLen) + lastLits for *seqBuf*. |
3129 | * Similar to another function in zstd_compress.c (determine_blockSize), |
3130 | * except it doesn't check for a block delimiter to end summation. |
3131 | * Removing the early exit allows the compiler to auto-vectorize (https://godbolt.org/z/cY1cajz9P). |
3132 | * This function can be deleted and replaced by determine_blockSize after we resolve issue #3456. */ |
3133 | static size_t ZSTD_fastSequenceLengthSum(ZSTD_Sequence const* seqBuf, size_t seqBufSize) { |
3134 | size_t matchLenSum, litLenSum, i; |
3135 | matchLenSum = 0; |
3136 | litLenSum = 0; |
3137 | for (i = 0; i < seqBufSize; i++) { |
3138 | litLenSum += seqBuf[i].litLength; |
3139 | matchLenSum += seqBuf[i].matchLength; |
3140 | } |
3141 | return litLenSum + matchLenSum; |
3142 | } |
3143 | |
3144 | typedef enum { ZSTDbss_compress, ZSTDbss_noCompress } ZSTD_buildSeqStore_e; |
3145 | |
3146 | static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) |
3147 | { |
3148 | ZSTD_matchState_t* const ms = &zc->blockState.matchState; |
3149 | DEBUGLOG(5, "ZSTD_buildSeqStore (srcSize=%zu)" , srcSize); |
3150 | assert(srcSize <= ZSTD_BLOCKSIZE_MAX); |
3151 | /* Assert that we have correctly flushed the ctx params into the ms's copy */ |
3152 | ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams); |
3153 | /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding |
3154 | * additional 1. We need to revisit and change this logic to be more consistent */ |
3155 | if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1+1) { |
3156 | if (zc->appliedParams.cParams.strategy >= ZSTD_btopt) { |
3157 | ZSTD_ldm_skipRawSeqStoreBytes(&zc->externSeqStore, srcSize); |
3158 | } else { |
3159 | ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch); |
3160 | } |
3161 | return ZSTDbss_noCompress; /* don't even attempt compression below a certain srcSize */ |
3162 | } |
3163 | ZSTD_resetSeqStore(&(zc->seqStore)); |
3164 | /* required for optimal parser to read stats from dictionary */ |
3165 | ms->opt.symbolCosts = &zc->blockState.prevCBlock->entropy; |
3166 | /* tell the optimal parser how we expect to compress literals */ |
3167 | ms->opt.literalCompressionMode = zc->appliedParams.literalCompressionMode; |
3168 | /* a gap between an attached dict and the current window is not safe, |
3169 | * they must remain adjacent, |
3170 | * and when that stops being the case, the dict must be unset */ |
3171 | assert(ms->dictMatchState == NULL || ms->loadedDictEnd == ms->window.dictLimit); |
3172 | |
3173 | /* limited update after a very long match */ |
3174 | { const BYTE* const base = ms->window.base; |
3175 | const BYTE* const istart = (const BYTE*)src; |
3176 | const U32 curr = (U32)(istart-base); |
3177 | if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */ |
3178 | if (curr > ms->nextToUpdate + 384) |
3179 | ms->nextToUpdate = curr - MIN(192, (U32)(curr - ms->nextToUpdate - 384)); |
3180 | } |
3181 | |
3182 | /* select and store sequences */ |
3183 | { ZSTD_dictMode_e const dictMode = ZSTD_matchState_dictMode(ms); |
3184 | size_t lastLLSize; |
3185 | { int i; |
3186 | for (i = 0; i < ZSTD_REP_NUM; ++i) |
3187 | zc->blockState.nextCBlock->rep[i] = zc->blockState.prevCBlock->rep[i]; |
3188 | } |
3189 | if (zc->externSeqStore.pos < zc->externSeqStore.size) { |
3190 | assert(zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_disable); |
3191 | |
3192 | /* External matchfinder + LDM is technically possible, just not implemented yet. |
3193 | * We need to revisit soon and implement it. */ |
3194 | RETURN_ERROR_IF( |
3195 | zc->appliedParams.useSequenceProducer, |
3196 | parameter_combination_unsupported, |
3197 | "Long-distance matching with external sequence producer enabled is not currently supported." |
3198 | ); |
3199 | |
3200 | /* Updates ldmSeqStore.pos */ |
3201 | lastLLSize = |
3202 | ZSTD_ldm_blockCompress(&zc->externSeqStore, |
3203 | ms, &zc->seqStore, |
3204 | zc->blockState.nextCBlock->rep, |
3205 | zc->appliedParams.useRowMatchFinder, |
3206 | src, srcSize); |
3207 | assert(zc->externSeqStore.pos <= zc->externSeqStore.size); |
3208 | } else if (zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) { |
3209 | rawSeqStore_t ldmSeqStore = kNullRawSeqStore; |
3210 | |
3211 | /* External matchfinder + LDM is technically possible, just not implemented yet. |
3212 | * We need to revisit soon and implement it. */ |
3213 | RETURN_ERROR_IF( |
3214 | zc->appliedParams.useSequenceProducer, |
3215 | parameter_combination_unsupported, |
3216 | "Long-distance matching with external sequence producer enabled is not currently supported." |
3217 | ); |
3218 | |
3219 | ldmSeqStore.seq = zc->ldmSequences; |
3220 | ldmSeqStore.capacity = zc->maxNbLdmSequences; |
3221 | /* Updates ldmSeqStore.size */ |
3222 | FORWARD_IF_ERROR(ZSTD_ldm_generateSequences(&zc->ldmState, &ldmSeqStore, |
3223 | &zc->appliedParams.ldmParams, |
3224 | src, srcSize), "" ); |
3225 | /* Updates ldmSeqStore.pos */ |
3226 | lastLLSize = |
3227 | ZSTD_ldm_blockCompress(&ldmSeqStore, |
3228 | ms, &zc->seqStore, |
3229 | zc->blockState.nextCBlock->rep, |
3230 | zc->appliedParams.useRowMatchFinder, |
3231 | src, srcSize); |
3232 | assert(ldmSeqStore.pos == ldmSeqStore.size); |
3233 | } else if (zc->appliedParams.useSequenceProducer) { |
3234 | assert( |
3235 | zc->externalMatchCtx.seqBufferCapacity >= ZSTD_sequenceBound(srcSize) |
3236 | ); |
3237 | assert(zc->externalMatchCtx.mFinder != NULL); |
3238 | |
3239 | { U32 const windowSize = (U32)1 << zc->appliedParams.cParams.windowLog; |
3240 | |
3241 | size_t const nbExternalSeqs = (zc->externalMatchCtx.mFinder)( |
3242 | zc->externalMatchCtx.mState, |
3243 | zc->externalMatchCtx.seqBuffer, |
3244 | zc->externalMatchCtx.seqBufferCapacity, |
3245 | src, srcSize, |
3246 | NULL, 0, /* dict and dictSize, currently not supported */ |
3247 | zc->appliedParams.compressionLevel, |
3248 | windowSize |
3249 | ); |
3250 | |
3251 | size_t const nbPostProcessedSeqs = ZSTD_postProcessSequenceProducerResult( |
3252 | zc->externalMatchCtx.seqBuffer, |
3253 | nbExternalSeqs, |
3254 | zc->externalMatchCtx.seqBufferCapacity, |
3255 | srcSize |
3256 | ); |
3257 | |
3258 | /* Return early if there is no error, since we don't need to worry about last literals */ |
3259 | if (!ZSTD_isError(nbPostProcessedSeqs)) { |
3260 | ZSTD_sequencePosition seqPos = {0,0,0}; |
3261 | size_t const seqLenSum = ZSTD_fastSequenceLengthSum(zc->externalMatchCtx.seqBuffer, nbPostProcessedSeqs); |
3262 | RETURN_ERROR_IF(seqLenSum > srcSize, externalSequences_invalid, "External sequences imply too large a block!" ); |
3263 | FORWARD_IF_ERROR( |
3264 | ZSTD_copySequencesToSeqStoreExplicitBlockDelim( |
3265 | zc, &seqPos, |
3266 | zc->externalMatchCtx.seqBuffer, nbPostProcessedSeqs, |
3267 | src, srcSize, |
3268 | zc->appliedParams.searchForExternalRepcodes |
3269 | ), |
3270 | "Failed to copy external sequences to seqStore!" |
3271 | ); |
3272 | ms->ldmSeqStore = NULL; |
3273 | DEBUGLOG(5, "Copied %lu sequences from external sequence producer to internal seqStore." , (unsigned long)nbExternalSeqs); |
3274 | return ZSTDbss_compress; |
3275 | } |
3276 | |
3277 | /* Propagate the error if fallback is disabled */ |
3278 | if (!zc->appliedParams.enableMatchFinderFallback) { |
3279 | return nbPostProcessedSeqs; |
3280 | } |
3281 | |
3282 | /* Fallback to software matchfinder */ |
3283 | { ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, |
3284 | zc->appliedParams.useRowMatchFinder, |
3285 | dictMode); |
3286 | ms->ldmSeqStore = NULL; |
3287 | DEBUGLOG( |
3288 | 5, |
3289 | "External sequence producer returned error code %lu. Falling back to internal parser." , |
3290 | (unsigned long)nbExternalSeqs |
3291 | ); |
3292 | lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize); |
3293 | } } |
3294 | } else { /* not long range mode and no external matchfinder */ |
3295 | ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, |
3296 | zc->appliedParams.useRowMatchFinder, |
3297 | dictMode); |
3298 | ms->ldmSeqStore = NULL; |
3299 | lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize); |
3300 | } |
3301 | { const BYTE* const lastLiterals = (const BYTE*)src + srcSize - lastLLSize; |
3302 | ZSTD_storeLastLiterals(&zc->seqStore, lastLiterals, lastLLSize); |
3303 | } } |
3304 | return ZSTDbss_compress; |
3305 | } |
3306 | |
3307 | static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc) |
3308 | { |
3309 | const seqStore_t* seqStore = ZSTD_getSeqStore(zc); |
3310 | const seqDef* seqStoreSeqs = seqStore->sequencesStart; |
3311 | size_t seqStoreSeqSize = seqStore->sequences - seqStoreSeqs; |
3312 | size_t seqStoreLiteralsSize = (size_t)(seqStore->lit - seqStore->litStart); |
3313 | size_t literalsRead = 0; |
3314 | size_t lastLLSize; |
3315 | |
3316 | ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex]; |
3317 | size_t i; |
3318 | repcodes_t updatedRepcodes; |
3319 | |
3320 | assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences); |
3321 | /* Ensure we have enough space for last literals "sequence" */ |
3322 | assert(zc->seqCollector.maxSequences >= seqStoreSeqSize + 1); |
3323 | ZSTD_memcpy(updatedRepcodes.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); |
3324 | for (i = 0; i < seqStoreSeqSize; ++i) { |
3325 | U32 rawOffset = seqStoreSeqs[i].offBase - ZSTD_REP_NUM; |
3326 | outSeqs[i].litLength = seqStoreSeqs[i].litLength; |
3327 | outSeqs[i].matchLength = seqStoreSeqs[i].mlBase + MINMATCH; |
3328 | outSeqs[i].rep = 0; |
3329 | |
3330 | if (i == seqStore->longLengthPos) { |
3331 | if (seqStore->longLengthType == ZSTD_llt_literalLength) { |
3332 | outSeqs[i].litLength += 0x10000; |
3333 | } else if (seqStore->longLengthType == ZSTD_llt_matchLength) { |
3334 | outSeqs[i].matchLength += 0x10000; |
3335 | } |
3336 | } |
3337 | |
3338 | if (seqStoreSeqs[i].offBase <= ZSTD_REP_NUM) { |
3339 | /* Derive the correct offset corresponding to a repcode */ |
3340 | outSeqs[i].rep = seqStoreSeqs[i].offBase; |
3341 | if (outSeqs[i].litLength != 0) { |
3342 | rawOffset = updatedRepcodes.rep[outSeqs[i].rep - 1]; |
3343 | } else { |
3344 | if (outSeqs[i].rep == 3) { |
3345 | rawOffset = updatedRepcodes.rep[0] - 1; |
3346 | } else { |
3347 | rawOffset = updatedRepcodes.rep[outSeqs[i].rep]; |
3348 | } |
3349 | } |
3350 | } |
3351 | outSeqs[i].offset = rawOffset; |
3352 | /* seqStoreSeqs[i].offset == offCode+1, and ZSTD_updateRep() expects offCode |
3353 | so we provide seqStoreSeqs[i].offset - 1 */ |
3354 | ZSTD_updateRep(updatedRepcodes.rep, |
3355 | seqStoreSeqs[i].offBase, |
3356 | seqStoreSeqs[i].litLength == 0); |
3357 | literalsRead += outSeqs[i].litLength; |
3358 | } |
3359 | /* Insert last literals (if any exist) in the block as a sequence with ml == off == 0. |
3360 | * If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker |
3361 | * for the block boundary, according to the API. |
3362 | */ |
3363 | assert(seqStoreLiteralsSize >= literalsRead); |
3364 | lastLLSize = seqStoreLiteralsSize - literalsRead; |
3365 | outSeqs[i].litLength = (U32)lastLLSize; |
3366 | outSeqs[i].matchLength = outSeqs[i].offset = outSeqs[i].rep = 0; |
3367 | seqStoreSeqSize++; |
3368 | zc->seqCollector.seqIndex += seqStoreSeqSize; |
3369 | } |
3370 | |
3371 | size_t ZSTD_sequenceBound(size_t srcSize) { |
3372 | return (srcSize / ZSTD_MINMATCH_MIN) + 1; |
3373 | } |
3374 | |
3375 | size_t ZSTD_generateSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs, |
3376 | size_t outSeqsSize, const void* src, size_t srcSize) |
3377 | { |
3378 | const size_t dstCapacity = ZSTD_compressBound(srcSize); |
3379 | void* dst = ZSTD_customMalloc(dstCapacity, ZSTD_defaultCMem); |
3380 | SeqCollector seqCollector; |
3381 | |
3382 | RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!" ); |
3383 | |
3384 | seqCollector.collectSequences = 1; |
3385 | seqCollector.seqStart = outSeqs; |
3386 | seqCollector.seqIndex = 0; |
3387 | seqCollector.maxSequences = outSeqsSize; |
3388 | zc->seqCollector = seqCollector; |
3389 | |
3390 | ZSTD_compress2(zc, dst, dstCapacity, src, srcSize); |
3391 | ZSTD_customFree(dst, ZSTD_defaultCMem); |
3392 | return zc->seqCollector.seqIndex; |
3393 | } |
3394 | |
3395 | size_t ZSTD_mergeBlockDelimiters(ZSTD_Sequence* sequences, size_t seqsSize) { |
3396 | size_t in = 0; |
3397 | size_t out = 0; |
3398 | for (; in < seqsSize; ++in) { |
3399 | if (sequences[in].offset == 0 && sequences[in].matchLength == 0) { |
3400 | if (in != seqsSize - 1) { |
3401 | sequences[in+1].litLength += sequences[in].litLength; |
3402 | } |
3403 | } else { |
3404 | sequences[out] = sequences[in]; |
3405 | ++out; |
3406 | } |
3407 | } |
3408 | return out; |
3409 | } |
3410 | |
3411 | /* Unrolled loop to read four size_ts of input at a time. Returns 1 if is RLE, 0 if not. */ |
3412 | static int ZSTD_isRLE(const BYTE* src, size_t length) { |
3413 | const BYTE* ip = src; |
3414 | const BYTE value = ip[0]; |
3415 | const size_t valueST = (size_t)((U64)value * 0x0101010101010101ULL); |
3416 | const size_t unrollSize = sizeof(size_t) * 4; |
3417 | const size_t unrollMask = unrollSize - 1; |
3418 | const size_t prefixLength = length & unrollMask; |
3419 | size_t i; |
3420 | if (length == 1) return 1; |
3421 | /* Check if prefix is RLE first before using unrolled loop */ |
3422 | if (prefixLength && ZSTD_count(ip+1, ip, ip+prefixLength) != prefixLength-1) { |
3423 | return 0; |
3424 | } |
3425 | for (i = prefixLength; i != length; i += unrollSize) { |
3426 | size_t u; |
3427 | for (u = 0; u < unrollSize; u += sizeof(size_t)) { |
3428 | if (MEM_readST(ip + i + u) != valueST) { |
3429 | return 0; |
3430 | } } } |
3431 | return 1; |
3432 | } |
3433 | |
3434 | /* Returns true if the given block may be RLE. |
3435 | * This is just a heuristic based on the compressibility. |
3436 | * It may return both false positives and false negatives. |
3437 | */ |
3438 | static int ZSTD_maybeRLE(seqStore_t const* seqStore) |
3439 | { |
3440 | size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart); |
3441 | size_t const nbLits = (size_t)(seqStore->lit - seqStore->litStart); |
3442 | |
3443 | return nbSeqs < 4 && nbLits < 10; |
3444 | } |
3445 | |
3446 | static void |
3447 | ZSTD_blockState_confirmRepcodesAndEntropyTables(ZSTD_blockState_t* const bs) |
3448 | { |
3449 | ZSTD_compressedBlockState_t* const tmp = bs->prevCBlock; |
3450 | bs->prevCBlock = bs->nextCBlock; |
3451 | bs->nextCBlock = tmp; |
3452 | } |
3453 | |
3454 | /* Writes the block header */ |
3455 | static void |
3456 | (void* op, size_t cSize, size_t blockSize, U32 lastBlock) |
3457 | { |
3458 | U32 const = cSize == 1 ? |
3459 | lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) : |
3460 | lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3); |
3461 | MEM_writeLE24(op, cBlockHeader); |
3462 | DEBUGLOG(3, "writeBlockHeader: cSize: %zu blockSize: %zu lastBlock: %u" , cSize, blockSize, lastBlock); |
3463 | } |
3464 | |
3465 | /** ZSTD_buildBlockEntropyStats_literals() : |
3466 | * Builds entropy for the literals. |
3467 | * Stores literals block type (raw, rle, compressed, repeat) and |
3468 | * huffman description table to hufMetadata. |
3469 | * Requires ENTROPY_WORKSPACE_SIZE workspace |
3470 | * @return : size of huffman description table, or an error code |
3471 | */ |
3472 | static size_t |
3473 | ZSTD_buildBlockEntropyStats_literals(void* const src, size_t srcSize, |
3474 | const ZSTD_hufCTables_t* prevHuf, |
3475 | ZSTD_hufCTables_t* nextHuf, |
3476 | ZSTD_hufCTablesMetadata_t* hufMetadata, |
3477 | const int literalsCompressionIsDisabled, |
3478 | void* workspace, size_t wkspSize, |
3479 | int hufFlags) |
3480 | { |
3481 | BYTE* const wkspStart = (BYTE*)workspace; |
3482 | BYTE* const wkspEnd = wkspStart + wkspSize; |
3483 | BYTE* const countWkspStart = wkspStart; |
3484 | unsigned* const countWksp = (unsigned*)workspace; |
3485 | const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned); |
3486 | BYTE* const nodeWksp = countWkspStart + countWkspSize; |
3487 | const size_t nodeWkspSize = (size_t)(wkspEnd - nodeWksp); |
3488 | unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX; |
3489 | unsigned huffLog = LitHufLog; |
3490 | HUF_repeat repeat = prevHuf->repeatMode; |
3491 | DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_literals (srcSize=%zu)" , srcSize); |
3492 | |
3493 | /* Prepare nextEntropy assuming reusing the existing table */ |
3494 | ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); |
3495 | |
3496 | if (literalsCompressionIsDisabled) { |
3497 | DEBUGLOG(5, "set_basic - disabled" ); |
3498 | hufMetadata->hType = set_basic; |
3499 | return 0; |
3500 | } |
3501 | |
3502 | /* small ? don't even attempt compression (speed opt) */ |
3503 | #ifndef COMPRESS_LITERALS_SIZE_MIN |
3504 | # define COMPRESS_LITERALS_SIZE_MIN 63 /* heuristic */ |
3505 | #endif |
3506 | { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN; |
3507 | if (srcSize <= minLitSize) { |
3508 | DEBUGLOG(5, "set_basic - too small" ); |
3509 | hufMetadata->hType = set_basic; |
3510 | return 0; |
3511 | } } |
3512 | |
3513 | /* Scan input and build symbol stats */ |
3514 | { size_t const largest = |
3515 | HIST_count_wksp (countWksp, &maxSymbolValue, |
3516 | (const BYTE*)src, srcSize, |
3517 | workspace, wkspSize); |
3518 | FORWARD_IF_ERROR(largest, "HIST_count_wksp failed" ); |
3519 | if (largest == srcSize) { |
3520 | /* only one literal symbol */ |
3521 | DEBUGLOG(5, "set_rle" ); |
3522 | hufMetadata->hType = set_rle; |
3523 | return 0; |
3524 | } |
3525 | if (largest <= (srcSize >> 7)+4) { |
3526 | /* heuristic: likely not compressible */ |
3527 | DEBUGLOG(5, "set_basic - no gain" ); |
3528 | hufMetadata->hType = set_basic; |
3529 | return 0; |
3530 | } } |
3531 | |
3532 | /* Validate the previous Huffman table */ |
3533 | if (repeat == HUF_repeat_check |
3534 | && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) { |
3535 | repeat = HUF_repeat_none; |
3536 | } |
3537 | |
3538 | /* Build Huffman Tree */ |
3539 | ZSTD_memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable)); |
3540 | huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue, nodeWksp, nodeWkspSize, nextHuf->CTable, countWksp, hufFlags); |
3541 | assert(huffLog <= LitHufLog); |
3542 | { size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp, |
3543 | maxSymbolValue, huffLog, |
3544 | nodeWksp, nodeWkspSize); |
3545 | FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp" ); |
3546 | huffLog = (U32)maxBits; |
3547 | } |
3548 | { /* Build and write the CTable */ |
3549 | size_t const newCSize = HUF_estimateCompressedSize( |
3550 | (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue); |
3551 | size_t const hSize = HUF_writeCTable_wksp( |
3552 | hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer), |
3553 | (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog, |
3554 | nodeWksp, nodeWkspSize); |
3555 | /* Check against repeating the previous CTable */ |
3556 | if (repeat != HUF_repeat_none) { |
3557 | size_t const oldCSize = HUF_estimateCompressedSize( |
3558 | (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue); |
3559 | if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) { |
3560 | DEBUGLOG(5, "set_repeat - smaller" ); |
3561 | ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); |
3562 | hufMetadata->hType = set_repeat; |
3563 | return 0; |
3564 | } } |
3565 | if (newCSize + hSize >= srcSize) { |
3566 | DEBUGLOG(5, "set_basic - no gains" ); |
3567 | ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); |
3568 | hufMetadata->hType = set_basic; |
3569 | return 0; |
3570 | } |
3571 | DEBUGLOG(5, "set_compressed (hSize=%u)" , (U32)hSize); |
3572 | hufMetadata->hType = set_compressed; |
3573 | nextHuf->repeatMode = HUF_repeat_check; |
3574 | return hSize; |
3575 | } |
3576 | } |
3577 | |
3578 | |
3579 | /* ZSTD_buildDummySequencesStatistics(): |
3580 | * Returns a ZSTD_symbolEncodingTypeStats_t with all encoding types as set_basic, |
3581 | * and updates nextEntropy to the appropriate repeatMode. |
3582 | */ |
3583 | static ZSTD_symbolEncodingTypeStats_t |
3584 | ZSTD_buildDummySequencesStatistics(ZSTD_fseCTables_t* nextEntropy) |
3585 | { |
3586 | ZSTD_symbolEncodingTypeStats_t stats = {set_basic, set_basic, set_basic, 0, 0, 0}; |
3587 | nextEntropy->litlength_repeatMode = FSE_repeat_none; |
3588 | nextEntropy->offcode_repeatMode = FSE_repeat_none; |
3589 | nextEntropy->matchlength_repeatMode = FSE_repeat_none; |
3590 | return stats; |
3591 | } |
3592 | |
3593 | /** ZSTD_buildBlockEntropyStats_sequences() : |
3594 | * Builds entropy for the sequences. |
3595 | * Stores symbol compression modes and fse table to fseMetadata. |
3596 | * Requires ENTROPY_WORKSPACE_SIZE wksp. |
3597 | * @return : size of fse tables or error code */ |
3598 | static size_t |
3599 | ZSTD_buildBlockEntropyStats_sequences( |
3600 | const seqStore_t* seqStorePtr, |
3601 | const ZSTD_fseCTables_t* prevEntropy, |
3602 | ZSTD_fseCTables_t* nextEntropy, |
3603 | const ZSTD_CCtx_params* cctxParams, |
3604 | ZSTD_fseCTablesMetadata_t* fseMetadata, |
3605 | void* workspace, size_t wkspSize) |
3606 | { |
3607 | ZSTD_strategy const strategy = cctxParams->cParams.strategy; |
3608 | size_t const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); |
3609 | BYTE* const ostart = fseMetadata->fseTablesBuffer; |
3610 | BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer); |
3611 | BYTE* op = ostart; |
3612 | unsigned* countWorkspace = (unsigned*)workspace; |
3613 | unsigned* entropyWorkspace = countWorkspace + (MaxSeq + 1); |
3614 | size_t entropyWorkspaceSize = wkspSize - (MaxSeq + 1) * sizeof(*countWorkspace); |
3615 | ZSTD_symbolEncodingTypeStats_t stats; |
3616 | |
3617 | DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_sequences (nbSeq=%zu)" , nbSeq); |
3618 | stats = nbSeq != 0 ? ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq, |
3619 | prevEntropy, nextEntropy, op, oend, |
3620 | strategy, countWorkspace, |
3621 | entropyWorkspace, entropyWorkspaceSize) |
3622 | : ZSTD_buildDummySequencesStatistics(nextEntropy); |
3623 | FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!" ); |
3624 | fseMetadata->llType = (symbolEncodingType_e) stats.LLtype; |
