zstd-ruby 1.4.4.0 → 1.4.5.0

data/ext/zstdruby/libzstd/compress/zstd_ldm.h

@@ -1,10 +1,11 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
  * LICENSE file in the root directory of this source tree) and the GPLv2 (found
  * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef ZSTD_LDM_H
@@ -15,7 +16,7 @@ extern "C" {
 #endif
 
 #include "zstd_compress_internal.h"   /* ldmParams_t, U32 */
-#include "zstd.h"   /* ZSTD_CCtx, size_t */
+#include "../zstd.h"   /* ZSTD_CCtx, size_t */
 
 /*-*************************************
 *  Long distance matching
@@ -23,6 +24,10 @@ extern "C" {
 
 #define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_LIMIT_DEFAULT
 
+void ZSTD_ldm_fillHashTable(
+            ldmState_t* state, const BYTE* ip,
+            const BYTE* iend, ldmParams_t const* params);
+
 /**
  * ZSTD_ldm_generateSequences():
  *

data/ext/zstdruby/libzstd/compress/zstd_opt.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -249,40 +249,6 @@ static U32 ZSTD_litLengthPrice(U32 const litLength, const optState_t* const optP
     }
 }
 
-/* ZSTD_litLengthContribution() :
- * @return ( cost(litlength) - cost(0) )
- * this value can then be added to rawLiteralsCost()
- * to provide a cost which is directly comparable to a match ending at same position */
-static int ZSTD_litLengthContribution(U32 const litLength, const optState_t* const optPtr, int optLevel)
-{
-    if (optPtr->priceType >= zop_predef) return (int)WEIGHT(litLength, optLevel);
-
-    /* dynamic statistics */
-    {   U32 const llCode = ZSTD_LLcode(litLength);
-        int const contribution = (int)(LL_bits[llCode] * BITCOST_MULTIPLIER)
-                               + (int)WEIGHT(optPtr->litLengthFreq[0], optLevel)   /* note: log2litLengthSum cancel out */
-                               - (int)WEIGHT(optPtr->litLengthFreq[llCode], optLevel);
-#if 1
-        return contribution;
-#else
-        return MAX(0, contribution); /* sometimes better, sometimes not ... */
-#endif
-    }
-}
-
-/* ZSTD_literalsContribution() :
- * creates a fake cost for the literals part of a sequence
- * which can be compared to the ending cost of a match
- * should a new match start at this position */
-static int ZSTD_literalsContribution(const BYTE* const literals, U32 const litLength,
-                                     const optState_t* const optPtr,
-                                     int optLevel)
-{
-    int const contribution = (int)ZSTD_rawLiteralsCost(literals, litLength, optPtr, optLevel)
-                           + ZSTD_litLengthContribution(litLength, optPtr, optLevel);
-    return contribution;
-}
-
 /* ZSTD_getMatchPrice() :
  * Provides the cost of the match part (offset + matchLength) of a sequence
  * Must be combined with ZSTD_fullLiteralsCost() to get the full cost of a sequence.
@@ -603,7 +569,10 @@ U32 ZSTD_insertBtAndGetAllMatches (
             U32 repLen = 0;
             assert(current >= dictLimit);
             if (repOffset-1 /* intentional overflow, discards 0 and -1 */ < current-dictLimit) {  /* equivalent to `current > repIndex >= dictLimit` */
-                if (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch)) {
+                /* We must validate the repcode offset because when we're using a dictionary the
+                 * valid offset range shrinks when the dictionary goes out of bounds.
+                 */
+                if ((repIndex >= windowLow) & (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch))) {
                     repLen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repOffset, iLimit) + minMatch;
                 }
             } else {  /* repIndex < dictLimit || repIndex >= current */
@@ -799,30 +768,6 @@ FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
 /*-*******************************
 *  Optimal parser
 *********************************/
-typedef struct repcodes_s {
-    U32 rep[3];
-} repcodes_t;
-
-static repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
-{
-    repcodes_t newReps;
-    if (offset >= ZSTD_REP_NUM) {  /* full offset */
-        newReps.rep[2] = rep[1];
-        newReps.rep[1] = rep[0];
-        newReps.rep[0] = offset - ZSTD_REP_MOVE;
-    } else {   /* repcode */
-        U32 const repCode = offset + ll0;
-        if (repCode > 0) {  /* note : if repCode==0, no change */
-            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
-            newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
-            newReps.rep[1] = rep[0];
-            newReps.rep[0] = currentOffset;
-        } else {   /* repCode == 0 */
-            memcpy(&newReps, rep, sizeof(newReps));
-        }
-    }
-    return newReps;
-}
 
