Offsets >= 32MB in 32-bits mode

diff --git a/lib/common/bitstream.h b/lib/common/bitstream.h
index 3a45244f84cd6cb8801699ad68a67330f3aa571a..d3873002ebd1a3944f8d47437bca8c0aebcd1b01 100644 (file)
--- a/lib/common/bitstream.h
+++ b/lib/common/bitstream.h
@@ -60,6 +60,9 @@ extern "C" {
 #  include <immintrin.h>   /* support for bextr (experimental) */
 #endif
 
+#define STREAM_ACCUMULATOR_MIN_32  25
+#define STREAM_ACCUMULATOR_MIN_64  57
+#define STREAM_ACCUMULATOR_MIN    ((U32)(MEM_32bits() ? STREAM_ACCUMULATOR_MIN_32 : STREAM_ACCUMULATOR_MIN_64))
 
 /*-******************************************
 *  bitStream encoding API (write forward)

diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
index ec758db00c37450781fecb79dd4af2efa43390ed..89f6575dcc487973028147503529c82c0d8a73c4 100644 (file)
--- a/lib/compress/zstd_compress.c
+++ b/lib/compress/zstd_compress.c
@@ -579,9 +579,9 @@ void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
 }
 
 
-size_t ZSTD_compressSequences(ZSTD_CCtx* zc,
+FORCE_INLINE size_t ZSTD_compressSequences_generic (ZSTD_CCtx* zc,
                               void* dst, size_t dstCapacity,
-                              size_t srcSize)
+                              size_t srcSize, int const longOffsets)
 {
     const seqStore_t* seqStorePtr = &(zc->seqStore);
     U32 count[MaxSeq+1];
@@ -716,7 +716,18 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* zc,
         if (MEM_32bits()) BIT_flushBits(&blockStream);
         BIT_addBits(&blockStream, sequences[nbSeq-1].matchLength, ML_bits[mlCodeTable[nbSeq-1]]);
         if (MEM_32bits()) BIT_flushBits(&blockStream);
-        BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
+        if (longOffsets) {
+            U32 const ofBits = ofCodeTable[nbSeq-1];
+            int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+            if (extraBits) {
+                BIT_addBits(&blockStream, sequences[nbSeq-1].offset, extraBits);
+                BIT_flushBits(&blockStream);
+            }
+            BIT_addBits(&blockStream, sequences[nbSeq-1].offset >> extraBits,
+                        ofBits - extraBits);
+        } else {
+            BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
+        }
         BIT_flushBits(&blockStream);
 
         {   size_t n;
@@ -738,7 +749,17 @@ size_t ZSTD_compressSequences(ZSTD_CCtx* zc,
                 if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
                 BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
                 if (MEM_32bits()) BIT_flushBits(&blockStream);                  /* (7)*/
-                BIT_addBits(&blockStream, sequences[n].offset, ofBits);         /* 31 */
+                if (longOffsets) {
+                    int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+                    if (extraBits) {
+                        BIT_addBits(&blockStream, sequences[n].offset, extraBits);
+                        BIT_flushBits(&blockStream);                            /* (7)*/
+                    }
+                    BIT_addBits(&blockStream, sequences[n].offset >> extraBits,
+                                ofBits - extraBits);                            /* 31 */
+                } else {
+                    BIT_addBits(&blockStream, sequences[n].offset, ofBits);     /* 31 */
+                }
                 BIT_flushBits(&blockStream);                                    /* (7)*/
         }   }
 
@@ -763,6 +784,16 @@ _check_compressibility:
     return op - ostart;
 }
 
+FORCE_INLINE size_t ZSTD_compressSequences (ZSTD_CCtx* zc,
+                              void* dst, size_t dstCapacity,
+                              size_t srcSize)
+{
+    if (zc->params.cParams.windowLog > STREAM_ACCUMULATOR_MIN) {
+        return ZSTD_compressSequences_generic(zc, dst, dstCapacity, srcSize, 1);
+    } else {
+        return ZSTD_compressSequences_generic(zc, dst, dstCapacity, srcSize, 0);
+    }
+}
 