3625 | fseMetadata->ofType = (symbolEncodingType_e) stats.Offtype; |
3626 | fseMetadata->mlType = (symbolEncodingType_e) stats.MLtype; |
3627 | fseMetadata->lastCountSize = stats.lastCountSize; |
3628 | return stats.size; |
3629 | } |
3630 | |
3631 | |
3632 | /** ZSTD_buildBlockEntropyStats() : |
3633 | * Builds entropy for the block. |
3634 | * Requires workspace size ENTROPY_WORKSPACE_SIZE |
3635 | * @return : 0 on success, or an error code |
3636 | * Note : also employed in superblock |
3637 | */ |
3638 | size_t ZSTD_buildBlockEntropyStats( |
3639 | const seqStore_t* seqStorePtr, |
3640 | const ZSTD_entropyCTables_t* prevEntropy, |
3641 | ZSTD_entropyCTables_t* nextEntropy, |
3642 | const ZSTD_CCtx_params* cctxParams, |
3643 | ZSTD_entropyCTablesMetadata_t* entropyMetadata, |
3644 | void* workspace, size_t wkspSize) |
3645 | { |
3646 | size_t const litSize = (size_t)(seqStorePtr->lit - seqStorePtr->litStart); |
3647 | int const huf_useOptDepth = (cctxParams->cParams.strategy >= HUF_OPTIMAL_DEPTH_THRESHOLD); |
3648 | int const hufFlags = huf_useOptDepth ? HUF_flags_optimalDepth : 0; |
3649 | |
3650 | entropyMetadata->hufMetadata.hufDesSize = |
3651 | ZSTD_buildBlockEntropyStats_literals(seqStorePtr->litStart, litSize, |
3652 | &prevEntropy->huf, &nextEntropy->huf, |
3653 | &entropyMetadata->hufMetadata, |
3654 | ZSTD_literalsCompressionIsDisabled(cctxParams), |
3655 | workspace, wkspSize, hufFlags); |
3656 | |
3657 | FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildBlockEntropyStats_literals failed" ); |
3658 | entropyMetadata->fseMetadata.fseTablesSize = |
3659 | ZSTD_buildBlockEntropyStats_sequences(seqStorePtr, |
3660 | &prevEntropy->fse, &nextEntropy->fse, |
3661 | cctxParams, |
3662 | &entropyMetadata->fseMetadata, |
3663 | workspace, wkspSize); |
3664 | FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildBlockEntropyStats_sequences failed" ); |
3665 | return 0; |
3666 | } |
3667 | |
3668 | /* Returns the size estimate for the literals section (header + content) of a block */ |
3669 | static size_t |
3670 | ZSTD_estimateBlockSize_literal(const BYTE* literals, size_t litSize, |
3671 | const ZSTD_hufCTables_t* huf, |
3672 | const ZSTD_hufCTablesMetadata_t* hufMetadata, |
3673 | void* workspace, size_t wkspSize, |
3674 | int writeEntropy) |
3675 | { |
3676 | unsigned* const countWksp = (unsigned*)workspace; |
3677 | unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX; |
3678 | size_t = 3 + (litSize >= 1 KB) + (litSize >= 16 KB); |
3679 | U32 singleStream = litSize < 256; |
3680 | |
3681 | if (hufMetadata->hType == set_basic) return litSize; |
3682 | else if (hufMetadata->hType == set_rle) return 1; |
3683 | else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) { |
3684 | size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize); |
3685 | if (ZSTD_isError(largest)) return litSize; |
3686 | { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue); |
3687 | if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize; |
3688 | if (!singleStream) cLitSizeEstimate += 6; /* multi-stream huffman uses 6-byte jump table */ |
3689 | return cLitSizeEstimate + literalSectionHeaderSize; |
3690 | } } |
3691 | assert(0); /* impossible */ |
3692 | return 0; |
3693 | } |
3694 | |
3695 | /* Returns the size estimate for the FSE-compressed symbols (of, ml, ll) of a block */ |
3696 | static size_t |
3697 | ZSTD_estimateBlockSize_symbolType(symbolEncodingType_e type, |
3698 | const BYTE* codeTable, size_t nbSeq, unsigned maxCode, |
3699 | const FSE_CTable* fseCTable, |
3700 | const U8* additionalBits, |
3701 | short const* defaultNorm, U32 defaultNormLog, U32 defaultMax, |
3702 | void* workspace, size_t wkspSize) |
3703 | { |
3704 | unsigned* const countWksp = (unsigned*)workspace; |
3705 | const BYTE* ctp = codeTable; |
3706 | const BYTE* const ctStart = ctp; |
3707 | const BYTE* const ctEnd = ctStart + nbSeq; |
3708 | size_t cSymbolTypeSizeEstimateInBits = 0; |
3709 | unsigned max = maxCode; |
3710 | |
3711 | HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */ |
3712 | if (type == set_basic) { |
3713 | /* We selected this encoding type, so it must be valid. */ |
3714 | assert(max <= defaultMax); |
3715 | (void)defaultMax; |
3716 | cSymbolTypeSizeEstimateInBits = ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max); |
3717 | } else if (type == set_rle) { |
3718 | cSymbolTypeSizeEstimateInBits = 0; |
3719 | } else if (type == set_compressed || type == set_repeat) { |
3720 | cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max); |
3721 | } |
3722 | if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) { |
3723 | return nbSeq * 10; |
3724 | } |
3725 | while (ctp < ctEnd) { |
3726 | if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp]; |
3727 | else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */ |
3728 | ctp++; |
3729 | } |
3730 | return cSymbolTypeSizeEstimateInBits >> 3; |
3731 | } |
3732 | |
3733 | /* Returns the size estimate for the sequences section (header + content) of a block */ |
3734 | static size_t |
3735 | ZSTD_estimateBlockSize_sequences(const BYTE* ofCodeTable, |
3736 | const BYTE* llCodeTable, |
3737 | const BYTE* mlCodeTable, |
3738 | size_t nbSeq, |
3739 | const ZSTD_fseCTables_t* fseTables, |
3740 | const ZSTD_fseCTablesMetadata_t* fseMetadata, |
3741 | void* workspace, size_t wkspSize, |
3742 | int writeEntropy) |
3743 | { |
3744 | size_t = 1 /* seqHead */ + 1 /* min seqSize size */ + (nbSeq >= 128) + (nbSeq >= LONGNBSEQ); |
3745 | size_t cSeqSizeEstimate = 0; |
3746 | cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, nbSeq, MaxOff, |
3747 | fseTables->offcodeCTable, NULL, |
3748 | OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff, |
3749 | workspace, wkspSize); |
3750 | cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->llType, llCodeTable, nbSeq, MaxLL, |
3751 | fseTables->litlengthCTable, LL_bits, |
3752 | LL_defaultNorm, LL_defaultNormLog, MaxLL, |
3753 | workspace, wkspSize); |
3754 | cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, nbSeq, MaxML, |
3755 | fseTables->matchlengthCTable, ML_bits, |
3756 | ML_defaultNorm, ML_defaultNormLog, MaxML, |
3757 | workspace, wkspSize); |
3758 | if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize; |
3759 | return cSeqSizeEstimate + sequencesSectionHeaderSize; |
3760 | } |
3761 | |
3762 | /* Returns the size estimate for a given stream of literals, of, ll, ml */ |
3763 | static size_t |
3764 | ZSTD_estimateBlockSize(const BYTE* literals, size_t litSize, |
3765 | const BYTE* ofCodeTable, |
3766 | const BYTE* llCodeTable, |
3767 | const BYTE* mlCodeTable, |
3768 | size_t nbSeq, |
3769 | const ZSTD_entropyCTables_t* entropy, |
3770 | const ZSTD_entropyCTablesMetadata_t* entropyMetadata, |
3771 | void* workspace, size_t wkspSize, |
3772 | int writeLitEntropy, int writeSeqEntropy) |
3773 | { |
3774 | size_t const literalsSize = ZSTD_estimateBlockSize_literal(literals, litSize, |
3775 | &entropy->huf, &entropyMetadata->hufMetadata, |
3776 | workspace, wkspSize, writeLitEntropy); |
3777 | size_t const seqSize = ZSTD_estimateBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable, |
3778 | nbSeq, &entropy->fse, &entropyMetadata->fseMetadata, |
3779 | workspace, wkspSize, writeSeqEntropy); |
3780 | return seqSize + literalsSize + ZSTD_blockHeaderSize; |
3781 | } |
3782 | |
3783 | /* Builds entropy statistics and uses them for blocksize estimation. |
3784 | * |
3785 | * @return: estimated compressed size of the seqStore, or a zstd error. |
3786 | */ |
3787 | static size_t |
3788 | ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(seqStore_t* seqStore, ZSTD_CCtx* zc) |
3789 | { |
3790 | ZSTD_entropyCTablesMetadata_t* const entropyMetadata = &zc->blockSplitCtx.entropyMetadata; |
3791 | DEBUGLOG(6, "ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize()" ); |
3792 | FORWARD_IF_ERROR(ZSTD_buildBlockEntropyStats(seqStore, |
3793 | &zc->blockState.prevCBlock->entropy, |
3794 | &zc->blockState.nextCBlock->entropy, |
3795 | &zc->appliedParams, |
3796 | entropyMetadata, |
3797 | zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE), "" ); |
3798 | return ZSTD_estimateBlockSize( |
3799 | seqStore->litStart, (size_t)(seqStore->lit - seqStore->litStart), |
3800 | seqStore->ofCode, seqStore->llCode, seqStore->mlCode, |
3801 | (size_t)(seqStore->sequences - seqStore->sequencesStart), |
3802 | &zc->blockState.nextCBlock->entropy, |
3803 | entropyMetadata, |
3804 | zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE, |
3805 | (int)(entropyMetadata->hufMetadata.hType == set_compressed), 1); |
3806 | } |
3807 | |
3808 | /* Returns literals bytes represented in a seqStore */ |
3809 | static size_t ZSTD_countSeqStoreLiteralsBytes(const seqStore_t* const seqStore) |
3810 | { |
3811 | size_t literalsBytes = 0; |
3812 | size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart); |
3813 | size_t i; |
3814 | for (i = 0; i < nbSeqs; ++i) { |
3815 | seqDef const seq = seqStore->sequencesStart[i]; |
3816 | literalsBytes += seq.litLength; |
3817 | if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_literalLength) { |
3818 | literalsBytes += 0x10000; |
3819 | } } |
3820 | return literalsBytes; |
3821 | } |
3822 | |
3823 | /* Returns match bytes represented in a seqStore */ |
3824 | static size_t ZSTD_countSeqStoreMatchBytes(const seqStore_t* const seqStore) |
3825 | { |
3826 | size_t matchBytes = 0; |
3827 | size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart); |
3828 | size_t i; |
3829 | for (i = 0; i < nbSeqs; ++i) { |
3830 | seqDef seq = seqStore->sequencesStart[i]; |
3831 | matchBytes += seq.mlBase + MINMATCH; |
3832 | if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_matchLength) { |
3833 | matchBytes += 0x10000; |
3834 | } } |
3835 | return matchBytes; |
3836 | } |
3837 | |
3838 | /* Derives the seqStore that is a chunk of the originalSeqStore from [startIdx, endIdx). |
3839 | * Stores the result in resultSeqStore. |
3840 | */ |
3841 | static void ZSTD_deriveSeqStoreChunk(seqStore_t* resultSeqStore, |
3842 | const seqStore_t* originalSeqStore, |
3843 | size_t startIdx, size_t endIdx) |
3844 | { |
3845 | *resultSeqStore = *originalSeqStore; |
3846 | if (startIdx > 0) { |
3847 | resultSeqStore->sequences = originalSeqStore->sequencesStart + startIdx; |
3848 | resultSeqStore->litStart += ZSTD_countSeqStoreLiteralsBytes(resultSeqStore); |
3849 | } |
3850 | |
3851 | /* Move longLengthPos into the correct position if necessary */ |
3852 | if (originalSeqStore->longLengthType != ZSTD_llt_none) { |
3853 | if (originalSeqStore->longLengthPos < startIdx || originalSeqStore->longLengthPos > endIdx) { |
3854 | resultSeqStore->longLengthType = ZSTD_llt_none; |
3855 | } else { |
3856 | resultSeqStore->longLengthPos -= (U32)startIdx; |
3857 | } |
3858 | } |
3859 | resultSeqStore->sequencesStart = originalSeqStore->sequencesStart + startIdx; |
3860 | resultSeqStore->sequences = originalSeqStore->sequencesStart + endIdx; |
3861 | if (endIdx == (size_t)(originalSeqStore->sequences - originalSeqStore->sequencesStart)) { |
3862 | /* This accounts for possible last literals if the derived chunk reaches the end of the block */ |
3863 | assert(resultSeqStore->lit == originalSeqStore->lit); |
3864 | } else { |
3865 | size_t const literalsBytes = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore); |
3866 | resultSeqStore->lit = resultSeqStore->litStart + literalsBytes; |
3867 | } |
3868 | resultSeqStore->llCode += startIdx; |
3869 | resultSeqStore->mlCode += startIdx; |
3870 | resultSeqStore->ofCode += startIdx; |
3871 | } |
3872 | |
3873 | /** |
3874 | * Returns the raw offset represented by the combination of offBase, ll0, and repcode history. |
3875 | * offBase must represent a repcode in the numeric representation of ZSTD_storeSeq(). |
3876 | */ |
3877 | static U32 |
3878 | ZSTD_resolveRepcodeToRawOffset(const U32 rep[ZSTD_REP_NUM], const U32 offBase, const U32 ll0) |
3879 | { |
3880 | U32 const adjustedRepCode = OFFBASE_TO_REPCODE(offBase) - 1 + ll0; /* [ 0 - 3 ] */ |
3881 | assert(OFFBASE_IS_REPCODE(offBase)); |
3882 | if (adjustedRepCode == ZSTD_REP_NUM) { |
3883 | assert(ll0); |
3884 | /* litlength == 0 and offCode == 2 implies selection of first repcode - 1 |
3885 | * This is only valid if it results in a valid offset value, aka > 0. |
3886 | * Note : it may happen that `rep[0]==1` in exceptional circumstances. |
3887 | * In which case this function will return 0, which is an invalid offset. |
3888 | * It's not an issue though, since this value will be |
3889 | * compared and discarded within ZSTD_seqStore_resolveOffCodes(). |
3890 | */ |
3891 | return rep[0] - 1; |
3892 | } |
3893 | return rep[adjustedRepCode]; |
3894 | } |
3895 | |
3896 | /** |
3897 | * ZSTD_seqStore_resolveOffCodes() reconciles any possible divergences in offset history that may arise |
3898 | * due to emission of RLE/raw blocks that disturb the offset history, |
3899 | * and replaces any repcodes within the seqStore that may be invalid. |
3900 | * |
3901 | * dRepcodes are updated as would be on the decompression side. |
3902 | * cRepcodes are updated exactly in accordance with the seqStore. |
3903 | * |
3904 | * Note : this function assumes seq->offBase respects the following numbering scheme : |
3905 | * 0 : invalid |
3906 | * 1-3 : repcode 1-3 |
3907 | * 4+ : real_offset+3 |
3908 | */ |
3909 | static void |
3910 | ZSTD_seqStore_resolveOffCodes(repcodes_t* const dRepcodes, repcodes_t* const cRepcodes, |
3911 | const seqStore_t* const seqStore, U32 const nbSeq) |
3912 | { |
3913 | U32 idx = 0; |
3914 | U32 const longLitLenIdx = seqStore->longLengthType == ZSTD_llt_literalLength ? seqStore->longLengthPos : nbSeq; |
3915 | for (; idx < nbSeq; ++idx) { |
3916 | seqDef* const seq = seqStore->sequencesStart + idx; |
3917 | U32 const ll0 = (seq->litLength == 0) && (idx != longLitLenIdx); |
3918 | U32 const offBase = seq->offBase; |
3919 | assert(offBase > 0); |
3920 | if (OFFBASE_IS_REPCODE(offBase)) { |
3921 | U32 const dRawOffset = ZSTD_resolveRepcodeToRawOffset(dRepcodes->rep, offBase, ll0); |
3922 | U32 const cRawOffset = ZSTD_resolveRepcodeToRawOffset(cRepcodes->rep, offBase, ll0); |
3923 | /* Adjust simulated decompression repcode history if we come across a mismatch. Replace |
3924 | * the repcode with the offset it actually references, determined by the compression |
3925 | * repcode history. |
3926 | */ |
3927 | if (dRawOffset != cRawOffset) { |
3928 | seq->offBase = OFFSET_TO_OFFBASE(cRawOffset); |
3929 | } |
3930 | } |
3931 | /* Compression repcode history is always updated with values directly from the unmodified seqStore. |
3932 | * Decompression repcode history may use modified seq->offset value taken from compression repcode history. |
3933 | */ |
3934 | ZSTD_updateRep(dRepcodes->rep, seq->offBase, ll0); |
3935 | ZSTD_updateRep(cRepcodes->rep, offBase, ll0); |
3936 | } |
3937 | } |
3938 | |
3939 | /* ZSTD_compressSeqStore_singleBlock(): |
3940 | * Compresses a seqStore into a block with a block header, into the buffer dst. |
3941 | * |
3942 | * Returns the total size of that block (including header) or a ZSTD error code. |
3943 | */ |
3944 | static size_t |
3945 | ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx* zc, |
3946 | const seqStore_t* const seqStore, |
3947 | repcodes_t* const dRep, repcodes_t* const cRep, |
3948 | void* dst, size_t dstCapacity, |
3949 | const void* src, size_t srcSize, |
3950 | U32 lastBlock, U32 isPartition) |
3951 | { |
3952 | const U32 rleMaxLength = 25; |
3953 | BYTE* op = (BYTE*)dst; |
3954 | const BYTE* ip = (const BYTE*)src; |
3955 | size_t cSize; |
3956 | size_t cSeqsSize; |
3957 | |
3958 | /* In case of an RLE or raw block, the simulated decompression repcode history must be reset */ |
3959 | repcodes_t const dRepOriginal = *dRep; |
3960 | DEBUGLOG(5, "ZSTD_compressSeqStore_singleBlock" ); |
3961 | if (isPartition) |
3962 | ZSTD_seqStore_resolveOffCodes(dRep, cRep, seqStore, (U32)(seqStore->sequences - seqStore->sequencesStart)); |
3963 | |
3964 | RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "Block header doesn't fit" ); |
3965 | cSeqsSize = ZSTD_entropyCompressSeqStore(seqStore, |
3966 | &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy, |
3967 | &zc->appliedParams, |
3968 | op + ZSTD_blockHeaderSize, dstCapacity - ZSTD_blockHeaderSize, |
3969 | srcSize, |
3970 | zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, |
3971 | zc->bmi2); |
3972 | FORWARD_IF_ERROR(cSeqsSize, "ZSTD_entropyCompressSeqStore failed!" ); |
3973 | |
3974 | if (!zc->isFirstBlock && |
3975 | cSeqsSize < rleMaxLength && |
3976 | ZSTD_isRLE((BYTE const*)src, srcSize)) { |
3977 | /* We don't want to emit our first block as a RLE even if it qualifies because |
3978 | * doing so will cause the decoder (cli only) to throw a "should consume all input error." |
3979 | * This is only an issue for zstd <= v1.4.3 |
3980 | */ |
3981 | cSeqsSize = 1; |
3982 | } |
3983 | |
3984 | if (zc->seqCollector.collectSequences) { |
3985 | ZSTD_copyBlockSequences(zc); |
3986 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); |
3987 | return 0; |
3988 | } |
3989 | |
3990 | if (cSeqsSize == 0) { |
3991 | cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock); |
3992 | FORWARD_IF_ERROR(cSize, "Nocompress block failed" ); |
3993 | DEBUGLOG(4, "Writing out nocompress block, size: %zu" , cSize); |
3994 | *dRep = dRepOriginal; /* reset simulated decompression repcode history */ |
3995 | } else if (cSeqsSize == 1) { |
3996 | cSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, srcSize, lastBlock); |
3997 | FORWARD_IF_ERROR(cSize, "RLE compress block failed" ); |
3998 | DEBUGLOG(4, "Writing out RLE block, size: %zu" , cSize); |
3999 | *dRep = dRepOriginal; /* reset simulated decompression repcode history */ |
4000 | } else { |
4001 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); |
4002 | writeBlockHeader(op, cSeqsSize, srcSize, lastBlock); |
4003 | cSize = ZSTD_blockHeaderSize + cSeqsSize; |
4004 | DEBUGLOG(4, "Writing out compressed block, size: %zu" , cSize); |
4005 | } |
4006 | |
4007 | if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) |
4008 | zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; |
4009 | |
4010 | return cSize; |
4011 | } |
4012 | |
4013 | /* Struct to keep track of where we are in our recursive calls. */ |
4014 | typedef struct { |
4015 | U32* splitLocations; /* Array of split indices */ |
4016 | size_t idx; /* The current index within splitLocations being worked on */ |
4017 | } seqStoreSplits; |
4018 | |
4019 | #define MIN_SEQUENCES_BLOCK_SPLITTING 300 |
4020 | |
4021 | /* Helper function to perform the recursive search for block splits. |
4022 | * Estimates the cost of seqStore prior to split, and estimates the cost of splitting the sequences in half. |
4023 | * If advantageous to split, then we recurse down the two sub-blocks. |
4024 | * If not, or if an error occurred in estimation, then we do not recurse. |
4025 | * |
4026 | * Note: The recursion depth is capped by a heuristic minimum number of sequences, |
4027 | * defined by MIN_SEQUENCES_BLOCK_SPLITTING. |
4028 | * In theory, this means the absolute largest recursion depth is 10 == log2(maxNbSeqInBlock/MIN_SEQUENCES_BLOCK_SPLITTING). |
4029 | * In practice, recursion depth usually doesn't go beyond 4. |
4030 | * |
4031 | * Furthermore, the number of splits is capped by ZSTD_MAX_NB_BLOCK_SPLITS. |
4032 | * At ZSTD_MAX_NB_BLOCK_SPLITS == 196 with the current existing blockSize |
4033 | * maximum of 128 KB, this value is actually impossible to reach. |
4034 | */ |
4035 | static void |
4036 | ZSTD_deriveBlockSplitsHelper(seqStoreSplits* splits, size_t startIdx, size_t endIdx, |
4037 | ZSTD_CCtx* zc, const seqStore_t* origSeqStore) |
4038 | { |
4039 | seqStore_t* const fullSeqStoreChunk = &zc->blockSplitCtx.fullSeqStoreChunk; |
4040 | seqStore_t* const firstHalfSeqStore = &zc->blockSplitCtx.firstHalfSeqStore; |
4041 | seqStore_t* const secondHalfSeqStore = &zc->blockSplitCtx.secondHalfSeqStore; |
4042 | size_t estimatedOriginalSize; |
4043 | size_t estimatedFirstHalfSize; |
4044 | size_t estimatedSecondHalfSize; |
4045 | size_t midIdx = (startIdx + endIdx)/2; |
4046 | |
4047 | DEBUGLOG(5, "ZSTD_deriveBlockSplitsHelper: startIdx=%zu endIdx=%zu" , startIdx, endIdx); |
4048 | assert(endIdx >= startIdx); |
4049 | if (endIdx - startIdx < MIN_SEQUENCES_BLOCK_SPLITTING || splits->idx >= ZSTD_MAX_NB_BLOCK_SPLITS) { |
4050 | DEBUGLOG(6, "ZSTD_deriveBlockSplitsHelper: Too few sequences (%zu)" , endIdx - startIdx); |
4051 | return; |
4052 | } |
4053 | ZSTD_deriveSeqStoreChunk(fullSeqStoreChunk, origSeqStore, startIdx, endIdx); |
4054 | ZSTD_deriveSeqStoreChunk(firstHalfSeqStore, origSeqStore, startIdx, midIdx); |
4055 | ZSTD_deriveSeqStoreChunk(secondHalfSeqStore, origSeqStore, midIdx, endIdx); |
4056 | estimatedOriginalSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(fullSeqStoreChunk, zc); |
4057 | estimatedFirstHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(firstHalfSeqStore, zc); |
4058 | estimatedSecondHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(secondHalfSeqStore, zc); |
4059 | DEBUGLOG(5, "Estimated original block size: %zu -- First half split: %zu -- Second half split: %zu" , |
4060 | estimatedOriginalSize, estimatedFirstHalfSize, estimatedSecondHalfSize); |
4061 | if (ZSTD_isError(estimatedOriginalSize) || ZSTD_isError(estimatedFirstHalfSize) || ZSTD_isError(estimatedSecondHalfSize)) { |
4062 | return; |
4063 | } |
4064 | if (estimatedFirstHalfSize + estimatedSecondHalfSize < estimatedOriginalSize) { |
4065 | DEBUGLOG(5, "split decided at seqNb:%zu" , midIdx); |
4066 | ZSTD_deriveBlockSplitsHelper(splits, startIdx, midIdx, zc, origSeqStore); |
4067 | splits->splitLocations[splits->idx] = (U32)midIdx; |
4068 | splits->idx++; |
4069 | ZSTD_deriveBlockSplitsHelper(splits, midIdx, endIdx, zc, origSeqStore); |
4070 | } |
4071 | } |
4072 | |
4073 | /* Base recursive function. |
4074 | * Populates a table with intra-block partition indices that can improve compression ratio. |
4075 | * |
4076 | * @return: number of splits made (which equals the size of the partition table - 1). |
4077 | */ |
