From: Stella Lau <laus@fb.com>
Date: Tue, 18 Jul 2017 01:13:09 +0000 (-0700)
Subject: Experiment with using a lag when hashing
X-Git-Tag: v1.3.1^2~12^2~7^2~22
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=fc41a8796493063c26103ca7ed32561dd2e649d0;p=thirdparty%2Fzstd.git

Experiment with using a lag when hashing
---

diff --git a/contrib/long_distance_matching/Makefile b/contrib/long_distance_matching/Makefile
index 47085022d..df4390157 100644
--- a/contrib/long_distance_matching/Makefile
+++ b/contrib/long_distance_matching/Makefile
@@ -25,7 +25,7 @@ LDFLAGS += -lzstd
 
 default: all
 
-all: main-basic main-circular-buffer
+all: main-basic main-circular-buffer main-lag
 
 main-basic : basic_table.c ldm.c main-ldm.c
 	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
@@ -33,9 +33,11 @@ main-basic : basic_table.c ldm.c main-ldm.c
 main-circular-buffer: circular_buffer_table.c ldm.c main-ldm.c
 	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
 
+main-lag: lag_table.c ldm.c main-ldm.c
+	$(CC) $(CPPFLAGS) $(CFLAGS) $^ $(LDFLAGS) -o $@
 
 clean:
 	@rm -f core *.o tmp* result* *.ldm *.ldm.dec \
-	main-basic main-circular-buffer
+	main-basic main-circular-buffer main-lag
 	@echo Cleaning completed
 
diff --git a/contrib/long_distance_matching/basic_table.c b/contrib/long_distance_matching/basic_table.c
index 859bf0618..893a4caf9 100644
--- a/contrib/long_distance_matching/basic_table.c
+++ b/contrib/long_distance_matching/basic_table.c
@@ -27,7 +27,6 @@ LDM_hashEntry *getBucket(const LDM_hashTable *table, const hash_t hash) {
   return table->entries + hash;
 }
 
-
 LDM_hashEntry *HASH_getEntryFromHash(
     const LDM_hashTable *table, const hash_t hash, const U32 checksum) {
   (void)checksum;
@@ -43,13 +42,10 @@ LDM_hashEntry *HASH_getValidEntry(const LDM_hashTable *table,
   (void)checksum;
   if ((*isValid)(pIn, entry->offset + table->offsetBase)) {
     return entry;
-  } else {
-    return NULL;
   }
+  return NULL;
 }
 
-
-
 void HASH_insert(LDM_hashTable *table,
                  const hash_t hash, const LDM_hashEntry entry) {
   *getBucket(table, hash) = entry;
diff --git a/contrib/long_distance_matching/circular_buffer_table.c b/contrib/long_distance_matching/circular_buffer_table.c
index f45f945ce..b578d2bf1 100644
--- a/contrib/long_distance_matching/circular_buffer_table.c
+++ b/contrib/long_distance_matching/circular_buffer_table.c
@@ -9,7 +9,7 @@
 //      refactor code to scale the number of elements appropriately.
 
 // Number of elements per hash bucket.
-#define HASH_BUCKET_SIZE_LOG 1  // MAX is 4 for now
+#define HASH_BUCKET_SIZE_LOG 0 // MAX is 4 for now
 #define HASH_BUCKET_SIZE (1 << (HASH_BUCKET_SIZE_LOG))
 
 struct LDM_hashTable {
@@ -19,6 +19,7 @@ struct LDM_hashTable {
   // Position corresponding to offset=0 in LDM_hashEntry.
   const BYTE *offsetBase;
   BYTE *bucketOffsets;     // Pointer to current insert position.
+
                            // Last insert was at bucketOffsets - 1?
 };
 
@@ -35,15 +36,6 @@ static LDM_hashEntry *getBucket(const LDM_hashTable *table, const hash_t hash) {
   return table->entries + (hash << HASH_BUCKET_SIZE_LOG);
 }
 