 
 static U32 ZSTD_totalLen(ZSTD_optimal_t sol)
@@ -839,7 +784,7 @@ listStats(const U32* table, int lastEltID)
     int enb;
     for (enb=0; enb < nbElts; enb++) {
         (void)table;
-        //RAWLOG(2, "%3i:%3i,  ", enb, table[enb]);
+        /* RAWLOG(2, "%3i:%3i,  ", enb, table[enb]); */
         RAWLOG(2, "%4i,", table[enb]);
     }
     RAWLOG(2, " \n");
@@ -894,7 +839,12 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
             { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
             opt[0].mlen = 0;  /* means is_a_literal */
             opt[0].litlen = litlen;
-            opt[0].price = ZSTD_literalsContribution(anchor, litlen, optStatePtr, optLevel);
+            /* We don't need to include the actual price of the literals because
+             * it is static for the duration of the forward pass, and is included
+             * in every price. We include the literal length to avoid negative
+             * prices when we subtract the previous literal length.
+             */
+            opt[0].price = ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
 
             /* large match -> immediate encoding */
             {   U32 const maxML = matches[nbMatches-1].len;
@@ -923,7 +873,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 for (matchNb = 0; matchNb < nbMatches; matchNb++) {
                     U32 const offset = matches[matchNb].off;
                     U32 const end = matches[matchNb].len;
-                    repcodes_t const repHistory = ZSTD_updateRep(rep, offset, ll0);
                     for ( ; pos <= end ; pos++ ) {
                         U32 const matchPrice = ZSTD_getMatchPrice(offset, pos, optStatePtr, optLevel);
                         U32 const sequencePrice = literalsPrice + matchPrice;
@@ -933,8 +882,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                         opt[pos].off = offset;
                         opt[pos].litlen = litlen;
                         opt[pos].price = sequencePrice;
-                        ZSTD_STATIC_ASSERT(sizeof(opt[pos].rep) == sizeof(repHistory));
-                        memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
                 }   }
                 last_pos = pos-1;
             }
@@ -961,7 +908,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                     opt[cur].off = 0;
                     opt[cur].litlen = litlen;
                     opt[cur].price = price;
-                    memcpy(opt[cur].rep, opt[cur-1].rep, sizeof(opt[cur].rep));
                 } else {
                     DEBUGLOG(7, "cPos:%zi==rPos:%u : literal would cost more (%.2f>%.2f) (hist:%u,%u,%u)",
                                 inr-istart, cur, ZSTD_fCost(price), ZSTD_fCost(opt[cur].price),
@@ -969,6 +915,21 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 }
             }
 
+            /* Set the repcodes of the current position. We must do it here
+             * because we rely on the repcodes of the 2nd to last sequence being
+             * correct to set the next chunks repcodes during the backward
+             * traversal.
+             */
+            ZSTD_STATIC_ASSERT(sizeof(opt[cur].rep) == sizeof(repcodes_t));
+            assert(cur >= opt[cur].mlen);
+            if (opt[cur].mlen != 0) {
+                U32 const prev = cur - opt[cur].mlen;
+                repcodes_t newReps = ZSTD_updateRep(opt[prev].rep, opt[cur].off, opt[cur].litlen==0);
+                memcpy(opt[cur].rep, &newReps, sizeof(repcodes_t));
+            } else {
+                memcpy(opt[cur].rep, opt[cur - 1].rep, sizeof(repcodes_t));
+            }
+
             /* last match must start at a minimum distance of 8 from oend */
             if (inr > ilimit) continue;
 
