git.sesse.net Git - remoteglot-book/blob - binloader.cpp

   1 //#define _GLIBCXX_PARALLEL
   2
   3 // Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS NUM_POS_PER_SUBSHARD
   4
   5 #include <stdio.h>
   6 #include <vector>
   7 #include <mtbl.h>
   8 #include <algorithm>
   9 #include <utility>
  10 #include <memory>
  11 #include <string>
  12 #include <unordered_set>
  13 #include <string.h>
  14 #include "count.pb.h"
  15 #include "arena.h"
  16 #include "hash.h"
  17
  18 #define DUMMY_TIMESTAMP 32503680000
  19
  20 using namespace std;
  21
  22 static inline int memcmp_different_len(const void *s1, size_t n1, const void *s2, size_t n2)
  23 {
  24         size_t shared_len = min(n1, n2);
  25         if (shared_len >= 8) {
  26                 uint64_t a1 = *(const uint64_t *)s1;
  27                 uint64_t a2 = *(const uint64_t *)s2;
  28                 if (a1 != a2) {
  29                         a1 = __builtin_bswap64(a1);
  30                         a2 = __builtin_bswap64(a2);
  31                         return (a1 < a2) ? -1 : 1;
  32                 }
  33         }
  34
  35         int s = memcmp(s1, s2, shared_len);
  36         if (s != 0) {
  37                 return s;
  38         }
  39
  40         return n2 - n1;
  41 }
  42
  43 enum Result { WHITE = 0, DRAW, BLACK };
  44 struct Element {
  45         char *bpfen;  // includes prev_board_hash
  46         int bpfen_len;
  47         char move[8];   // Na1xc3+
  48         Result result;
  49         int opening_num, white_elo, black_elo;
  50         int file_num;
  51         time_t timestamp;
  52         long start_position;
  53
  54         bool operator< (const Element& other) const {
  55                 int s = memcmp_different_len(bpfen, bpfen_len, other.bpfen, other.bpfen_len);
  56                 return s < 0;
  57         }
  58 };
  59
  60 struct ShardData {
  61         vector<Element> elems;
  62         unique_ptr<Arena> arena;  // Used to allocate bpfen.
  63         int num_written_subshards = 0;
  64 };
  65
  66 void write_subshard(const char *basename, ShardData* shard, int bucket)
  67 {
  68         string buf;  // Keep allocated.
  69         char filename[256];
  70         snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
  71                 basename, bucket, shard->num_written_subshards++);
  72         printf("Writing SSTable %s...\n", filename);
  73
  74         sort(shard->elems.begin(), shard->elems.end());
  75
  76         mtbl_writer_options* wopt = mtbl_writer_options_init();
  77         mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
  78         mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
  79         Count c;
  80         unordered_set<string> moves;
  81         for (size_t i = 0; i < shard->elems.size(); ++i) {
  82                 const Element &e = shard->elems[i];
  83                 if (e.result == WHITE) {
  84                         c.set_white(c.white() + 1);
  85                 } else if (e.result == DRAW) {
  86                         c.set_draw(c.draw() + 1);
  87                 } else if (e.result == BLACK) {
  88                         c.set_black(c.black() + 1);
  89                 }
  90                 if (e.white_elo >= 100 && e.black_elo >= 100) {
  91                         c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
  92                         c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
  93                         c.set_num_elo(c.num_elo() + 1);
  94                 }
  95                 if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
  96                         if (e.timestamp != DUMMY_TIMESTAMP) {
  97                                 c.set_first_timestamp(e.timestamp);
  98                         }
  99                         c.set_opening_num(e.opening_num);
 100                         c.set_pgn_file_num(e.file_num);
 101                         c.set_pgn_start_position(e.start_position);
 102                 }
 103                 if (!moves.count(e.move)) {
 104                         moves.insert(e.move);
 105                         c.add_move(e.move);
 106                 }
 107                 if (i == shard->elems.size() - 1 ||
 108                     e.bpfen_len != shard->elems[i + 1].bpfen_len ||
 109                     memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
 110                         c.SerializeToString(&buf);
 111                         mtbl_res res = mtbl_writer_add(mtbl,
 112                                 (const uint8_t *)e.bpfen, e.bpfen_len,
 113                                 (const uint8_t *)buf.data(), buf.size());
 114                         assert(res == mtbl_res_success);
 115                         c = Count();
 116                         moves.clear();
 117                 }
 118         }
 119         mtbl_writer_destroy(&mtbl);
 120
 121         shard->elems.clear();
 122         shard->arena.reset(new Arena);
 123 }
 124
 125 int main(int argc, char **argv)
 126 {
 127         int num_buckets = atoi(argv[argc - 2]);
 128         size_t num_pos_per_subshard = atoi(argv[argc - 1]);  // 500000 is a reasonable value.
