X-Git-Url: https://mattmccutchen.net/rsync/rsync.git/blobdiff_plain/37a56445764da046aada79adea49715b1f1b3ed0..fdd10da6f7995296f6bb5ee3dd7231d6dbc20b5e:/match.c diff --git a/match.c b/match.c index 89e5c529..72947d0a 100644 --- a/match.c +++ b/match.c @@ -1,88 +1,69 @@ /* - Copyright (C) Andrew Tridgell 1996 - Copyright (C) Paul Mackerras 1996 - - This program is free software; you can redistribute it and/or modify - it under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2 of the License, or - (at your option) any later version. - - This program is distributed in the hope that it will be useful, - but WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - GNU General Public License for more details. - - You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. -*/ + * Block matching used by the file-transfer code. + * + * Copyright (C) 1996 Andrew Tridgell + * Copyright (C) 1996 Paul Mackerras + * Copyright (C) 2003, 2004, 2005, 2006 Wayne Davison + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License along + * with this program; if not, write to the Free Software Foundation, Inc., + * 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. + */ #include "rsync.h" extern int verbose; -extern int am_server; extern int do_progress; extern int checksum_seed; +extern int append_mode; int updating_basis_file; -typedef unsigned short tag; - -#define TABLESIZE (1<<16) -#define NULL_TAG (-1) - static int false_alarms; -static int tag_hits; +static int hash_hits; static int matches; static int64 data_transfer; static int total_false_alarms; -static int total_tag_hits; +static int total_hash_hits; static int total_matches; extern struct stats stats; -struct target { - tag t; - int32 i; -}; - -static struct target *targets; +#define TABLESIZE (1<<16) -static int32 *tag_table; - -#define gettag2(s1,s2) (((s1) + (s2)) & 0xFFFF) -#define gettag(sum) gettag2((sum)&0xFFFF,(sum)>>16) - -static int compare_targets(struct target *t1,struct target *t2) -{ - return (int)t1->t - (int)t2->t; -} +static int32 *hash_table; +#define SUM2HASH2(s1,s2) (((s1) + (s2)) & 0xFFFF) +#define SUM2HASH(sum) SUM2HASH2((sum)&0xFFFF,(sum)>>16) static void build_hash_table(struct sum_struct *s) { int32 i; - if (!tag_table) - tag_table = new_array(int32, TABLESIZE); + if (!hash_table) { + hash_table = new_array(int32, TABLESIZE); + if (!hash_table) + out_of_memory("build_hash_table"); + } - targets = new_array(struct target, s->count); - if (!tag_table || !targets) - out_of_memory("build_hash_table"); + memset(hash_table, 0xFF, TABLESIZE * sizeof hash_table[0]); for (i = 0; i < s->count; i++) { - targets[i].i = i; - targets[i].t = gettag(s->sums[i].sum1); + uint32 t = SUM2HASH(s->sums[i].sum1); + s->sums[i].chain = hash_table[t]; + hash_table[t] = i; } - - qsort(targets,s->count,sizeof(targets[0]),(int (*)())compare_targets); - - for (i = 0; i < TABLESIZE; i++) - tag_table[i] = NULL_TAG; - - for (i = s->count; i-- > 0; ) - tag_table[targets[i].t] = i; } @@ -127,18 +108,13 @@ static void matched(int f, struct sum_struct *s, struct map_struct *buf, sum_update(map_ptr(buf, last_match + j, n1), n1); } - if (i >= 0) last_match = offset + s->sums[i].len; else last_match = offset; - if (buf && do_progress) { + if (buf && do_progress) show_progress(last_match, buf->file_size); - - if (i == -1) - end_progress(buf->file_size); - } } @@ -181,20 +157,22 @@ static void hash_search(int f,struct sum_struct *s, } do { - tag t = gettag2(s1,s2); int done_csum2 = 0; - int32 j = tag_table[t]; + int32 i; - if (verbose > 4) - rprintf(FINFO,"offset=%.0f sum=%08x\n",(double)offset,sum); + if (verbose > 4) { + rprintf(FINFO, "offset=%.0f sum=%04x%04x\n", + (double)offset, s2 & 0xFFFF, s1 & 0xFFFF); + } - if (j == NULL_TAG) - goto null_tag; + i = hash_table[SUM2HASH2(s1,s2)]; + if (i < 0) + goto null_hash; sum = (s1 & 0xffff) | (s2 << 16); - tag_hits++; + hash_hits++; do { - int32 l, i = targets[j].i; + int32 l; if (sum != s->sums[i].sum1) continue; @@ -210,9 +188,11 @@ static void hash_search(int f,struct sum_struct *s, && !