X-Git-Url: https://mattmccutchen.net/rsync/rsync.git/blobdiff_plain/a16bbc3990cc3c9ef1228508246f9f6840d9ffa2..c95da96a0c51c66c8cb2eff97b768a717d9e0c79:/match.c diff --git a/match.c b/match.c index 7f6e7088..272025aa 100644 --- a/match.c +++ b/match.c @@ -29,18 +29,18 @@ extern int remote_version; typedef unsigned short tag; #define TABLESIZE (1<<16) -#define NULL_TAG ((tag)-1) +#define NULL_TAG (-1) static int false_alarms; static int tag_hits; static int matches; -static int data_transfer; +static int64 data_transfer; static int total_false_alarms; static int total_tag_hits; static int total_matches; -static int64 total_data_transfer; +extern struct stats stats; struct target { tag t; @@ -49,7 +49,7 @@ struct target { static struct target *targets; -static tag *tag_table; +static int *tag_table; #define gettag2(s1,s2) (((s1) + (s2)) & 0xFFFF) #define gettag(sum) gettag2((sum)&0xFFFF,(sum)>>16) @@ -65,7 +65,7 @@ static void build_hash_table(struct sum_struct *s) int i; if (!tag_table) - tag_table = (tag *)malloc(sizeof(tag)*TABLESIZE); + tag_table = (int *)malloc(sizeof(tag_table[0])*TABLESIZE); targets = (struct target *)malloc(sizeof(targets[0])*s->count); if (!tag_table || !targets) @@ -87,27 +87,26 @@ static void build_hash_table(struct sum_struct *s) } -static off_t last_match; +static OFF_T last_match; static void matched(int f,struct sum_struct *s,struct map_struct *buf, - off_t offset,int i) + OFF_T offset,int i) { - off_t n = offset - last_match; + OFF_T n = offset - last_match; int j; if (verbose > 2 && i >= 0) - fprintf(FINFO,"match at %d last_match=%d j=%d len=%d n=%d\n", + rprintf(FINFO,"match at %d last_match=%d j=%d len=%d n=%d\n", (int)offset,(int)last_match,i,(int)s->sums[i].len,(int)n); - send_token(f,i,buf,last_match,n,i==-1?0:s->sums[i].len); + send_token(f,i,buf,last_match,n,i<0?0:s->sums[i].len); data_transfer += n; - if (n > 0) - write_flush(f); - - if (i >= 0) + if (i >= 0) { + stats.matched_data += s->sums[i].len; n += s->sums[i].len; + } for (j=0;j 2) - fprintf(FINFO,"hash search b=%d len=%d\n",s->n,(int)len); + rprintf(FINFO,"hash search b=%d len=%d\n",s->n,(int)len); k = MIN(len, s->n); - map = (signed char *)map_ptr(buf,0,k); + map = (schar *)map_ptr(buf,0,k); sum = get_checksum1((char *)map, k); s1 = sum & 0xFFFF; s2 = sum >> 16; if (verbose > 3) - fprintf(FINFO, "sum=%.8x k=%d\n", sum, k); + rprintf(FINFO, "sum=%.8x k=%d\n", sum, k); offset = 0; end = len + 1 - s->sums[s->count-1].len; if (verbose > 3) - fprintf(FINFO,"hash search s->n=%d len=%d count=%d\n", + rprintf(FINFO,"hash search s->n=%d len=%d count=%d\n", s->n,(int)len,s->count); do { @@ -159,7 +158,7 @@ static void hash_search(int f,struct sum_struct *s, j = tag_table[t]; if (verbose > 4) - fprintf(FINFO,"offset=%d sum=%08x\n",(int)offset,sum); + rprintf(FINFO,"offset=%d sum=%08x\n",(int)offset,sum); if (j == NULL_TAG) { goto null_tag; @@ -173,12 +172,12 @@ static void hash_search(int f,struct sum_struct *s, if (sum != s->sums[i].sum1) continue; if (verbose > 3) - fprintf(FINFO,"potential match at %d target=%d %d sum=%08x\n", + rprintf(FINFO,"potential