 #if 0 /* for debug */
 #  define STORESEQ_DEBUG

diff --git a/lib/decompress/zstd_decompress.c b/lib/decompress/zstd_decompress.c
index 2646c802883ecb1390e81688c756b81a0e8467d8..e39bf42bf859bfb16c1699d5b96c8c1337818ec8 100644 (file)
--- a/lib/decompress/zstd_decompress.c
+++ b/lib/decompress/zstd_decompress.c
@@ -1144,7 +1144,7 @@ static size_t ZSTD_decompressSequences(
 }
 
 
-static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState)
+FORCE_INLINE seq_t ZSTD_decodeSequenceLong_generic(seqState_t* seqState, int const longOffsets)
 {
     seq_t seq;
 
@@ -1179,8 +1179,15 @@ static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState)
         if (!ofCode)
             offset = 0;
         else {
-            offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
-            if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            if (longOffsets) {
+                int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN);
+                offset = OF_base[ofCode] + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
+                if (MEM_32bits() || extraBits) BIT_reloadDStream(&seqState->DStream);
+                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
+            } else {
+                offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            }
         }
 
         if (ofCode <= 1) {
@@ -1224,6 +1231,14 @@ static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState)
     return seq;
 }
 
+static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState, unsigned const windowSize) {
+    if (ZSTD_highbit32(windowSize) > STREAM_ACCUMULATOR_MIN) {
+        return ZSTD_decodeSequenceLong_generic(seqState, 1);
+    } else {
+        return ZSTD_decodeSequenceLong_generic(seqState, 0);
+    }
+}
+
 FORCE_INLINE
 size_t ZSTD_execSequenceLong(BYTE* op,
                                 BYTE* const oend, seq_t sequence,
@@ -1321,6 +1336,7 @@ static size_t ZSTD_decompressSequencesLong(
     const BYTE* const base = (const BYTE*) (dctx->base);
     const BYTE* const vBase = (const BYTE*) (dctx->vBase);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+    unsigned const windowSize = dctx->fParams.windowSize;
     int nbSeq;
 
     /* Build Decoding Tables */
@@ -1350,13 +1366,13 @@ static size_t ZSTD_decompressSequencesLong(
 
         /* prepare in advance */
         for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && seqNb<seqAdvance; seqNb++) {
-            sequences[seqNb] = ZSTD_decodeSequenceLong(&seqState);
+            sequences[seqNb] = ZSTD_decodeSequenceLong(&seqState, windowSize);
         }
         if (seqNb<seqAdvance) return ERROR(corruption_detected);
 
         /* decode and decompress */
         for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && seqNb<nbSeq ; seqNb++) {
-            seq_t const sequence = ZSTD_decodeSequenceLong(&seqState);
+            seq_t const sequence = ZSTD_decodeSequenceLong(&seqState, windowSize);
             size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STOSEQ_MASK], &litPtr, litEnd, base, vBase, dictEnd);
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
             ZSTD_PREFETCH(sequence.match);

diff --git a/lib/zstd.h b/lib/zstd.h
index 7cca369092d6f6b792767968654e4c34bc600591..f462d0997bbc9b91df86ac9f73f052a683a9fe99 100644 (file)
--- a/lib/zstd.h
+++ b/lib/zstd.h
@@ -343,7 +343,7 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 #define ZSTD_CONTENTSIZE_UNKNOWN (0ULL - 1)
 #define ZSTD_CONTENTSIZE_ERROR   (0ULL - 2)
 
-#define ZSTD_WINDOWLOG_MAX_32  25
+#define ZSTD_WINDOWLOG_MAX_32  27
 #define ZSTD_WINDOWLOG_MAX_64  27
 #define ZSTD_WINDOWLOG_MAX    ((U32)(MEM_32bits() ? ZSTD_WINDOWLOG_MAX_32 : ZSTD_WINDOWLOG_MAX_64))
 #define ZSTD_WINDOWLOG_MIN     10