 129
 130         vector<ShardData> shards;
 131         shards.resize(num_buckets);
 132
 133         for (int i = 0; i < num_buckets; ++i) {
 134                 shards[i].elems.reserve(num_pos_per_subshard);
 135                 shards[i].arena.reset(new Arena);
 136         }
 137
 138         size_t num_elems = 0;
 139         for (int i = 1; i < argc - 3; ++i) {
 140                 FILE *fp;
 141                 if (strcmp(argv[i], "-") == 0) {
 142                         fp = stdin;
 143                 } else {
 144                         fp = fopen(argv[i], "rb");
 145                         if (fp == NULL) {
 146                                 perror(argv[i]);
 147                                 exit(1);
 148                         }
 149                 }
 150                 for ( ;; ) {
 151                         char bpfen[256];
 152
 153                         int bpfen_len = getc(fp);
 154                         if (bpfen_len == -1) {
 155                                 break;
 156                         }
 157                         if (bpfen_len >= int(sizeof(bpfen))) {
 158                                 fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
 159                 //              exit(1);
 160                                 break;
 161                         }
 162                         if (fread(bpfen, bpfen_len, 1, fp) != 1) {
 163                                 perror("fread()");
 164                 //              exit(1);
 165                                 break;
 166                         }
 167
 168                         int r = getc(fp);
 169                         if (r == -1) {
 170                                 perror("getc()");
 171                                 //exit(1);
 172                                 break;
 173                         }
 174
 175                         int opening_num, white_elo, black_elo, file_num;
 176                         time_t timestamp;
 177                         long start_position;
 178                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
 179                                 perror("fread()");
 180                                 //exit(1);
 181                                 break;
 182                         }
 183                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
 184                                 perror("fread()");
 185                                 //exit(1);
 186                                 break;
 187                         }
 188                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
 189                                 perror("fread()");
 190                                 //exit(1);
 191                                 break;
 192                         }
 193                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
 194                                 perror("fread()");
 195                                 //exit(1);
 196                                 break;
 197                         }
 198                         if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
 199                                 perror("fread()");
 200                                 //exit(1);
 201                                 break;
 202                         }
 203                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
 204                                 perror("fread()");
 205                                 //exit(1);
 206                                 break;
 207                         }
 208
 209
 210                         char move[8];
 211                         int l = getc(fp);
 212                         if (l == -1) {
 213                                 break;
 214                         }
 215                         if (l >= int(sizeof(move))) {
 216                                 fprintf(stderr, "Overlong move (%d bytes)\n", l);
 217                 //              exit(1);
 218                                 break;
 219                         }
 220                         if (fread(&move[0], l, 1, fp) != 1) {
 221                                 perror("fread()");
 222                 //              exit(1);
 223                                 break;
 224                         }
 225                         move[l] = 0;
 226
 227                         int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
 228                         Element e;
 229                         e.bpfen = shards[bucket].arena->alloc(bpfen_len);
 230                         memcpy(e.bpfen, bpfen, bpfen_len);
 231                         e.bpfen_len = bpfen_len;
 232                         strcpy(e.move, move);
 233                         e.result = Result(r);
 234                         e.opening_num = opening_num;
 235                         e.white_elo = white_elo;
 236                         e.black_elo = black_elo;
 237                         e.file_num = file_num;
 238                         e.timestamp = timestamp;
 239                         e.start_position = start_position;
 240                         shards[bucket].elems.push_back(e);
 241                         ++num_elems;
 242
 243                         if (shards[bucket].elems.size() >= num_pos_per_subshard) {
 244                                 write_subshard(argv[argc - 3], &shards[bucket], bucket);
 245                                 shards[bucket].elems.reserve(num_pos_per_subshard);
 246                         }
 247                 }
 248                 fclose(fp);
 249
 250                 printf("Read %ld elems\n", num_elems);
 251         }
 252
 253         for (int i = 0; i < num_buckets; ++i) {
 254                 write_subshard(argv[argc - 3], &shards[i], i);
 255         }
 256 }
 257