2 * UCW Library -- Universal Sorter: Common Declarations
4 * (c) 2007 Martin Mares <mj@ucw.cz>
6 * This software may be freely distributed and used according to the terms
7 * of the GNU Lesser General Public License.
10 #ifndef _UCW_SORTER_COMMON_H
11 #define _UCW_SORTER_COMMON_H
13 #include "lib/clists.h"
15 /* Configuration, some of the variables are used by the old sorter, too. */
16 extern uns sorter_trace, sorter_presort_bufsize, sorter_stream_bufsize;
17 extern uns sorter_debug, sorter_min_radix_bits, sorter_max_radix_bits;
18 extern uns sorter_min_multiway_bits, sorter_max_multiway_bits;
19 extern u64 sorter_bufsize;
20 extern struct fb_params sorter_fb_params;
22 #define SORT_TRACE(x...) do { if (sorter_trace) msg(L_DEBUG, x); } while(0)
23 #define SORT_XTRACE(level, x...) do { if (sorter_trace >= level) msg(L_DEBUG, x); } while(0)
26 SORT_DEBUG_NO_PRESORT = 1,
27 SORT_DEBUG_NO_JOIN = 2,
28 SORT_DEBUG_KEEP_BUCKETS = 4,
29 SORT_DEBUG_NO_RADIX = 8,
30 SORT_DEBUG_NO_MULTIWAY = 16,
36 struct fastbuf *in_fb;
37 struct fastbuf *out_fb;
46 int (*custom_presort)(struct fastbuf *dest, void *buf, size_t bufsize);
48 // Take as much as possible from the source bucket, sort it in memory and dump to destination bucket.
49 // Return 1 if there is more data available in the source bucket.
50 int (*internal_sort)(struct sort_context *ctx, struct sort_bucket *in, struct sort_bucket *out, struct sort_bucket *out_only);
52 // Estimate how much input data from `b' will fit in the internal sorting buffer.
53 u64 (*internal_estimate)(struct sort_context *ctx, struct sort_bucket *b);
55 // Two-way split/merge: merge up to 2 source buckets to up to 2 destination buckets.
56 // Bucket arrays are NULL-terminated.
57 void (*twoway_merge)(struct sort_context *ctx, struct sort_bucket **ins, struct sort_bucket **outs);
59 // Multi-way merge: merge an arbitrary number of source buckets to a single destination bucket.
60 void (*multiway_merge)(struct sort_context *ctx, struct sort_bucket **ins, struct sort_bucket *out);
62 // Radix split according to hash function
63 void (*radix_split)(struct sort_context *ctx, struct sort_bucket *in, struct sort_bucket **outs, uns bitpos, uns numbits);
65 // State variables of internal_sort
70 timestamp_t start_time;
72 uns total_int_time, total_pre_time, total_ext_time;
75 void sorter_run(struct sort_context *ctx);
79 void *sorter_alloc(struct sort_context *ctx, uns size);
80 void sorter_prepare_buf(struct sort_context *ctx);
81 void sorter_alloc_buf(struct sort_context *ctx);
82 void sorter_free_buf(struct sort_context *ctx);
88 struct sort_context *ctx;
92 u64 size; // Size in bytes (not valid when writing)
93 uns runs; // Number of runs, 0 if not sorted
94 uns hash_bits; // Remaining bits of the hash function
95 byte *ident; // Identifier used in debug messages
98 enum sort_bucket_flags {
99 SBF_FINAL = 1, // This bucket corresponds to the final output file (always 1 run)
100 SBF_SOURCE = 2, // Contains the source file (always 0 runs)
101 SBF_CUSTOM_PRESORT = 4, // Contains source to read via custom presorter
102 SBF_OPEN_WRITE = 256, // We are currently writing to the fastbuf
103 SBF_OPEN_READ = 512, // We are reading from the fastbuf
104 SBF_DESTROYED = 1024, // Already done with, no further references allowed
105 SBF_SWAPPED_OUT = 2048, // Swapped out to a named file
108 struct sort_bucket *sbuck_new(struct sort_context *ctx);
109 void sbuck_drop(struct sort_bucket *b);
110 int sbuck_have(struct sort_bucket *b);
111 int sbuck_has_file(struct sort_bucket *b);
112 sh_off_t sbuck_size(struct sort_bucket *b);
113 struct fastbuf *sbuck_read(struct sort_bucket *b);
114 struct fastbuf *sbuck_write(struct sort_bucket *b);
115 void sbuck_swap_out(struct sort_bucket *b);