4078 | static size_t ZSTD_deriveBlockSplits(ZSTD_CCtx* zc, U32 partitions[], U32 nbSeq) |
4079 | { |
4080 | seqStoreSplits splits; |
4081 | splits.splitLocations = partitions; |
4082 | splits.idx = 0; |
4083 | if (nbSeq <= 4) { |
4084 | DEBUGLOG(5, "ZSTD_deriveBlockSplits: Too few sequences to split (%u <= 4)" , nbSeq); |
4085 | /* Refuse to try and split anything with less than 4 sequences */ |
4086 | return 0; |
4087 | } |
4088 | ZSTD_deriveBlockSplitsHelper(&splits, 0, nbSeq, zc, &zc->seqStore); |
4089 | splits.splitLocations[splits.idx] = nbSeq; |
4090 | DEBUGLOG(5, "ZSTD_deriveBlockSplits: final nb partitions: %zu" , splits.idx+1); |
4091 | return splits.idx; |
4092 | } |
4093 | |
4094 | /* ZSTD_compressBlock_splitBlock(): |
4095 | * Attempts to split a given block into multiple blocks to improve compression ratio. |
4096 | * |
4097 | * Returns combined size of all blocks (which includes headers), or a ZSTD error code. |
4098 | */ |
4099 | static size_t |
4100 | ZSTD_compressBlock_splitBlock_internal(ZSTD_CCtx* zc, |
4101 | void* dst, size_t dstCapacity, |
4102 | const void* src, size_t blockSize, |
4103 | U32 lastBlock, U32 nbSeq) |
4104 | { |
4105 | size_t cSize = 0; |
4106 | const BYTE* ip = (const BYTE*)src; |
4107 | BYTE* op = (BYTE*)dst; |
4108 | size_t i = 0; |
4109 | size_t srcBytesTotal = 0; |
4110 | U32* const partitions = zc->blockSplitCtx.partitions; /* size == ZSTD_MAX_NB_BLOCK_SPLITS */ |
4111 | seqStore_t* const nextSeqStore = &zc->blockSplitCtx.nextSeqStore; |
4112 | seqStore_t* const currSeqStore = &zc->blockSplitCtx.currSeqStore; |
4113 | size_t const numSplits = ZSTD_deriveBlockSplits(zc, partitions, nbSeq); |
4114 | |
4115 | /* If a block is split and some partitions are emitted as RLE/uncompressed, then repcode history |
4116 | * may become invalid. In order to reconcile potentially invalid repcodes, we keep track of two |
4117 | * separate repcode histories that simulate repcode history on compression and decompression side, |
4118 | * and use the histories to determine whether we must replace a particular repcode with its raw offset. |
4119 | * |
4120 | * 1) cRep gets updated for each partition, regardless of whether the block was emitted as uncompressed |
4121 | * or RLE. This allows us to retrieve the offset value that an invalid repcode references within |
4122 | * a nocompress/RLE block. |
4123 | * 2) dRep gets updated only for compressed partitions, and when a repcode gets replaced, will use |
4124 | * the replacement offset value rather than the original repcode to update the repcode history. |
4125 | * dRep also will be the final repcode history sent to the next block. |
4126 | * |
4127 | * See ZSTD_seqStore_resolveOffCodes() for more details. |
4128 | */ |
4129 | repcodes_t dRep; |
4130 | repcodes_t cRep; |
4131 | ZSTD_memcpy(dRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); |
4132 | ZSTD_memcpy(cRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); |
4133 | ZSTD_memset(nextSeqStore, 0, sizeof(seqStore_t)); |
4134 | |
4135 | DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)" , |
4136 | (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, |
4137 | (unsigned)zc->blockState.matchState.nextToUpdate); |
4138 | |
4139 | if (numSplits == 0) { |
4140 | size_t cSizeSingleBlock = |
4141 | ZSTD_compressSeqStore_singleBlock(zc, &zc->seqStore, |
4142 | &dRep, &cRep, |
4143 | op, dstCapacity, |
4144 | ip, blockSize, |
4145 | lastBlock, 0 /* isPartition */); |
4146 | FORWARD_IF_ERROR(cSizeSingleBlock, "Compressing single block from splitBlock_internal() failed!" ); |
4147 | DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal: No splits" ); |
4148 | assert(zc->blockSize <= ZSTD_BLOCKSIZE_MAX); |
4149 | assert(cSizeSingleBlock <= zc->blockSize + ZSTD_blockHeaderSize); |
4150 | return cSizeSingleBlock; |
4151 | } |
4152 | |
4153 | ZSTD_deriveSeqStoreChunk(currSeqStore, &zc->seqStore, 0, partitions[0]); |
4154 | for (i = 0; i <= numSplits; ++i) { |
4155 | size_t cSizeChunk; |
4156 | U32 const lastPartition = (i == numSplits); |
4157 | U32 lastBlockEntireSrc = 0; |
4158 | |
4159 | size_t srcBytes = ZSTD_countSeqStoreLiteralsBytes(currSeqStore) + ZSTD_countSeqStoreMatchBytes(currSeqStore); |
4160 | srcBytesTotal += srcBytes; |
4161 | if (lastPartition) { |
4162 | /* This is the final partition, need to account for possible last literals */ |
4163 | srcBytes += blockSize - srcBytesTotal; |
4164 | lastBlockEntireSrc = lastBlock; |
4165 | } else { |
4166 | ZSTD_deriveSeqStoreChunk(nextSeqStore, &zc->seqStore, partitions[i], partitions[i+1]); |
4167 | } |
4168 | |
4169 | cSizeChunk = ZSTD_compressSeqStore_singleBlock(zc, currSeqStore, |
4170 | &dRep, &cRep, |
4171 | op, dstCapacity, |
4172 | ip, srcBytes, |
4173 | lastBlockEntireSrc, 1 /* isPartition */); |
4174 | DEBUGLOG(5, "Estimated size: %zu vs %zu : actual size" , |
4175 | ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(currSeqStore, zc), cSizeChunk); |
4176 | FORWARD_IF_ERROR(cSizeChunk, "Compressing chunk failed!" ); |
4177 | |
4178 | ip += srcBytes; |
4179 | op += cSizeChunk; |
4180 | dstCapacity -= cSizeChunk; |
4181 | cSize += cSizeChunk; |
4182 | *currSeqStore = *nextSeqStore; |
4183 | assert(cSizeChunk <= zc->blockSize + ZSTD_blockHeaderSize); |
4184 | } |
4185 | /* cRep and dRep may have diverged during the compression. |
4186 | * If so, we use the dRep repcodes for the next block. |
4187 | */ |
4188 | ZSTD_memcpy(zc->blockState.prevCBlock->rep, dRep.rep, sizeof(repcodes_t)); |
4189 | return cSize; |
4190 | } |
4191 | |
4192 | static size_t |
4193 | ZSTD_compressBlock_splitBlock(ZSTD_CCtx* zc, |
4194 | void* dst, size_t dstCapacity, |
4195 | const void* src, size_t srcSize, U32 lastBlock) |
4196 | { |
4197 | U32 nbSeq; |
4198 | size_t cSize; |
4199 | DEBUGLOG(4, "ZSTD_compressBlock_splitBlock" ); |
4200 | assert(zc->appliedParams.useBlockSplitter == ZSTD_ps_enable); |
4201 | |
4202 | { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize); |
4203 | FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed" ); |
4204 | if (bss == ZSTDbss_noCompress) { |
4205 | if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) |
4206 | zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; |
4207 | cSize = ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock); |
4208 | FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed" ); |
4209 | DEBUGLOG(4, "ZSTD_compressBlock_splitBlock: Nocompress block" ); |
4210 | return cSize; |
4211 | } |
4212 | nbSeq = (U32)(zc->seqStore.sequences - zc->seqStore.sequencesStart); |
4213 | } |
4214 | |
4215 | cSize = ZSTD_compressBlock_splitBlock_internal(zc, dst, dstCapacity, src, srcSize, lastBlock, nbSeq); |
4216 | FORWARD_IF_ERROR(cSize, "Splitting blocks failed!" ); |
4217 | return cSize; |
4218 | } |
4219 | |
4220 | static size_t |
4221 | ZSTD_compressBlock_internal(ZSTD_CCtx* zc, |
4222 | void* dst, size_t dstCapacity, |
4223 | const void* src, size_t srcSize, U32 frame) |
4224 | { |
4225 | /* This is an estimated upper bound for the length of an rle block. |
4226 | * This isn't the actual upper bound. |
4227 | * Finding the real threshold needs further investigation. |
4228 | */ |
4229 | const U32 rleMaxLength = 25; |
4230 | size_t cSize; |
4231 | const BYTE* ip = (const BYTE*)src; |
4232 | BYTE* op = (BYTE*)dst; |
4233 | DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)" , |
4234 | (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, |
4235 | (unsigned)zc->blockState.matchState.nextToUpdate); |
4236 | |
4237 | { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize); |
4238 | FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed" ); |
4239 | if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; } |
4240 | } |
4241 | |
4242 | if (zc->seqCollector.collectSequences) { |
4243 | ZSTD_copyBlockSequences(zc); |
4244 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); |
4245 | return 0; |
4246 | } |
4247 | |
4248 | /* encode sequences and literals */ |
4249 | cSize = ZSTD_entropyCompressSeqStore(&zc->seqStore, |
4250 | &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy, |
4251 | &zc->appliedParams, |
4252 | dst, dstCapacity, |
4253 | srcSize, |
4254 | zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, |
4255 | zc->bmi2); |
4256 | |
4257 | if (frame && |
4258 | /* We don't want to emit our first block as a RLE even if it qualifies because |
4259 | * doing so will cause the decoder (cli only) to throw a "should consume all input error." |
4260 | * This is only an issue for zstd <= v1.4.3 |
4261 | */ |
4262 | !zc->isFirstBlock && |
4263 | cSize < rleMaxLength && |
4264 | ZSTD_isRLE(ip, srcSize)) |
4265 | { |
4266 | cSize = 1; |
4267 | op[0] = ip[0]; |
4268 | } |
4269 | |
4270 | out: |
4271 | if (!ZSTD_isError(cSize) && cSize > 1) { |
4272 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); |
4273 | } |
4274 | /* We check that dictionaries have offset codes available for the first |
4275 | * block. After the first block, the offcode table might not have large |
4276 | * enough codes to represent the offsets in the data. |
4277 | */ |
4278 | if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) |
4279 | zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; |
4280 | |
4281 | return cSize; |
4282 | } |
4283 | |
4284 | static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc, |
4285 | void* dst, size_t dstCapacity, |
4286 | const void* src, size_t srcSize, |
4287 | const size_t bss, U32 lastBlock) |
4288 | { |
4289 | DEBUGLOG(6, "Attempting ZSTD_compressSuperBlock()" ); |
4290 | if (bss == ZSTDbss_compress) { |
4291 | if (/* We don't want to emit our first block as a RLE even if it qualifies because |
4292 | * doing so will cause the decoder (cli only) to throw a "should consume all input error." |
4293 | * This is only an issue for zstd <= v1.4.3 |
4294 | */ |
4295 | !zc->isFirstBlock && |
4296 | ZSTD_maybeRLE(&zc->seqStore) && |
4297 | ZSTD_isRLE((BYTE const*)src, srcSize)) |
4298 | { |
4299 | return ZSTD_rleCompressBlock(dst, dstCapacity, *(BYTE const*)src, srcSize, lastBlock); |
4300 | } |
4301 | /* Attempt superblock compression. |
4302 | * |
4303 | * Note that compressed size of ZSTD_compressSuperBlock() is not bound by the |
4304 | * standard ZSTD_compressBound(). This is a problem, because even if we have |
4305 | * space now, taking an extra byte now could cause us to run out of space later |
4306 | * and violate ZSTD_compressBound(). |
4307 | * |
4308 | * Define blockBound(blockSize) = blockSize + ZSTD_blockHeaderSize. |
4309 | * |
4310 | * In order to respect ZSTD_compressBound() we must attempt to emit a raw |
4311 | * uncompressed block in these cases: |
4312 | * * cSize == 0: Return code for an uncompressed block. |
4313 | * * cSize == dstSize_tooSmall: We may have expanded beyond blockBound(srcSize). |
4314 | * ZSTD_noCompressBlock() will return dstSize_tooSmall if we are really out of |
4315 | * output space. |
4316 | * * cSize >= blockBound(srcSize): We have expanded the block too much so |
4317 | * emit an uncompressed block. |
4318 | */ |
4319 | { size_t const cSize = |
4320 | ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock); |
4321 | if (cSize != ERROR(dstSize_tooSmall)) { |
4322 | size_t const maxCSize = |
4323 | srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy); |
4324 | FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed" ); |
4325 | if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) { |
4326 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); |
4327 | return cSize; |
4328 | } |
4329 | } |
4330 | } |
4331 | } /* if (bss == ZSTDbss_compress)*/ |
4332 | |
4333 | DEBUGLOG(6, "Resorting to ZSTD_noCompressBlock()" ); |
4334 | /* Superblock compression failed, attempt to emit a single no compress block. |
4335 | * The decoder will be able to stream this block since it is uncompressed. |
4336 | */ |
4337 | return ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock); |
4338 | } |
4339 | |
4340 | static size_t ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx* zc, |
4341 | void* dst, size_t dstCapacity, |
4342 | const void* src, size_t srcSize, |
4343 | U32 lastBlock) |
4344 | { |
4345 | size_t cSize = 0; |
4346 | const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize); |
4347 | DEBUGLOG(5, "ZSTD_compressBlock_targetCBlockSize (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u, srcSize=%zu)" , |
4348 | (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate, srcSize); |
4349 | FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed" ); |
4350 | |
4351 | cSize = ZSTD_compressBlock_targetCBlockSize_body(zc, dst, dstCapacity, src, srcSize, bss, lastBlock); |
4352 | FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize_body failed" ); |
4353 | |
4354 | if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) |
4355 | zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; |
4356 | |
4357 | return cSize; |
4358 | } |
4359 | |
4360 | static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, |
4361 | ZSTD_cwksp* ws, |
4362 | ZSTD_CCtx_params const* params, |
4363 | void const* ip, |
4364 | void const* iend) |
4365 | { |
4366 | U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy); |
4367 | U32 const maxDist = (U32)1 << params->cParams.windowLog; |
4368 | if (ZSTD_window_needOverflowCorrection(ms->window, cycleLog, maxDist, ms->loadedDictEnd, ip, iend)) { |
4369 | U32 const correction = ZSTD_window_correctOverflow(&ms->window, cycleLog, maxDist, ip); |
4370 | ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30); |
4371 | ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30); |
4372 | ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31); |
4373 | ZSTD_cwksp_mark_tables_dirty(ws); |
4374 | ZSTD_reduceIndex(ms, params, correction); |
4375 | ZSTD_cwksp_mark_tables_clean(ws); |
4376 | if (ms->nextToUpdate < correction) ms->nextToUpdate = 0; |
4377 | else ms->nextToUpdate -= correction; |
4378 | /* invalidate dictionaries on overflow correction */ |
4379 | ms->loadedDictEnd = 0; |
4380 | ms->dictMatchState = NULL; |
4381 | } |
4382 | } |
4383 | |
4384 | /*! ZSTD_compress_frameChunk() : |
4385 | * Compress a chunk of data into one or multiple blocks. |
4386 | * All blocks will be terminated, all input will be consumed. |
4387 | * Function will issue an error if there is not enough `dstCapacity` to hold the compressed content. |
4388 | * Frame is supposed already started (header already produced) |
4389 | * @return : compressed size, or an error code |
4390 | */ |
4391 | static size_t ZSTD_compress_frameChunk(ZSTD_CCtx* cctx, |
4392 | void* dst, size_t dstCapacity, |
4393 | const void* src, size_t srcSize, |
4394 | U32 lastFrameChunk) |
4395 | { |
4396 | size_t blockSize = cctx->blockSize; |
4397 | size_t remaining = srcSize; |
4398 | const BYTE* ip = (const BYTE*)src; |
4399 | BYTE* const ostart = (BYTE*)dst; |
4400 | BYTE* op = ostart; |
4401 | U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog; |
4402 | |
4403 | assert(cctx->appliedParams.cParams.windowLog <= ZSTD_WINDOWLOG_MAX); |
4404 | |
4405 | DEBUGLOG(4, "ZSTD_compress_frameChunk (blockSize=%u)" , (unsigned)blockSize); |
4406 | if (cctx->appliedParams.fParams.checksumFlag && srcSize) |
4407 | XXH64_update(&cctx->xxhState, src, srcSize); |
4408 | |
4409 | while (remaining) { |
4410 | ZSTD_matchState_t* const ms = &cctx->blockState.matchState; |
4411 | U32 const lastBlock = lastFrameChunk & (blockSize >= remaining); |
4412 | |
4413 | /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding |
4414 | * additional 1. We need to revisit and change this logic to be more consistent */ |
4415 | RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE + 1, |
4416 | dstSize_tooSmall, |
4417 | "not enough space to store compressed block" ); |
4418 | if (remaining < blockSize) blockSize = remaining; |
4419 | |
4420 | ZSTD_overflowCorrectIfNeeded( |
4421 | ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize); |
4422 | ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState); |
4423 | ZSTD_window_enforceMaxDist(&ms->window, ip, maxDist, &ms->loadedDictEnd, &ms->dictMatchState); |
4424 | |
4425 | /* Ensure hash/chain table insertion resumes no sooner than lowlimit */ |
4426 | if (ms->nextToUpdate < ms->window.lowLimit) ms->nextToUpdate = ms->window.lowLimit; |
4427 | |
4428 | { size_t cSize; |
4429 | if (ZSTD_useTargetCBlockSize(&cctx->appliedParams)) { |
4430 | cSize = ZSTD_compressBlock_targetCBlockSize(cctx, op, dstCapacity, ip, blockSize, lastBlock); |
4431 | FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed" ); |
4432 | assert(cSize > 0); |
4433 | assert(cSize <= blockSize + ZSTD_blockHeaderSize); |
4434 | } else if (ZSTD_blockSplitterEnabled(&cctx->appliedParams)) { |
4435 | cSize = ZSTD_compressBlock_splitBlock(cctx, op, dstCapacity, ip, blockSize, lastBlock); |
4436 | FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_splitBlock failed" ); |
4437 | assert(cSize > 0 || cctx->seqCollector.collectSequences == 1); |
4438 | } else { |
4439 | cSize = ZSTD_compressBlock_internal(cctx, |
4440 | op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize, |
4441 | ip, blockSize, 1 /* frame */); |
4442 | FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_internal failed" ); |
4443 | |
4444 | if (cSize == 0) { /* block is not compressible */ |
4445 | cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); |
4446 | FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed" ); |
4447 | } else { |
4448 | U32 const = cSize == 1 ? |
4449 | lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) : |
4450 | lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3); |
4451 | MEM_writeLE24(op, cBlockHeader); |
4452 | cSize += ZSTD_blockHeaderSize; |
4453 | } |
4454 | } /* if (ZSTD_useTargetCBlockSize(&cctx->appliedParams))*/ |
4455 | |
4456 | |
4457 | ip += blockSize; |
4458 | assert(remaining >= blockSize); |
4459 | remaining -= blockSize; |
4460 | op += cSize; |
4461 | assert(dstCapacity >= cSize); |
4462 | dstCapacity -= cSize; |
4463 | cctx->isFirstBlock = 0; |
4464 | DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u" , |
4465 | (unsigned)cSize); |
4466 | } } |
4467 | |
4468 | if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending; |
4469 | return (size_t)(op-ostart); |
4470 | } |
4471 | |
4472 | |
4473 | static size_t (void* dst, size_t dstCapacity, |
4474 | const ZSTD_CCtx_params* params, U64 pledgedSrcSize, U32 dictID) |
4475 | { BYTE* const op = (BYTE*)dst; |
4476 | U32 const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536); /* 0-3 */ |
4477 | U32 const dictIDSizeCode = params->fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength; /* 0-3 */ |
4478 | U32 const checksumFlag = params->fParams.checksumFlag>0; |
4479 | U32 const windowSize = (U32)1 << params->cParams.windowLog; |
4480 | U32 const singleSegment = params->fParams.contentSizeFlag && (windowSize >= pledgedSrcSize); |
4481 | BYTE const windowLogByte = (BYTE)((params->cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3); |
4482 | U32 const fcsCode = params->fParams.contentSizeFlag ? |
4483 | (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0; /* 0-3 */ |
4484 | BYTE const = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) ); |
4485 | size_t pos=0; |
4486 | |
4487 | assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)); |
4488 | RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall, |
4489 | "dst buf is too small to fit worst-case frame header size." ); |
4490 | DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u" , |
4491 | !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode); |
4492 | if (params->format == ZSTD_f_zstd1) { |
4493 | MEM_writeLE32(dst, ZSTD_MAGICNUMBER); |
4494 | pos = 4; |
4495 | } |
4496 | op[pos++] = frameHeaderDescriptionByte; |
4497 | if (!singleSegment) op[pos++] = windowLogByte; |
4498 | switch(dictIDSizeCode) |
4499 | { |
4500 | default: |
4501 | assert(0); /* impossible */ |
4502 | ZSTD_FALLTHROUGH; |
4503 | case 0 : break; |
4504 | case 1 : op[pos] = (BYTE)(dictID); pos++; break; |
4505 | case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break; |
4506 | case 3 : MEM_writeLE32(op+pos, dictID); pos+=4; break; |
4507 | } |
4508 | switch(fcsCode) |
4509 | { |
4510 | default: |
4511 | assert(0); /* impossible */ |
4512 | ZSTD_FALLTHROUGH; |
4513 | case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break; |
4514 | case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break; |
4515 | case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break; |
4516 | case 3 : MEM_writeLE64(op+pos, (U64)(pledgedSrcSize)); pos+=8; break; |
4517 | } |
4518 | return pos; |
4519 | } |
4520 | |
4521 | /* ZSTD_writeSkippableFrame_advanced() : |
4522 | * Writes out a skippable frame with the specified magic number variant (16 are supported), |
4523 | * from ZSTD_MAGIC_SKIPPABLE_START to ZSTD_MAGIC_SKIPPABLE_START+15, and the desired source data. |
4524 | * |
4525 | * Returns the total number of bytes written, or a ZSTD error code. |
4526 | */ |
4527 | size_t ZSTD_writeSkippableFrame(void* dst, size_t dstCapacity, |
4528 | const void* src, size_t srcSize, unsigned magicVariant) { |
4529 | BYTE* op = (BYTE*)dst; |
4530 | RETURN_ERROR_IF(dstCapacity < srcSize + ZSTD_SKIPPABLEHEADERSIZE /* Skippable frame overhead */, |
4531 | dstSize_tooSmall, "Not enough room for skippable frame" ); |
4532 | RETURN_ERROR_IF(srcSize > (unsigned)0xFFFFFFFF, srcSize_wrong, "Src size too large for skippable frame" ); |
4533 | RETURN_ERROR_IF(magicVariant > 15, parameter_outOfBound, "Skippable frame magic number variant not supported" ); |
4534 | |
4535 | MEM_writeLE32(op, (U32)(ZSTD_MAGIC_SKIPPABLE_START + magicVariant)); |
4536 | MEM_writeLE32(op+4, (U32)srcSize); |
4537 | ZSTD_memcpy(op+8, src, srcSize); |
4538 | return srcSize + ZSTD_SKIPPABLEHEADERSIZE; |
4539 | } |
4540 | |
4541 | /* ZSTD_writeLastEmptyBlock() : |
4542 | * output an empty Block with end-of-frame mark to complete a frame |
4543 | * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h)) |
4544 | * or an error code if `dstCapacity` is too small (<ZSTD_blockHeaderSize) |
4545 | */ |
4546 | size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity) |