@@ -1009,7 +970,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                 /* set prices using matches found at position == cur */
                 for (matchNb = 0; matchNb < nbMatches; matchNb++) {
                     U32 const offset = matches[matchNb].off;
-                    repcodes_t const repHistory = ZSTD_updateRep(opt[cur].rep, offset, ll0);
                     U32 const lastML = matches[matchNb].len;
                     U32 const startML = (matchNb>0) ? matches[matchNb-1].len+1 : minMatch;
                     U32 mlen;
@@ -1029,8 +989,6 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                             opt[pos].off = offset;
                             opt[pos].litlen = litlen;
                             opt[pos].price = price;
-                            ZSTD_STATIC_ASSERT(sizeof(opt[pos].rep) == sizeof(repHistory));
-                            memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
                         } else {
                             DEBUGLOG(7, "rPos:%u (ml=%2u) => new price is worse (%.2f>=%.2f)",
                                         pos, mlen, ZSTD_fCost(price), ZSTD_fCost(opt[pos].price));
@@ -1046,6 +1004,17 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
 _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
         assert(opt[0].mlen == 0);
 
+        /* Set the next chunk's repcodes based on the repcodes of the beginning
+         * of the last match, and the last sequence. This avoids us having to
+         * update them while traversing the sequences.
+         */
+        if (lastSequence.mlen != 0) {
+            repcodes_t reps = ZSTD_updateRep(opt[cur].rep, lastSequence.off, lastSequence.litlen==0);
+            memcpy(rep, &reps, sizeof(reps));
+        } else {
+            memcpy(rep, opt[cur].rep, sizeof(repcodes_t));
+        }
+
         {   U32 const storeEnd = cur + 1;
             U32 storeStart = storeEnd;
             U32 seqPos = cur;
@@ -1082,20 +1051,6 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
                         continue;   /* will finish */
                     }
 
-                    /* repcodes update : like ZSTD_updateRep(), but update in place */
-                    if (offCode >= ZSTD_REP_NUM) {  /* full offset */
-                        rep[2] = rep[1];
-                        rep[1] = rep[0];
-                        rep[0] = offCode - ZSTD_REP_MOVE;
-                    } else {   /* repcode */
-                        U32 const repCode = offCode + (llen==0);
-                        if (repCode) {  /* note : if repCode==0, no change */
-                            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
-                            if (repCode >= 2) rep[2] = rep[1];
-                            rep[1] = rep[0];
-                            rep[0] = currentOffset;
-                    }   }
-
                     assert(anchor + llen <= iend);
                     ZSTD_updateStats(optStatePtr, llen, anchor, offCode, mlen);
                     ZSTD_storeSeq(seqStore, llen, anchor, iend, offCode, mlen-MINMATCH);
@@ -1104,7 +1059,6 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
             }   }
             ZSTD_setBasePrices(optStatePtr, optLevel);
         }
-
     }   /* while (ip < ilimit) */
 
     /* Return the last literals size */

data/ext/zstdruby/libzstd/compress/zstd_opt.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the

data/ext/zstdruby/libzstd/compress/zstdmt_compress.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -22,9 +22,9 @@
 /* ======   Dependencies   ====== */
 #include <string.h>      /* memcpy, memset */
 #include <limits.h>      /* INT_MAX, UINT_MAX */
-#include "mem.h"         /* MEM_STATIC */
-#include "pool.h"        /* threadpool */
-#include "threading.h"   /* mutex */
+#include "../common/mem.h"         /* MEM_STATIC */
+#include "../common/pool.h"        /* threadpool */
+#include "../common/threading.h"   /* mutex */
 #include "zstd_compress_internal.h"  /* MIN, ERROR, ZSTD_*, ZSTD_highbit32 */
 #include "zstd_ldm.h"
 #include "zstdmt_compress.h"
@@ -461,7 +461,13 @@ typedef struct {
     ZSTD_window_t ldmWindow;  /* A thread-safe copy of ldmState.window */
 } serialState_t;
 
