Merge pull request #580 from facebook/speedStream

Improve streaming decompression speed

lib/decompress/zstd_decompress.c

@@ -1469,8 +1469,7 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
         if (ZSTD_isError(headerSize)) return headerSize;
 
         /* Frame Header */
-        {
+        {   size_t const ret = ZSTD_getFrameParams(&fParams, ip, remainingSize);
-            size_t const ret = ZSTD_getFrameParams(&fParams, ip, remainingSize);
             if (ZSTD_isError(ret)) return ret;
             if (ret > 0) return ERROR(srcSize_wrong);
         }
@@ -1503,7 +1502,7 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
 }
 
 /*! ZSTD_decompressFrame() :
-*   `dctx` must be properly initialized */
+*   @dctx must be properly initialized */
 static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
                                  void* dst, size_t dstCapacity,
                                  const void** srcPtr, size_t *srcSizePtr)
@@ -1570,7 +1569,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         remainingSize -= 4;
     }
 
-    // Allow caller to get size read
+    /* Allow caller to get size read */
     *srcPtr = ip;
     *srcSizePtr = remainingSize;
     return op-ostart;
@@ -2100,7 +2099,7 @@ size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
 {
     /* pass content and size in case legacy frames are encountered */
     return ZSTD_decompressMultiFrame(dctx, dst, dstCapacity, src, srcSize,
-                                     ddict->dictContent, ddict->dictSize,
+                                     NULL, 0,
                                      ddict);
 }
 
@@ -2301,6 +2300,21 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                     break;
             }   }
 
+            /* check for single-pass mode opportunity */
+            if (zds->fParams.frameContentSize && zds->fParams.windowSize /* skippable frame if == 0 */
+                && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
+                size_t const cSize = ZSTD_findFrameCompressedSize(istart, iend-istart);
+                if (cSize <= (size_t)(iend-istart)) {
+                    size_t const decompressedSize = ZSTD_decompress_usingDDict(zds->dctx, op, oend-op, istart, cSize, zds->ddict);
+                    if (ZSTD_isError(decompressedSize)) return decompressedSize;
+                    ip = istart + cSize;
+                    op += decompressedSize;
+                    zds->dctx->expected = 0;
+                    zds->stage = zdss_init;
+                    someMoreWork = 0;
+                    break;
+            }   }
+
             /* Consume header */
             ZSTD_refDDict(zds->dctx, zds->ddict);
             {   size_t const h1Size = ZSTD_nextSrcSizeToDecompress(zds->dctx);  /* == ZSTD_frameHeaderSize_prefix */

tests/fuzzer.c

@@ -108,6 +108,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     void* const CNBuffer = malloc(CNBuffSize);
     void* const compressedBuffer = malloc(ZSTD_compressBound(CNBuffSize));
     void* const decodedBuffer = malloc(CNBuffSize);
+    ZSTD_DCtx* dctx = ZSTD_createDCtx();
     int testResult = 0;
     U32 testNb=0;
     size_t cSize;
@@ -155,6 +156,16 @@ static int basicUnitTests(U32 seed, double compressibility)
     }   }
     DISPLAYLEVEL(4, "OK \n");
 
+    DISPLAYLEVEL(4, "test%3i : decompress with null dict : ", testNb++);
+    { size_t const r = ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, NULL, 0);
+      if (r != CNBuffSize) goto _output_error; }
+    DISPLAYLEVEL(4, "OK \n");
+
+    DISPLAYLEVEL(4, "test%3i : decompress with null DDict : ", testNb++);
+    { size_t const r = ZSTD_decompress_usingDDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, NULL);
+      if (r != CNBuffSize) goto _output_error; }
+    DISPLAYLEVEL(4, "OK \n");
+
     DISPLAYLEVEL(4, "test%3i : decompress with 1 missing byte : ", testNb++);
     { size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize-1);
       if (!ZSTD_isError(r)) goto _output_error;
@@ -210,7 +221,6 @@ static int basicUnitTests(U32 seed, double compressibility)
     /* Dictionary and CCtx Duplication tests */
     {   ZSTD_CCtx* const ctxOrig = ZSTD_createCCtx();
         ZSTD_CCtx* const ctxDuplicated = ZSTD_createCCtx();
-        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         static const size_t dictSize = 551;
 
         DISPLAYLEVEL(4, "test%3i : copy context too soon : ", testNb++);
@@ -283,12 +293,10 @@ static int basicUnitTests(U32 seed, double compressibility)
 
         ZSTD_freeCCtx(ctxOrig);
         ZSTD_freeCCtx(ctxDuplicated);
-        ZSTD_freeDCtx(dctx);
     }
 
     /* Dictionary and dictBuilder tests */
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
-        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         size_t dictSize = 16 KB;
         void* dictBuffer = malloc(dictSize);
         size_t const totalSampleSize = 1 MB;
@@ -370,14 +378,12 @@ static int basicUnitTests(U32 seed, double compressibility)
         DISPLAYLEVEL(4, "OK \n");
 
         ZSTD_freeCCtx(cctx);
-        ZSTD_freeDCtx(dctx);
         free(dictBuffer);
         free(samplesSizes);
     }
 
     /* COVER dictionary builder tests */
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
-        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         size_t dictSize = 16 KB;
         size_t optDictSize = dictSize;
         void* dictBuffer = malloc(dictSize);
@@ -414,7 +420,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         memset(&params, 0, sizeof(params));
         params.steps = 4;
         optDictSize = COVER_optimizeTrainFromBuffer(dictBuffer, optDictSize,
-                                                    CNBuffer, samplesSizes, nbSamples,
+                                                    CNBuffer, samplesSizes, nbSamples / 4,
                                                     &params);
         if (ZDICT_isError(optDictSize)) goto _output_error;
         DISPLAYLEVEL(4, "OK, created dictionary of size %u \n", (U32)optDictSize);
@@ -425,7 +431,6 @@ static int basicUnitTests(U32 seed, double compressibility)
         DISPLAYLEVEL(4, "OK : %u \n", dictID);
 
         ZSTD_freeCCtx(cctx);
-        ZSTD_freeDCtx(dctx);
         free(dictBuffer);
         free(samplesSizes);
     }
@@ -445,7 +450,6 @@ static int basicUnitTests(U32 seed, double compressibility)
 
     /* block API tests */
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
-        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         static const size_t dictSize = 65 KB;
         static const size_t blockSize = 100 KB;   /* won't cause pb with small dict size */
         size_t cSize2;

tests/zstreamtest.c

@@ -218,7 +218,7 @@ static int basicUnitTests(U32 seed, double compressibility, ZSTD_customMem custo
     outBuff.pos = 0;
     { size_t const r = ZSTD_decompressStream(zd, &outBuff, &inBuff);
       if (r != 0) goto _output_error; }
-    if (outBuff.pos != 0) goto _output_error;   /* skippable frame len is 0 */
+    if (outBuff.pos != 0) goto _output_error;   /* skippable frame output len is 0 */
     DISPLAYLEVEL(3, "OK \n");
 
     /* Basic decompression test */