4547 | { |
4548 | RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, |
4549 | "dst buf is too small to write frame trailer empty block." ); |
4550 | { U32 const = 1 /*lastBlock*/ + (((U32)bt_raw)<<1); /* 0 size */ |
4551 | MEM_writeLE24(dst, cBlockHeader24); |
4552 | return ZSTD_blockHeaderSize; |
4553 | } |
4554 | } |
4555 | |
4556 | size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq) |
4557 | { |
4558 | RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong, |
4559 | "wrong cctx stage" ); |
4560 | RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable, |
4561 | parameter_unsupported, |
4562 | "incompatible with ldm" ); |
4563 | cctx->externSeqStore.seq = seq; |
4564 | cctx->externSeqStore.size = nbSeq; |
4565 | cctx->externSeqStore.capacity = nbSeq; |
4566 | cctx->externSeqStore.pos = 0; |
4567 | cctx->externSeqStore.posInSequence = 0; |
4568 | return 0; |
4569 | } |
4570 | |
4571 | |
4572 | static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx, |
4573 | void* dst, size_t dstCapacity, |
4574 | const void* src, size_t srcSize, |
4575 | U32 frame, U32 lastFrameChunk) |
4576 | { |
4577 | ZSTD_matchState_t* const ms = &cctx->blockState.matchState; |
4578 | size_t fhSize = 0; |
4579 | |
4580 | DEBUGLOG(5, "ZSTD_compressContinue_internal, stage: %u, srcSize: %u" , |
4581 | cctx->stage, (unsigned)srcSize); |
4582 | RETURN_ERROR_IF(cctx->stage==ZSTDcs_created, stage_wrong, |
4583 | "missing init (ZSTD_compressBegin)" ); |
4584 | |
4585 | if (frame && (cctx->stage==ZSTDcs_init)) { |
4586 | fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, |
4587 | cctx->pledgedSrcSizePlusOne-1, cctx->dictID); |
4588 | FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed" ); |
4589 | assert(fhSize <= dstCapacity); |
4590 | dstCapacity -= fhSize; |
4591 | dst = (char*)dst + fhSize; |
4592 | cctx->stage = ZSTDcs_ongoing; |
4593 | } |
4594 | |
4595 | if (!srcSize) return fhSize; /* do not generate an empty block if no input */ |
4596 | |
4597 | if (!ZSTD_window_update(&ms->window, src, srcSize, ms->forceNonContiguous)) { |
4598 | ms->forceNonContiguous = 0; |
4599 | ms->nextToUpdate = ms->window.dictLimit; |
4600 | } |
4601 | if (cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) { |
4602 | ZSTD_window_update(&cctx->ldmState.window, src, srcSize, /* forceNonContiguous */ 0); |
4603 | } |
4604 | |
4605 | if (!frame) { |
4606 | /* overflow check and correction for block mode */ |
4607 | ZSTD_overflowCorrectIfNeeded( |
4608 | ms, &cctx->workspace, &cctx->appliedParams, |
4609 | src, (BYTE const*)src + srcSize); |
4610 | } |
4611 | |
4612 | DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)" , (unsigned)cctx->blockSize); |
4613 | { size_t const cSize = frame ? |
4614 | ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) : |
4615 | ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */); |
4616 | FORWARD_IF_ERROR(cSize, "%s" , frame ? "ZSTD_compress_frameChunk failed" : "ZSTD_compressBlock_internal failed" ); |
4617 | cctx->consumedSrcSize += srcSize; |
4618 | cctx->producedCSize += (cSize + fhSize); |
4619 | assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0)); |
4620 | if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */ |
4621 | ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1); |
4622 | RETURN_ERROR_IF( |
4623 | cctx->consumedSrcSize+1 > cctx->pledgedSrcSizePlusOne, |
4624 | srcSize_wrong, |
4625 | "error : pledgedSrcSize = %u, while realSrcSize >= %u" , |
4626 | (unsigned)cctx->pledgedSrcSizePlusOne-1, |
4627 | (unsigned)cctx->consumedSrcSize); |
4628 | } |
4629 | return cSize + fhSize; |
4630 | } |
4631 | } |
4632 | |
4633 | size_t ZSTD_compressContinue_public(ZSTD_CCtx* cctx, |
4634 | void* dst, size_t dstCapacity, |
4635 | const void* src, size_t srcSize) |
4636 | { |
4637 | DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)" , (unsigned)srcSize); |
4638 | return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1 /* frame mode */, 0 /* last chunk */); |
4639 | } |
4640 | |
4641 | /* NOTE: Must just wrap ZSTD_compressContinue_public() */ |
4642 | size_t ZSTD_compressContinue(ZSTD_CCtx* cctx, |
4643 | void* dst, size_t dstCapacity, |
4644 | const void* src, size_t srcSize) |
4645 | { |
4646 | return ZSTD_compressContinue_public(cctx, dst, dstCapacity, src, srcSize); |
4647 | } |
4648 | |
4649 | static size_t ZSTD_getBlockSize_deprecated(const ZSTD_CCtx* cctx) |
4650 | { |
4651 | ZSTD_compressionParameters const cParams = cctx->appliedParams.cParams; |
4652 | assert(!ZSTD_checkCParams(cParams)); |
4653 | return MIN(cctx->appliedParams.maxBlockSize, (size_t)1 << cParams.windowLog); |
4654 | } |
4655 | |
4656 | /* NOTE: Must just wrap ZSTD_getBlockSize_deprecated() */ |
4657 | size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx) |
4658 | { |
4659 | return ZSTD_getBlockSize_deprecated(cctx); |
4660 | } |
4661 | |
4662 | /* NOTE: Must just wrap ZSTD_compressBlock_deprecated() */ |
4663 | size_t ZSTD_compressBlock_deprecated(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
4664 | { |
4665 | DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u" , (unsigned)srcSize); |
4666 | { size_t const blockSizeMax = ZSTD_getBlockSize_deprecated(cctx); |
4667 | RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong, "input is larger than a block" ); } |
4668 | |
4669 | return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */); |
4670 | } |
4671 | |
4672 | /* NOTE: Must just wrap ZSTD_compressBlock_deprecated() */ |
4673 | size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) |
4674 | { |
4675 | return ZSTD_compressBlock_deprecated(cctx, dst, dstCapacity, src, srcSize); |
4676 | } |
4677 | |
4678 | /*! ZSTD_loadDictionaryContent() : |
4679 | * @return : 0, or an error code |
4680 | */ |
4681 | static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms, |
4682 | ldmState_t* ls, |
4683 | ZSTD_cwksp* ws, |
4684 | ZSTD_CCtx_params const* params, |
4685 | const void* src, size_t srcSize, |
4686 | ZSTD_dictTableLoadMethod_e dtlm, |
4687 | ZSTD_tableFillPurpose_e tfp) |
4688 | { |
4689 | const BYTE* ip = (const BYTE*) src; |
4690 | const BYTE* const iend = ip + srcSize; |
4691 | int const loadLdmDict = params->ldmParams.enableLdm == ZSTD_ps_enable && ls != NULL; |
4692 | |
4693 | /* Assert that the ms params match the params we're being given */ |
4694 | ZSTD_assertEqualCParams(params->cParams, ms->cParams); |
4695 | |
4696 | { /* Ensure large dictionaries can't cause index overflow */ |
4697 | |
4698 | /* Allow the dictionary to set indices up to exactly ZSTD_CURRENT_MAX. |
4699 | * Dictionaries right at the edge will immediately trigger overflow |
4700 | * correction, but I don't want to insert extra constraints here. |
4701 | */ |
4702 | U32 maxDictSize = ZSTD_CURRENT_MAX - ZSTD_WINDOW_START_INDEX; |
4703 | |
4704 | int const CDictTaggedIndices = ZSTD_CDictIndicesAreTagged(¶ms->cParams); |
4705 | if (CDictTaggedIndices && tfp == ZSTD_tfp_forCDict) { |
4706 | /* Some dictionary matchfinders in zstd use "short cache", |
4707 | * which treats the lower ZSTD_SHORT_CACHE_TAG_BITS of each |
4708 | * CDict hashtable entry as a tag rather than as part of an index. |
4709 | * When short cache is used, we need to truncate the dictionary |
4710 | * so that its indices don't overlap with the tag. */ |
4711 | U32 const shortCacheMaxDictSize = (1u << (32 - ZSTD_SHORT_CACHE_TAG_BITS)) - ZSTD_WINDOW_START_INDEX; |
4712 | maxDictSize = MIN(maxDictSize, shortCacheMaxDictSize); |
4713 | assert(!loadLdmDict); |
4714 | } |
4715 | |
4716 | /* If the dictionary is too large, only load the suffix of the dictionary. */ |
4717 | if (srcSize > maxDictSize) { |
4718 | ip = iend - maxDictSize; |
4719 | src = ip; |
4720 | srcSize = maxDictSize; |
4721 | } |
4722 | } |
4723 | |
4724 | if (srcSize > ZSTD_CHUNKSIZE_MAX) { |
4725 | /* We must have cleared our windows when our source is this large. */ |
4726 | assert(ZSTD_window_isEmpty(ms->window)); |
4727 | if (loadLdmDict) assert(ZSTD_window_isEmpty(ls->window)); |
4728 | } |
4729 | ZSTD_window_update(&ms->window, src, srcSize, /* forceNonContiguous */ 0); |
4730 | |
4731 | DEBUGLOG(4, "ZSTD_loadDictionaryContent(): useRowMatchFinder=%d" , (int)params->useRowMatchFinder); |
4732 | |
4733 | if (loadLdmDict) { /* Load the entire dict into LDM matchfinders. */ |
4734 | ZSTD_window_update(&ls->window, src, srcSize, /* forceNonContiguous */ 0); |
4735 | ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base); |
4736 | ZSTD_ldm_fillHashTable(ls, ip, iend, ¶ms->ldmParams); |
4737 | } |
4738 | |
4739 | /* If the dict is larger than we can reasonably index in our tables, only load the suffix. */ |
4740 | if (params->cParams.strategy < ZSTD_btultra) { |
4741 | U32 maxDictSize = 8U << MIN(MAX(params->cParams.hashLog, params->cParams.chainLog), 28); |
4742 | if (srcSize > maxDictSize) { |
4743 | ip = iend - maxDictSize; |
4744 | src = ip; |
4745 | srcSize = maxDictSize; |
4746 | } |
4747 | } |
4748 | |
4749 | ms->nextToUpdate = (U32)(ip - ms->window.base); |
4750 | ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base); |
4751 | ms->forceNonContiguous = params->deterministicRefPrefix; |
4752 | |
4753 | if (srcSize <= HASH_READ_SIZE) return 0; |
4754 | |
4755 | ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, iend); |
4756 | |
4757 | switch(params->cParams.strategy) |
4758 | { |
4759 | case ZSTD_fast: |
4760 | ZSTD_fillHashTable(ms, iend, dtlm, tfp); |
4761 | break; |
4762 | case ZSTD_dfast: |
4763 | ZSTD_fillDoubleHashTable(ms, iend, dtlm, tfp); |
4764 | break; |
4765 | |
4766 | case ZSTD_greedy: |
4767 | case ZSTD_lazy: |
4768 | case ZSTD_lazy2: |
4769 | assert(srcSize >= HASH_READ_SIZE); |
4770 | if (ms->dedicatedDictSearch) { |
4771 | assert(ms->chainTable != NULL); |
4772 | ZSTD_dedicatedDictSearch_lazy_loadDictionary(ms, iend-HASH_READ_SIZE); |
4773 | } else { |
4774 | assert(params->useRowMatchFinder != ZSTD_ps_auto); |
4775 | if (params->useRowMatchFinder == ZSTD_ps_enable) { |
4776 | size_t const tagTableSize = ((size_t)1 << params->cParams.hashLog); |
4777 | ZSTD_memset(ms->tagTable, 0, tagTableSize); |
4778 | ZSTD_row_update(ms, iend-HASH_READ_SIZE); |
4779 | DEBUGLOG(4, "Using row-based hash table for lazy dict" ); |
4780 | } else { |
4781 | ZSTD_insertAndFindFirstIndex(ms, iend-HASH_READ_SIZE); |
4782 | DEBUGLOG(4, "Using chain-based hash table for lazy dict" ); |
4783 | } |
4784 | } |
4785 | break; |
4786 | |
4787 | case ZSTD_btlazy2: /* we want the dictionary table fully sorted */ |
4788 | case ZSTD_btopt: |
4789 | case ZSTD_btultra: |
4790 | case ZSTD_btultra2: |
4791 | assert(srcSize >= HASH_READ_SIZE); |
4792 | ZSTD_updateTree(ms, iend-HASH_READ_SIZE, iend); |
4793 | break; |
4794 | |
4795 | default: |
4796 | assert(0); /* not possible : not a valid strategy id */ |
4797 | } |
4798 | |
4799 | ms->nextToUpdate = (U32)(iend - ms->window.base); |
4800 | return 0; |
4801 | } |
4802 | |
4803 | |
4804 | /* Dictionaries that assign zero probability to symbols that show up causes problems |
4805 | * when FSE encoding. Mark dictionaries with zero probability symbols as FSE_repeat_check |
4806 | * and only dictionaries with 100% valid symbols can be assumed valid. |
4807 | */ |
4808 | static FSE_repeat ZSTD_dictNCountRepeat(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) |
4809 | { |
4810 | U32 s; |
4811 | if (dictMaxSymbolValue < maxSymbolValue) { |
4812 | return FSE_repeat_check; |
4813 | } |
4814 | for (s = 0; s <= maxSymbolValue; ++s) { |
4815 | if (normalizedCounter[s] == 0) { |
4816 | return FSE_repeat_check; |
4817 | } |
4818 | } |
4819 | return FSE_repeat_valid; |
4820 | } |
4821 | |
4822 | size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, |
4823 | const void* const dict, size_t dictSize) |
4824 | { |
4825 | short offcodeNCount[MaxOff+1]; |
4826 | unsigned offcodeMaxValue = MaxOff; |
4827 | const BYTE* dictPtr = (const BYTE*)dict; /* skip magic num and dict ID */ |
4828 | const BYTE* const dictEnd = dictPtr + dictSize; |
4829 | dictPtr += 8; |
4830 | bs->entropy.huf.repeatMode = HUF_repeat_check; |
4831 | |
4832 | { unsigned maxSymbolValue = 255; |
4833 | unsigned hasZeroWeights = 1; |
4834 | size_t const = HUF_readCTable((HUF_CElt*)bs->entropy.huf.CTable, &maxSymbolValue, dictPtr, |
4835 | dictEnd-dictPtr, &hasZeroWeights); |
4836 | |
4837 | /* We only set the loaded table as valid if it contains all non-zero |
4838 | * weights. Otherwise, we set it to check */ |
4839 | if (!hasZeroWeights) |
4840 | bs->entropy.huf.repeatMode = HUF_repeat_valid; |
4841 | |
4842 | RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted, "" ); |
4843 | RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted, "" ); |
4844 | dictPtr += hufHeaderSize; |
4845 | } |
4846 | |
4847 | { unsigned offcodeLog; |
4848 | size_t const = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr); |
4849 | RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "" ); |
4850 | RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "" ); |
4851 | /* fill all offset symbols to avoid garbage at end of table */ |
4852 | RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( |
4853 | bs->entropy.fse.offcodeCTable, |
4854 | offcodeNCount, MaxOff, offcodeLog, |
4855 | workspace, HUF_WORKSPACE_SIZE)), |
4856 | dictionary_corrupted, "" ); |
4857 | /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */ |
4858 | dictPtr += offcodeHeaderSize; |
4859 | } |
4860 | |
4861 | { short matchlengthNCount[MaxML+1]; |
4862 | unsigned matchlengthMaxValue = MaxML, matchlengthLog; |
4863 | size_t const = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr); |
4864 | RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "" ); |
4865 | RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "" ); |
4866 | RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( |
4867 | bs->entropy.fse.matchlengthCTable, |
4868 | matchlengthNCount, matchlengthMaxValue, matchlengthLog, |
4869 | workspace, HUF_WORKSPACE_SIZE)), |
4870 | dictionary_corrupted, "" ); |
4871 | bs->entropy.fse.matchlength_repeatMode = ZSTD_dictNCountRepeat(matchlengthNCount, matchlengthMaxValue, MaxML); |
4872 | dictPtr += matchlengthHeaderSize; |
4873 | } |
4874 | |
4875 | { short litlengthNCount[MaxLL+1]; |
4876 | unsigned litlengthMaxValue = MaxLL, litlengthLog; |
4877 | size_t const = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr); |
4878 | RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "" ); |
4879 | RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "" ); |
4880 | RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( |
4881 | bs->entropy.fse.litlengthCTable, |
4882 | litlengthNCount, litlengthMaxValue, litlengthLog, |
4883 | workspace, HUF_WORKSPACE_SIZE)), |
4884 | dictionary_corrupted, "" ); |
4885 | bs->entropy.fse.litlength_repeatMode = ZSTD_dictNCountRepeat(litlengthNCount, litlengthMaxValue, MaxLL); |
4886 | dictPtr += litlengthHeaderSize; |
4887 | } |
4888 | |
4889 | RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "" ); |
4890 | bs->rep[0] = MEM_readLE32(dictPtr+0); |
4891 | bs->rep[1] = MEM_readLE32(dictPtr+4); |
4892 | bs->rep[2] = MEM_readLE32(dictPtr+8); |
4893 | dictPtr += 12; |
4894 | |
4895 | { size_t const dictContentSize = (size_t)(dictEnd - dictPtr); |
4896 | U32 offcodeMax = MaxOff; |
4897 | if (dictContentSize <= ((U32)-1) - 128 KB) { |
4898 | U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */ |
4899 | offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */ |
4900 | } |
4901 | /* All offset values <= dictContentSize + 128 KB must be representable for a valid table */ |
4902 | bs->entropy.fse.offcode_repeatMode = ZSTD_dictNCountRepeat(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)); |
4903 | |
4904 | /* All repCodes must be <= dictContentSize and != 0 */ |
4905 | { U32 u; |
4906 | for (u=0; u<3; u++) { |
4907 | RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, "" ); |
4908 | RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, "" ); |
4909 | } } } |
4910 | |
4911 | return dictPtr - (const BYTE*)dict; |
4912 | } |
4913 | |
4914 | /* Dictionary format : |
4915 | * See : |
4916 | * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#dictionary-format |
4917 | */ |
4918 | /*! ZSTD_loadZstdDictionary() : |
4919 | * @return : dictID, or an error code |
4920 | * assumptions : magic number supposed already checked |
4921 | * dictSize supposed >= 8 |
4922 | */ |
4923 | static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs, |
4924 | ZSTD_matchState_t* ms, |
4925 | ZSTD_cwksp* ws, |
4926 | ZSTD_CCtx_params const* params, |
4927 | const void* dict, size_t dictSize, |
4928 | ZSTD_dictTableLoadMethod_e dtlm, |
4929 | ZSTD_tableFillPurpose_e tfp, |
4930 | void* workspace) |
4931 | { |
4932 | const BYTE* dictPtr = (const BYTE*)dict; |
4933 | const BYTE* const dictEnd = dictPtr + dictSize; |
4934 | size_t dictID; |
4935 | size_t eSize; |
4936 | ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog))); |
4937 | assert(dictSize >= 8); |
4938 | assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY); |
4939 | |
4940 | dictID = params->fParams.noDictIDFlag ? 0 : MEM_readLE32(dictPtr + 4 /* skip magic number */ ); |
4941 | eSize = ZSTD_loadCEntropy(bs, workspace, dict, dictSize); |
4942 | FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed" ); |
4943 | dictPtr += eSize; |
4944 | |
4945 | { |
4946 | size_t const dictContentSize = (size_t)(dictEnd - dictPtr); |
4947 | FORWARD_IF_ERROR(ZSTD_loadDictionaryContent( |
4948 | ms, NULL, ws, params, dictPtr, dictContentSize, dtlm, tfp), "" ); |
4949 | } |
4950 | return dictID; |
4951 | } |
4952 | |
4953 | /** ZSTD_compress_insertDictionary() : |
4954 | * @return : dictID, or an error code */ |
4955 | static size_t |
4956 | ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs, |
4957 | ZSTD_matchState_t* ms, |
4958 | ldmState_t* ls, |
4959 | ZSTD_cwksp* ws, |
4960 | const ZSTD_CCtx_params* params, |
4961 | const void* dict, size_t dictSize, |
4962 | ZSTD_dictContentType_e dictContentType, |
4963 | ZSTD_dictTableLoadMethod_e dtlm, |
4964 | ZSTD_tableFillPurpose_e tfp, |
4965 | void* workspace) |
4966 | { |
4967 | DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)" , (U32)dictSize); |
4968 | if ((dict==NULL) || (dictSize<8)) { |
4969 | RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "" ); |
4970 | return 0; |
4971 | } |
4972 | |
4973 | ZSTD_reset_compressedBlockState(bs); |
4974 | |
4975 | /* dict restricted modes */ |
4976 | if (dictContentType == ZSTD_dct_rawContent) |
4977 | return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm, tfp); |
4978 | |
4979 | if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) { |
4980 | if (dictContentType == ZSTD_dct_auto) { |
4981 | DEBUGLOG(4, "raw content dictionary detected" ); |
4982 | return ZSTD_loadDictionaryContent( |
4983 | ms, ls, ws, params, dict, dictSize, dtlm, tfp); |
4984 | } |
4985 | RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "" ); |
4986 | assert(0); /* impossible */ |
4987 | } |
4988 | |
4989 | /* dict as full zstd dictionary */ |
4990 | return ZSTD_loadZstdDictionary( |
4991 | bs, ms, ws, params, dict, dictSize, dtlm, tfp, workspace); |
4992 | } |
4993 | |
4994 | #define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB) |
4995 | #define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6ULL) |
4996 | |
4997 | /*! ZSTD_compressBegin_internal() : |
4998 | * Assumption : either @dict OR @cdict (or none) is non-NULL, never both |
4999 | * @return : 0, or an error code */ |
5000 | static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx, |
5001 | const void* dict, size_t dictSize, |
5002 | ZSTD_dictContentType_e dictContentType, |
5003 | ZSTD_dictTableLoadMethod_e dtlm, |
5004 | const ZSTD_CDict* cdict, |
5005 | const ZSTD_CCtx_params* params, U64 pledgedSrcSize, |
5006 | ZSTD_buffered_policy_e zbuff) |
5007 | { |
5008 | size_t const dictContentSize = cdict ? cdict->dictContentSize : dictSize; |
5009 | #if ZSTD_TRACE |
5010 | cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0; |
5011 | #endif |
5012 | DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u" , params->cParams.windowLog); |
5013 | /* params are supposed to be fully validated at this point */ |
5014 | assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams))); |
5015 | assert(!((dict) && (cdict))); /* either dict or cdict, not both */ |
5016 | if ( (cdict) |
5017 | && (cdict->dictContentSize > 0) |
5018 | && ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF |
5019 | || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER |
5020 | || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN |
5021 | || cdict->compressionLevel == 0) |
5022 | && (params->attachDictPref != ZSTD_dictForceLoad) ) { |
5023 | return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff); |
5024 | } |
5025 | |
5026 | FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize, |
5027 | dictContentSize, |
5028 | ZSTDcrp_makeClean, zbuff) , "" ); |
5029 | { size_t const dictID = cdict ? |
5030 | ZSTD_compress_insertDictionary( |
5031 | cctx->blockState.prevCBlock, &cctx->blockState.matchState, |
5032 | &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent, |
5033 | cdict->dictContentSize, cdict->dictContentType, dtlm, |
5034 | ZSTD_tfp_forCCtx, cctx->entropyWorkspace) |
5035 | : ZSTD_compress_insertDictionary( |
5036 | cctx->blockState.prevCBlock, &cctx->blockState.matchState, |
5037 | &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize, |
5038 | dictContentType, dtlm, ZSTD_tfp_forCCtx, cctx->entropyWorkspace); |
5039 | FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed" ); |
5040 | assert(dictID <= UINT_MAX); |
5041 | cctx->dictID = (U32)dictID; |
5042 | cctx->dictContentSize = dictContentSize; |
5043 | } |
5044 | return 0; |
5045 | } |
5046 | |
5047 | size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx, |
5048 | const void* dict, size_t dictSize, |
5049 | ZSTD_dictContentType_e dictContentType, |
5050 | ZSTD_dictTableLoadMethod_e dtlm, |
5051 | const ZSTD_CDict* cdict, |
5052 | const ZSTD_CCtx_params* params, |
5053 | unsigned long long pledgedSrcSize) |
5054 | { |
5055 | DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u" , params->cParams.windowLog); |