-static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool* seqPool, ZSTD_CCtx_params params, size_t jobSize)
+static int
+ZSTDMT_serialState_reset(serialState_t* serialState,
+                         ZSTDMT_seqPool* seqPool,
+                         ZSTD_CCtx_params params,
+                         size_t jobSize,
+                         const void* dict, size_t const dictSize,
+                         ZSTD_dictContentType_e dictContentType)
 {
     /* Adjust parameters */
     if (params.ldmParams.enableLdm) {
@@ -490,8 +496,7 @@ static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool*
         /* Size the seq pool tables */
         ZSTDMT_setNbSeq(seqPool, ZSTD_ldm_getMaxNbSeq(params.ldmParams, jobSize));
         /* Reset the window */
-        ZSTD_window_clear(&serialState->ldmState.window);
-        serialState->ldmWindow = serialState->ldmState.window;
+        ZSTD_window_init(&serialState->ldmState.window);
         /* Resize tables and output space if necessary. */
         if (serialState->ldmState.hashTable == NULL || serialState->params.ldmParams.hashLog < hashLog) {
             ZSTD_free(serialState->ldmState.hashTable, cMem);
@@ -506,7 +511,24 @@ static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool*
         /* Zero the tables */
         memset(serialState->ldmState.hashTable, 0, hashSize);
         memset(serialState->ldmState.bucketOffsets, 0, bucketSize);
+
+        /* Update window state and fill hash table with dict */
+        serialState->ldmState.loadedDictEnd = 0;
+        if (dictSize > 0) {
+            if (dictContentType == ZSTD_dct_rawContent) {
+                BYTE const* const dictEnd = (const BYTE*)dict + dictSize;
+                ZSTD_window_update(&serialState->ldmState.window, dict, dictSize);
+                ZSTD_ldm_fillHashTable(&serialState->ldmState, (const BYTE*)dict, dictEnd, &params.ldmParams);
+                serialState->ldmState.loadedDictEnd = params.forceWindow ? 0 : (U32)(dictEnd - serialState->ldmState.window.base);
+            } else {
+                /* don't even load anything */
+            }
+        }
+
+        /* Initialize serialState's copy of ldmWindow. */
+        serialState->ldmWindow = serialState->ldmState.window;
     }
+
     serialState->params = params;
     serialState->params.jobSize = (U32)jobSize;
     return 0;
@@ -1054,7 +1076,7 @@ static ZSTD_CCtx_params ZSTDMT_initJobCCtxParams(const ZSTD_CCtx_params* params)
 static size_t ZSTDMT_resize(ZSTDMT_CCtx* mtctx, unsigned nbWorkers)
 {
     if (POOL_resize(mtctx->factory, nbWorkers)) return ERROR(memory_allocation);
-    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbWorkers) );
+    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbWorkers) , "");
     mtctx->bufPool = ZSTDMT_expandBufferPool(mtctx->bufPool, nbWorkers);
     if (mtctx->bufPool == NULL) return ERROR(memory_allocation);
     mtctx->cctxPool = ZSTDMT_expandCCtxPool(mtctx->cctxPool, nbWorkers);
@@ -1076,7 +1098,7 @@ void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_p
     DEBUGLOG(5, "ZSTDMT_updateCParams_whileCompressing (level:%i)",
                 compressionLevel);
     mtctx->params.compressionLevel = compressionLevel;
-    {   ZSTD_compressionParameters cParams = ZSTD_getCParamsFromCCtxParams(cctxParams, 0, 0);
+    {   ZSTD_compressionParameters cParams = ZSTD_getCParamsFromCCtxParams(cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, 0);
         cParams.windowLog = saved_wlog;
         mtctx->params.cParams = cParams;
     }
@@ -1235,7 +1257,8 @@ ZSTDMT_computeNbJobs(const ZSTD_CCtx_params* params, size_t srcSize, unsigned nb
 /* ZSTDMT_compress_advanced_internal() :
  * This is a blocking function : it will only give back control to caller after finishing its compression job.
  */
-static size_t ZSTDMT_compress_advanced_internal(
+static size_t
+ZSTDMT_compress_advanced_internal(
                 ZSTDMT_CCtx* mtctx,
                 void* dst, size_t dstCapacity,
           const void* src, size_t srcSize,
@@ -1267,10 +1290,11 @@ static size_t ZSTDMT_compress_advanced_internal(
 
     assert(avgJobSize >= 256 KB);  /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), required to compress directly into Dst (no additional buffer) */
     ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(avgJobSize) );
-    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, avgJobSize))
+    /* LDM doesn't even try to load the dictionary in single-ingestion mode */
+    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, avgJobSize, NULL, 0, ZSTD_dct_auto))
         return ERROR(memory_allocation);
 