-/*
-static LDM_hashEntry *getLastInsertFromHash(const LDM_hashTable *table,
-                                            const hash_t hash) {
-  LDM_hashEntry *bucket = getBucket(table, hash);
-  BYTE offset = (table->bucketOffsets[hash] - 1) & (HASH_BUCKET_SIZE - 1);
-  return bucket + offset;
-}
-*/
-
 LDM_hashEntry *HASH_getValidEntry(const LDM_hashTable *table,
                                   const hash_t hash,
                                   const U32 checksum,
@@ -53,7 +45,12 @@ LDM_hashEntry *HASH_getValidEntry(const LDM_hashTable *table,
   LDM_hashEntry *cur = bucket;
   // TODO: in order of recency?
   for (; cur < bucket + HASH_BUCKET_SIZE; ++cur) {
-    // CHeck checksum for faster check.
+    /*
+    if (cur->checksum == 0 && cur->offset == 0) {
+      return NULL;
+    }
+    */
+    // Check checksum for faster check.
     if (cur->checksum == checksum &&
         (*isValid)(pIn, cur->offset + table->offsetBase)) {
       return cur;
@@ -62,7 +59,6 @@ LDM_hashEntry *HASH_getValidEntry(const LDM_hashTable *table,
   return NULL;
 }
 
-
 LDM_hashEntry *HASH_getEntryFromHash(const LDM_hashTable *table,
                                      const hash_t hash,
                                      const U32 checksum) {
diff --git a/contrib/long_distance_matching/ldm.c b/contrib/long_distance_matching/ldm.c
index bf54842f6..dedbf79a9 100644
--- a/contrib/long_distance_matching/ldm.c
+++ b/contrib/long_distance_matching/ldm.c
@@ -5,7 +5,7 @@
 #include <string.h>
 
 // Insert every (HASH_ONLY_EVERY + 1) into the hash table.
-#define HASH_ONLY_EVERY 31
+#define HASH_ONLY_EVERY 15
 
 #define LDM_HASHLOG (LDM_MEMORY_USAGE-2)
 #define LDM_HASHTABLESIZE (1 << (LDM_MEMORY_USAGE))
@@ -18,6 +18,10 @@
 
 #define COMPUTE_STATS
 #define CHECKSUM_CHAR_OFFSET 10
+
+#define LAG 0
+
+//#define HASH_CHECK
 //#define RUN_CHECKS
 //#define LDM_DEBUG
 
@@ -79,6 +83,10 @@ struct LDM_CCtx {
 
   unsigned step;                      // ip step, should be 1.
 
+  const BYTE *lagIp;
+  hash_t lagHash;
+  U32 lagSum;
+
   // DEBUG
   const BYTE *DEBUG_setNextHash;
 };
@@ -253,6 +261,17 @@ static void setNextHash(LDM_CCtx *cctx) {
   cctx->nextPosHashed = cctx->nextIp;
   cctx->nextHash = checksumToHash(cctx->nextSum);
 
+#if LAG
+  if (cctx->ip - cctx->ibase > LAG) {
+//    printf("LAG %zu\n", cctx->ip - cctx->lagIp);
+    cctx->lagSum = updateChecksum(
+      cctx->lagSum, LDM_HASH_LENGTH,
+      cctx->lagIp[0], cctx->lagIp[LDM_HASH_LENGTH]);
+    cctx->lagIp++;
+    cctx->lagHash = checksumToHash(cctx->lagSum);
+  }
+#endif
+
 #ifdef RUN_CHECKS
   check = getChecksum(cctx->nextIp, LDM_HASH_LENGTH);
 