5056 | /* compression parameters verification and optimization */ |
5057 | FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) , "" ); |
5058 | return ZSTD_compressBegin_internal(cctx, |
5059 | dict, dictSize, dictContentType, dtlm, |
5060 | cdict, |
5061 | params, pledgedSrcSize, |
5062 | ZSTDb_not_buffered); |
5063 | } |
5064 | |
5065 | /*! ZSTD_compressBegin_advanced() : |
5066 | * @return : 0, or an error code */ |
5067 | size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, |
5068 | const void* dict, size_t dictSize, |
5069 | ZSTD_parameters params, unsigned long long pledgedSrcSize) |
5070 | { |
5071 | ZSTD_CCtx_params cctxParams; |
5072 | ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, ZSTD_NO_CLEVEL); |
5073 | return ZSTD_compressBegin_advanced_internal(cctx, |
5074 | dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, |
5075 | NULL /*cdict*/, |
5076 | &cctxParams, pledgedSrcSize); |
5077 | } |
5078 | |
5079 | static size_t |
5080 | ZSTD_compressBegin_usingDict_deprecated(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel) |
5081 | { |
5082 | ZSTD_CCtx_params cctxParams; |
5083 | { ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_noAttachDict); |
5084 | ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel); |
5085 | } |
5086 | DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)" , (unsigned)dictSize); |
5087 | return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL, |
5088 | &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered); |
5089 | } |
5090 | |
5091 | size_t |
5092 | ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel) |
5093 | { |
5094 | return ZSTD_compressBegin_usingDict_deprecated(cctx, dict, dictSize, compressionLevel); |
5095 | } |
5096 | |
5097 | size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel) |
5098 | { |
5099 | return ZSTD_compressBegin_usingDict_deprecated(cctx, NULL, 0, compressionLevel); |
5100 | } |
5101 | |
5102 | |
5103 | /*! ZSTD_writeEpilogue() : |
5104 | * Ends a frame. |
5105 | * @return : nb of bytes written into dst (or an error code) */ |
5106 | static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity) |
5107 | { |
5108 | BYTE* const ostart = (BYTE*)dst; |
5109 | BYTE* op = ostart; |
5110 | size_t fhSize = 0; |
5111 | |
5112 | DEBUGLOG(4, "ZSTD_writeEpilogue" ); |
5113 | RETURN_ERROR_IF(cctx->stage == ZSTDcs_created, stage_wrong, "init missing" ); |
5114 | |
5115 | /* special case : empty frame */ |
5116 | if (cctx->stage == ZSTDcs_init) { |
5117 | fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0); |
5118 | FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed" ); |
5119 | dstCapacity -= fhSize; |
5120 | op += fhSize; |
5121 | cctx->stage = ZSTDcs_ongoing; |
5122 | } |
5123 | |
5124 | if (cctx->stage != ZSTDcs_ending) { |
5125 | /* write one last empty block, make it the "last" block */ |
5126 | U32 const = 1 /* last block */ + (((U32)bt_raw)<<1) + 0; |
5127 | RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for epilogue" ); |
5128 | MEM_writeLE32(op, cBlockHeader24); |
5129 | op += ZSTD_blockHeaderSize; |
5130 | dstCapacity -= ZSTD_blockHeaderSize; |
5131 | } |
5132 | |
5133 | if (cctx->appliedParams.fParams.checksumFlag) { |
5134 | U32 const checksum = (U32) XXH64_digest(&cctx->xxhState); |
5135 | RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum" ); |
5136 | DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X" , (unsigned)checksum); |
5137 | MEM_writeLE32(op, checksum); |
5138 | op += 4; |
5139 | } |
5140 | |
5141 | cctx->stage = ZSTDcs_created; /* return to "created but no init" status */ |
5142 | return op-ostart; |
5143 | } |
5144 | |
5145 | void ZSTD_CCtx_trace(ZSTD_CCtx* cctx, size_t ) |
5146 | { |
5147 | #if ZSTD_TRACE |
5148 | if (cctx->traceCtx && ZSTD_trace_compress_end != NULL) { |
5149 | int const streaming = cctx->inBuffSize > 0 || cctx->outBuffSize > 0 || cctx->appliedParams.nbWorkers > 0; |
5150 | ZSTD_Trace trace; |
5151 | ZSTD_memset(&trace, 0, sizeof(trace)); |
5152 | trace.version = ZSTD_VERSION_NUMBER; |
5153 | trace.streaming = streaming; |
5154 | trace.dictionaryID = cctx->dictID; |
5155 | trace.dictionarySize = cctx->dictContentSize; |
5156 | trace.uncompressedSize = cctx->consumedSrcSize; |
5157 | trace.compressedSize = cctx->producedCSize + extraCSize; |
5158 | trace.params = &cctx->appliedParams; |
5159 | trace.cctx = cctx; |
5160 | ZSTD_trace_compress_end(cctx->traceCtx, &trace); |
5161 | } |
5162 | cctx->traceCtx = 0; |
5163 | #else |
5164 | (void)cctx; |
5165 | (void)extraCSize; |
5166 | #endif |
5167 | } |
5168 | |
5169 | size_t ZSTD_compressEnd_public(ZSTD_CCtx* cctx, |
5170 | void* dst, size_t dstCapacity, |
5171 | const void* src, size_t srcSize) |
5172 | { |
5173 | size_t endResult; |
5174 | size_t const cSize = ZSTD_compressContinue_internal(cctx, |
5175 | dst, dstCapacity, src, srcSize, |
5176 | 1 /* frame mode */, 1 /* last chunk */); |
5177 | FORWARD_IF_ERROR(cSize, "ZSTD_compressContinue_internal failed" ); |
5178 | endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize); |
5179 | FORWARD_IF_ERROR(endResult, "ZSTD_writeEpilogue failed" ); |
5180 | assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0)); |
5181 | if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */ |
5182 | ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1); |
5183 | DEBUGLOG(4, "end of frame : controlling src size" ); |
5184 | RETURN_ERROR_IF( |
5185 | cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1, |
5186 | srcSize_wrong, |
5187 | "error : pledgedSrcSize = %u, while realSrcSize = %u" , |
5188 | (unsigned)cctx->pledgedSrcSizePlusOne-1, |
5189 | (unsigned)cctx->consumedSrcSize); |
5190 | } |
5191 | ZSTD_CCtx_trace(cctx, endResult); |
5192 | return cSize + endResult; |
5193 | } |
5194 | |
5195 | /* NOTE: Must just wrap ZSTD_compressEnd_public() */ |
5196 | size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, |
5197 | void* dst, size_t dstCapacity, |
5198 | const void* src, size_t srcSize) |
5199 | { |
5200 | return ZSTD_compressEnd_public(cctx, dst, dstCapacity, src, srcSize); |
5201 | } |
5202 | |
5203 | size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx, |
5204 | void* dst, size_t dstCapacity, |
5205 | const void* src, size_t srcSize, |
5206 | const void* dict,size_t dictSize, |
5207 | ZSTD_parameters params) |
5208 | { |
5209 | DEBUGLOG(4, "ZSTD_compress_advanced" ); |
5210 | FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "" ); |
5211 | ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, ZSTD_NO_CLEVEL); |
5212 | return ZSTD_compress_advanced_internal(cctx, |
5213 | dst, dstCapacity, |
5214 | src, srcSize, |
5215 | dict, dictSize, |
5216 | &cctx->simpleApiParams); |
5217 | } |
5218 | |
5219 | /* Internal */ |
5220 | size_t ZSTD_compress_advanced_internal( |
5221 | ZSTD_CCtx* cctx, |
5222 | void* dst, size_t dstCapacity, |
5223 | const void* src, size_t srcSize, |
5224 | const void* dict,size_t dictSize, |
5225 | const ZSTD_CCtx_params* params) |
5226 | { |
5227 | DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)" , (unsigned)srcSize); |
5228 | FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx, |
5229 | dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL, |
5230 | params, srcSize, ZSTDb_not_buffered) , "" ); |
5231 | return ZSTD_compressEnd_public(cctx, dst, dstCapacity, src, srcSize); |
5232 | } |
5233 | |
5234 | size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx, |
5235 | void* dst, size_t dstCapacity, |
5236 | const void* src, size_t srcSize, |
5237 | const void* dict, size_t dictSize, |
5238 | int compressionLevel) |
5239 | { |
5240 | { |
5241 | ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0, ZSTD_cpm_noAttachDict); |
5242 | assert(params.fParams.contentSizeFlag == 1); |
5243 | ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT: compressionLevel); |
5244 | } |
5245 | DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)" , (unsigned)srcSize); |
5246 | return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctx->simpleApiParams); |
5247 | } |
5248 | |
5249 | size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx, |
5250 | void* dst, size_t dstCapacity, |
5251 | const void* src, size_t srcSize, |
5252 | int compressionLevel) |
5253 | { |
5254 | DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)" , (unsigned)srcSize); |
5255 | assert(cctx != NULL); |
5256 | return ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel); |
5257 | } |
5258 | |
5259 | size_t ZSTD_compress(void* dst, size_t dstCapacity, |
5260 | const void* src, size_t srcSize, |
5261 | int compressionLevel) |
5262 | { |
5263 | size_t result; |
5264 | #if ZSTD_COMPRESS_HEAPMODE |
5265 | ZSTD_CCtx* cctx = ZSTD_createCCtx(); |
5266 | RETURN_ERROR_IF(!cctx, memory_allocation, "ZSTD_createCCtx failed" ); |
5267 | result = ZSTD_compressCCtx(cctx, dst, dstCapacity, src, srcSize, compressionLevel); |
5268 | ZSTD_freeCCtx(cctx); |
5269 | #else |
5270 | ZSTD_CCtx ctxBody; |
5271 | ZSTD_initCCtx(&ctxBody, ZSTD_defaultCMem); |
5272 | result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel); |
5273 | ZSTD_freeCCtxContent(&ctxBody); /* can't free ctxBody itself, as it's on stack; free only heap content */ |
5274 | #endif |
5275 | return result; |
5276 | } |
5277 | |
5278 | |
5279 | /* ===== Dictionary API ===== */ |
5280 | |
5281 | /*! ZSTD_estimateCDictSize_advanced() : |
5282 | * Estimate amount of memory that will be needed to create a dictionary with following arguments */ |
5283 | size_t ZSTD_estimateCDictSize_advanced( |
5284 | size_t dictSize, ZSTD_compressionParameters cParams, |
5285 | ZSTD_dictLoadMethod_e dictLoadMethod) |
5286 | { |
5287 | DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u" , (unsigned)sizeof(ZSTD_CDict)); |
5288 | return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) |
5289 | + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) |
5290 | /* enableDedicatedDictSearch == 1 ensures that CDict estimation will not be too small |
5291 | * in case we are using DDS with row-hash. */ |
5292 | + ZSTD_sizeof_matchState(&cParams, ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams), |
5293 | /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0) |
5294 | + (dictLoadMethod == ZSTD_dlm_byRef ? 0 |
5295 | : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *)))); |
5296 | } |
5297 | |
5298 | size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel) |
5299 | { |
5300 | ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); |
5301 | return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy); |
5302 | } |
5303 | |
5304 | size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict) |
5305 | { |
5306 | if (cdict==NULL) return 0; /* support sizeof on NULL */ |
5307 | DEBUGLOG(5, "sizeof(*cdict) : %u" , (unsigned)sizeof(*cdict)); |
5308 | /* cdict may be in the workspace */ |
5309 | return (cdict->workspace.workspace == cdict ? 0 : sizeof(*cdict)) |
5310 | + ZSTD_cwksp_sizeof(&cdict->workspace); |
5311 | } |
5312 | |
5313 | static size_t ZSTD_initCDict_internal( |
5314 | ZSTD_CDict* cdict, |
5315 | const void* dictBuffer, size_t dictSize, |
5316 | ZSTD_dictLoadMethod_e dictLoadMethod, |
5317 | ZSTD_dictContentType_e dictContentType, |
5318 | ZSTD_CCtx_params params) |
5319 | { |
5320 | DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)" , (unsigned)dictContentType); |
5321 | assert(!ZSTD_checkCParams(params.cParams)); |
5322 | cdict->matchState.cParams = params.cParams; |
5323 | cdict->matchState.dedicatedDictSearch = params.enableDedicatedDictSearch; |
5324 | if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) { |
5325 | cdict->dictContent = dictBuffer; |
5326 | } else { |
5327 | void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*))); |
5328 | RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!" ); |
5329 | cdict->dictContent = internalBuffer; |
5330 | ZSTD_memcpy(internalBuffer, dictBuffer, dictSize); |
5331 | } |
5332 | cdict->dictContentSize = dictSize; |
5333 | cdict->dictContentType = dictContentType; |
5334 | |
5335 | cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE); |
5336 | |
5337 | |
5338 | /* Reset the state to no dictionary */ |
5339 | ZSTD_reset_compressedBlockState(&cdict->cBlockState); |
5340 | FORWARD_IF_ERROR(ZSTD_reset_matchState( |
5341 | &cdict->matchState, |
5342 | &cdict->workspace, |
5343 | ¶ms.cParams, |
5344 | params.useRowMatchFinder, |
5345 | ZSTDcrp_makeClean, |
5346 | ZSTDirp_reset, |
5347 | ZSTD_resetTarget_CDict), "" ); |
5348 | /* (Maybe) load the dictionary |
5349 | * Skips loading the dictionary if it is < 8 bytes. |
5350 | */ |
5351 | { params.compressionLevel = ZSTD_CLEVEL_DEFAULT; |
5352 | params.fParams.contentSizeFlag = 1; |
5353 | { size_t const dictID = ZSTD_compress_insertDictionary( |
5354 | &cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace, |
5355 | ¶ms, cdict->dictContent, cdict->dictContentSize, |
5356 | dictContentType, ZSTD_dtlm_full, ZSTD_tfp_forCDict, cdict->entropyWorkspace); |
5357 | FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed" ); |
5358 | assert(dictID <= (size_t)(U32)-1); |
5359 | cdict->dictID = (U32)dictID; |
5360 | } |
5361 | } |
5362 | |
5363 | return 0; |
5364 | } |
5365 | |
5366 | static ZSTD_CDict* ZSTD_createCDict_advanced_internal(size_t dictSize, |
5367 | ZSTD_dictLoadMethod_e dictLoadMethod, |
5368 | ZSTD_compressionParameters cParams, |
5369 | ZSTD_paramSwitch_e useRowMatchFinder, |
5370 | U32 enableDedicatedDictSearch, |
5371 | ZSTD_customMem customMem) |
5372 | { |
5373 | if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; |
5374 | |
5375 | { size_t const workspaceSize = |
5376 | ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) + |
5377 | ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) + |
5378 | ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, enableDedicatedDictSearch, /* forCCtx */ 0) + |
5379 | (dictLoadMethod == ZSTD_dlm_byRef ? 0 |
5380 | : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*)))); |
5381 | void* const workspace = ZSTD_customMalloc(workspaceSize, customMem); |
5382 | ZSTD_cwksp ws; |
5383 | ZSTD_CDict* cdict; |
5384 | |
5385 | if (!workspace) { |
5386 | ZSTD_customFree(workspace, customMem); |
5387 | return NULL; |
5388 | } |
5389 | |
5390 | ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_dynamic_alloc); |
5391 | |
5392 | cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict)); |
5393 | assert(cdict != NULL); |
5394 | ZSTD_cwksp_move(&cdict->workspace, &ws); |
5395 | cdict->customMem = customMem; |
5396 | cdict->compressionLevel = ZSTD_NO_CLEVEL; /* signals advanced API usage */ |
5397 | cdict->useRowMatchFinder = useRowMatchFinder; |
5398 | return cdict; |
5399 | } |
5400 | } |
5401 | |
5402 | ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize, |
5403 | ZSTD_dictLoadMethod_e dictLoadMethod, |
5404 | ZSTD_dictContentType_e dictContentType, |
5405 | ZSTD_compressionParameters cParams, |
5406 | ZSTD_customMem customMem) |
5407 | { |
5408 | ZSTD_CCtx_params cctxParams; |
5409 | ZSTD_memset(&cctxParams, 0, sizeof(cctxParams)); |
5410 | ZSTD_CCtxParams_init(&cctxParams, 0); |
5411 | cctxParams.cParams = cParams; |
5412 | cctxParams.customMem = customMem; |
5413 | return ZSTD_createCDict_advanced2( |
5414 | dictBuffer, dictSize, |
5415 | dictLoadMethod, dictContentType, |
5416 | &cctxParams, customMem); |
5417 | } |
5418 | |
5419 | ZSTD_CDict* ZSTD_createCDict_advanced2( |
5420 | const void* dict, size_t dictSize, |
5421 | ZSTD_dictLoadMethod_e dictLoadMethod, |
5422 | ZSTD_dictContentType_e dictContentType, |
5423 | const ZSTD_CCtx_params* originalCctxParams, |
5424 | ZSTD_customMem customMem) |
5425 | { |
5426 | ZSTD_CCtx_params cctxParams = *originalCctxParams; |
5427 | ZSTD_compressionParameters cParams; |
5428 | ZSTD_CDict* cdict; |
5429 | |
5430 | DEBUGLOG(3, "ZSTD_createCDict_advanced2, mode %u" , (unsigned)dictContentType); |
5431 | if (!customMem.customAlloc ^ !customMem.customFree) return NULL; |
5432 | |
5433 | if (cctxParams.enableDedicatedDictSearch) { |
5434 | cParams = ZSTD_dedicatedDictSearch_getCParams( |
5435 | cctxParams.compressionLevel, dictSize); |
5436 | ZSTD_overrideCParams(&cParams, &cctxParams.cParams); |
5437 | } else { |
5438 | cParams = ZSTD_getCParamsFromCCtxParams( |
5439 | &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); |
5440 | } |
5441 | |
5442 | if (!ZSTD_dedicatedDictSearch_isSupported(&cParams)) { |
5443 | /* Fall back to non-DDSS params */ |
5444 | cctxParams.enableDedicatedDictSearch = 0; |
5445 | cParams = ZSTD_getCParamsFromCCtxParams( |
5446 | &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); |
5447 | } |
5448 | |
5449 | DEBUGLOG(3, "ZSTD_createCDict_advanced2: DDS: %u" , cctxParams.enableDedicatedDictSearch); |
5450 | cctxParams.cParams = cParams; |
5451 | cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams); |
5452 | |
5453 | cdict = ZSTD_createCDict_advanced_internal(dictSize, |
5454 | dictLoadMethod, cctxParams.cParams, |
5455 | cctxParams.useRowMatchFinder, cctxParams.enableDedicatedDictSearch, |
5456 | customMem); |
5457 | |
5458 | if (ZSTD_isError( ZSTD_initCDict_internal(cdict, |
5459 | dict, dictSize, |
5460 | dictLoadMethod, dictContentType, |
5461 | cctxParams) )) { |
5462 | ZSTD_freeCDict(cdict); |
5463 | return NULL; |
5464 | } |
5465 | |
5466 | return cdict; |
5467 | } |
5468 | |
5469 | ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel) |
5470 | { |
5471 | ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); |
5472 | ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize, |
5473 | ZSTD_dlm_byCopy, ZSTD_dct_auto, |
5474 | cParams, ZSTD_defaultCMem); |
5475 | if (cdict) |
5476 | cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel; |
5477 | return cdict; |
5478 | } |
5479 | |
5480 | ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel) |
5481 | { |
5482 | ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); |
5483 | ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize, |
5484 | ZSTD_dlm_byRef, ZSTD_dct_auto, |
5485 | cParams, ZSTD_defaultCMem); |
5486 | if (cdict) |
5487 | cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel; |
5488 | return cdict; |
5489 | } |
5490 | |
5491 | size_t ZSTD_freeCDict(ZSTD_CDict* cdict) |
5492 | { |
5493 | if (cdict==NULL) return 0; /* support free on NULL */ |
5494 | { ZSTD_customMem const cMem = cdict->customMem; |
5495 | int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict); |
5496 | ZSTD_cwksp_free(&cdict->workspace, cMem); |
5497 | if (!cdictInWorkspace) { |
5498 | ZSTD_customFree(cdict, cMem); |
5499 | } |
5500 | return 0; |
5501 | } |
5502 | } |
5503 | |
5504 | /*! ZSTD_initStaticCDict_advanced() : |
5505 | * Generate a digested dictionary in provided memory area. |
5506 | * workspace: The memory area to emplace the dictionary into. |
5507 | * Provided pointer must 8-bytes aligned. |
5508 | * It must outlive dictionary usage. |
5509 | * workspaceSize: Use ZSTD_estimateCDictSize() |
5510 | * to determine how large workspace must be. |
5511 | * cParams : use ZSTD_getCParams() to transform a compression level |
5512 | * into its relevants cParams. |
5513 | * @return : pointer to ZSTD_CDict*, or NULL if error (size too small) |
5514 | * Note : there is no corresponding "free" function. |
5515 | * Since workspace was allocated externally, it must be freed externally. |
5516 | */ |
5517 | const ZSTD_CDict* ZSTD_initStaticCDict( |
5518 | void* workspace, size_t workspaceSize, |
5519 | const void* dict, size_t dictSize, |
5520 | ZSTD_dictLoadMethod_e dictLoadMethod, |
5521 | ZSTD_dictContentType_e dictContentType, |
5522 | ZSTD_compressionParameters cParams) |
5523 | { |
5524 | ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams); |
5525 | /* enableDedicatedDictSearch == 1 ensures matchstate is not too small in case this CDict will be used for DDS + row hash */ |
5526 | size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0); |
5527 | size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) |
5528 | + (dictLoadMethod == ZSTD_dlm_byRef ? 0 |
5529 | : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*)))) |
5530 | + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) |
5531 | + matchStateSize; |
5532 | ZSTD_CDict* cdict; |
5533 | ZSTD_CCtx_params params; |
5534 | |
5535 | if ((size_t)workspace & 7) return NULL; /* 8-aligned */ |
5536 | |
5537 | { |
5538 | ZSTD_cwksp ws; |
5539 | ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc); |
5540 | cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict)); |
5541 | if (cdict == NULL) return NULL; |
5542 | ZSTD_cwksp_move(&cdict->workspace, &ws); |
5543 | } |
5544 | |
5545 | DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u" , |
5546 | (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize)); |
5547 | if (workspaceSize < neededSize) return NULL; |
5548 | |
5549 | ZSTD_CCtxParams_init(¶ms, 0); |
5550 | params.cParams = cParams; |
5551 | params.useRowMatchFinder = useRowMatchFinder; |
5552 | cdict->useRowMatchFinder = useRowMatchFinder; |
5553 | cdict->compressionLevel = ZSTD_NO_CLEVEL; |
5554 | |
5555 | if (ZSTD_isError( ZSTD_initCDict_internal(cdict, |
5556 | dict, dictSize, |
5557 | dictLoadMethod, dictContentType, |
5558 | params) )) |
5559 | return NULL; |
5560 | |
5561 | return cdict; |
5562 | } |
5563 | |
5564 | ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict) |
5565 | { |
5566 | assert(cdict != NULL); |
5567 | return cdict->matchState.cParams; |
5568 | } |
5569 | |
5570 | /*! ZSTD_getDictID_fromCDict() : |
5571 | * Provides the dictID of the dictionary loaded into `cdict`. |
5572 | * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty. |
5573 | * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */ |
5574 | unsigned ZSTD_getDictID_fromCDict(const ZSTD_CDict* cdict) |
5575 | { |
5576 | if (cdict==NULL) return 0; |
5577 | return cdict->dictID; |
5578 | } |
5579 | |
5580 | /* ZSTD_compressBegin_usingCDict_internal() : |
5581 | * Implementation of various ZSTD_compressBegin_usingCDict* functions. |
5582 | */ |
5583 | static size_t ZSTD_compressBegin_usingCDict_internal( |
5584 | ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, |