-    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbJobs) );  /* only expands if necessary */
+    FORWARD_IF_ERROR( ZSTDMT_expandJobsTable(mtctx, nbJobs) , "");  /* only expands if necessary */
 
     {   unsigned u;
         for (u=0; u<nbJobs; u++) {
@@ -1403,7 +1427,7 @@ size_t ZSTDMT_initCStream_internal(
 
     /* init */
     if (params.nbWorkers != mtctx->params.nbWorkers)
-        FORWARD_IF_ERROR( ZSTDMT_resize(mtctx, params.nbWorkers) );
+        FORWARD_IF_ERROR( ZSTDMT_resize(mtctx, params.nbWorkers) , "");
 
     if (params.jobSize != 0 && params.jobSize < ZSTDMT_JOBSIZE_MIN) params.jobSize = ZSTDMT_JOBSIZE_MIN;
     if (params.jobSize > (size_t)ZSTDMT_JOBSIZE_MAX) params.jobSize = (size_t)ZSTDMT_JOBSIZE_MAX;
@@ -1500,7 +1524,8 @@ size_t ZSTDMT_initCStream_internal(
     mtctx->allJobsCompleted = 0;
     mtctx->consumed = 0;
     mtctx->produced = 0;
-    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, mtctx->targetSectionSize))
+    if (ZSTDMT_serialState_reset(&mtctx->serial, mtctx->seqPool, params, mtctx->targetSectionSize,
+                                 dict, dictSize, dictContentType))
         return ERROR(memory_allocation);
     return 0;
 }
@@ -1714,9 +1739,11 @@ static size_t ZSTDMT_flushProduced(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, u
             assert(mtctx->doneJobID < mtctx->nextJobID);
             assert(cSize >= mtctx->jobs[wJobID].dstFlushed);
             assert(mtctx->jobs[wJobID].dstBuff.start != NULL);
-            memcpy((char*)output->dst + output->pos,
-                   (const char*)mtctx->jobs[wJobID].dstBuff.start + mtctx->jobs[wJobID].dstFlushed,
-                   toFlush);
+            if (toFlush > 0) {
+                memcpy((char*)output->dst + output->pos,
+                    (const char*)mtctx->jobs[wJobID].dstBuff.start + mtctx->jobs[wJobID].dstFlushed,
+                    toFlush);
+            }
             output->pos += toFlush;
             mtctx->jobs[wJobID].dstFlushed += toFlush;  /* can write : this value is only used by mtctx */
 
@@ -1786,7 +1813,7 @@ static int ZSTDMT_isOverlapped(buffer_t buffer, range_t range)
     BYTE const* const bufferStart = (BYTE const*)buffer.start;
     BYTE const* const bufferEnd = bufferStart + buffer.capacity;
     BYTE const* const rangeStart = (BYTE const*)range.start;
-    BYTE const* const rangeEnd = rangeStart + range.size;
+    BYTE const* const rangeEnd = range.size != 0 ? rangeStart + range.size : rangeStart;
 
     if (rangeStart == NULL || bufferStart == NULL)
         return 0;
@@ -2060,7 +2087,7 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
       || ((endOp == ZSTD_e_end) && (!mtctx->frameEnded)) ) {   /* must finish the frame with a zero-size block */
         size_t const jobSize = mtctx->inBuff.filled;
         assert(mtctx->inBuff.filled <= mtctx->targetSectionSize);
-        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, jobSize, endOp) );
+        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, jobSize, endOp) , "");
     }
 
     /* check for potential compressed data ready to be flushed */
@@ -2074,7 +2101,7 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
 
 size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
-    FORWARD_IF_ERROR( ZSTDMT_compressStream_generic(mtctx, output, input, ZSTD_e_continue) );
+    FORWARD_IF_ERROR( ZSTDMT_compressStream_generic(mtctx, output, input, ZSTD_e_continue) , "");
 
     /* recommended next input size : fill current input buffer */
     return mtctx->targetSectionSize - mtctx->inBuff.filled;   /* note : could be zero when input buffer is fully filled and no more availability to create new job */
@@ -2091,7 +2118,7 @@ static size_t ZSTDMT_flushStream_internal(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* ou
       || ((endFrame==ZSTD_e_end) && !mtctx->frameEnded)) {  /* need a last 0-size block to end frame */
            DEBUGLOG(5, "ZSTDMT_flushStream_internal : create a new job (%u bytes, end:%u)",
                         (U32)srcSize, (U32)endFrame);
-        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, srcSize, endFrame) );
+        FORWARD_IF_ERROR( ZSTDMT_createCompressionJob(mtctx, srcSize, endFrame) , "");
     }
 
