Split block compresser out of long range matcher

* `ZSTD_ldm_generateSequences()` generates the LDM sequences and
  stores them in a table. It should work with any chunk size, but
  is currently only called one block at a time.
* `ZSTD_ldm_blockCompress()` emits the pre-defined sequences, and
  instead of encoding the literals directly, it passes them to a
  secondary block compressor. The code to handle chunk sizes greater
  than the block size is currently commented out, since it is unused.
  The next PR will uncomment exercise this code.
* During optimal parsing, ensure LDM `minMatchLength` is at least
  `targetLength`. Also don't emit repcode matches in the LDM block
  compressor. Enabling the LDM with the optimal parser now actually improves
  the compression ratio.
* The compression ratio is very similar to before. It is very slightly
  different, because the repcode handling is slightly different. If I remove
  immediate repcode checking in both branches the compressed size is exactly
  the same.
* The speed looks to be the same or better than before.

Up Next (in a separate PR)
--------------------------

Allow sequence generation to happen prior to compression, and produce more
than a block worth of sequences. Expose some API for zstdmt to consume.
This will test out some currently untested code in
`ZSTD_ldm_blockCompress()`.

lib/compress/zstd_ldm.h

@@ -24,34 +24,57 @@ extern "C" {
 #define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_DEFAULTMAX
 #define ZSTD_LDM_HASHEVERYLOG_NOTSET 9999
 
-/** ZSTD_compressBlock_ldm_generic() :
+/**
+ * ZSTD_ldm_generateSequences():
  *
- *  This is a block compressor intended for long distance matching.
+ * Generates the sequences using the long distance match finder.
+ * The sequences completely parse a prefix of the source, but leave off the last
+ * literals. Returns the number of sequences generated into `sequences`.
  *
- *  The function searches for matches of length at least
- *  ldmParams.minMatchLength using a hash table in cctx->ldmState.
- *  Matches can be at a distance of up to cParams.windowLog.
+ * NOTE: The source may be any size, assuming it doesn't overflow the hash table
+ * indices, and the output sequences table is large enough..
+ */
+size_t ZSTD_ldm_generateSequences(
+        ldmState_t* ldms, rawSeq* sequences, ZSTD_matchState_t const* ms,
+        ldmParams_t const* params, void const* src, size_t srcSize,
+        int const extDict);
+
+/**
+ * ZSTD_ldm_blockCompress():
  *
- *  Upon finding a match, the unmatched literals are compressed using a
- *  ZSTD_blockCompressor (depending on the strategy in the compression
- *  parameters), which stores the matched sequences. The "long distance"
- *  match is then stored with the remaining literals from the
- *  ZSTD_blockCompressor. */
-size_t ZSTD_compressBlock_ldm(
-        ldmState_t* ldms, ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
-        ZSTD_CCtx_params const* params, void const* src, size_t srcSize);
+ * Compresses a block using the predefined sequences, along with a secondary
+ * block compressor. The literals section of every sequence is passed to the
+ * secondary block compressor, and those sequences are interspersed with the
+ * predefined sequences. Returns the length of the last literals.
+ * `nbSeq` is the number of sequences available in `sequences`.
+ *
+ * NOTE: The source must be at most the maximum block size, but the predefined
+ * sequences can be any size, and may be longer than the block. In the case that
+ * they are longer than the block, the last sequences may need to be split into
+ * two. We handle that case correctly, and update `sequences` and `nbSeq`
+ * appropriately.
+ */
+size_t ZSTD_ldm_blockCompress(rawSeq const* sequences, size_t nbSeq,
+    ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+    ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
+    int const extDict);
 
-size_t ZSTD_compressBlock_ldm_extDict(
-        ldmState_t* ldms, ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
-        ZSTD_CCtx_params const* params, void const* src, size_t srcSize);
 
 /** ZSTD_ldm_initializeParameters() :
  *  Initialize the long distance matching parameters to their default values. */
 size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm);
 
 /** ZSTD_ldm_getTableSize() :
- *  Estimate the space needed for long distance matching tables. */
-size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog);
+ *  Estimate the space needed for long distance matching tables or 0 if LDM is
+ *  disabled.
+ */
+size_t ZSTD_ldm_getTableSize(ldmParams_t params);
+
+/** ZSTD_ldm_getSeqSpace() :
+ *  Return an upper bound on the number of sequences that can be produced by
+ *  the long distance matcher, or 0 if LDM is disabled.
+ */
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
 
 /** ZSTD_ldm_getTableSize() :
  *  Return prime8bytes^(minMatchLength-1) */
@@ -62,8 +85,12 @@ U64 ZSTD_ldm_getHashPower(U32 minMatchLength);
  *  windowLog and params->hashLog.
  *
  *  Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
- *  params->hashLog if it is not). */
-void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog);
+ *  params->hashLog if it is not).
+ *
+ *  Ensures that the minMatchLength >= targetLength during optimal parsing.
+ */
+void ZSTD_ldm_adjustParameters(ldmParams_t* params,
+                               ZSTD_compressionParameters const* cParams);
 
 #if defined (__cplusplus)
 }