5585 | ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize) |
5586 | { |
5587 | ZSTD_CCtx_params cctxParams; |
5588 | DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_internal" ); |
5589 | RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!" ); |
5590 | /* Initialize the cctxParams from the cdict */ |
5591 | { |
5592 | ZSTD_parameters params; |
5593 | params.fParams = fParams; |
5594 | params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF |
5595 | || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER |
5596 | || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN |
5597 | || cdict->compressionLevel == 0 ) ? |
5598 | ZSTD_getCParamsFromCDict(cdict) |
5599 | : ZSTD_getCParams(cdict->compressionLevel, |
5600 | pledgedSrcSize, |
5601 | cdict->dictContentSize); |
5602 | ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, cdict->compressionLevel); |
5603 | } |
5604 | /* Increase window log to fit the entire dictionary and source if the |
5605 | * source size is known. Limit the increase to 19, which is the |
5606 | * window log for compression level 1 with the largest source size. |
5607 | */ |
5608 | if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) { |
5609 | U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19); |
5610 | U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1; |
5611 | cctxParams.cParams.windowLog = MAX(cctxParams.cParams.windowLog, limitedSrcLog); |
5612 | } |
5613 | return ZSTD_compressBegin_internal(cctx, |
5614 | NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, |
5615 | cdict, |
5616 | &cctxParams, pledgedSrcSize, |
5617 | ZSTDb_not_buffered); |
5618 | } |
5619 | |
5620 | |
5621 | /* ZSTD_compressBegin_usingCDict_advanced() : |
5622 | * This function is DEPRECATED. |
5623 | * cdict must be != NULL */ |
5624 | size_t ZSTD_compressBegin_usingCDict_advanced( |
5625 | ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, |
5626 | ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize) |
5627 | { |
5628 | return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, pledgedSrcSize); |
5629 | } |
5630 | |
5631 | /* ZSTD_compressBegin_usingCDict() : |
5632 | * cdict must be != NULL */ |
5633 | size_t ZSTD_compressBegin_usingCDict_deprecated(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) |
5634 | { |
5635 | ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; |
5636 | return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN); |
5637 | } |
5638 | |
5639 | size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) |
5640 | { |
5641 | return ZSTD_compressBegin_usingCDict_deprecated(cctx, cdict); |
5642 | } |
5643 | |
5644 | /*! ZSTD_compress_usingCDict_internal(): |
5645 | * Implementation of various ZSTD_compress_usingCDict* functions. |
5646 | */ |
5647 | static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx, |
5648 | void* dst, size_t dstCapacity, |
5649 | const void* src, size_t srcSize, |
5650 | const ZSTD_CDict* cdict, ZSTD_frameParameters fParams) |
5651 | { |
5652 | FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, srcSize), "" ); /* will check if cdict != NULL */ |
5653 | return ZSTD_compressEnd_public(cctx, dst, dstCapacity, src, srcSize); |
5654 | } |
5655 | |
5656 | /*! ZSTD_compress_usingCDict_advanced(): |
5657 | * This function is DEPRECATED. |
5658 | */ |
5659 | size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx, |
5660 | void* dst, size_t dstCapacity, |
5661 | const void* src, size_t srcSize, |
5662 | const ZSTD_CDict* cdict, ZSTD_frameParameters fParams) |
5663 | { |
5664 | return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams); |
5665 | } |
5666 | |
5667 | /*! ZSTD_compress_usingCDict() : |
5668 | * Compression using a digested Dictionary. |
5669 | * Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times. |
5670 | * Note that compression parameters are decided at CDict creation time |
5671 | * while frame parameters are hardcoded */ |
5672 | size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx, |
5673 | void* dst, size_t dstCapacity, |
5674 | const void* src, size_t srcSize, |
5675 | const ZSTD_CDict* cdict) |
5676 | { |
5677 | ZSTD_frameParameters const fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; |
5678 | return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams); |
5679 | } |
5680 | |
5681 | |
5682 | |
5683 | /* ****************************************************************** |
5684 | * Streaming |
5685 | ********************************************************************/ |
5686 | |
5687 | ZSTD_CStream* ZSTD_createCStream(void) |
5688 | { |
5689 | DEBUGLOG(3, "ZSTD_createCStream" ); |
5690 | return ZSTD_createCStream_advanced(ZSTD_defaultCMem); |
5691 | } |
5692 | |
5693 | ZSTD_CStream* ZSTD_initStaticCStream(void *workspace, size_t workspaceSize) |
5694 | { |
5695 | return ZSTD_initStaticCCtx(workspace, workspaceSize); |
5696 | } |
5697 | |
5698 | ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem) |
5699 | { /* CStream and CCtx are now same object */ |
5700 | return ZSTD_createCCtx_advanced(customMem); |
5701 | } |
5702 | |
5703 | size_t ZSTD_freeCStream(ZSTD_CStream* zcs) |
5704 | { |
5705 | return ZSTD_freeCCtx(zcs); /* same object */ |
5706 | } |
5707 | |
5708 | |
5709 | |
5710 | /*====== Initialization ======*/ |
5711 | |
5712 | size_t ZSTD_CStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX; } |
5713 | |
5714 | size_t ZSTD_CStreamOutSize(void) |
5715 | { |
5716 | return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ; |
5717 | } |
5718 | |
5719 | static ZSTD_cParamMode_e ZSTD_getCParamMode(ZSTD_CDict const* cdict, ZSTD_CCtx_params const* params, U64 pledgedSrcSize) |
5720 | { |
5721 | if (cdict != NULL && ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) |
5722 | return ZSTD_cpm_attachDict; |
5723 | else |
5724 | return ZSTD_cpm_noAttachDict; |
5725 | } |
5726 | |
5727 | /* ZSTD_resetCStream(): |
5728 | * pledgedSrcSize == 0 means "unknown" */ |
5729 | size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss) |
5730 | { |
5731 | /* temporary : 0 interpreted as "unknown" during transition period. |
5732 | * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN. |
5733 | * 0 will be interpreted as "empty" in the future. |
5734 | */ |
5735 | U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss; |
5736 | DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u" , (unsigned)pledgedSrcSize); |
5737 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5738 | FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "" ); |
5739 | return 0; |
5740 | } |
5741 | |
5742 | /*! ZSTD_initCStream_internal() : |
5743 | * Note : for lib/compress only. Used by zstdmt_compress.c. |
5744 | * Assumption 1 : params are valid |
5745 | * Assumption 2 : either dict, or cdict, is defined, not both */ |
5746 | size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs, |
5747 | const void* dict, size_t dictSize, const ZSTD_CDict* cdict, |
5748 | const ZSTD_CCtx_params* params, |
5749 | unsigned long long pledgedSrcSize) |
5750 | { |
5751 | DEBUGLOG(4, "ZSTD_initCStream_internal" ); |
5752 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5753 | FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "" ); |
5754 | assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams))); |
5755 | zcs->requestedParams = *params; |
5756 | assert(!((dict) && (cdict))); /* either dict or cdict, not both */ |
5757 | if (dict) { |
5758 | FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "" ); |
5759 | } else { |
5760 | /* Dictionary is cleared if !cdict */ |
5761 | FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "" ); |
5762 | } |
5763 | return 0; |
5764 | } |
5765 | |
5766 | /* ZSTD_initCStream_usingCDict_advanced() : |
5767 | * same as ZSTD_initCStream_usingCDict(), with control over frame parameters */ |
5768 | size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, |
5769 | const ZSTD_CDict* cdict, |
5770 | ZSTD_frameParameters fParams, |
5771 | unsigned long long pledgedSrcSize) |
5772 | { |
5773 | DEBUGLOG(4, "ZSTD_initCStream_usingCDict_advanced" ); |
5774 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5775 | FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "" ); |
5776 | zcs->requestedParams.fParams = fParams; |
5777 | FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "" ); |
5778 | return 0; |
5779 | } |
5780 | |
5781 | /* note : cdict must outlive compression session */ |
5782 | size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict) |
5783 | { |
5784 | DEBUGLOG(4, "ZSTD_initCStream_usingCDict" ); |
5785 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5786 | FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "" ); |
5787 | return 0; |
5788 | } |
5789 | |
5790 | |
5791 | /* ZSTD_initCStream_advanced() : |
5792 | * pledgedSrcSize must be exact. |
5793 | * if srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN. |
5794 | * dict is loaded with default parameters ZSTD_dct_auto and ZSTD_dlm_byCopy. */ |
5795 | size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, |
5796 | const void* dict, size_t dictSize, |
5797 | ZSTD_parameters params, unsigned long long pss) |
5798 | { |
5799 | /* for compatibility with older programs relying on this behavior. |
5800 | * Users should now specify ZSTD_CONTENTSIZE_UNKNOWN. |
5801 | * This line will be removed in the future. |
5802 | */ |
5803 | U64 const pledgedSrcSize = (pss==0 && params.fParams.contentSizeFlag==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss; |
5804 | DEBUGLOG(4, "ZSTD_initCStream_advanced" ); |
5805 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5806 | FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "" ); |
5807 | FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "" ); |
5808 | ZSTD_CCtxParams_setZstdParams(&zcs->requestedParams, ¶ms); |
5809 | FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "" ); |
5810 | return 0; |
5811 | } |
5812 | |
5813 | size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel) |
5814 | { |
5815 | DEBUGLOG(4, "ZSTD_initCStream_usingDict" ); |
5816 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5817 | FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "" ); |
5818 | FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "" ); |
5819 | return 0; |
5820 | } |
5821 | |
5822 | size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pss) |
5823 | { |
5824 | /* temporary : 0 interpreted as "unknown" during transition period. |
5825 | * Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN. |
5826 | * 0 will be interpreted as "empty" in the future. |
5827 | */ |
5828 | U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss; |
5829 | DEBUGLOG(4, "ZSTD_initCStream_srcSize" ); |
5830 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5831 | FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "" ); |
5832 | FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "" ); |
5833 | FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "" ); |
5834 | return 0; |
5835 | } |
5836 | |
5837 | size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel) |
5838 | { |
5839 | DEBUGLOG(4, "ZSTD_initCStream" ); |
5840 | FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "" ); |
5841 | FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "" ); |
5842 | FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "" ); |
5843 | return 0; |
5844 | } |
5845 | |
5846 | /*====== Compression ======*/ |
5847 | |
5848 | static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx) |
5849 | { |
5850 | if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { |
5851 | return cctx->blockSize - cctx->stableIn_notConsumed; |
5852 | } |
5853 | assert(cctx->appliedParams.inBufferMode == ZSTD_bm_buffered); |
5854 | { size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos; |
5855 | if (hintInSize==0) hintInSize = cctx->blockSize; |
5856 | return hintInSize; |
5857 | } |
5858 | } |
5859 | |
5860 | /** ZSTD_compressStream_generic(): |
5861 | * internal function for all *compressStream*() variants |
5862 | * @return : hint size for next input to complete ongoing block */ |
5863 | static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, |
5864 | ZSTD_outBuffer* output, |
5865 | ZSTD_inBuffer* input, |
5866 | ZSTD_EndDirective const flushMode) |
5867 | { |
5868 | const char* const istart = (assert(input != NULL), (const char*)input->src); |
5869 | const char* const iend = (istart != NULL) ? istart + input->size : istart; |
5870 | const char* ip = (istart != NULL) ? istart + input->pos : istart; |
5871 | char* const ostart = (assert(output != NULL), (char*)output->dst); |
5872 | char* const oend = (ostart != NULL) ? ostart + output->size : ostart; |
5873 | char* op = (ostart != NULL) ? ostart + output->pos : ostart; |
5874 | U32 someMoreWork = 1; |
5875 | |
5876 | /* check expectations */ |
5877 | DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%i, srcSize = %zu" , (int)flushMode, input->size - input->pos); |
5878 | assert(zcs != NULL); |
5879 | if (zcs->appliedParams.inBufferMode == ZSTD_bm_stable) { |
5880 | assert(input->pos >= zcs->stableIn_notConsumed); |
5881 | input->pos -= zcs->stableIn_notConsumed; |
5882 | ip -= zcs->stableIn_notConsumed; |
5883 | zcs->stableIn_notConsumed = 0; |
5884 | } |
5885 | if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) { |
5886 | assert(zcs->inBuff != NULL); |
5887 | assert(zcs->inBuffSize > 0); |
5888 | } |
5889 | if (zcs->appliedParams.outBufferMode == ZSTD_bm_buffered) { |
5890 | assert(zcs->outBuff != NULL); |
5891 | assert(zcs->outBuffSize > 0); |
5892 | } |
5893 | if (input->src == NULL) assert(input->size == 0); |
5894 | assert(input->pos <= input->size); |
5895 | if (output->dst == NULL) assert(output->size == 0); |
5896 | assert(output->pos <= output->size); |
5897 | assert((U32)flushMode <= (U32)ZSTD_e_end); |
5898 | |
5899 | while (someMoreWork) { |
5900 | switch(zcs->streamStage) |
5901 | { |
5902 | case zcss_init: |
5903 | RETURN_ERROR(init_missing, "call ZSTD_initCStream() first!" ); |
5904 | |
5905 | case zcss_load: |
5906 | if ( (flushMode == ZSTD_e_end) |
5907 | && ( (size_t)(oend-op) >= ZSTD_compressBound(iend-ip) /* Enough output space */ |
5908 | || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) /* OR we are allowed to return dstSizeTooSmall */ |
5909 | && (zcs->inBuffPos == 0) ) { |
5910 | /* shortcut to compression pass directly into output buffer */ |
5911 | size_t const cSize = ZSTD_compressEnd_public(zcs, |
5912 | op, oend-op, ip, iend-ip); |
5913 | DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u" , (unsigned)cSize); |
5914 | FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed" ); |
5915 | ip = iend; |
5916 | op += cSize; |
5917 | zcs->frameEnded = 1; |
5918 | ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only); |
5919 | someMoreWork = 0; break; |
5920 | } |
5921 | /* complete loading into inBuffer in buffered mode */ |
5922 | if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) { |
5923 | size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos; |
5924 | size_t const loaded = ZSTD_limitCopy( |
5925 | zcs->inBuff + zcs->inBuffPos, toLoad, |
5926 | ip, iend-ip); |
5927 | zcs->inBuffPos += loaded; |
5928 | if (ip) ip += loaded; |
5929 | if ( (flushMode == ZSTD_e_continue) |
5930 | && (zcs->inBuffPos < zcs->inBuffTarget) ) { |
5931 | /* not enough input to fill full block : stop here */ |
5932 | someMoreWork = 0; break; |
5933 | } |
5934 | if ( (flushMode == ZSTD_e_flush) |
5935 | && (zcs->inBuffPos == zcs->inToCompress) ) { |
5936 | /* empty */ |
5937 | someMoreWork = 0; break; |
5938 | } |
5939 | } else { |
5940 | assert(zcs->appliedParams.inBufferMode == ZSTD_bm_stable); |
5941 | if ( (flushMode == ZSTD_e_continue) |
5942 | && ( (size_t)(iend - ip) < zcs->blockSize) ) { |
5943 | /* can't compress a full block : stop here */ |
5944 | zcs->stableIn_notConsumed = (size_t)(iend - ip); |
5945 | ip = iend; /* pretend to have consumed input */ |
5946 | someMoreWork = 0; break; |
5947 | } |
5948 | if ( (flushMode == ZSTD_e_flush) |
5949 | && (ip == iend) ) { |
5950 | /* empty */ |
5951 | someMoreWork = 0; break; |
5952 | } |
5953 | } |
5954 | /* compress current block (note : this stage cannot be stopped in the middle) */ |
5955 | DEBUGLOG(5, "stream compression stage (flushMode==%u)" , flushMode); |
5956 | { int const inputBuffered = (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered); |
5957 | void* cDst; |
5958 | size_t cSize; |
5959 | size_t oSize = oend-op; |
5960 | size_t const iSize = inputBuffered ? zcs->inBuffPos - zcs->inToCompress |
5961 | : MIN((size_t)(iend - ip), zcs->blockSize); |
5962 | if (oSize >= ZSTD_compressBound(iSize) || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) |
5963 | cDst = op; /* compress into output buffer, to skip flush stage */ |
5964 | else |
5965 | cDst = zcs->outBuff, oSize = zcs->outBuffSize; |
5966 | if (inputBuffered) { |
5967 | unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend); |
5968 | cSize = lastBlock ? |
5969 | ZSTD_compressEnd_public(zcs, cDst, oSize, |
5970 | zcs->inBuff + zcs->inToCompress, iSize) : |
5971 | ZSTD_compressContinue_public(zcs, cDst, oSize, |
5972 | zcs->inBuff + zcs->inToCompress, iSize); |
5973 | FORWARD_IF_ERROR(cSize, "%s" , lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed" ); |
5974 | zcs->frameEnded = lastBlock; |
5975 | /* prepare next block */ |
5976 | zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize; |
5977 | if (zcs->inBuffTarget > zcs->inBuffSize) |
5978 | zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize; |
5979 | DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u" , |
5980 | (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize); |
5981 | if (!lastBlock) |
5982 | assert(zcs->inBuffTarget <= zcs->inBuffSize); |
5983 | zcs->inToCompress = zcs->inBuffPos; |
5984 | } else { /* !inputBuffered, hence ZSTD_bm_stable */ |
5985 | unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip + iSize == iend); |
5986 | cSize = lastBlock ? |
5987 | ZSTD_compressEnd_public(zcs, cDst, oSize, ip, iSize) : |
5988 | ZSTD_compressContinue_public(zcs, cDst, oSize, ip, iSize); |
5989 | /* Consume the input prior to error checking to mirror buffered mode. */ |
5990 | if (ip) ip += iSize; |
5991 | FORWARD_IF_ERROR(cSize, "%s" , lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed" ); |
5992 | zcs->frameEnded = lastBlock; |
5993 | if (lastBlock) assert(ip == iend); |
5994 | } |
5995 | if (cDst == op) { /* no need to flush */ |
5996 | op += cSize; |
5997 | if (zcs->frameEnded) { |
5998 | DEBUGLOG(5, "Frame completed directly in outBuffer" ); |
5999 | someMoreWork = 0; |
6000 | ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only); |
6001 | } |
6002 | break; |
6003 | } |
6004 | zcs->outBuffContentSize = cSize; |
6005 | zcs->outBuffFlushedSize = 0; |
6006 | zcs->streamStage = zcss_flush; /* pass-through to flush stage */ |
6007 | } |
6008 | ZSTD_FALLTHROUGH; |
6009 | case zcss_flush: |
6010 | DEBUGLOG(5, "flush stage" ); |
6011 | assert(zcs->appliedParams.outBufferMode == ZSTD_bm_buffered); |
6012 | { size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize; |
6013 | size_t const flushed = ZSTD_limitCopy(op, (size_t)(oend-op), |
6014 | zcs->outBuff + zcs->outBuffFlushedSize, toFlush); |
6015 | DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u" , |
6016 | (unsigned)toFlush, (unsigned)(oend-op), (unsigned)flushed); |
6017 | if (flushed) |
6018 | op += flushed; |
6019 | zcs->outBuffFlushedSize += flushed; |
6020 | if (toFlush!=flushed) { |
6021 | /* flush not fully completed, presumably because dst is too small */ |
6022 | assert(op==oend); |
6023 | someMoreWork = 0; |
6024 | break; |
6025 | } |
6026 | zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0; |
6027 | if (zcs->frameEnded) { |
6028 | DEBUGLOG(5, "Frame completed on flush" ); |
6029 | someMoreWork = 0; |
6030 | ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only); |
6031 | break; |
6032 | } |
6033 | zcs->streamStage = zcss_load; |
6034 | break; |
6035 | } |
6036 | |
6037 | default: /* impossible */ |
6038 | assert(0); |
6039 | } |
6040 | } |
6041 | |
6042 | input->pos = ip - istart; |
6043 | output->pos = op - ostart; |
6044 | if (zcs->frameEnded) return 0; |
6045 | return ZSTD_nextInputSizeHint(zcs); |
6046 | } |
6047 | |
6048 | static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx) |
6049 | { |
6050 | #ifdef ZSTD_MULTITHREAD |
6051 | if (cctx->appliedParams.nbWorkers >= 1) { |
6052 | assert(cctx->mtctx != NULL); |
6053 | return ZSTDMT_nextInputSizeHint(cctx->mtctx); |
6054 | } |
6055 | #endif |
6056 | return ZSTD_nextInputSizeHint(cctx); |
6057 | |
6058 | } |
6059 | |
6060 | size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input) |
6061 | { |
6062 | FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) , "" ); |
6063 | return ZSTD_nextInputSizeHint_MTorST(zcs); |
6064 | } |
6065 | |
6066 | /* After a compression call set the expected input/output buffer. |
6067 | * This is validated at the start of the next compression call. |
6068 | */ |
6069 | static void |
6070 | ZSTD_setBufferExpectations(ZSTD_CCtx* cctx, const ZSTD_outBuffer* output, const ZSTD_inBuffer* input) |
6071 | { |
6072 | DEBUGLOG(5, "ZSTD_setBufferExpectations (for advanced stable in/out modes)" ); |
6073 | if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { |
6074 | cctx->expectedInBuffer = *input; |
6075 | } |
6076 | if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) { |
6077 | cctx->expectedOutBufferSize = output->size - output->pos; |
6078 | } |
6079 | } |
6080 | |
6081 | /* Validate that the input/output buffers match the expectations set by |
6082 | * ZSTD_setBufferExpectations. |
6083 | */ |
6084 | static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx, |
6085 | ZSTD_outBuffer const* output, |
6086 | ZSTD_inBuffer const* input, |
6087 | ZSTD_EndDirective endOp) |
6088 | { |
6089 | if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { |
6090 | ZSTD_inBuffer const expect = cctx->expectedInBuffer; |
6091 | if (expect.src != input->src || expect.pos != input->pos) |
6092 | RETURN_ERROR(stabilityCondition_notRespected, "ZSTD_c_stableInBuffer enabled but input differs!" ); |