     /* check if there is any data available to flush */

data/ext/zstdruby/libzstd/compress/zstdmt_compress.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -40,7 +40,7 @@
 /* ===   Dependencies   === */
 #include <stddef.h>                /* size_t */
 #define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_parameters */
-#include "zstd.h"            /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
+#include "../zstd.h"            /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
 
 
 /* ===   Constants   === */

data/ext/zstdruby/libzstd/decompress/huf_decompress.c

@@ -1,47 +1,27 @@
 /* ******************************************************************
-   huff0 huffman decoder,
-   part of Finite State Entropy library
-   Copyright (C) 2013-present, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ * huff0 huffman decoder,
+ * part of Finite State Entropy library
+ * Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
+ *
+ *  You can contact the author at :
+ *  - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
 ****************************************************************** */
 
 /* **************************************************************
 *  Dependencies
 ****************************************************************/
 #include <string.h>     /* memcpy, memset */
-#include "compiler.h"
-#include "bitstream.h"  /* BIT_* */
-#include "fse.h"        /* to compress headers */
+#include "../common/compiler.h"
+#include "../common/bitstream.h"  /* BIT_* */
+#include "../common/fse.h"        /* to compress headers */
 #define HUF_STATIC_LINKING_ONLY
-#include "huf.h"
-#include "error_private.h"
+#include "../common/huf.h"
+#include "../common/error_private.h"
 
 /* **************************************************************
 *  Macros
@@ -61,9 +41,6 @@
 *  Error Management
 ****************************************************************/
 #define HUF_isError ERR_isError
-#ifndef CHECK_F
-#define CHECK_F(f) { size_t const err_ = (f); if (HUF_isError(err_)) return err_; }
-#endif
 
 
 /* **************************************************************
@@ -181,17 +158,29 @@ size_t HUF_readDTableX1_wksp(HUF_DTable* DTable, const void* src, size_t srcSize
 
     /* fill DTable */
     {   U32 n;
-        for (n=0; n<nbSymbols; n++) {
-            U32 const w = huffWeight[n];
-            U32 const length = (1 << w) >> 1;
-            U32 u;
+        size_t const nEnd = nbSymbols;
+        for (n=0; n<nEnd; n++) {
+            size_t const w = huffWeight[n];
+            size_t const length = (1 << w) >> 1;
+            size_t const uStart = rankVal[w];
+            size_t const uEnd = uStart + length;
+            size_t u;
             HUF_DEltX1 D;
-            D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
-            for (u = rankVal[w]; u < rankVal[w] + length; u++)
-                dt[u] = D;
-            rankVal[w] += length;
-    }   }
-
+            D.byte = (BYTE)n;
+            D.nbBits = (BYTE)(tableLog + 1 - w);
+            rankVal[w] = (U32)uEnd;
+            if (length < 4) {
+                /* Use length in the loop bound so the compiler knows it is short. */
+                for (u = 0; u < length; ++u)
+                    dt[uStart + u] = D;
+            } else {
+                /* Unroll the loop 4 times, we know it is a power of 2. */
+                for (u = uStart; u < uEnd; u += 4) {
+                    dt[u + 0] = D;
+                    dt[u + 1] = D;
+                    dt[u + 2] = D;
+                    dt[u + 3] = D;
+    }   }   }   }
     return iSize;
 }
 
@@ -282,6 +271,7 @@ HUF_decompress4X1_usingDTable_internal_body(
     {   const BYTE* const istart = (const BYTE*) cSrc;
         BYTE* const ostart = (BYTE*) dst;
         BYTE* const oend = ostart + dstSize;
+        BYTE* const olimit = oend - 3;
         const void* const dtPtr = DTable + 1;
         const HUF_DEltX1* const dt = (const HUF_DEltX1*)dtPtr;
 