match at %d target=%d %d sum=%08x\n", (int)offset,j,i,sum); if (!done_csum2) { int l = MIN(s->n,len-offset); - map = (signed char *)map_ptr(buf,offset,l); + map = (schar *)map_ptr(buf,offset,l); get_checksum2((char *)map,l,sum2); done_csum2 = 1; } @@ -191,7 +190,7 @@ static void hash_search(int f,struct sum_struct *s, matched(f,s,buf,offset,i); offset += s->sums[i].len - 1; k = MIN((len-offset), s->n); - map = (signed char *)map_ptr(buf,offset,k); + map = (schar *)map_ptr(buf,offset,k); sum = get_checksum1((char *)map, k); s1 = sum & 0xFFFF; s2 = sum >> 16; @@ -201,7 +200,7 @@ static void hash_search(int f,struct sum_struct *s, null_tag: /* Trim off the first byte from the checksum */ - map = (signed char *)map_ptr(buf,offset,k+1); + map = (schar *)map_ptr(buf,offset,k+1); s1 -= map[0] + CHAR_OFFSET; s2 -= k * (map[0]+CHAR_OFFSET); @@ -212,7 +211,17 @@ static void hash_search(int f,struct sum_struct *s, } else { --k; } - + + /* By matching early we avoid re-reading the + data 3 times in the case where a token + match comes a long way after last + match. The 3 reads are caused by the + running match, the checksum update and the + literal send. */ + if (offset-last_match >= CHUNK_SIZE+s->n && + (end-offset > CHUNK_SIZE)) { + matched(f,s,buf,offset - s->n, -2); + } } while (++offset < end); matched(f,s,buf,len,-1); @@ -220,7 +229,7 @@ static void hash_search(int f,struct sum_struct *s, } -void match_sums(int f,struct sum_struct *s,struct map_struct *buf,off_t len) +void match_sums(int f,struct sum_struct *s,struct map_struct *buf,OFF_T len) { char file_sum[MD4_SUM_LENGTH]; @@ -236,12 +245,12 @@ void match_sums(int f,struct sum_struct *s,struct map_struct *buf,off_t len) build_hash_table(s); if (verbose > 2) - fprintf(FINFO,"built hash table\n"); + rprintf(FINFO,"built hash table\n"); hash_search(f,s,buf,len); if (verbose > 2) - fprintf(FINFO,"done hash search\n"); + rprintf(FINFO,"done hash search\n"); } else { matched(f,s,buf,len,-1); } @@ -250,7 +259,7 @@ void match_sums(int f,struct sum_struct *s,struct map_struct *buf,off_t len) if (remote_version >= 14) { if (verbose > 2) - fprintf(FINFO,"sending file_sum\n"); + rprintf(FINFO,"sending file_sum\n"); write_buf(f,file_sum,MD4_SUM_LENGTH); } @@ -260,13 +269,13 @@ void match_sums(int f,struct sum_struct *s,struct map_struct *buf,off_t len) } if (verbose > 2) - fprintf(FINFO, "false_alarms=%d tag_hits=%d matches=%d\n", + rprintf(FINFO, "false_alarms=%d tag_hits=%d matches=%d\n", false_alarms, tag_hits, matches); total_tag_hits += tag_hits; total_false_alarms += false_alarms; total_matches += matches; - total_data_transfer += data_transfer; + stats.literal_data += data_transfer; } void match_report(void) @@ -274,8 +283,9 @@ void match_report(void) if (verbose <= 1) return; - fprintf(FINFO, - "total: matches=%d tag_hits=%d false_alarms=%d data=%ld\n", + rprintf(FINFO, + "total: matches=%d tag_hits=%d false_alarms=%d data=%.0f\n", total_matches,total_tag_hits, - total_false_alarms,(long)total_data_transfer); + total_false_alarms, + (double)stats.literal_data); }