AuroraMiddleware/zstd
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Document the ZSTD_c_enableDedicatedDictSearch Parameter

Browse Source
This commit is contained in:
W. Felix Handte 2020-08-18 16:58:11 -04:00
parent 5390fee4f7
commit ad9f98ac3f
1 changed files with 45 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
lib/zstd.h
View File

@ -1546,7 +1546,51 @@ ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* pre
* but compression ratio may regress significantly if guess considerably underestimates */
#define ZSTD_c_srcSizeHint ZSTD_c_experimentalParam7
/* TODO: document.
/* Controls whether the new and experimental "dedicated dictionary search
* structure" can be used.
*
* How to use it:
*
* When using a CDict, whether to use this feature or not is controlled at
* CDict creation, and it must be set in a CCtxParams set passed into that
* construction. A compression will then use the feature or not based on how
* the CDict was constructed; the value of this param, set in the CCtx, will
* have no effect.
*
* However, when a dictionary buffer is passed into a CCtx, such as via
* ZSTD_CCtx_loadDictionary(), this param can be set on the CCtx to control
* whether the CDict that is created internally can use the feature or not.
*
* What it does:
*
* Normally, the internal data structures of the CDict are analogous to what
* would be stored in a CCtx after compressing the contents of a dictionary.
* To an approximation, a compression using a dictionary can then use those
* data structures to simply continue what is effectively a streaming
* compression where the simulated compression of the dictionary left off.
* Which is to say, the search structures in the CDict are normally the same
* format as in the CCtx.
*
* It is possible to do better, since the CDict is not like a CCtx: the search
* structures are written once during CDict creation, and then are only read
* after that, while the search structures in the CCtx are both read and
* written as the compression goes along. This means we can choose a search
* structure for the dictionary that is read-optimized.
*
* This feature enables the use of that different structure. Note that this
* means that the CDict tables can no longer be copied into the CCtx, so
* the dict attachment mode ZSTD_dictForceCopy will no longer be useable. The
* dictionary can only be attached or reloaded.
*
* Effects:
*
* This will only have any effect when the selected ZSTD_strategy
* implementation supports this feature. Currently, that's limited to
* ZSTD_greedy, ZSTD_lazy, and ZSTD_lazy2.
*
* In general, you should expect compression to be faster, and CDict creation
* to be slightly slower. Eventually, we will probably make this mode the
* default.
*/
#define ZSTD_c_enableDedicatedDictSearch ZSTD_c_experimentalParam8