@@ -306,9 +296,9 @@ HUF_decompress4X1_usingDTable_internal_body(
         BYTE* op2 = opStart2;
         BYTE* op3 = opStart3;
         BYTE* op4 = opStart4;
-        U32 endSignal = BIT_DStream_unfinished;
         DTableDesc const dtd = HUF_getDTableDesc(DTable);
         U32 const dtLog = dtd.tableLog;
+        U32 endSignal = 1;
 
         if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
         CHECK_F( BIT_initDStream(&bitD1, istart1, length1) );
@@ -317,8 +307,7 @@ HUF_decompress4X1_usingDTable_internal_body(
         CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
 
         /* up to 16 symbols per loop (4 symbols per stream) in 64-bit mode */
-        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
-        while ( (endSignal==BIT_DStream_unfinished) && (op4<(oend-3)) ) {
+        for ( ; (endSignal) & (op4 < olimit) ; ) {
             HUF_DECODE_SYMBOLX1_2(op1, &bitD1);
             HUF_DECODE_SYMBOLX1_2(op2, &bitD2);
             HUF_DECODE_SYMBOLX1_2(op3, &bitD3);
@@ -335,10 +324,10 @@ HUF_decompress4X1_usingDTable_internal_body(
             HUF_DECODE_SYMBOLX1_0(op2, &bitD2);
             HUF_DECODE_SYMBOLX1_0(op3, &bitD3);
             HUF_DECODE_SYMBOLX1_0(op4, &bitD4);
-            BIT_reloadDStream(&bitD1);
-            BIT_reloadDStream(&bitD2);
-            BIT_reloadDStream(&bitD3);
-            BIT_reloadDStream(&bitD4);
+            endSignal &= BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished;
+            endSignal &= BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished;
+            endSignal &= BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished;
+            endSignal &= BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished;
         }
 
         /* check corruption */
@@ -757,7 +746,6 @@ HUF_decompress1X2_usingDTable_internal_body(
     return dstSize;
 }
 
-
 FORCE_INLINE_TEMPLATE size_t
 HUF_decompress4X2_usingDTable_internal_body(
           void* dst,  size_t dstSize,
@@ -769,6 +757,7 @@ HUF_decompress4X2_usingDTable_internal_body(
     {   const BYTE* const istart = (const BYTE*) cSrc;
         BYTE* const ostart = (BYTE*) dst;
         BYTE* const oend = ostart + dstSize;
+        BYTE* const olimit = oend - (sizeof(size_t)-1);
         const void* const dtPtr = DTable+1;
         const HUF_DEltX2* const dt = (const HUF_DEltX2*)dtPtr;
 
@@ -793,7 +782,7 @@ HUF_decompress4X2_usingDTable_internal_body(
         BYTE* op2 = opStart2;
         BYTE* op3 = opStart3;
         BYTE* op4 = opStart4;
-        U32 endSignal;
+        U32 endSignal = 1;
         DTableDesc const dtd = HUF_getDTableDesc(DTable);
         U32 const dtLog = dtd.tableLog;
 
@@ -804,8 +793,29 @@ HUF_decompress4X2_usingDTable_internal_body(
         CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
 
         /* 16-32 symbols per loop (4-8 symbols per stream) */
-        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
-        for ( ; (endSignal==BIT_DStream_unfinished) & (op4<(oend-(sizeof(bitD4.bitContainer)-1))) ; ) {
+        for ( ; (endSignal) & (op4 < olimit); ) {
+#if defined(__clang__) && (defined(__x86_64__) || defined(__i386__))
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
+            endSignal &= BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished;
+            endSignal &= BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished;
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
+            endSignal &= BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished;
+            endSignal &= BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished;
+#else
             HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
             HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
             HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
@@ -822,8 +832,12 @@ HUF_decompress4X2_usingDTable_internal_body(
             HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
             HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
             HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
-
-            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+            endSignal = (U32)LIKELY(
+                        (BIT_reloadDStreamFast(&bitD1) == BIT_DStream_unfinished)
+                      & (BIT_reloadDStreamFast(&bitD2) == BIT_DStream_unfinished)
+                      & (BIT_reloadDStreamFast(&bitD3) == BIT_DStream_unfinished)
+                      & (BIT_reloadDStreamFast(&bitD4) == BIT_DStream_unfinished));
+#endif
         }
 
         /* check corruption */