6093 | } |
6094 | (void)endOp; |
6095 | if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) { |
6096 | size_t const outBufferSize = output->size - output->pos; |
6097 | if (cctx->expectedOutBufferSize != outBufferSize) |
6098 | RETURN_ERROR(stabilityCondition_notRespected, "ZSTD_c_stableOutBuffer enabled but output size differs!" ); |
6099 | } |
6100 | return 0; |
6101 | } |
6102 | |
6103 | static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, |
6104 | ZSTD_EndDirective endOp, |
6105 | size_t inSize) |
6106 | { |
6107 | ZSTD_CCtx_params params = cctx->requestedParams; |
6108 | ZSTD_prefixDict const prefixDict = cctx->prefixDict; |
6109 | FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , "" ); /* Init the local dict if present. */ |
6110 | ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */ |
6111 | assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */ |
6112 | if (cctx->cdict && !cctx->localDict.cdict) { |
6113 | /* Let the cdict's compression level take priority over the requested params. |
6114 | * But do not take the cdict's compression level if the "cdict" is actually a localDict |
6115 | * generated from ZSTD_initLocalDict(). |
6116 | */ |
6117 | params.compressionLevel = cctx->cdict->compressionLevel; |
6118 | } |
6119 | DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage" ); |
6120 | if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = inSize + 1; /* auto-determine pledgedSrcSize */ |
6121 | |
6122 | { size_t const dictSize = prefixDict.dict |
6123 | ? prefixDict.dictSize |
6124 | : (cctx->cdict ? cctx->cdict->dictContentSize : 0); |
6125 | ZSTD_cParamMode_e const mode = ZSTD_getCParamMode(cctx->cdict, ¶ms, cctx->pledgedSrcSizePlusOne - 1); |
6126 | params.cParams = ZSTD_getCParamsFromCCtxParams( |
6127 | ¶ms, cctx->pledgedSrcSizePlusOne-1, |
6128 | dictSize, mode); |
6129 | } |
6130 | |
6131 | params.useBlockSplitter = ZSTD_resolveBlockSplitterMode(params.useBlockSplitter, ¶ms.cParams); |
6132 | params.ldmParams.enableLdm = ZSTD_resolveEnableLdm(params.ldmParams.enableLdm, ¶ms.cParams); |
6133 | params.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params.useRowMatchFinder, ¶ms.cParams); |
6134 | params.validateSequences = ZSTD_resolveExternalSequenceValidation(params.validateSequences); |
6135 | params.maxBlockSize = ZSTD_resolveMaxBlockSize(params.maxBlockSize); |
6136 | params.searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(params.searchForExternalRepcodes, params.compressionLevel); |
6137 | |
6138 | #ifdef ZSTD_MULTITHREAD |
6139 | /* If external matchfinder is enabled, make sure to fail before checking job size (for consistency) */ |
6140 | RETURN_ERROR_IF( |
6141 | params.useSequenceProducer == 1 && params.nbWorkers >= 1, |
6142 | parameter_combination_unsupported, |
6143 | "External sequence producer isn't supported with nbWorkers >= 1" |
6144 | ); |
6145 | |
6146 | if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) { |
6147 | params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */ |
6148 | } |
6149 | if (params.nbWorkers > 0) { |
6150 | #if ZSTD_TRACE |
6151 | cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0; |
6152 | #endif |
6153 | /* mt context creation */ |
6154 | if (cctx->mtctx == NULL) { |
6155 | DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u" , |
6156 | params.nbWorkers); |
6157 | cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem, cctx->pool); |
6158 | RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!" ); |
6159 | } |
6160 | /* mt compression */ |
6161 | DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u" , params.nbWorkers); |
6162 | FORWARD_IF_ERROR( ZSTDMT_initCStream_internal( |
6163 | cctx->mtctx, |
6164 | prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, |
6165 | cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , "" ); |
6166 | cctx->dictID = cctx->cdict ? cctx->cdict->dictID : 0; |
6167 | cctx->dictContentSize = cctx->cdict ? cctx->cdict->dictContentSize : prefixDict.dictSize; |
6168 | cctx->consumedSrcSize = 0; |
6169 | cctx->producedCSize = 0; |
6170 | cctx->streamStage = zcss_load; |
6171 | cctx->appliedParams = params; |
6172 | } else |
6173 | #endif /* ZSTD_MULTITHREAD */ |
6174 | { U64 const pledgedSrcSize = cctx->pledgedSrcSizePlusOne - 1; |
6175 | assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams))); |
6176 | FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx, |
6177 | prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, ZSTD_dtlm_fast, |
6178 | cctx->cdict, |
6179 | ¶ms, pledgedSrcSize, |
6180 | ZSTDb_buffered) , "" ); |
6181 | assert(cctx->appliedParams.nbWorkers == 0); |
6182 | cctx->inToCompress = 0; |
6183 | cctx->inBuffPos = 0; |
6184 | if (cctx->appliedParams.inBufferMode == ZSTD_bm_buffered) { |
6185 | /* for small input: avoid automatic flush on reaching end of block, since |
6186 | * it would require to add a 3-bytes null block to end frame |
6187 | */ |
6188 | cctx->inBuffTarget = cctx->blockSize + (cctx->blockSize == pledgedSrcSize); |
6189 | } else { |
6190 | cctx->inBuffTarget = 0; |
6191 | } |
6192 | cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0; |
6193 | cctx->streamStage = zcss_load; |
6194 | cctx->frameEnded = 0; |
6195 | } |
6196 | return 0; |
6197 | } |
6198 | |
6199 | /* @return provides a minimum amount of data remaining to be flushed from internal buffers |
6200 | */ |
6201 | size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, |
6202 | ZSTD_outBuffer* output, |
6203 | ZSTD_inBuffer* input, |
6204 | ZSTD_EndDirective endOp) |
6205 | { |
6206 | DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u " , (unsigned)endOp); |
6207 | /* check conditions */ |
6208 | RETURN_ERROR_IF(output->pos > output->size, dstSize_tooSmall, "invalid output buffer" ); |
6209 | RETURN_ERROR_IF(input->pos > input->size, srcSize_wrong, "invalid input buffer" ); |
6210 | RETURN_ERROR_IF((U32)endOp > (U32)ZSTD_e_end, parameter_outOfBound, "invalid endDirective" ); |
6211 | assert(cctx != NULL); |
6212 | |
6213 | /* transparent initialization stage */ |
6214 | if (cctx->streamStage == zcss_init) { |
6215 | size_t const inputSize = input->size - input->pos; /* no obligation to start from pos==0 */ |
6216 | size_t const totalInputSize = inputSize + cctx->stableIn_notConsumed; |
6217 | if ( (cctx->requestedParams.inBufferMode == ZSTD_bm_stable) /* input is presumed stable, across invocations */ |
6218 | && (endOp == ZSTD_e_continue) /* no flush requested, more input to come */ |
6219 | && (totalInputSize < ZSTD_BLOCKSIZE_MAX) ) { /* not even reached one block yet */ |
6220 | if (cctx->stableIn_notConsumed) { /* not the first time */ |
6221 | /* check stable source guarantees */ |
6222 | RETURN_ERROR_IF(input->src != cctx->expectedInBuffer.src, stabilityCondition_notRespected, "stableInBuffer condition not respected: wrong src pointer" ); |
6223 | RETURN_ERROR_IF(input->pos != cctx->expectedInBuffer.size, stabilityCondition_notRespected, "stableInBuffer condition not respected: externally modified pos" ); |
6224 | } |
6225 | /* pretend input was consumed, to give a sense forward progress */ |
6226 | input->pos = input->size; |
6227 | /* save stable inBuffer, for later control, and flush/end */ |
6228 | cctx->expectedInBuffer = *input; |
6229 | /* but actually input wasn't consumed, so keep track of position from where compression shall resume */ |
6230 | cctx->stableIn_notConsumed += inputSize; |
6231 | /* don't initialize yet, wait for the first block of flush() order, for better parameters adaptation */ |
6232 | return ZSTD_FRAMEHEADERSIZE_MIN(cctx->requestedParams.format); /* at least some header to produce */ |
6233 | } |
6234 | FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, endOp, totalInputSize), "compressStream2 initialization failed" ); |
6235 | ZSTD_setBufferExpectations(cctx, output, input); /* Set initial buffer expectations now that we've initialized */ |
6236 | } |
6237 | /* end of transparent initialization stage */ |
6238 | |
6239 | FORWARD_IF_ERROR(ZSTD_checkBufferStability(cctx, output, input, endOp), "invalid buffers" ); |
6240 | /* compression stage */ |
6241 | #ifdef ZSTD_MULTITHREAD |
6242 | if (cctx->appliedParams.nbWorkers > 0) { |
6243 | size_t flushMin; |
6244 | if (cctx->cParamsChanged) { |
6245 | ZSTDMT_updateCParams_whileCompressing(cctx->mtctx, &cctx->requestedParams); |
6246 | cctx->cParamsChanged = 0; |
6247 | } |
6248 | if (cctx->stableIn_notConsumed) { |
6249 | assert(cctx->appliedParams.inBufferMode == ZSTD_bm_stable); |
6250 | /* some early data was skipped - make it available for consumption */ |
6251 | assert(input->pos >= cctx->stableIn_notConsumed); |
6252 | input->pos -= cctx->stableIn_notConsumed; |
6253 | cctx->stableIn_notConsumed = 0; |
6254 | } |
6255 | for (;;) { |
6256 | size_t const ipos = input->pos; |
6257 | size_t const opos = output->pos; |
6258 | flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp); |
6259 | cctx->consumedSrcSize += (U64)(input->pos - ipos); |
6260 | cctx->producedCSize += (U64)(output->pos - opos); |
6261 | if ( ZSTD_isError(flushMin) |
6262 | || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */ |
6263 | if (flushMin == 0) |
6264 | ZSTD_CCtx_trace(cctx, 0); |
6265 | ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only); |
6266 | } |
6267 | FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed" ); |
6268 | |
6269 | if (endOp == ZSTD_e_continue) { |
6270 | /* We only require some progress with ZSTD_e_continue, not maximal progress. |
6271 | * We're done if we've consumed or produced any bytes, or either buffer is |
6272 | * full. |
6273 | */ |
6274 | if (input->pos != ipos || output->pos != opos || input->pos == input->size || output->pos == output->size) |
6275 | break; |
6276 | } else { |
6277 | assert(endOp == ZSTD_e_flush || endOp == ZSTD_e_end); |
6278 | /* We require maximal progress. We're done when the flush is complete or the |
6279 | * output buffer is full. |
6280 | */ |
6281 | if (flushMin == 0 || output->pos == output->size) |
6282 | break; |
6283 | } |
6284 | } |
6285 | DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic" ); |
6286 | /* Either we don't require maximum forward progress, we've finished the |
6287 | * flush, or we are out of output space. |
6288 | */ |
6289 | assert(endOp == ZSTD_e_continue || flushMin == 0 || output->pos == output->size); |
6290 | ZSTD_setBufferExpectations(cctx, output, input); |
6291 | return flushMin; |
6292 | } |
6293 | #endif /* ZSTD_MULTITHREAD */ |
6294 | FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "" ); |
6295 | DEBUGLOG(5, "completed ZSTD_compressStream2" ); |
6296 | ZSTD_setBufferExpectations(cctx, output, input); |
6297 | return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */ |
6298 | } |
6299 | |
6300 | size_t ZSTD_compressStream2_simpleArgs ( |
6301 | ZSTD_CCtx* cctx, |
6302 | void* dst, size_t dstCapacity, size_t* dstPos, |
6303 | const void* src, size_t srcSize, size_t* srcPos, |
6304 | ZSTD_EndDirective endOp) |
6305 | { |
6306 | ZSTD_outBuffer output; |
6307 | ZSTD_inBuffer input; |
6308 | output.dst = dst; |
6309 | output.size = dstCapacity; |
6310 | output.pos = *dstPos; |
6311 | input.src = src; |
6312 | input.size = srcSize; |
6313 | input.pos = *srcPos; |
6314 | /* ZSTD_compressStream2() will check validity of dstPos and srcPos */ |
6315 | { size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp); |
6316 | *dstPos = output.pos; |
6317 | *srcPos = input.pos; |
6318 | return cErr; |
6319 | } |
6320 | } |
6321 | |
6322 | size_t ZSTD_compress2(ZSTD_CCtx* cctx, |
6323 | void* dst, size_t dstCapacity, |
6324 | const void* src, size_t srcSize) |
6325 | { |
6326 | ZSTD_bufferMode_e const originalInBufferMode = cctx->requestedParams.inBufferMode; |
6327 | ZSTD_bufferMode_e const originalOutBufferMode = cctx->requestedParams.outBufferMode; |
6328 | DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)" , (unsigned)srcSize); |
6329 | ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only); |
6330 | /* Enable stable input/output buffers. */ |
6331 | cctx->requestedParams.inBufferMode = ZSTD_bm_stable; |
6332 | cctx->requestedParams.outBufferMode = ZSTD_bm_stable; |
6333 | { size_t oPos = 0; |
6334 | size_t iPos = 0; |
6335 | size_t const result = ZSTD_compressStream2_simpleArgs(cctx, |
6336 | dst, dstCapacity, &oPos, |
6337 | src, srcSize, &iPos, |
6338 | ZSTD_e_end); |
6339 | /* Reset to the original values. */ |
6340 | cctx->requestedParams.inBufferMode = originalInBufferMode; |
6341 | cctx->requestedParams.outBufferMode = originalOutBufferMode; |
6342 | |
6343 | FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed" ); |
6344 | if (result != 0) { /* compression not completed, due to lack of output space */ |
6345 | assert(oPos == dstCapacity); |
6346 | RETURN_ERROR(dstSize_tooSmall, "" ); |
6347 | } |
6348 | assert(iPos == srcSize); /* all input is expected consumed */ |
6349 | return oPos; |
6350 | } |
6351 | } |
6352 | |
6353 | /* ZSTD_validateSequence() : |
6354 | * @offCode : is presumed to follow format required by ZSTD_storeSeq() |
6355 | * @returns a ZSTD error code if sequence is not valid |
6356 | */ |
6357 | static size_t |
6358 | ZSTD_validateSequence(U32 offCode, U32 matchLength, U32 minMatch, |
6359 | size_t posInSrc, U32 windowLog, size_t dictSize, int useSequenceProducer) |
6360 | { |
6361 | U32 const windowSize = 1u << windowLog; |
6362 | /* posInSrc represents the amount of data the decoder would decode up to this point. |
6363 | * As long as the amount of data decoded is less than or equal to window size, offsets may be |
6364 | * larger than the total length of output decoded in order to reference the dict, even larger than |
6365 | * window size. After output surpasses windowSize, we're limited to windowSize offsets again. |
6366 | */ |
6367 | size_t const offsetBound = posInSrc > windowSize ? (size_t)windowSize : posInSrc + (size_t)dictSize; |
6368 | size_t const matchLenLowerBound = (minMatch == 3 || useSequenceProducer) ? 3 : 4; |
6369 | RETURN_ERROR_IF(offCode > OFFSET_TO_OFFBASE(offsetBound), externalSequences_invalid, "Offset too large!" ); |
6370 | /* Validate maxNbSeq is large enough for the given matchLength and minMatch */ |
6371 | RETURN_ERROR_IF(matchLength < matchLenLowerBound, externalSequences_invalid, "Matchlength too small for the minMatch" ); |
6372 | return 0; |
6373 | } |
6374 | |
6375 | /* Returns an offset code, given a sequence's raw offset, the ongoing repcode array, and whether litLength == 0 */ |
6376 | static U32 ZSTD_finalizeOffBase(U32 rawOffset, const U32 rep[ZSTD_REP_NUM], U32 ll0) |
6377 | { |
6378 | U32 offBase = OFFSET_TO_OFFBASE(rawOffset); |
6379 | |
6380 | if (!ll0 && rawOffset == rep[0]) { |
6381 | offBase = REPCODE1_TO_OFFBASE; |
6382 | } else if (rawOffset == rep[1]) { |
6383 | offBase = REPCODE_TO_OFFBASE(2 - ll0); |
6384 | } else if (rawOffset == rep[2]) { |
6385 | offBase = REPCODE_TO_OFFBASE(3 - ll0); |
6386 | } else if (ll0 && rawOffset == rep[0] - 1) { |
6387 | offBase = REPCODE3_TO_OFFBASE; |
6388 | } |
6389 | return offBase; |
6390 | } |
6391 | |
6392 | size_t |
6393 | ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, |
6394 | ZSTD_sequencePosition* seqPos, |
6395 | const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, |
6396 | const void* src, size_t blockSize, |
6397 | ZSTD_paramSwitch_e externalRepSearch) |
6398 | { |
6399 | U32 idx = seqPos->idx; |
6400 | U32 const startIdx = idx; |
6401 | BYTE const* ip = (BYTE const*)(src); |
6402 | const BYTE* const iend = ip + blockSize; |
6403 | repcodes_t updatedRepcodes; |
6404 | U32 dictSize; |
6405 | |
6406 | DEBUGLOG(5, "ZSTD_copySequencesToSeqStoreExplicitBlockDelim (blockSize = %zu)" , blockSize); |
6407 | |
6408 | if (cctx->cdict) { |
6409 | dictSize = (U32)cctx->cdict->dictContentSize; |
6410 | } else if (cctx->prefixDict.dict) { |
6411 | dictSize = (U32)cctx->prefixDict.dictSize; |
6412 | } else { |
6413 | dictSize = 0; |
6414 | } |
6415 | ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); |
6416 | for (; idx < inSeqsSize && (inSeqs[idx].matchLength != 0 || inSeqs[idx].offset != 0); ++idx) { |
6417 | U32 const litLength = inSeqs[idx].litLength; |
6418 | U32 const matchLength = inSeqs[idx].matchLength; |
6419 | U32 offBase; |
6420 | |
6421 | if (externalRepSearch == ZSTD_ps_disable) { |
6422 | offBase = OFFSET_TO_OFFBASE(inSeqs[idx].offset); |
6423 | } else { |
6424 | U32 const ll0 = (litLength == 0); |
6425 | offBase = ZSTD_finalizeOffBase(inSeqs[idx].offset, updatedRepcodes.rep, ll0); |
6426 | ZSTD_updateRep(updatedRepcodes.rep, offBase, ll0); |
6427 | } |
6428 | |
6429 | DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)" , offBase, matchLength, litLength); |
6430 | if (cctx->appliedParams.validateSequences) { |
6431 | seqPos->posInSrc += litLength + matchLength; |
6432 | FORWARD_IF_ERROR(ZSTD_validateSequence(offBase, matchLength, cctx->appliedParams.cParams.minMatch, seqPos->posInSrc, |
6433 | cctx->appliedParams.cParams.windowLog, dictSize, cctx->appliedParams.useSequenceProducer), |
6434 | "Sequence validation failed" ); |
6435 | } |
6436 | RETURN_ERROR_IF(idx - seqPos->idx >= cctx->seqStore.maxNbSeq, externalSequences_invalid, |
6437 | "Not enough memory allocated. Try adjusting ZSTD_c_minMatch." ); |
6438 | ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offBase, matchLength); |
6439 | ip += matchLength + litLength; |
6440 | } |
6441 | |
6442 | /* If we skipped repcode search while parsing, we need to update repcodes now */ |
6443 | assert(externalRepSearch != ZSTD_ps_auto); |
6444 | assert(idx >= startIdx); |
6445 | if (externalRepSearch == ZSTD_ps_disable && idx != startIdx) { |
6446 | U32* const rep = updatedRepcodes.rep; |
6447 | U32 lastSeqIdx = idx - 1; /* index of last non-block-delimiter sequence */ |
6448 | |
6449 | if (lastSeqIdx >= startIdx + 2) { |
6450 | rep[2] = inSeqs[lastSeqIdx - 2].offset; |
6451 | rep[1] = inSeqs[lastSeqIdx - 1].offset; |
6452 | rep[0] = inSeqs[lastSeqIdx].offset; |
6453 | } else if (lastSeqIdx == startIdx + 1) { |
6454 | rep[2] = rep[0]; |
6455 | rep[1] = inSeqs[lastSeqIdx - 1].offset; |
6456 | rep[0] = inSeqs[lastSeqIdx].offset; |
6457 | } else { |
6458 | assert(lastSeqIdx == startIdx); |
6459 | rep[2] = rep[1]; |
6460 | rep[1] = rep[0]; |
6461 | rep[0] = inSeqs[lastSeqIdx].offset; |
6462 | } |
6463 | } |
6464 | |
6465 | ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t)); |
6466 | |
6467 | if (inSeqs[idx].litLength) { |
6468 | DEBUGLOG(6, "Storing last literals of size: %u" , inSeqs[idx].litLength); |
6469 | ZSTD_storeLastLiterals(&cctx->seqStore, ip, inSeqs[idx].litLength); |
6470 | ip += inSeqs[idx].litLength; |
6471 | seqPos->posInSrc += inSeqs[idx].litLength; |
6472 | } |
6473 | RETURN_ERROR_IF(ip != iend, externalSequences_invalid, "Blocksize doesn't agree with block delimiter!" ); |
6474 | seqPos->idx = idx+1; |
6475 | return 0; |
6476 | } |
6477 | |
6478 | size_t |
6479 | ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos, |
6480 | const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, |
6481 | const void* src, size_t blockSize, ZSTD_paramSwitch_e externalRepSearch) |
6482 | { |
6483 | U32 idx = seqPos->idx; |
6484 | U32 startPosInSequence = seqPos->posInSequence; |
6485 | U32 endPosInSequence = seqPos->posInSequence + (U32)blockSize; |
6486 | size_t dictSize; |
6487 | BYTE const* ip = (BYTE const*)(src); |
6488 | BYTE const* iend = ip + blockSize; /* May be adjusted if we decide to process fewer than blockSize bytes */ |
6489 | repcodes_t updatedRepcodes; |
6490 | U32 bytesAdjustment = 0; |
6491 | U32 finalMatchSplit = 0; |
6492 | |
6493 | /* TODO(embg) support fast parsing mode in noBlockDelim mode */ |
6494 | (void)externalRepSearch; |
6495 | |
6496 | if (cctx->cdict) { |
6497 | dictSize = cctx->cdict->dictContentSize; |
6498 | } else if (cctx->prefixDict.dict) { |
6499 | dictSize = cctx->prefixDict.dictSize; |
6500 | } else { |
6501 | dictSize = 0; |
6502 | } |
6503 | DEBUGLOG(5, "ZSTD_copySequencesToSeqStoreNoBlockDelim: idx: %u PIS: %u blockSize: %zu" , idx, startPosInSequence, blockSize); |
6504 | DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)" , idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); |
6505 | ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); |
6506 | while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) { |
6507 | const ZSTD_Sequence currSeq = inSeqs[idx]; |
6508 | U32 litLength = currSeq.litLength; |
6509 | U32 matchLength = currSeq.matchLength; |
6510 | U32 const rawOffset = currSeq.offset; |
6511 | U32 offBase; |
6512 | |
6513 | /* Modify the sequence depending on where endPosInSequence lies */ |
6514 | if (endPosInSequence >= currSeq.litLength + currSeq.matchLength) { |
6515 | if (startPosInSequence >= litLength) { |
6516 | startPosInSequence -= litLength; |
6517 | litLength = 0; |
6518 | matchLength -= startPosInSequence; |
6519 | } else { |
6520 | litLength -= startPosInSequence; |
6521 | } |
6522 | /* Move to the next sequence */ |
6523 | endPosInSequence -= currSeq.litLength + currSeq.matchLength; |
6524 | startPosInSequence = 0; |
6525 | } else { |
6526 | /* This is the final (partial) sequence we're adding from inSeqs, and endPosInSequence |
6527 | does not reach the end of the match. So, we have to split the sequence */ |
6528 | DEBUGLOG(6, "Require a split: diff: %u, idx: %u PIS: %u" , |
6529 | currSeq.litLength + currSeq.matchLength - endPosInSequence, idx, endPosInSequence); |
6530 | if (endPosInSequence > litLength) { |
6531 | U32 firstHalfMatchLength; |
6532 | litLength = startPosInSequence >= litLength ? 0 : litLength - startPosInSequence; |
6533 | firstHalfMatchLength = endPosInSequence - startPosInSequence - litLength; |
6534 | if (matchLength > blockSize && firstHalfMatchLength >= cctx->appliedParams.cParams.minMatch) { |
6535 | /* Only ever split the match if it is larger than the block size */ |
6536 | U32 secondHalfMatchLength = currSeq.matchLength + currSeq.litLength - endPosInSequence; |
6537 | if (secondHalfMatchLength < cctx->appliedParams.cParams.minMatch) { |
6538 | /* Move the endPosInSequence backward so that it creates match of minMatch length */ |