@@ -270,18 +289,6 @@ static void setNextHash(LDM_CCtx *cctx) {
 
 static void putHashOfCurrentPositionFromHash(
     LDM_CCtx *cctx, hash_t hash, U32 sum) {
-  /*
-#ifdef COMPUTE_STATS
-  if (cctx->stats.numHashInserts < HASH_getSize(cctx->hashTable)) {
-    U32 offset = HASH_getEntryFromHash(cctx->hashTable, hash)->offset;
-    cctx->stats.numHashInserts++;
-    if (offset != 0 && !LDM_isValidMatch(cctx->ip, offset + cctx->ibase)) {
-      cctx->stats.numCollisions++;
-    }
-  }
-#endif
-*/
-
   // Hash only every HASH_ONLY_EVERY times, based on cctx->ip.
   // Note: this works only when cctx->step is 1.
   if (((cctx->ip - cctx->ibase) & HASH_ONLY_EVERY) == HASH_ONLY_EVERY) {
@@ -289,8 +296,19 @@ static void putHashOfCurrentPositionFromHash(
     const LDM_hashEntry entry = { cctx->ip - cctx->ibase ,
                                   MEM_read32(cctx->ip) };
                                   */
+#if LAG
+    // TODO: off by 1, but whatever
+    if (cctx->lagIp - cctx->ibase > 0) {
+      const LDM_hashEntry entry = { cctx->lagIp - cctx->ibase, cctx->lagSum };
+      HASH_insert(cctx->hashTable, cctx->lagHash, entry);
+    } else {
+      const LDM_hashEntry entry = { cctx->ip - cctx->ibase, sum };
+      HASH_insert(cctx->hashTable, hash, entry);
+    }
+#else
     const LDM_hashEntry entry = { cctx->ip - cctx->ibase, sum };
     HASH_insert(cctx->hashTable, hash, entry);
+#endif
   }
 
   cctx->lastPosHashed = cctx->ip;
@@ -331,15 +349,6 @@ static void LDM_putHashOfCurrentPosition(LDM_CCtx *cctx) {
   putHashOfCurrentPositionFromHash(cctx, hash, sum);
 }
 
-/**
- * Returns the position of the entry at hashTable[hash].
- */
-/*
-static const BYTE *getPositionOnHash(const LDM_CCtx *cctx, const hash_t hash) {
-  return HASH_getEntryFromHash(cctx->hashTable, hash)->offset + cctx->ibase;
-}
-*/
-
 U32 LDM_countMatchLength(const BYTE *pIn, const BYTE *pMatch,
                          const BYTE *pInLimit) {
   const BYTE * const pStart = pIn;
@@ -431,12 +440,20 @@ static int LDM_findBestMatch(LDM_CCtx *cctx, const BYTE **match) {
     if (cctx->ip > cctx->imatchLimit) {
       return 1;
     }
-
+#ifdef HASH_CHECK
+    entry = HASH_getEntryFromHash(cctx->hashTable, h, sum);
+#else
     entry = HASH_getValidEntry(cctx->hashTable, h, sum, cctx->ip,
                                &LDM_isValidMatch);
+#endif
 
     if (entry != NULL) {
       *match = entry->offset + cctx->ibase;
+#ifdef HASH_CHECK
+      if (!LDM_isValidMatch(cctx->ip, *match)) {
+        entry = NULL;
+      }
+#endif
     }
     putHashOfCurrentPositionFromHash(cctx, h, sum);
   }
@@ -508,6 +525,12 @@ size_t LDM_compress(const void *src, size_t srcSize,
   /* Hash the first position and put it into the hash table. */
   LDM_putHashOfCurrentPosition(&cctx);
 
+#if LAG
+  cctx.lagIp = cctx.ip;
+  cctx.lagHash = cctx.lastHash;
+  cctx.lagSum = cctx.lastSum;
+#endif
+
   /**
    * Find a match.
    * If no more matches can be found (i.e. the length of the remaining input
@@ -575,7 +598,7 @@ size_t LDM_compress(const void *src, size_t srcSize,
 