(s->sums[i].flags & SUMFLG_SAME_OFFSET)) continue; - if (verbose > 3) - rprintf(FINFO,"potential match at %.0f target=%.0f %.0f sum=%08x\n", - (double)offset,(double)j,(double)i,sum); + if (verbose > 3) { + rprintf(FINFO, + "potential match at %.0f i=%ld sum=%08x\n", + (double)offset, (long)i, sum); + } if (!done_csum2) { map = (schar *)map_ptr(buf,offset,l); @@ -229,8 +209,8 @@ static void hash_search(int f,struct sum_struct *s, * one with an identical offset, so we prefer that over * the following want_i optimization. */ if (updating_basis_file) { - do { - int32 i2 = targets[j].i; + int32 i2; + for (i2 = i; i2 >= 0; i2 = s->sums[i2].chain) { if (s->sums[i2].offset != offset) continue; if (i2 != i) { @@ -245,7 +225,7 @@ static void hash_search(int f,struct sum_struct *s, * both the sender and the receiver. */ s->sums[i].flags |= SUMFLG_SAME_OFFSET; goto set_want_i; - } while (++j < s->count && targets[j].t == t); + } } /* we've found a match, but now check to see @@ -271,9 +251,9 @@ static void hash_search(int f,struct sum_struct *s, s2 = sum >> 16; matches++; break; - } while (++j < s->count && targets[j].t == t); + } while ((i = s->sums[i].chain) >= 0); - null_tag: + null_hash: backup = offset - last_match; /* We sometimes read 1 byte prior to last_match... */ if (backup < 0) @@ -328,12 +308,31 @@ void match_sums(int f, struct sum_struct *s, struct map_struct *buf, OFF_T len) last_match = 0; false_alarms = 0; - tag_hits = 0; + hash_hits = 0; matches = 0; data_transfer = 0; sum_init(checksum_seed); + if (append_mode) { + OFF_T j = 0; + for (j = CHUNK_SIZE; j < s->flength; j += CHUNK_SIZE) { + if (buf && do_progress) + show_progress(last_match, buf->file_size); + sum_update(map_ptr(buf, last_match, CHUNK_SIZE), + CHUNK_SIZE); + last_match = j; + } + if (last_match < s->flength) { + int32 len = s->flength - last_match; + if (buf && do_progress) + show_progress(last_match, buf->file_size); + sum_update(map_ptr(buf, last_match, len), len); + last_match = s->flength; + } + s->count = 0; + } + if (len > 0 && s->count > 0) { build_hash_table(s); @@ -347,7 +346,7 @@ void match_sums(int f, struct sum_struct *s, struct map_struct *buf, OFF_T len) } else { OFF_T j; /* by doing this in pieces we avoid too many seeks */ - for (j = CHUNK_SIZE; j < len; j += CHUNK_SIZE) + for (j = last_match + CHUNK_SIZE; j < len; j += CHUNK_SIZE) matched(f, s, buf, j, -2); matched(f, s, buf, len, -1); } @@ -361,16 +360,11 @@ void match_sums(int f, struct sum_struct *s, struct map_struct *buf, OFF_T len) rprintf(FINFO,"sending file_sum\n"); write_buf(f,file_sum,MD4_SUM_LENGTH); - if (targets) { - free(targets); - targets=NULL; - } - if (verbose > 2) - rprintf(FINFO, "false_alarms=%d tag_hits=%d matches=%d\n", - false_alarms, tag_hits, matches); + rprintf(FINFO, "false_alarms=%d hash_hits=%d matches=%d\n", + false_alarms, hash_hits, matches); - total_tag_hits += tag_hits; + total_hash_hits += hash_hits; total_false_alarms += false_alarms; total_matches += matches; stats.literal_data += data_transfer; @@ -382,8 +376,7 @@ void match_report(void) return; rprintf(FINFO, - "total: matches=%d tag_hits=%d false_alarms=%d data=%.0f\n", - total_matches,total_tag_hits, - total_false_alarms, + "total: matches=%d hash_hits=%d false_alarms=%d data=%.0f\n", + total_matches, total_hash_hits, total_false_alarms, (double)stats.literal_data); }