6539 | endPosInSequence -= cctx->appliedParams.cParams.minMatch - secondHalfMatchLength; |
6540 | bytesAdjustment = cctx->appliedParams.cParams.minMatch - secondHalfMatchLength; |
6541 | firstHalfMatchLength -= bytesAdjustment; |
6542 | } |
6543 | matchLength = firstHalfMatchLength; |
6544 | /* Flag that we split the last match - after storing the sequence, exit the loop, |
6545 | but keep the value of endPosInSequence */ |
6546 | finalMatchSplit = 1; |
6547 | } else { |
6548 | /* Move the position in sequence backwards so that we don't split match, and break to store |
6549 | * the last literals. We use the original currSeq.litLength as a marker for where endPosInSequence |
6550 | * should go. We prefer to do this whenever it is not necessary to split the match, or if doing so |
6551 | * would cause the first half of the match to be too small |
6552 | */ |
6553 | bytesAdjustment = endPosInSequence - currSeq.litLength; |
6554 | endPosInSequence = currSeq.litLength; |
6555 | break; |
6556 | } |
6557 | } else { |
6558 | /* This sequence ends inside the literals, break to store the last literals */ |
6559 | break; |
6560 | } |
6561 | } |
6562 | /* Check if this offset can be represented with a repcode */ |
6563 | { U32 const ll0 = (litLength == 0); |
6564 | offBase = ZSTD_finalizeOffBase(rawOffset, updatedRepcodes.rep, ll0); |
6565 | ZSTD_updateRep(updatedRepcodes.rep, offBase, ll0); |
6566 | } |
6567 | |
6568 | if (cctx->appliedParams.validateSequences) { |
6569 | seqPos->posInSrc += litLength + matchLength; |
6570 | FORWARD_IF_ERROR(ZSTD_validateSequence(offBase, matchLength, cctx->appliedParams.cParams.minMatch, seqPos->posInSrc, |
6571 | cctx->appliedParams.cParams.windowLog, dictSize, cctx->appliedParams.useSequenceProducer), |
6572 | "Sequence validation failed" ); |
6573 | } |
6574 | DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)" , offBase, matchLength, litLength); |
6575 | RETURN_ERROR_IF(idx - seqPos->idx >= cctx->seqStore.maxNbSeq, externalSequences_invalid, |
6576 | "Not enough memory allocated. Try adjusting ZSTD_c_minMatch." ); |
6577 | ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offBase, matchLength); |
6578 | ip += matchLength + litLength; |
6579 | if (!finalMatchSplit) |
6580 | idx++; /* Next Sequence */ |
6581 | } |
6582 | DEBUGLOG(5, "Ending seq: idx: %u (of: %u ml: %u ll: %u)" , idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); |
6583 | assert(idx == inSeqsSize || endPosInSequence <= inSeqs[idx].litLength + inSeqs[idx].matchLength); |
6584 | seqPos->idx = idx; |
6585 | seqPos->posInSequence = endPosInSequence; |
6586 | ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t)); |
6587 | |
6588 | iend -= bytesAdjustment; |
6589 | if (ip != iend) { |
6590 | /* Store any last literals */ |
6591 | U32 lastLLSize = (U32)(iend - ip); |
6592 | assert(ip <= iend); |
6593 | DEBUGLOG(6, "Storing last literals of size: %u" , lastLLSize); |
6594 | ZSTD_storeLastLiterals(&cctx->seqStore, ip, lastLLSize); |
6595 | seqPos->posInSrc += lastLLSize; |
6596 | } |
6597 | |
6598 | return bytesAdjustment; |
6599 | } |
6600 | |
6601 | typedef size_t (*ZSTD_sequenceCopier) (ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos, |
6602 | const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, |
6603 | const void* src, size_t blockSize, ZSTD_paramSwitch_e externalRepSearch); |
6604 | static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode) |
6605 | { |
6606 | ZSTD_sequenceCopier sequenceCopier = NULL; |
6607 | assert(ZSTD_cParam_withinBounds(ZSTD_c_blockDelimiters, mode)); |
6608 | if (mode == ZSTD_sf_explicitBlockDelimiters) { |
6609 | return ZSTD_copySequencesToSeqStoreExplicitBlockDelim; |
6610 | } else if (mode == ZSTD_sf_noBlockDelimiters) { |
6611 | return ZSTD_copySequencesToSeqStoreNoBlockDelim; |
6612 | } |
6613 | assert(sequenceCopier != NULL); |
6614 | return sequenceCopier; |
6615 | } |
6616 | |
6617 | /* Discover the size of next block by searching for the delimiter. |
6618 | * Note that a block delimiter **must** exist in this mode, |
6619 | * otherwise it's an input error. |
6620 | * The block size retrieved will be later compared to ensure it remains within bounds */ |
6621 | static size_t |
6622 | blockSize_explicitDelimiter(const ZSTD_Sequence* inSeqs, size_t inSeqsSize, ZSTD_sequencePosition seqPos) |
6623 | { |
6624 | int end = 0; |
6625 | size_t blockSize = 0; |
6626 | size_t spos = seqPos.idx; |
6627 | DEBUGLOG(6, "blockSize_explicitDelimiter : seq %zu / %zu" , spos, inSeqsSize); |
6628 | assert(spos <= inSeqsSize); |
6629 | while (spos < inSeqsSize) { |
6630 | end = (inSeqs[spos].offset == 0); |
6631 | blockSize += inSeqs[spos].litLength + inSeqs[spos].matchLength; |
6632 | if (end) { |
6633 | if (inSeqs[spos].matchLength != 0) |
6634 | RETURN_ERROR(externalSequences_invalid, "delimiter format error : both matchlength and offset must be == 0" ); |
6635 | break; |
6636 | } |
6637 | spos++; |
6638 | } |
6639 | if (!end) |
6640 | RETURN_ERROR(externalSequences_invalid, "Reached end of sequences without finding a block delimiter" ); |
6641 | return blockSize; |
6642 | } |
6643 | |
6644 | /* More a "target" block size */ |
6645 | static size_t blockSize_noDelimiter(size_t blockSize, size_t remaining) |
6646 | { |
6647 | int const lastBlock = (remaining <= blockSize); |
6648 | return lastBlock ? remaining : blockSize; |
6649 | } |
6650 | |
6651 | static size_t determine_blockSize(ZSTD_sequenceFormat_e mode, |
6652 | size_t blockSize, size_t remaining, |
6653 | const ZSTD_Sequence* inSeqs, size_t inSeqsSize, ZSTD_sequencePosition seqPos) |
6654 | { |
6655 | DEBUGLOG(6, "determine_blockSize : remainingSize = %zu" , remaining); |
6656 | if (mode == ZSTD_sf_noBlockDelimiters) |
6657 | return blockSize_noDelimiter(blockSize, remaining); |
6658 | { size_t const explicitBlockSize = blockSize_explicitDelimiter(inSeqs, inSeqsSize, seqPos); |
6659 | FORWARD_IF_ERROR(explicitBlockSize, "Error while determining block size with explicit delimiters" ); |
6660 | if (explicitBlockSize > blockSize) |
6661 | RETURN_ERROR(externalSequences_invalid, "sequences incorrectly define a too large block" ); |
6662 | if (explicitBlockSize > remaining) |
6663 | RETURN_ERROR(externalSequences_invalid, "sequences define a frame longer than source" ); |
6664 | return explicitBlockSize; |
6665 | } |
6666 | } |
6667 | |
6668 | /* Compress, block-by-block, all of the sequences given. |
6669 | * |
6670 | * Returns the cumulative size of all compressed blocks (including their headers), |
6671 | * otherwise a ZSTD error. |
6672 | */ |
6673 | static size_t |
6674 | ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, |
6675 | void* dst, size_t dstCapacity, |
6676 | const ZSTD_Sequence* inSeqs, size_t inSeqsSize, |
6677 | const void* src, size_t srcSize) |
6678 | { |
6679 | size_t cSize = 0; |
6680 | size_t remaining = srcSize; |
6681 | ZSTD_sequencePosition seqPos = {0, 0, 0}; |
6682 | |
6683 | BYTE const* ip = (BYTE const*)src; |
6684 | BYTE* op = (BYTE*)dst; |
6685 | ZSTD_sequenceCopier const sequenceCopier = ZSTD_selectSequenceCopier(cctx->appliedParams.blockDelimiters); |
6686 | |
6687 | DEBUGLOG(4, "ZSTD_compressSequences_internal srcSize: %zu, inSeqsSize: %zu" , srcSize, inSeqsSize); |
6688 | /* Special case: empty frame */ |
6689 | if (remaining == 0) { |
6690 | U32 const = 1 /* last block */ + (((U32)bt_raw)<<1); |
6691 | RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "No room for empty frame block header" ); |
6692 | MEM_writeLE32(op, cBlockHeader24); |
6693 | op += ZSTD_blockHeaderSize; |
6694 | dstCapacity -= ZSTD_blockHeaderSize; |
6695 | cSize += ZSTD_blockHeaderSize; |
6696 | } |
6697 | |
6698 | while (remaining) { |
6699 | size_t compressedSeqsSize; |
6700 | size_t cBlockSize; |
6701 | size_t additionalByteAdjustment; |
6702 | size_t blockSize = determine_blockSize(cctx->appliedParams.blockDelimiters, |
6703 | cctx->blockSize, remaining, |
6704 | inSeqs, inSeqsSize, seqPos); |
6705 | U32 const lastBlock = (blockSize == remaining); |
6706 | FORWARD_IF_ERROR(blockSize, "Error while trying to determine block size" ); |
6707 | assert(blockSize <= remaining); |
6708 | ZSTD_resetSeqStore(&cctx->seqStore); |
6709 | DEBUGLOG(5, "Working on new block. Blocksize: %zu (total:%zu)" , blockSize, (ip - (const BYTE*)src) + blockSize); |
6710 | |
6711 | additionalByteAdjustment = sequenceCopier(cctx, &seqPos, inSeqs, inSeqsSize, ip, blockSize, cctx->appliedParams.searchForExternalRepcodes); |
6712 | FORWARD_IF_ERROR(additionalByteAdjustment, "Bad sequence copy" ); |
6713 | blockSize -= additionalByteAdjustment; |
6714 | |
6715 | /* If blocks are too small, emit as a nocompress block */ |
6716 | /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding |
6717 | * additional 1. We need to revisit and change this logic to be more consistent */ |
6718 | if (blockSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1+1) { |
6719 | cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); |
6720 | FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed" ); |
6721 | DEBUGLOG(5, "Block too small, writing out nocompress block: cSize: %zu" , cBlockSize); |
6722 | cSize += cBlockSize; |
6723 | ip += blockSize; |
6724 | op += cBlockSize; |
6725 | remaining -= blockSize; |
6726 | dstCapacity -= cBlockSize; |
6727 | continue; |
6728 | } |
6729 | |
6730 | RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "not enough dstCapacity to write a new compressed block" ); |
6731 | compressedSeqsSize = ZSTD_entropyCompressSeqStore(&cctx->seqStore, |
6732 | &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy, |
6733 | &cctx->appliedParams, |
6734 | op + ZSTD_blockHeaderSize /* Leave space for block header */, dstCapacity - ZSTD_blockHeaderSize, |
6735 | blockSize, |
6736 | cctx->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, |
6737 | cctx->bmi2); |
6738 | FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed" ); |
6739 | DEBUGLOG(5, "Compressed sequences size: %zu" , compressedSeqsSize); |
6740 | |
6741 | if (!cctx->isFirstBlock && |
6742 | ZSTD_maybeRLE(&cctx->seqStore) && |
6743 | ZSTD_isRLE(ip, blockSize)) { |
6744 | /* We don't want to emit our first block as a RLE even if it qualifies because |
6745 | * doing so will cause the decoder (cli only) to throw a "should consume all input error." |
6746 | * This is only an issue for zstd <= v1.4.3 |
6747 | */ |
6748 | compressedSeqsSize = 1; |
6749 | } |
6750 | |
6751 | if (compressedSeqsSize == 0) { |
6752 | /* ZSTD_noCompressBlock writes the block header as well */ |
6753 | cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); |
6754 | FORWARD_IF_ERROR(cBlockSize, "ZSTD_noCompressBlock failed" ); |
6755 | DEBUGLOG(5, "Writing out nocompress block, size: %zu" , cBlockSize); |
6756 | } else if (compressedSeqsSize == 1) { |
6757 | cBlockSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, blockSize, lastBlock); |
6758 | FORWARD_IF_ERROR(cBlockSize, "ZSTD_rleCompressBlock failed" ); |
6759 | DEBUGLOG(5, "Writing out RLE block, size: %zu" , cBlockSize); |
6760 | } else { |
6761 | U32 ; |
6762 | /* Error checking and repcodes update */ |
6763 | ZSTD_blockState_confirmRepcodesAndEntropyTables(&cctx->blockState); |
6764 | if (cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) |
6765 | cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; |
6766 | |
6767 | /* Write block header into beginning of block*/ |
6768 | cBlockHeader = lastBlock + (((U32)bt_compressed)<<1) + (U32)(compressedSeqsSize << 3); |
6769 | MEM_writeLE24(op, cBlockHeader); |
6770 | cBlockSize = ZSTD_blockHeaderSize + compressedSeqsSize; |
6771 | DEBUGLOG(5, "Writing out compressed block, size: %zu" , cBlockSize); |
6772 | } |
6773 | |
6774 | cSize += cBlockSize; |
6775 | |
6776 | if (lastBlock) { |
6777 | break; |
6778 | } else { |
6779 | ip += blockSize; |
6780 | op += cBlockSize; |
6781 | remaining -= blockSize; |
6782 | dstCapacity -= cBlockSize; |
6783 | cctx->isFirstBlock = 0; |
6784 | } |
6785 | DEBUGLOG(5, "cSize running total: %zu (remaining dstCapacity=%zu)" , cSize, dstCapacity); |
6786 | } |
6787 | |
6788 | DEBUGLOG(4, "cSize final total: %zu" , cSize); |
6789 | return cSize; |
6790 | } |
6791 | |
6792 | size_t ZSTD_compressSequences(ZSTD_CCtx* cctx, |
6793 | void* dst, size_t dstCapacity, |
6794 | const ZSTD_Sequence* inSeqs, size_t inSeqsSize, |
6795 | const void* src, size_t srcSize) |
6796 | { |
6797 | BYTE* op = (BYTE*)dst; |
6798 | size_t cSize = 0; |
6799 | size_t compressedBlocksSize = 0; |
6800 | size_t = 0; |
6801 | |
6802 | /* Transparent initialization stage, same as compressStream2() */ |
6803 | DEBUGLOG(4, "ZSTD_compressSequences (dstCapacity=%zu)" , dstCapacity); |
6804 | assert(cctx != NULL); |
6805 | FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed" ); |
6806 | /* Begin writing output, starting with frame header */ |
6807 | frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity, &cctx->appliedParams, srcSize, cctx->dictID); |
6808 | op += frameHeaderSize; |
6809 | dstCapacity -= frameHeaderSize; |
6810 | cSize += frameHeaderSize; |
6811 | if (cctx->appliedParams.fParams.checksumFlag && srcSize) { |
6812 | XXH64_update(&cctx->xxhState, src, srcSize); |
6813 | } |
6814 | /* cSize includes block header size and compressed sequences size */ |
6815 | compressedBlocksSize = ZSTD_compressSequences_internal(cctx, |
6816 | op, dstCapacity, |
6817 | inSeqs, inSeqsSize, |
6818 | src, srcSize); |
6819 | FORWARD_IF_ERROR(compressedBlocksSize, "Compressing blocks failed!" ); |
6820 | cSize += compressedBlocksSize; |
6821 | dstCapacity -= compressedBlocksSize; |
6822 | |
6823 | if (cctx->appliedParams.fParams.checksumFlag) { |
6824 | U32 const checksum = (U32) XXH64_digest(&cctx->xxhState); |
6825 | RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum" ); |
6826 | DEBUGLOG(4, "Write checksum : %08X" , (unsigned)checksum); |
6827 | MEM_writeLE32((char*)dst + cSize, checksum); |
6828 | cSize += 4; |
6829 | } |
6830 | |
6831 | DEBUGLOG(4, "Final compressed size: %zu" , cSize); |
6832 | return cSize; |
6833 | } |
6834 | |
6835 | /*====== Finalize ======*/ |
6836 | |
6837 | static ZSTD_inBuffer inBuffer_forEndFlush(const ZSTD_CStream* zcs) |
6838 | { |
6839 | const ZSTD_inBuffer nullInput = { NULL, 0, 0 }; |
6840 | const int stableInput = (zcs->appliedParams.inBufferMode == ZSTD_bm_stable); |
6841 | return stableInput ? zcs->expectedInBuffer : nullInput; |
6842 | } |
6843 | |
6844 | /*! ZSTD_flushStream() : |
6845 | * @return : amount of data remaining to flush */ |
6846 | size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output) |
6847 | { |
6848 | ZSTD_inBuffer input = inBuffer_forEndFlush(zcs); |
6849 | input.size = input.pos; /* do not ingest more input during flush */ |
6850 | return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush); |
6851 | } |
6852 | |
6853 | |
6854 | size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output) |
6855 | { |
6856 | ZSTD_inBuffer input = inBuffer_forEndFlush(zcs); |
6857 | size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end); |
6858 | FORWARD_IF_ERROR(remainingToFlush , "ZSTD_compressStream2(,,ZSTD_e_end) failed" ); |
6859 | if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */ |
6860 | /* single thread mode : attempt to calculate remaining to flush more precisely */ |
6861 | { size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE; |
6862 | size_t const checksumSize = (size_t)(zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4); |
6863 | size_t const toFlush = remainingToFlush + lastBlockSize + checksumSize; |
6864 | DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u" , (unsigned)toFlush); |
6865 | return toFlush; |
6866 | } |
6867 | } |
6868 | |
6869 | |
6870 | /*-===== Pre-defined compression levels =====-*/ |
6871 | #include "clevels.h" |
6872 | |
6873 | int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; } |
6874 | int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; } |
6875 | int ZSTD_defaultCLevel(void) { return ZSTD_CLEVEL_DEFAULT; } |
6876 | |
6877 | static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel, size_t const dictSize) |
6878 | { |
6879 | ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, 0, dictSize, ZSTD_cpm_createCDict); |
6880 | switch (cParams.strategy) { |
6881 | case ZSTD_fast: |
6882 | case ZSTD_dfast: |
6883 | break; |
6884 | case ZSTD_greedy: |
6885 | case ZSTD_lazy: |
6886 | case ZSTD_lazy2: |
6887 | cParams.hashLog += ZSTD_LAZY_DDSS_BUCKET_LOG; |
6888 | break; |
6889 | case ZSTD_btlazy2: |
6890 | case ZSTD_btopt: |
6891 | case ZSTD_btultra: |
6892 | case ZSTD_btultra2: |
6893 | break; |
6894 | } |
6895 | return cParams; |
6896 | } |
6897 | |
6898 | static int ZSTD_dedicatedDictSearch_isSupported( |
6899 | ZSTD_compressionParameters const* cParams) |
6900 | { |
6901 | return (cParams->strategy >= ZSTD_greedy) |
6902 | && (cParams->strategy <= ZSTD_lazy2) |
6903 | && (cParams->hashLog > cParams->chainLog) |
6904 | && (cParams->chainLog <= 24); |
6905 | } |
6906 | |
6907 | /** |
6908 | * Reverses the adjustment applied to cparams when enabling dedicated dict |
6909 | * search. This is used to recover the params set to be used in the working |
6910 | * context. (Otherwise, those tables would also grow.) |
6911 | */ |
6912 | static void ZSTD_dedicatedDictSearch_revertCParams( |
6913 | ZSTD_compressionParameters* cParams) { |
6914 | switch (cParams->strategy) { |
6915 | case ZSTD_fast: |
6916 | case ZSTD_dfast: |
6917 | break; |
6918 | case ZSTD_greedy: |
6919 | case ZSTD_lazy: |
6920 | case ZSTD_lazy2: |
6921 | cParams->hashLog -= ZSTD_LAZY_DDSS_BUCKET_LOG; |
6922 | if (cParams->hashLog < ZSTD_HASHLOG_MIN) { |
6923 | cParams->hashLog = ZSTD_HASHLOG_MIN; |
6924 | } |
6925 | break; |
6926 | case ZSTD_btlazy2: |
6927 | case ZSTD_btopt: |
6928 | case ZSTD_btultra: |
6929 | case ZSTD_btultra2: |
6930 | break; |
6931 | } |
6932 | } |
6933 | |
6934 | static U64 ZSTD_getCParamRowSize(U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) |
6935 | { |
6936 | switch (mode) { |
6937 | case ZSTD_cpm_unknown: |
6938 | case ZSTD_cpm_noAttachDict: |
6939 | case ZSTD_cpm_createCDict: |
6940 | break; |
6941 | case ZSTD_cpm_attachDict: |
6942 | dictSize = 0; |
6943 | break; |
6944 | default: |
6945 | assert(0); |
6946 | break; |
6947 | } |
6948 | { int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN; |
6949 | size_t const addedSize = unknown && dictSize > 0 ? 500 : 0; |
6950 | return unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize; |
6951 | } |
6952 | } |
6953 | |
6954 | /*! ZSTD_getCParams_internal() : |
6955 | * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize. |
6956 | * Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown. |
6957 | * Use dictSize == 0 for unknown or unused. |
6958 | * Note: `mode` controls how we treat the `dictSize`. See docs for `ZSTD_cParamMode_e`. */ |
6959 | static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) |
6960 | { |
6961 | U64 const rSize = ZSTD_getCParamRowSize(srcSizeHint, dictSize, mode); |
6962 | U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB); |
6963 | int row; |
6964 | DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)" , compressionLevel); |
6965 | |
6966 | /* row */ |
6967 | if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */ |
6968 | else if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */ |
6969 | else if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL; |
6970 | else row = compressionLevel; |
6971 | |
6972 | { ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row]; |
6973 | DEBUGLOG(5, "ZSTD_getCParams_internal selected tableID: %u row: %u strat: %u" , tableID, row, (U32)cp.strategy); |
6974 | /* acceleration factor */ |
6975 | if (compressionLevel < 0) { |
6976 | int const clampedCompressionLevel = MAX(ZSTD_minCLevel(), compressionLevel); |
6977 | cp.targetLength = (unsigned)(-clampedCompressionLevel); |
6978 | } |
6979 | /* refine parameters based on srcSize & dictSize */ |
6980 | return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode, ZSTD_ps_auto); |
6981 | } |
6982 | } |
6983 | |
6984 | /*! ZSTD_getCParams() : |
6985 | * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize. |
6986 | * Size values are optional, provide 0 if not known or unused */ |
6987 | ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) |
6988 | { |
6989 | if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN; |
6990 | return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown); |
6991 | } |
6992 | |
6993 | /*! ZSTD_getParams() : |
6994 | * same idea as ZSTD_getCParams() |
6995 | * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`). |
6996 | * Fields of `ZSTD_frameParameters` are set to default values */ |
6997 | static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) { |
6998 | ZSTD_parameters params; |
6999 | ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, mode); |
7000 | DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)" , compressionLevel); |
7001 | ZSTD_memset(¶ms, 0, sizeof(params)); |
7002 | params.cParams = cParams; |
7003 | params.fParams.contentSizeFlag = 1; |
7004 | return params; |
7005 | } |
7006 | |
7007 | /*! ZSTD_getParams() : |
7008 | * same idea as ZSTD_getCParams() |
7009 | * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`). |
7010 | * Fields of `ZSTD_frameParameters` are set to default values */ |
7011 | ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) { |
7012 | if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN; |
7013 | return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown); |
7014 | } |
7015 | |
7016 | void ZSTD_registerSequenceProducer( |
7017 | ZSTD_CCtx* zc, void* mState, |
7018 | ZSTD_sequenceProducer_F* mFinder |
7019 | ) { |
7020 | if (mFinder != NULL) { |
7021 | ZSTD_externalMatchCtx emctx; |
7022 | emctx.mState = mState; |
7023 | emctx.mFinder = mFinder; |
7024 | emctx.seqBuffer = NULL; |
7025 | emctx.seqBufferCapacity = 0; |
7026 | zc->externalMatchCtx = emctx; |
7027 | zc->requestedParams.useSequenceProducer = 1; |
7028 | } else { |
7029 | ZSTD_memset(&zc->externalMatchCtx, 0, sizeof(zc->externalMatchCtx)); |
7030 | zc->requestedParams.useSequenceProducer = 0; |
7031 | } |
7032 | } |
7033 | |