   /* Encode the last literals (no more matches). */
   {
-    const size_t lastRun = cctx.iend - cctx.anchor;
+    const U32 lastRun = cctx.iend - cctx.anchor;
     BYTE *pToken = cctx.op++;
     LDM_encodeLiteralLengthAndLiterals(&cctx, pToken, lastRun);
   }
diff --git a/contrib/long_distance_matching/ldm.h b/contrib/long_distance_matching/ldm.h
index 6d97bd560..6d7c4af27 100644
--- a/contrib/long_distance_matching/ldm.h
+++ b/contrib/long_distance_matching/ldm.h
@@ -10,8 +10,8 @@
 #define LDM_HEADER_SIZE ((LDM_COMPRESS_SIZE)+(LDM_DECOMPRESS_SIZE))
 #define LDM_OFFSET_SIZE 4
 
-// Defines the size of the hash table.
-#define LDM_MEMORY_USAGE 20
+// Defines the size of the hash table (currently the number of elements).
+#define LDM_MEMORY_USAGE 12
 
 #define LDM_WINDOW_SIZE_LOG 30
 #define LDM_WINDOW_SIZE (1 << (LDM_WINDOW_SIZE_LOG))
diff --git a/contrib/long_distance_matching/ldm_hashtable.h b/contrib/long_distance_matching/ldm_hashtable.h
index 88d19ae20..83a9ed27a 100644
--- a/contrib/long_distance_matching/ldm_hashtable.h
+++ b/contrib/long_distance_matching/ldm_hashtable.h
@@ -3,34 +3,54 @@
 
 #include "mem.h"
 
+// TODO: clean up comments
+
 typedef U32 hash_t;
 
 typedef struct LDM_hashEntry {
-  U32 offset;
+  U32 offset;   // TODO: Replace with pointer?
   U32 checksum;
 } LDM_hashEntry;
 
 typedef struct LDM_hashTable LDM_hashTable;
 
-// TODO: rename functions
-// TODO: comments
-
+/**
+ * Create a hash table with size hash buckets.
+ * LDM_hashEntry.offset is added to offsetBase to calculate pMatch in
+ * HASH_getValidEntry.
+ */
 LDM_hashTable *HASH_createTable(U32 size, const BYTE *offsetBase);
 
-//TODO: unneeded?
+/**
+ * Returns an LDM_hashEntry from the table that matches the checksum.
+ * Returns NULL if one does not exist.
+ */
 LDM_hashEntry *HASH_getEntryFromHash(const LDM_hashTable *table,
                                      const hash_t hash,
                                      const U32 checksum);
 
+/**
+ *  Gets a valid entry that matches the checksum. A valid entry is defined by
+ *  *isValid.
+ *
+ *  The function finds an entry matching the checksum, computes pMatch as
+ *  offset + table.offsetBase, and calls isValid.
+ */
 LDM_hashEntry *HASH_getValidEntry(const LDM_hashTable *table,
                                   const hash_t hash,
                                   const U32 checksum,
                                   const BYTE *pIn,
                                   int (*isValid)(const BYTE *pIn, const BYTE *pMatch));
 
+/**
+ * Insert an LDM_hashEntry into the bucket corresponding to hash.
+ */
 void HASH_insert(LDM_hashTable *table, const hash_t hash,
-                        const LDM_hashEntry entry);
+                 const LDM_hashEntry entry);
 
+/**
+ * Return the number of distinct hash buckets.
+ */
 U32 HASH_getSize(const LDM_hashTable *table);
 
 void HASH_destroyTable(LDM_hashTable *table);
diff --git a/contrib/long_distance_matching/main-ldm.c b/contrib/long_distance_matching/main-ldm.c
index ea6375ba7..a379d3a6d 100644
--- a/contrib/long_distance_matching/main-ldm.c
+++ b/contrib/long_distance_matching/main-ldm.c
@@ -163,7 +163,6 @@ static int decompress(const char *fname, const char *oname) {
   outSize = LDM_decompress(
       src + LDM_HEADER_SIZE, statbuf.st_size - LDM_HEADER_SIZE,
       dst, decompressedSize);
-
   printf("Ret size out: %zu\n", outSize);
   ftruncate(fdout, outSize);